Thermal Management of Telecommunications Equipment

THERMAL MANAGEMENT
OF TELECOMMUNICATIONS
EQUIPMENT
THERMAL MANAGEMENT
OF TELECOMMUNICATIONS
EQUIPMENT
L. T. Yeh
R. C. Chu
ASME Press Book Series on Electronic Packaging

Series Editor Dereje Agonafer
© 2013, ASME, 2 Park Avenue, New York, NY 10016, USA (www.asme.org)
All rights reserved. Printed in the United States of America. Except as permitted
under the United States Copyright Act of 1976, no part of this publication may be
reproduced or distributed in any form or by any means, or stored in a database or
retrieval system, without the prior written permission of the publisher.
Information contained in this work has been obtained by the American Society
of Mechanical Engineers from sources believed to be reliable. However, neither
ASME nor its authors or editors guarantee the accuracy or completeness of any
information published in this work. Neither ASME nor its authors and editors shall
be responsible for any errors, omissions, or damages arising out of the use of this
information. The work is published with the understanding that ASME and its
authors and editors are supplying information but are not attempting to render
engineering or other professional services. If such engineering or professional ser-
vices are required, the assistance of an appropriate professional should be sought.
ASME shall not be responsible for statements or opinions advanced in papers

or . . . printed in its publications (B7.1.3). Statement from the Bylaws.
For authorization to photocopy material for internal or personal use under those
circumstances not falling within the fair use provisions of the Copyright Act, con-
tact the Copyright Clearance Center (CCC), 222 Rosewood Drive, Danvers, MA
01923, tel: 978-750-8400, www.copyright.com.
Requests for special permission or bulk reproduction should be addressed to the

ASME Publishing Department, or submitted online at: http://www.asme.org/shop/
books/book-proposals/permissions
ASME Press books are available at special quantity discounts to use as premi-
ums or for use in corporate training programs. For more information, contact
Special Sales at customercare@asme.org
Library of Congress Cataloging-in-Publication Data
Yeh, L.-T. (Lian-Tuu), 1944-

Thermal management of telecommunications equipment / L.T. Yeh, Ph.D., P.E.,
R.C. Chu. pages cm
Includes bibliographical references and index.
ISBN 978-0-7918-6020-5 (alk. paper)
1. Electronic apparatus and appliances--Thermal properties. 2. Telecommunica-
tion--Equipment and supplies--Cooling. 3. Telecommunication--Equipment and
supplies--Protection. 4. Heat--Transmission. I. Chu, R. C. (Richard C.), 1933- II.
Title.
TK7870.25.Y443 2013
621.382--dc23
2013027405
Table Of Contents
List of Figures ix
Foreword xv
Preface xvii
Chapter 1 Introduction 1
Chapter 2 Component and Printed Circuit Board 5

2.1 COMPONENT.................................................................................... 5
2.1.1 Chip Packaging Technology.............................................................5
2.1.2 Thermal Management......................................................................7
2.1.3 Advanced Packaging Technology..................................................11
2.2 PRINTED CIRCUIT BOARD.............................................................. 14
2.2.1 Construction....................................................................................14
2.2.2 Equivalent Thermal Conductivity...................................................15
2.2.2.1 Planar Thermal Conductivity, Kxy. ..................................15
2.2.2.2 Normal Thermal Conductivity, Kz...................................17
2.2.3 Thermal Consideration...................................................................19
Chapter 3 Fans and Air Handling Systems 21

3.1 FAN TYPES�� 21
3.2 FAN PERFORMANCE AND FAN LAWS�� 22
3.3 FAN INLET AND EXIT CONDITIONS�� 23
vi • Table of Contents
3.4 FAN AND SYSTEM PRESSURE�� 26

3.5 FAN ACOUSTIC NOISE�� 26
3.6 FAN RELIABILITY�� 27
Chapter 4 Air Cooling and Heat Sinks 31

4.1 NATURAL CONVECTION�� 31
4.2 FORCED CONVECTION�� 35
4.2.1 Heat Sink Thermal Performance��35
4.2.1.1 Average Heat Sink Thermal Resistance��36
4.2.1.2 Base Spreading Thermal Resistance��36
4.2.2 Factors Affecting Heat Sink Thermal Performance��38
4.2.2.1 Flow Bypass and Leakage��38
4.2.2.2 Fin Configurations��39
4.2.2.3 Mass Flow Rates��44
4.2.3 Heat Sink Thermal Enhancement��44
4.2.3.1 Improved Fin Thermal Performance, Ro��45
4.2.3.2 Enhanced Base Thermal Spreading, Rc��45
4.2.4 Materials and Manufacture Processes��49
Chapter 5 System Thermal Analysis—Rack/Cabinet 53

(Part I)
5.1 KEY ELEMENTS IN SYSTEMS�� 53
5.1.1 EMI Plates��53
5.1.2 Air Filters��55
5.1.3 Fans��57
5.2 SYSTEM LEVEL THERMAL ANALYSIS�� 58
5.3 PRACTICAL EXAMPLES�� 60
5.4 THERMAL DESIGN PROCESS�� 68
Chapter 6 System Thermal Analysis—Rack/Cabinet 71

(Part II)
6.1 EFFECTS OF SHELF CONFIGURATION ON SHELF FLOW�� 71
6.1.1 Effects of Shelf Height��72
6.1.2 Effects of Shelf Depth��75
6.1.3 Design Optimization Procedure��77
6.1.4 Accuracy of Results��80
Table of Contents • vii
6.2 VARIOUS SUBJECTS RELATED TO SYSTEM THERMAL DESIGN�� 86

6.2.1 Effects of Series Fan Trays��86
6.2.2 Factors Affecting Component Temperature��88
6.2.2.1 Thermal Interface Material (TIM)��88
6.2.2.2 Thermal Conductivity of Printed Circuit Board��92
6.2.2.3 Baffles in Inlet/Exit Section��92
6.2.2.4 Fillers in Card Cage��93
6.3 THERMAL ENHANCEMENT�� 98
6.4 THERMAL CONSIDERATION IN DATA CENTER OR
CENTRAL OFFICE�� 100
Chapter 7 System Thermal Analysis—Small Box 103

7.1 ACTIVE COOLED BOXES�� 103
7.2 PASSIVE COOLED BOXES�� 105
7.2.1 Market Survey��106
7.2.2 Case Studies��108
7.2.3 Other Applications��110
7.2.3.1 Table Top Deployment��110
7.2.3.2 Boxes Stack-Up��111
Chapter 8 Outdoor Electronic Equipment 115

8.1 GROUND MOUNTED OUTDOOR ELECTRONIC EQUIPMENT�� 115
8.1.1 Thermal Designs of Outdoor Electronic Equipment��115
8.1.2 Thermal Design of Outside Plant (OSP)��121
8.1.2.1 Conventional Heat Exchanger OSP��122
8.1.2.2 Geo-Thermal Heat Exchanger OSP��123
8.1.2.3 Direct Air Cooled OSP��126
8.1.2.4 Natural Convection OSP��128
8.2 TOWER/POLE MOUNTED OUTDOOR
ELECTRONIC EQUIPMENT�� 130
8.2.1 Thermal Design of Remote Radio Head��132
Chapter 9 Liquid Cooled Systems 141

9.1 CONVECTION AND BOILING HEAT TRANSFER�� 141
9.1.1 Single-Phase Forced Convection��141
9.1.2 Boiling Heat Transfer��144
viii • Table of Contents
9.2 COOLANT SELECTION�� 146

9.3 LIQUID COOLING TECHNOLOGIES�� 149
9.3.1 Direct Immersion Cooling��149
9.3.2 Indirect Liquid Cooling��152
9.4 INDUSTRY TRENDS IN LIQUID COOLED ELECTRONIC
EQUIPMENT�� 161
9.5 SYSTEM THERMAL DESIGN�� 168
9.5.1 Direct Liquid Cooling��168
9.5.2 Indirect Liquid Cooling��170
9.5.3 General Design Considerations��171
Chapter 10 Thermal Management and

Energy Efficiency 177
10.1 POWER DENSITY OF TELECOMMUNICATION
EQUIPMENT�� 177
10.2 CRITICAL THERMAL ISSUES�� 180
10.2.1 Equipment Energy Efficiency��180
10.2.2 Equipment Acoustic Noise��180
10.3 THERMAL MANAGEMENT AND ROADMAP�� 181
10.3.1 Chip Level Thermal Management��182
10.3.2 System Level Thermal Management��187
10.4 THERMAL DESIGN OF LIQUID COOLED
TELECOM EQUIPMENT�� 191
10.5 ENERGY EFFICIENCY�� 196
10.6 ALTERNATIVE ENERGY�� 198
10.6.1 Thermoelectric Power Generations��198
10.6.2 Fuel Cell Technology for Power Generation��201
10.7 FUTURE RESEARCH NEEDS�� 202
10.8 SUMMARY AND CONCLUSION�� 211
Index 215
About the Authors 225

LIST OF FIGURES
1.1 Typical telecommunication cabinet and box configuration........................2

1.2 Typical outside plant and remote radio head..............................................3
2.1 Typical flip chip package with heat sink......................................................6
2.2 Common solder bumping technologies.......................................................6
2.3 Thermal representation of flip chip and heat sink......................................7
2.4 Thermal conductivity and CTE of various materials..................................9
2.5 Cross-section of package in wind tunnel...................................................10
2.6 Schematic diagram of various thermal enhancements................................... 10
2.7 Heat conduction spreading in die..............................................................12
2.8 IBM 3D packages.........................................................................................12
2.9 Cross-section of CFD thermal model.........................................................13
2.10 Simulation and measured data for four hot spots in top layer....................... 13
2.11 Typical construction of a PCB....................................................................14
2.12 Simplified board construction for thermal model....................................16
3.1 Method of obtaining fan performance curves...........................................22
3.2 Typical fan curve and system operation point...........................................23
3.3 Effect of blade location on fan performance.............................................24
3.4 Effect of blade location on fan performance with fan adjacent
to restricted area.........................................................................................24
3.5 Fan performance curves for various plenum depth ratio (P/D)..................... 25
3.6 Causes of fan failure....................................................................................28
4.1 A vertical straight fin heat sink...................................................................32
4.2 Natural convection loss for heat sink width of 40 mm.............................33
4.5 Typical thermal performance of a heat sink..............................................35
4.6 Schematic view of local resistance or temperature
variation of heat sink..................................................................................37
4.7 Normalized local temperature rise with various size of heat
sources.........................................................................................................38
4.8 Effect of heat source location on heat sink thermal
resistance.....................................................................................................38
x • List of Figures
4.9 Typical straight fin heat sink.......................................................................39

4.10 Flow bypass over heat sinks.......................................................................39
4.11 Heat sink flow leakage........................................................................................ 39
4.12 Various types of heat sinks.................................................................................. 40
4.13 Configuration of single fin of cell fin heat sink................................................. 41
4.14 Channel cross section with flow bypass............................................................ 41
4.15 Average heat transfer coefficient of heat sink without flow bypass................ 41
4.16 Average thermal resistance of heat sink without flow bypass......................... 42
4.17 Average heat transfer coefficient of heat sink with flow bypass..................... 43
4.18 Average thermal resistance of heat sink with flow bypass.............................. 43
4.19 Thermal resistance for pin fin and discrete continuous heat sinks................ 44
4.20 Various types of vapor chamber heat sinks...................................................... 47
4.21 Measured heat sink thermal resistance.........................................................48
5.1 A typical telecom rack and sketch of shelf.................................................54
5.2 Loss coefficient versus fraction of opening for a thin plate......................55
5.3 Typical honeycomb structures....................................................................55
5.4a Performance curve of a clean air filter.............................................................. 56
5.4b Pressure drop versus air flow of typical air filters.....................................56
5.5 Sketch of typical rack and its flow resistance............................................59
5.6 Telecommunication rack with two identical shelves....................................... 60
5.7 Sketch of shelf.............................................................................................61
5.8 Odd-shaped opening at upper section of front door........................................ 62
5.9 System thermal model................................................................................62
5.10 Fans locations...................................................................................................... 63
5.11a Card cage sketch and board flow rates for front card cage............................. 63
5.11b Card cage sketch and board flow rates for rear card cage.............................. 64
5.12 Sketch of shelf under consideration.................................................................. 65
5.13 Air flow rates versus filter locations with four fans in operation.................... 65
5.14 Air flow rates versus filter locations for one fan failure................................... 66
5.15 Examples of fan damper/cover........................................................................... 67
5.16 Telecommunication rack thermal design process.....................................68
6.1 Sketch of typical rack and its flow resistance............................................72
6.2 Sketch of shelf with various inlet/exit height.............................................73
6.3 Inlet/exit height versus system flow resistance..........................................74
6.4 System flow rate ratios over inlet/exit height ratios..................................74
6.5 Sketch of shelf with various depth.............................................................75
6.6 Shelf depth versus shelf flow resistance.....................................................76
6.7 System flow rate ratios over shelf depth ratios.........................................76
6.8 Effect of inlet/exit section height on system flow rate..............................77
6.9 Effect of depth on system flow rate............................................................77
6.10 A telecommunication shelf under consideration......................................79
6.11 System under consideration............................................................................... 81
6.12 Inlet/exit section height versus system flow rate.............................................. 82
6.13 Comparison between graphical and CFD results for
R1278 = 0.596 Rt................................................................................................. 83
6.14 Inlet/exit section height versus system flow rate.............................................. 84
6.15 Comparison between graphical and CFD results
for R1278 = 0.237 Rt........................................................................................... 84
List of Figures • xi
6.16 Shelf flow ratios versus shelf depth ratios......................................................... 85

6.17 Comparison between graphical and CFD results for
R5 = 0.44 Rt.......................................................................................................... 86
6.18 Performance curves of fans in series and parallel............................................ 87
6.19 Fan trays in series................................................................................................ 87
6.20 Comparison of flow rate between Cases A and B............................................. 88
6.21 Sketch of heat sink and component.................................................................. 89
6.22 Measured and calculated thermal resistance of an epoxy joint...................... 90
6.23 Thermal pad joint thermal resistance versus pressure.................................... 90
6.24 Thermal resistance over various pad joints...................................................... 92
6.25 Sketch of shelf with baffles................................................................................. 93
6.26 System under consideration............................................................................... 94
6.27 Top view of fully filled card cage and three-board card cage.......................... 94
6.28 A passive cooled 4U shelf.................................................................................... 95
6.29 Results for fully heated boards in shelf............................................................. 96
6.30 Results for three-heated-board card cage with
zero power on fillers............................................................................................ 97
6.31 Results for three-heated-board card cage without fillers................................ 97
6.32 Modularized board.............................................................................................. 99
6.33 Large heat sink for multiple chips...................................................................100
6.34 Equipment layout in data centers....................................................................101
6.35 Hot exit air circulation..............................................................................101
7.1 Sketch of forced air cooling of box..........................................................104
7.2 1 U box with forced air cooling................................................................105
7.3 Vertical boards in low profile shelf...........................................................105
7.4 Cross-section view of vertical boards in low profile shelf..............................106
7.5 1 U box in telecommunication rack.........................................................107
7.6 Box internal layout....................................................................................108
7.7 Various heat sinks on board in box..........................................................109
7.8 Temperature distribution over top surface of box...................................109
7.9 Component temperatures on board in box..............................................110
7.10 Component temperatures for boxes with/without plate................................111
7.11 Component temperatures versus space between box and
table top.................................................................................................... 112
7.12 Component temperature in single and multiple
boxes stack-up on steel table top.....................................................................112
7.13 Component temperatures in single and three
boxes stack-up....................................................................................................113
7.14 Effects of gap size on component temperatures in boxes.............................114
8.1 Sketch of system under consideration.....................................................116
8.2 Internal and external air flow paths.........................................................117
8.3 Hot dry daily cycle of air temperature and solar flux.................................117
8.4 Internal air temperature history...............................................................118
8.5 Heat pipe heat exchanger.........................................................................119
8.6 Flow patterns inside enclosure.................................................................120
8.7 Schematic of double wall enclosure.........................................................120
8.8 External view of typical OSP cabinet.......................................................121
8.9 Basic thermal configurations in outdoor enclosure................................122
xii • List of Figures
8.10 CAD model of cabinet.......................................................................................123

8.11 Geothermal OSP................................................................................................124
8.12 Typical soil temperature profiles......................................................................125
8.13 Dust protection of various types of air filters..................................................127
8.14 Filtered vent thermal system concept..............................................................127
8.15 External natural convection for OSP cabinet.................................................128
8.16 Counter flow heat exchanger for OSP in passive cooling..............................129
8.17 Estimated cooling capacity of natural convection OSP................................130
8.18 Picture of tower mounted devices....................................................................132
8.19 Typical remote radio head................................................................................133
8.20 Various types of finned heat sinks....................................................................134
8.21 Active heat spreader heat sinks........................................................................134
8.22 Extruded heat sink thermal solutions.............................................................135
8.23 Basic thermosyphone models..........................................................................136
8.24 Passive cooling with closed-loop thermosypon..............................................137
8.25 A typical ONT with cover open........................................................................138
8.26 Basic configuration of ONT..........................................................................138
9.1 Prandtl number effect on boundary layers..............................................142
9.2 Liquid at constant pressure heating process...........................................144
9.3 Boiling curves for pool and flow boiling..................................................145
9.4 Pool boiling data for a plain tube.............................................................150
9.5 IBM liquid encapsulated module (LEM).................................................151
9.6 Boiling curve with DNC and DNB............................................................151
9.7 Sketch of a counter flow system for electronic cooling.................................152
9.8 Cold wall temperature distribution for unidirectional and
counter flow...............................................................................................153
9.9 Microchannel cooled silicon chip.............................................................154
9.10 Microchannel cold plate and round heater....................................................154
9.11 Temperatures of microchannel cold plate......................................................155
9.12 Swirl cold plate..................................................................................................156
9.13 Thermal performance of swirl cold.................................................................157
9.14 IBM microchannel cooled silicon chip...........................................................158
9.15 Thin rectangular offset fins...............................................................................158
9.16 Friction factor and Nusselt number for offset fins in
narrow passage...........................................................................................159
9.17 CHIC heat exchanger design concept..............................................................160
9.18 Laser diode/cooler stack....................................................................................161
9.19 IBM liquid cooled thermal conduction module.............................................162
9.20 Liquid cooling of bare chip with flexible heat pipe.................................162
9.21 Liquid cooling of bare chip with flexible heat pipe extension
into liquid flow...........................................................................................163
9.22 Fujitsu FACOM M-780 computer.....................................................................163
9.23 NEC SX-2 cooling system.................................................................................164
9.24 Liquid cooled clod plate for RF systems.........................................................165
9.25 IBM Power 575 system.....................................................................................165
9.26 Fujitsu High End Server (GS8900)..................................................................166
9.27 IBM Zurich liquid cooled board......................................................................167
9.28 Scheme of a microchannel heat sink...............................................................167
List of Figures • xiii
9.29A Passive direct liquid cooling concepts.............................................................169

9.29B Active direct liquid cooling concepts..............................................................169
9.30 Redundant liquid cooling system....................................................................172
9.31 Saturation pressure versus temperatures
for water and FC-72..........................................................................................172
9.32 Saturation pressure versus temperatures
for Freon-12 and ammonia..............................................................................173
9.33 Estimated flow rates required of dielectric
fluids for given system power...................................................................174
10.1 Cooling cost versus heat dissipation...............................................................178
10.2 Maximum chip power trends in telecommunication
applications........................................................................................................178
10.3 Maximum chip heat flux in telecommunication applications.....................179
10.4 Thermal loading at facility level for various products...................................179
10.5 Thermal technology map..................................................................................181
10.6 Thermal paths in chip package with heat sink..............................................183
10.7 Sketch of chip packaging and heat sink.........................................................183
10.8 Thermal conductivity and CTE of various materials....................................185
10.9 3D IC package (IBM Zurich Lab).............................................................185
10.10 SFP module/cage on board..............................................................................186
10.11 Thermal models for XFP..................................................................................186
10.12 Heat sink and SFP/XFP cage............................................................................187
10.13 Example of series of SFP modules on board..................................................188
10.14 Recommended system maximum pressure drop versus flow rates.............189
10.15 IBM water–air heat exchangers in cabinets...................................................190
10.16 IBM rear door heat exchanger.........................................................................190
10.17 Alcatel-lucent enhanced cooling in a sealed cabinet.....................................192
10.18 Cisco liquid cooled rack : back of chassis exposed........................................193
10.19 General liquid cooling loop in equipment......................................................193
10.20 Active liquid cooled rack with facility water..................................................194
10.21 Active liquid cooled rack with self supplied coolant......................................195
10.22 Sketch of board assembly connection to system...........................................195
10.23 Typical data center energy breakdown............................................................197
10.24 Ratio of 3-year site cost to 1 U server costs....................................................197
10.25 Cascade effect of energy savings from equipment to data center................198
10.26 Basic sketch of thermoelectric power generator............................................199
10.27 Possible locations of thermoelectric power
generators in data center..........................................................................200
10.28 Direct application of fuel cell energy to OSP operation................................201
10.29 Carbon nanotubes bump on electrodes and substrate..................................203
10.30 Piezoelectric fan and its application to heat sink..........................................205
10.31 Thermal performance comparison with
piezoelectric fan...............................................................................................205
10.32 Schematic of two experimental orientations.................................................206
10.33 Thermal performance of best cases of horizontal
and vertical orientations..................................................................................206
10.34 Thermal performance comparison between
piezoelectric and small axial fans...................................................................207
xiv • List of Figures
10.35 Schematic diagram of impinging synthetic jet...............................................208

10.36 Synthetic jet actuator operation in open ambient environment..................208
10.37 Thermal performance comparisons with synthetic jet..................................209
10.38 Schematic of test rig for a pair of synthetic jets.............................................210
FOREWORD
Miniaturization and convergence are key drivers in microelectronics and tele-

communications in general. Cooling of telecommunication systems is quite a
challenge as the scales vary from hand held devices to large systems. This is a
comprehensive book that addresses both the thermal and mechanical reliabilities
in the design of telecommunication systems. It is important to note that both
authors have a significant industrial experience and as such, the guidance pro-
vided in this book stems from a significant practical experience.
It is very sad that recently, Dr. Richard C. Chu has passed away. Dr. Chu has
been recognized by many as the “father of electronic cooling.” Dr. Chu’s inven-
tion achievements are documented in over 200 patents and over 150 patent dis-
closure publications. A prolific technical innovator in cooling technology, many
of his patents and publications have been recognized outside of IBM as industry
standards for the cooling of electronic equipment. Dr. Chu has achieved numer-
ous honors, including the ASME Heat Transfer Memorial Award in 1986, elec-
tion to the National Academy of Engineering in 1987 and his appointment to
Academia Sinica, Republic of China in 1996. He served as the first president of
the IBM Academy of Technology and was recognized by the Organization of Chi-
nese Americans (OCA) as Asian American Engineer of the Year in 2006. In 2010,
Dr. Chu received his 66th Invention Achievement Award, at that time a record for
IBM Poughkeepsie. Dr. Chu had a lifelong passion for the academic community,
as shown by his active involvement in IBM’s university research program which
resulted in over 80 thesis-related publications. Dick retired in 2011 after 51 years
of dedicated service to IBM. I worked for Dick Chu for 15 years at IBM and
consider him my mentor. He is by far more than anyone else responsible for my
career path guiding me during every fork in the road. Likewise, Dr. Yeh considers
Dick his mentor and dedicates this work to Dick’s memory.
The book can be used as an elective senior level or graduate level textbook. It is
also a must for engineers in the field who are always led to finding best practices
with the ever shrinking of lead time for designing systems.
Dereje Agonafer, ASME & AAAS Fellow

Jenkins Garrett, Professor
Site Director, NSF IUCRC Center in Energy Efficient Systems
Director, Electronics, MEMS & Nanoelectronics Systems Packaging Center
University of Texas at Arlington
Editor in Chief, ASME Book Series on Electronic Packaging
PREFACE
The need to develop reliable microelectronic devices capable of operating at high

speeds with complex functionality requires a better understanding of the factors
that govern the thermal performance of electronics. With an increased demand
on system reliability and performance combined with miniaturization of the
devices, thermal consideration has become a crucial factor in the design of elec-
tronic packages, from chip to system levels.
The authors understand that the challenge in the field of thermal manage-
ment of an electronic system resides not only with very high performance and
high heat dissipation devices, but also with the intermediate and lower power
devices where improved reliability objectives require cooler operation of chips.
The authors further realize that no one design method is best suited for all appli-
cations. It is common to employ several different heat transfer modes simulta-
neously in a system. Therefore, the authors published their first book entitled
“Thermal Management of Microelectronic Equipment” in 2002 to address a wide
range of subjects related to various heat transfer technologies that can be utilized
for thermal design of electronic equipment.
In the recent years, telecommunication product development has gone
through tremendous changes due to demands of high internet speeds and capac-
ities as well as system reliability. These changes result in a significant increase
in system power requirements which makes thermal design much more difficult
and challenging. Furthermore, equipment power consumption is continuously
increased at a fast pace. Since the current telecommunication systems are still
focusing air cooling, large high-performance fans are becoming a must for high
power systems in order to provide sufficient air flow rates. In addition to system
thermal issues, two major concerns about these large fans are power consump-
tion and acoustic noise of the fans. The increased system power results in a
significant increase in power consumption and the cost of powering and cooling
of equipment as well as the data centers in the last few years. Therefore, energy
efficiency at the telecommunication equipment level and in central office/data
center becomes the top priority of equipment manufacturers and operators of
the facilities.
This book covers all types of telecommunication equipment, including
indoor and outdoor systems. There are ten chapters in the book. Among them,
Chapters 2–4 focusing on the understanding of the physics of individual com-
ponents (or parts) such as chip package, heat sink, fan and air handling system.
xviii • Preface
Chapters 5–7 provide basic methodology for system thermal design on any indoor
system. The equipment under consideration includes large telecom racks or cabi-
nets, and small box configurations. Both active and passive cooling is employed,
depending on system configuration and power. The detailed descriptions of the
system thermal design for any outdoor equipment are included in Chapter 8.
The outdoor systems are generally classified into two groups, the ground/vehicle
mounted and tower/pole mounted equipment. All outdoor equipments are sub-
jected to extreme weather conditions, high humidity, high ambient temperature
and solar radiation. At the same time, they are expected to operate with high
reliability. A large variety of cooling techniques are available in cooling of the out-
door electronics as well as telecommunications equipment. These include passive
and active forced air cooling, the air conditioners or heat pumps, thermosyphons
and phase change materials (PCMs). To pave the way for future high power tele-
com systems, the available liquid cooling technologies are presented in Chapter 9.
Finally, Chapter 10 serves as a summary of the entire book which includes the
listing the future research needs. An extensive discussion of the energy efficiency
at the equipment and data centers, including the utilization of alternative energy
is also included in this chapter.
This book places a great deal of emphasis on providing practical solutions to
the thermal issues facing in today’s telecommunication industry. Therefore, all
materials presented in the book are based on authors’ experience in design of
all types of the telecommunication equipment. This book is developed to serve
many types of readers. For graduate students, this book will guide them by prac-
tical approaches to solve real world problems that are of a vast complexity. For
professional engineers, this book will provide them with valuable and resourceful
references.
L. T. Yeh
R. C. Chu
Chapter 1
Introduction
Telecommunication, which is defined as the use of information transmission over

some significant distances to communicate, employs the electrical devices and
the use of radio and microwave communications as well as the fiber optics and
their associated electronics plus the use of the orbiting satellites and the Inter-
net. A telecommunication network is a collection of terminals, links, and nodes
which connect together so that communications can be made between users of
the terminals. Networks include the circuit switching or message switching. The
terminals in the network must have a unique address so the messages or connec-
tions of addresses in the network can be routed to the proper terminal/recipients.
The telecommunication network generally consists of several basic components
which include terminals, telecommunications processors, telecommunications
channels, computers, and telecommunications control software.
The telecommunication equipment can be deplored in the indoor and the
outdoor environments depending on its applications. Based on the functionality
and performance, the indoor systems include, but are not limited to, the switch/
router, the optical network, the metro-ethernet, the access, and the core network.
A brief description of them is as follows:
Router: The network router is a device that forwards and routes data
packet along networks. Two basic functions of a router are: (a) to insure
information does not go where it is not needed and (b) to make sure
information to the intended destination. A router is often included as
part of a network switch.
Switch: The switch is a device that channels incoming data from any of
multiple input ports to the specific output port that will take the data
toward its intended destination. There are two types of switching meth-
ods, circuit-switching and packet-switching. The former is really a
dedicated and continuously connected path for its duration. In other
words, a network’s path can be used exclusively for certain duration
for tow or more parties and then switched for use to another set of
parties. For example, an ordinary voice phone call generally is utiliz-
ing a circuit-switching. Most of today data using digital signal is sent
over network that use packet-switching which allows users to share the
same paths at the same time. In addition, the specific route a data unit
travels can also be varied as conditions change, while access services are
specified sets of information transfer capabilities furnished to users at
2 • THERMAL MANAGEMENT OF TELECOMMUNICATIONS EQUIPMENT
t elecommunications network points-of-termination (POTs) to provide

access to network transport services.
A core network is the central part of a telecommunication network that pro-

vides various services to customers who are connected by the access network. All
of these systems have the same basic mechanical configuration. For the mechani-
cal configurations, the equipment is generally classified into the rack/cabinet and
box configurations as shown in Figure 1.1.
The explored growth of the internet usage in recent years has made the tele-
communication product development go through tremendous changes due to
demands of high internet speeds and capacities. These changes result in a sig-
nificant increase in the system power requirements as did in microelectronic sys-
tems. Therefore, thermal management becomes most critical factor in the design
of any telecommunication equipment.
Though liquid cooling has long been adopted in the computer industry, air
cooling is still the main cooling scheme in the telecommunication industry. The
key components in any telecommunication system are the microelectronics, the
printed circuit boards (PCBs), and the fans. Other relevant standards are listed
in the References.
The outdoor systems include the outside plant (OSP) and the remote radio head
(RRH). The former includes both landline and the wireless applications, while the
latter is generally for the wireless applications. Both systems are illustrated in Fig-
ure 1.2. The RRH is a pole or tower mounted equipment. Therefore, the size and
the weight are the critical factors to be considered in the design of the RRH.
The design and operation of the telecommunication equipment is mainly
governed by GR-63 [1] and GR-487 [2] for the indoor and outdoor systems,
respectively.
Figure 1.1 Typical telecommunication cabinet and box configuration.

Introduction • 3
Figure 1.2 Typical outside plant and remote radio head.
REFERENCES
1. Telcordia Technologies Generic Requirements, GR-63-CORE, Issue 3, March,

2006 (NEBS Requirements: Physical Protection).
2. Telcordia Technologies Generic Requirement, GR-487-CORE, Issue 3, April,
2009 (Generic Requirements for Electronics Equipment Cabinets).
3. Telcordia Technologies Generic Requirement, GR-3028-CORE, Issue 1,
December, 2001 (Thermal Management in Telecommunications Central
Offices).
4. Telcordia Technologies Generic Requirement, GR-950-CORE, Issue 2, Decem-
ber, 1998 (Generic Requirements for Optical Network Unit (ONU)).
5. Telcordia Technologies Generic Requirement, GR-3108-CORE, Issue 1, July,
2004 (General Requirements for Network Equipment in the Outside Plant
(OSP)).
Chapter 2
Component and Printed

Circuit Board
2.1 COMPONENT
Since the invention of an integrated circuit (IC) in the latter part of the 1950s, the
progress of development of new integrated circuits has been advanced very rap-
idly. Generally, heat is generated from the transistors on a small, thin, and fragile
silicon die or chip inside the package. The chip package that is the housing for the
silicon die serves to protect the chip from the environment and to facilitate the
handling during the manufacturing process as well as chip interconnection.
2.1.1 Chip Packaging Technology
The most common methods of chip level interconnects are wire bonding, tape
automated bonding (TAB), and solder bumping. Among them, the solder bump-
ing flip chip provides the highest packaging density with less packaging delay.
The main discussion here is to focus at the flip chip package because of its
explosive growth in recent years. The flip chip packaging, which is also known
as Controlled Collapse Chip Connection (C4), is a method for interconnecting
semiconductor devices such as IC chips to external circuitry with solder bumps.
The flip chip packaging has several advantages over the wire bond packaging and
these are as follows:
1. Higher I/O density

2. Better electric performance (i.e., for high speed or high frequency
design)
3. Smaller package and lower cost
4. Improved thermal management
Figure 2.1 shows a typical flip chip package with a heat sink. The flip chip is a
chip packaging technique in which the active area of the chip is flipped over fac-
ing downward. The most common bumping technologies are that the metalized
(solder) pads are deposited on the top side of the wafer. The chip is flipped over
Figure 2.1 Typical flip chip package with heat sink.
with its top side facing down and is soldered to the substrate are as illustrated in
Figure 2.2 [1]. The solder ball and column are made of high melting point (> 300°C)
solders such as 95/5 and 90/10 lead/tin compositions, which do not reflow during
package or board assembly.
An epoxy is employed to underfill the gap between the flip chip and the sub-
strate. The filling epoxy completely covers the flip chip interreconnections and
makes a strong bond to both the device and the substrate. The underfill is needed
because the shear stress generated by the CTE (coefficient of thermal expansion)
mismatch between the chip and the substrate will lead to the solder joint fail-
ure due to the fatigue. The underfill enhances the fatigue life of the solder inter-
connection by as much as 10 times and also provides the protection from the
environments.
Figure 2.2 Common solder bumping technologies.

Component and Printed Circuit Board • 7
2.1.2 Thermal Management
As indicated in Figure 2.1, the majority of heat generated from the chip flows to
the lid (or case) and then to the heat sink and finally heat is transferred to the air
stream over the heat sink. On the other hand, the minority of heat is conducted
from the chip to the printed circuit board, and through the spreading over the
board, heat is finally transferred to the air stream over the board.
The thermal representation of the flip chip package as illustrated in Figure 2.1 is
given in Figure 2.3. The thermal interface material (TIM) as shown in Figure 2.1 is
directly applied on the back side of the silicon die which provides a good thermal
path to the lid, and then to the heat sink, and finally to the ambient air stream. The
material properties of a TIM must be low thermal resistance, low stress, and good
adhesion. As chip power increases, the role of the TIM for the flip chip package
becomes increasingly important.
Reference 2 performs CFD analysis to identify the key design parameters for
TIM in different applications. The following design guidelines are recommended
based on the analysis results:
F igure 2.3 Thermal representation of flip chip and heat sink.

1. For chip power less than 10 W, the selection of TIM is less stringent
because thermal performance of packages is not very sensitive to the
changes to thermal conductivity, area coverage and thickness of TIM
as evidenced below
TIM thickness: £ 500 mm
TIM coverage: ³ 25% of die area
TIM thermal conductivity: ³ 1 W/m °C after degradation
2. For chip power ranging between 10 and 30 W, where a heat sink is
required, TIM must be carefully selected as described below
TIM coverage: ³ 50% of die area
3. For chip power ranging between 30 and 100 W, TIM selection becomes
critical and must meet the following minimum requirements
TIM coverage: ³ 75% of die area (recommendation of 100% coverage
for power greater than 50 W)
Because of the short thermal path, the primary heat transfer path from the flip
chip is through the TIM on the back side of the die to the lid as shown in Figure 2.1.
The total thermal resistance from the die to the lid is referred to as the thermal
resistance from the junction to case (qj-c). On the other hand, the total thermal
resistance from the die to the board is referred to as the thermal resistance from
the junction to board (qj-b) which is generally several times larger than qj-c. This
is the reason that the heat transfer from the die to the board is the secondary heat
transfer path. The heat transfer rate from the die to the case in the primary path
is typically greater than 75% of the chip power. The more effective the heat sink,
the higher the heat transfer rate through the primary heat transfer path is.
The lid is a critical part of the package. The lid provides the structural support
to the heat sink and also protects the package. In addition, the lid basically acts
as a heat spreader to reduce the non-uniform temperature distribution over the
lid which can lower the temperature gradient across the interface between the
package and the heat sink. Lids can be made from many high thermal conduc-
tivity materials such as copper, aluminum, AlSiC, AIN, CuW, and even alumina.
However, it is critical for the lid to have matching CTE as closer to Silicon die or
Gallium Arsenide (GaAs) chip as possible. Figure 2.4 shows the thermal conduc-
tivity and CTE of various materials [3].
Ramakrishna and Lee [4] conducted a detailed study to examine the effects of
various parameters on thermal performance of a flip-chip-plastic ball grid array (FC-
PBGA) package which is illustrated in Figure 2.5. The figure shows the cross-section
view of the package in a wind tunnel in the simulations. A schematic diagram of various
enhancements under consideration is given in Figure 2.6. Though the figure includes
all enhancements, however, all of them are studied individually and separately.
Traditional thermal performance metrics used to specify package thermal
performance are the thermal resistance of junction-to-air (qj-a), thermal resis-
tance of junction-to-board (qj-b), and thermal resistance of junction-to-case (qj-c)
which are defined below
Figure 2.4 Thermal conductivity and CTE of various materials.
θ j - a = (Tj,max - Ta) / P (2.1)
θ j - b = (Tj,max - Tpcb) / P (2.2)
θ j - c = (Tj,max - Tc) / P (2.3)
where Ta, Tc, Tj, and Tpcb are the temperature of inlet air, component case,
component junction, and board, respectively, and P is component (package)
power.
The parameters under consideration in the analysis are as follows:
a. Effect of C4 and underfill thermal conductivity

b. Effect of thermal balls and PTHs in PCB
c. Effect of heat spreader with and without overmolding
The analysis was performed under both natural convection and forced convec-
tion. The parametric study includes die area of 48.4 and 118.2 mm2, substrate
F igure 2.5 Cross-section of package in wind tunnel.
sizes of 25 to 35 mm and the free stream velocity 0.5 to 2 m/sec for forced convec-
tion air cooling. The goal of the analysis is to obtain the package thermal resis-
tances as described in Equations (2.1) to (2.3) which are then used to evaluate the
effects of various parameters on package thermal performance.
Based on this study, the following conclusions can be made:
a. Effect of C4 and underfill thermal conductivity

– The effect of combined thermal conductivity of the C4 and underfiller
materials in range between 0.6 and 10 W/m °C on package (FC-PBGA)
thermal performance, (e.g., qj-a, qj-b, and qj-c) is negligible for its value
greater than 3 W/m °C and some limiting improvements for those with
Figure 2.6 Schematic diagram of various thermal enhancements.

thermal conductivity smaller than 3 W/m °C. 0.6 to 10 W/m °C is the

underfill thermal conductivity.
–The thermal performance is significantly improved (decreasing in pack-
age thermal resistance) with increasing the die size or air velocity.
b. Effect of thermal balls and PTHs in PCB
–The C5 (controlled collapse chip connection carrier) connections between
the package and the board provide a low heat flow path. However, ther-
mal balls (C5) connected to the PTHs in the printed circuit board can
decrease package thermal resistance by about 10% to 15% as the number
of thermal balls and PTHs increase zero to 9 ´ 9 on 1.27 mm pitch. The
effect of the die size on this thermal enhancement is more profound on
the junction to board thermal resistance (qj-b) than on the junction to air
resistance (qj-a).
c. Effect of heat spreader with and without overmolding
– The heat spreader on the back of the die decreases qj-a by about 7% in
natural convection and by 28% in forced convection with air velocity at
2 m/sec for the die size of 48.4 mm2. The enhancement provided by a heat
spreader (without overmold) in natural convection for large die sizes is
so small that its use is not justified in view of the added cost of materials
and assembly.
– The overmolded die with a heat spreader on the top of the overmold
provides better thermal enhancement than a heat spreader on the die
for air velocities up to 1 m/sec. The heat spreader without overmold per-
forms slightly better than that with heat spreader plus overmold for the
air velocity greater than 1 m/sec.
The silicon die generally includes multiple heat sources. The size and power of
each heat source are different which results in the local hot spots. The shape of
the heat source on the silicon die actually has a great effect on the junction tem-
peratures of the chip. However, it seems that no study has been made to address
this issue.
The sketch as illustrated in Figure 2.7 will be used as an example for discus-
sion. The figure includes three heat sources on a single die. All of them have the
same area but with different length of the perimeter. Heat generated at the source
area is first conducted through the die thickness, then the TIM, and finally to the
lid. Along these heat conduction paths, heat is also spread out laterally within
each layer. Therefore, the larger the perimeter, the better the heat spreading will
be because of the larger lateral area for heat conduction.
2.1.3 Advanced Packaging Technology
Due to the increased power requirement, the vertical integration of the circuit die
which is also referred to as “3D Packaging” has received considerable attention
in the recent years. One of the examples of 3D package is proposed by IBM [5] as
shown in Figure 2.8. Thermally, the vertically stack chip packages are even more
demanding than the traditional 2D chips. Heat is generally conducted through
F igure 2.7 Heat conduction spreading in die.
the multiple interfaces and layers of the different materials to the case (lid) and to
the board in the traditional 2D package. The total thermal resistance is too large
to meet the temperature limits for the interlayer of the chip package due to limit-
ing thermal paths from the interlayers to the ambient.
The major challenge in the thermal design of the 3D chip package is how to
remove heat from interior layers of the dies. For these high power 3D chips, the
traditional air cooling techniques are not possible because of extremely poor ther-
mal properties of the air. It will require unrealistically high airflow rates for effec-
tive air cooling. The microchannel water cooling on the integrated circuits was
F igure 2.8 IBM 3D packages.

F igure 2.9 Cross-section of CFD thermal model.
first proposed by Tuckerman and Pease [6]. The backside heat removal using the
microchannel cooling technology has been shown effective for the single layer die.
However, it will not provide adequate cooling for the interlayer dies. The interlayer
integrated water cooling as shown in Figure 2.8 is a very promising approach.
As can be seen in Figure 2.8, the thermal design is based on the interlayer inte-
grated cooling concept. The cooling network, which includes micro pin-fins and
microchannel heat transfer structures, is integrated between the layers. The authors
not only perform experimental work but also develop the porous medium model to
simplify thermal model as illustrated in Figure 2.9 for simulations. The compari-
sons between the simulation and measurement are given in Figure 2.10. The simu-
lations can reproduce the experimental measurements from the test vehicle. The
Figure 2.10 Simulation and measured data for four hot spots in
top layer.
junction temperature needs to be distinguished from the hot spot temperature. The
former is the maximum temperature along streamwise directed line, in the symme-
try plane, lying between silicon slab and hot spot layers. The latter is the line aver-
aged temperature on the centerline along y axis in Figure 2.9 lying in the central
plane of the hot spot and therefore will be lower than the junction temperature.
Cautions must be taken in the design of the multilayer integrated water-cooled
chips to prevent any water leakage in the package. The short circuit is developed
while in contact with water. From an electric performance point view, the dielec-
tric fluids should be employed in all electric systems. Unfortunately, the thermal
performance of the dielectric fluids is much less effective than that of using water,
which must be treated in order to be used in the cooling system.
2.2 PRINTED CIRCUIT BOARD
The next level of packaging is the circuit board. The printed circuit board (PCB)
which is also often referred to as the printed wiring board (PWB) is a major ele-
ment of any electronic equipment. This is because the PCB provides a mounting
surface and structural support for the electronic components (or chip packages)
and wiring channels to serve as conduits for chip-to-chip connections. The PCB
is also frequently considered as the primary field replaceable unit and a test bed
to provide accessible points in making circuit checks.
2.2.1 Construction
A typical printed circuit board consists of a multilayer of dielectric material and

copper film, as shown in Figure 2.11. The copper layers are used to provide the
chip-to-chip connections for transmitting signals, power, and grounding purpose.
The basic concern in the thermal analysis is how to model the board accurately.
Although the majority of material of a PCB (probably over 95% by volume) are
insulation materials which have a very low thermal conductivity, one can never
omit the contribution due to the copper layers in the thermal analysis. Any omis-
sion of the copper layer in the analysis will result in excessive high component
F igure 2.11 Typical construction of a PCB.

temperatures, especially for the case of conduction-cooled PCB. The thermal con-
ductivity of the cooper and dielectric materials are listed in Table 2.1.
2.2.2 Equivalent Thermal Conductivity
It is not practical to model the board exactly as shown in Figure 2.11. Therefore,
the simplified configuration given in Figure 2.12 is considered. The equivalent
thermal conductivities of a printed circuit board (PCB) can be determined in the
following ways [3].
Planar Thermal Conductivity, Kxy
The laminated layers are considered as parallel paths in calculation of the equiva-
lent thermal conductivity. For a parallel network, the total resistance is
1 1 1 1
= + + + … (2.4)
Rxy R1 R2 R3
where
Ri = thermal resistance for the ith layer, i = 1, 2, 3, ...

Li
=
i Ai ci
k
Li = length of the ith layer
Ai = cross-sectional area of the ith layer (normal to heat flow)
ki = thermal conductivity of the ith layer
ci = fraction of the total coverage for the ith layer
(e.g., c = 0.8 if the cut-out for a given layer is 20%)
Rearranging Equation (2.9), we have

K xy A ckA  ckA 
= Σ i i i  + Σ i i i  (2.5)
L  L i in  Li cu
Table 2.1 Thermal Conductivity of Basic Materials

for PCB
Material W/m °C But/hr-ft-°F
Copper layer 386.00 223.180

Dielectric layer
G10 0.30 0.173
FR-4 0.35 0.202
Polyimide 0.52 0.301
Figure 2.12 Simplified board construction for thermal model.
where “in” and “cu” represent the dielectric insulation material and the copper,
respectively. A is the total cross-sectional area over the thickness of a PCB, i.e.,
A = t (thickness) w (width). Since L = L1 = L2 = L3 = ...,
Equation (2.5) can be rewritten as follows:
æc k A ö æc k A ö
K xy = å ç i i i ÷ + å ç i i i ÷ . (2.6)
è A ø in è A ø cu
Since ci is equal to one for the insulation layers, Equation (2.6) can be further
simplified to the following form:
 t′   t′ 
K xy = kin  in  + kcu  cu  (2.7)
t
 pcb   tpcb 
where
t¢in = total effective thickness of the dielectric material

= tpcb – t¢cu
t¢cu = total effective thickness of the copper film
= S (ci tc)cu
tpcb = total thickness of the printed circuit board,
= tin + tcu.
Normal Thermal Conductivity, Kz
Heat flow is normal to each layer, and it is assumed that the resistances are in
series as follows:
Rz = R1 + R2 + R3 + … (2.8)
or
t pcb æ ti ö æ ti ö
= åç + åç .
K z Az è c i k i Ai ÷ø in è c i k i Ai ÷ø cu (2.9)
Since Az = Ai, the equivalent thermal conductivity in the z-direction is
t pcb
Kz = .
æ ti ö æ ti ö (2.10)
å çè c i k i ÷ø å çè c i k i ÷ø
+
in cu
Again, since ci is equal to one for insulation layers, we have
′ / kin + tcu
K z = tpcb /(tin ′ / kcu ). (2.11)
Graebner [7] made detailed measurements of the conductivities for the planar
and the normal directions of the board. The conductivities for copper and glass-
epoxy are 350 and 0.59 W/m °C, respectively. The board averaged conductivities
are correlated into the following equations
K xy = 350 (t c¢ u / t pcb ) + 0.8 (2.12)
and
K z = [1.69 (1 - t cu ¢ / t pcb )]-1

¢ / t pcb ) + 0.0026 (t cu (2.13)
Later, Graebner and Azar [8] presented empirical equations given below to
estimate the thermal conductivities of a PCB. It should be noted that the value of
385 is the thermal conductivity of copper.
′ / tpcb ) + 0.87
K xy = 385 (tcu (2.14)
and
K z = [3.23 (1 − tcu ′ / tpcb )]−1

′ / tpcb ) + 0.0026 (tcu (2.15)
where the unit of thermal conductivity is W/m °C.

It should be noted that all of the above equations do not include the effects due
to the vias and any local non-uniform conditions. The above equations simply
consider the copper content in the board.
Table 2.2 Board Construction Information for a 14 Layers Board
Input layer Input tin (in) Input tcu (in) Input cu coverage Net t’cu (in)
1 0.0036 0.0022 0.28 0.000616

2 0.002 0.0013 0.61 0.000793
3 0.0031 0.0013 0.66 0.000858
4 0.004 0.0006 0.05 0.00003
5 0.0031 0.0006 0.03 0.000018
6 0.004 0.0013 0.62 0.000806
7 0.0031 0.0006 0.05 0.00003
8 0.004 0.0006 0.03 0.000018
9 0.0031 0.0013 0.63 0.000819
10 0.004 0.0006 0.04 0.000024
11 0.0031 0.0006 0.02 0.000012
12 0.002 0.0013 0.66 0.000858
13 0.0036 0.0013 0.61 0.000793
14 0.0022 0.23 0.000506
sum 0.0427 0.0158 0.006181
Notes:
tpcb (in) = 0.0427 + 0.0158 = 0.0585
t’in (in) = 0.058 – 0.006181 = 0.052319
Example 2.1
To compute the equivalent thermal conductivities for Kxy and Kz with the input
listed in Table 2.2. Table 2.3 includes the comparisons with the present predic-
tions which are based on Equations (2.6) and (2.10) and the results obtained
from Equations (2.12) through (2.15). The results from Equations (2.6) and (2.10)
compare extremely well with those obtained from the above empirical equations
(Equations (2.13) to (2.15)).
Recently, several PCB thermal design tools are available. One of them is
ANSYS TASPCB. The users can directly import all the relevant design data from
the ECAD tool, including component data, traces, power and ground planes and
vias. A built-in component library provides 3D geometry and thermal data to
automatically create a 3D board model which is then solved thermally with aid of
CFD analysis. The software can actually predict the local thermal conductivities
at all three directions at any point of the board.
Table 2.3 Comparisons of Board Thermal Conductivities
(Kcu = 385 Kin = 0.3) (Kcu = 350 Kin = 0.59)

Present Graebner/Azar (8) Present Graebner (7)
Kxy (W/m−°C) 40.964668 41.54838 37.508 37.78034
Kz (W/m−°C) 0.335411 0.346141 0.65957 0.6615
2.2.3 Thermal Consideration
One of the critical functions of the board is to spread the heat generated from
the components. In fact, the board is the largest heat sink for all components on
the board in the case of air cooling. To enhance heat conduction across the board
thickness, thermal vias are often utilized. There are two types of the thermal vias.
One is unfilled and another is filled thermal via. However, the thermal advantage
of the filled thermal via over unfilled one is limited.
The thermally enhanced PCB has been available for a long time. However, this
type of the PCB is generally applied to the power converters, motor controls and
LED lighting and is available from Laird Technologies and Bergquist.
REFERENCES
1. Bezuk, S., “Flip Chip Challenges”, Kyocera America Inc.

2. Yuan, T.D., Pan, H.-Y., Li, Y., “Thermal Interface Material (TIM) Design Guid-
ance for Flip Chip BGA Package Thermal Performance”, Altera, San Jose, CA,
2004.
3. Yeh, L.T., and Chu, R.C., Thermal Management of Microelectronic Equip-
ment, ASME Press, 2002.
4. Ramakrishna, K., and Lee, T.-Y.T., “Evaluation of Thermal Enhancements to
Flip-Chip-Plastic Ball Grid Array (FC-PBGA)”, Journal of Electronic Packag-
ing, Vol. 129, 2004.
5. Alfieri, F., Tiwari, M.K., Zinovik, I., Poulikakos, D., Brunschwiler, T., and
Michel, B., “3D Integrated Water Cooling of A Composite Multilayer Stack of
Chips”, Proceedings of 14th International Heat Transfer Conference, Wash-
ington, DC., USA, August 8–13, 2010.
6. Tuckerman, D.B. and Pease, F.F., “High Performance Heat Sinking for VLSI”,
IEEE Electron Device Letter EDL-2, 1981.
7. Graebner, J.E., “Thermal Conductivity of Printed Wiring Boards”, Electronic
Cooling, Vol. 1, No. 2, 1995.
8. Graebner, J.E., and Azar, K., “Thermal Conductivity Measurements in Printed
Wiring Boards”, Journal of Heat Transfer, Vol. 119, 1997.
9. Lee, T.-Y. Tom and Mahalingam, M., “Thermal Limits of Flip Chip Package-
Experimentally Validated, CFD Supported Case Studies”, IEEE Transaction on
Components, Packaging, and Manufacturing Technology, vol. 20, No. 1, 1997.
Chapter 3
Fans and Air Handling

Systems
The fan is the heart of any forced air cooling system. The fan propeller converts
torque from the motor to increase the static pressure across the rotor of the fan
and to energize the air particles. A fan is frequently employed in electronic equip-
ment to provide the airflow required for cooling electronics. Direct air cooling is
a primary example of using a fan for thermal control of electronic systems.
3.1 FAN TYPES
Fans are generally classified into two types, namely, centrifugal (blower) and axial
fans, depending on the direct of airflow through the impeller.
1. Centrifugal Fans
– receive air at its axis of rotation and exhaust air at its periphery in a
direction normal to the rotation axis;
– produce pressure from two sources (i) centrifugal force created by rotat-
ing air column and (ii) kinetic energy imparted to air as it leaves the
impeller; and
– create low flow rate against a high resistance (pressure).
2. Axial Fans
– deliver air in a direction parallel to the fan blade axis;
– produce pressure from the change in velocity passing through the impel-
ler; and
– provide a high airflow rate but tend to work against low pressure.
The axial fan can further be divided into three categories as follows:
a. Propeller fans are the simplest type and are designed to move large vol-
umes of air at low velocity and develop low static pressure.
b. Tube axial fans are similar to the propeller fan but provide a higher
total pressure capability than the propeller fans.
c. Vane axial fans are the same as the tube axial but have vanes to
straighten the swirling flow created as the air is accelerated which
results in improved efficiency and noise characteristics.
Basically, tube axial and vane axial fans fill the requirements between centrifu-
gal and propeller fans.
3.2 FAN PERFORMANCE AND FAN LAWS
The fan performance or characteristic curve is generated through a series of tests

as illustrated in Figure 3.1 [1]. The fan is tested from the shutoff conditions to
nearly free delivery conditions. At the shutoff point, the duct connected to the fan
outlet is completely blocked. On the other hand, the outlet resistance of the duct
is reduced to zero at the free delivery point. Additional test points between these
two limiting conditions are obtained when various flow restrictions are placed at
the end of the duct to simulate various operating conditions on the fan.
Fans designed to be used with a duct are generally tested with a length of duct
between the fan and the measurement station. The purpose of this length of duct
is to smooth out the flow as it exits the fan and to provide stable and uniform flow
conditions at the measurement station. The measured pressures are then cor-
rected back to the fan outlet conditions. Fans designed to be used without a duct
are tested without the duct.
As air moves through a system, the pressure is decreased because of expan-
sions, contractions, bends, other structure blockages, and friction. For a system
under consideration, the pressure drop is a function of the flow rate, therefore,
a system resistance curve can be generated. The intersection of the fan perfor-
mance and the system resistance curves is the operation point of the system, as
given in Figure 3.2 [2].
Figure 3.2 also shows the static efficiency, (h = Qp/P), which is defined as the
ratio of the static pressure power to the input power of a fan. The maximum static
Figure 3.1 Method of obtaining fan performance curves.

Fans and Air Handling Systems • 23
Figure 3.2 Typical fan curve and system operation point.
efficiency identifies the ideal combination of the flow rate and pressure of the fan.
Fans should be operated as close to this point as possible if they are to achieve
optimum results as given in the figure.
The fan laws relate the performance variables for any dynamically similar
series of fans. The variables involved are fan size (D), rotational speed (N), gas
density (r), volumetric flow rate (Q), pressure (p), power (P), and efficiency (h).
The relationships among the variables are as follows:
Q = ϕ ND3 or Q1 = Q2( D1/D2)3 ( N1/N 2) (3.1)
p = ψ ρ N 2 D2 or p1 = p2( D1/D2)2 ( N1/N 2)2 (3.2)
P = λ ρN 3 D5 or P1 = P 2( D1/D2)5 ( N1/N 2)3 (3.3)
where j, y, and l are the constants for geometrically and dynamically similar
operation and are also referred to as the flow coefficient, pressure coefficient, and
power coefficient, respectively.
3.3 FAN INLET AND EXIT CONDITIONS
Fan performance is strongly depending on the flow conditions at the inlet and exit
of the fan. The air must enter the fan housing as uniformly as possible in order
to achieve a better fan performance. In addition, the entering air also requires a
sufficient length to reach the developed flow. Therefore, the fan blade within the
housing of an axial fan is preferred being located at near the exit end of the fan.
Example of the effect of the blade location on fan performance is illustrated in Fig-
ure 3.3 [3]. Figure 3.4 shows the effect of the blade location on the fan performance
when the fan is located adjacent to a restricted area such as a 90° bend [3].
Hill and Hill [5] have studied experimentally the effects of electronic enclosure
on fan performance. Two axial fans with the same speed (3100 RPM) were tested.
The fan characteristics are presented in Table 3.1.
The ratio of plenum depth, P to fan tip shroud diameter, D was selected as the
parameter for comparison. In addition, the flow rate was normalized based on
the free flow rates listed in Table 3.1. The plenum wall has various opening areas.
The test results for the 35% plenum area opening are given in Figure 3.5. The
findings from the tests can be summarized as follows:
Figure 3.3 Effect of blade location on fan performance.
Figure 3.4 Effect of blade location on fan performance with fan adjacent
to restricted area.
Table 3.1 Characteristics of Tested Fans
Fan A Fan B
Tip shroud diameter, D (mm) 77.1 112.9

Hub diameter, d (mm) 36.1 52.3
Thickness, t (mm) 25.0 38.9
Speed, N (RPM) 3100 3100
Input power (W) 2.2 5.6
Max. static pressure, Ps (mm wg.) 4.09 7.11
Free flow, (L/s) 17.7 47.5
1. The evacuating flow direction is more susceptible to degraded fan per-

formance at low P/D ratios than the pressurizing flow direction.
2. Larger fan (Fan B) is not affected by the lower P/D ratios as much as
smaller fan (Fan A) for either flow direction.
3. The results reveal that P/D ratios greater than 1 do not yield signifi-
cantly better fan performance than a P/D ratio of unity.
Figure 3.5 Fan performance curves for various plenum depth ratio
(P/D).
3.4 FAN AND SYSTEM PRESSURE
Fan total pressure is the true indication of the energy to the airstream by a fan.
Therefore, the total pressure should be used for the fan selection and air distribu-
tion system design. The system pressure loss is defined as the summation of all
the individual total pressures in the entire air distribution system.
The system resistance is the sum of various pressure drops in the air deliv-
ery system. In addition to the friction loss, the system pressure drop is caused
by other losses such as inlet, exit, change of cross section (expansion and con-
traction), turns, and elbows. The pressure loss in the above items is typically
expressed by the velocity head in the following form.
2
Dp = k rV /(2gc ) (3.4)
where k is the loss coefficient, and its value for various configurations can often
be found in the handbook of hydraulics [4]. gc is 32.2 lb ft/lbf sec2.
Frequently, the pressure drop in an airflow system is expressed in the unit of
“in of water.” Equation (3.4) for dry air at standard conditions (14.696 psia and
70 °F with air density of 0.075 lbm/ft3) can be rewritten for this unit as follows
D p = k (V/4005)2 (3.5)
where Dp is “in of water,” and V is ft/min.

If air is not at the standard conditions, Equation (3.5) becomes as follows:
D p = k ρ (V/1097)2 (3.6)
The density of air is a function of the pressure and temperature which in turn
is a function of the altitudes.
3.5 FAN ACOUSTIC NOISE

Acoustic noise which is more or less a purely subjective personal phenomena has
two basic characteristics. One is the physical phenomenon which can be measured,
and another one is the psychoacoustic characteristics which are to judge the effect
of noise on human beings. The acoustic noise can be measured in terms of “sound
pressure” and “sound power” which are two distinct but commonly confused charac-
teristics of sound. However, both share the same unit of measure, the decibel (dB).
“Sound pressure,” which is measurable, is a pressure disturbance in the atmo-
sphere whose intensity is influenced not only by the strength of the source but
also by the surroundings and the distance from the source to the receiver. The
sound pressure is what the ears hear and what the sound meters measure. On the
other hand, “Sound power” is the acoustic energy emitted by the sound source.
The sound power is an absolute value and is not affected by the environment.
The sound power cannot be measured directly and is calculated from the sound
pressure measurement.
Fan noise is a function of the fan design, volume flow rate, total pressure, and
efficiency. The most efficient operating range for a given line of fans is generally
the quietest. Fan noise is expressed by the sound power levels in eight octave
bands. It should be noted that low exit velocity does not necessarily guarantee
quiet operation. Furthermore, noise comparisons of different types of fan or fans
from different manufacturers cannot be made by comparing the fan rotational or
tip speeds, and the only valid basis for comparison are the actual sound power
levels generated by fans.
The fan noise will be affected by the system in which fan operates. Maling and
Yeager [6] provide the following guideline to minimize the fan noise
1. No obstruction should be located within one fan diameter from the fan
inlet for axial fans so that the airflow entering the fan inlet is as spa-
tially uniform as possible.
2. Pulling air over equipment being cooled usually causes undesirable tur-
bulence at the (exhaust) fan inlet and increases noise level.
3. Designing the system to have the lowest possible static pressure rise for
the required airflow so that the fan can operate at low tip speeds and
reduce the noise level.
3.6 FAN RELIABILITY
The fan is the heart of any forced air cooling system and therefore has been a
main contributor to the failure of many electronic systems. Just as microelectron-
ics, fan failure can be classified into “catastrophic failures” in which fan is not
running due to hardware failures and “software failures” in which fan can still
be functioning but fails to meet performance requirements. Since a fan consists
of both mechanical and electrical parts, the major fan failure modes as stated by
Tian [7] include mechanical failures, electrical failures, failures due to installa-
tion, acoustic failures, failures due to poor serviceability, and poor cooling perfor-
mance failures and the pareto of the fan failures is given in Figure 3.6.
Fans can fail in many ways, and failures can also be defined differently by the
different vendors and/or applications. Two metrics of fan reliability commonly
quoted by vendors are the L2 life and L10 life which are the second and tenth
percentiles under some assumed fan life distribution. In other words, L2 and L10
indicate the time points at which 2% and 10% of the fans will fail, respectively.
The L2 and L10 can be determined by the following equations [8]
L 2 = a (0.02020)1/b (3.7a)
L10 = a (0.10536)1/b (3.7b)
where a is the characteristic life, and b is the shape parameter.

The shape parameters for Weibull model are generally greater than 1, which means
that a fan failure increases with age. Reference 7 indicates in the high reliability/
F igure 3.6 Causes of fan failure.
availability products, such as servers, a more useful metrics is L1 life that means the
life when 1% of products fail. L1 can be expressed as follows:
L1= α (0.01005)1/β (3.7c)
Vendors also often use the mean time to failure (MTTF) as the reliability indi-
cator. For the Weibull distribution, MTTF can be expressed as follows:
MTTF = αG(1+1/ β ) (3.8)
where G denotes the Gamma function.

It is often confused MTTF with MTBF (mean time between failures). It should
be noted that the MTBF should only be used in a repairable system setting in
order to understand the system’s maintenance needs and the service cost.
The life of most fans is generally limited by the bearing life because the
bearing failure is the major failure mechanism in most fans. The bearing life
is strongly depending on the composition and reliability of the bearing lubri-
cants (e.g., greases). Kim and Claassen [8] recommend to limiting the bearing
temperature rise to 10°C or less where the fan reliability is critical. The authors
further recommend the Booser grease life equation as given below for the fan
life estimation:
logL10 = 2.6 + (Kt/Tbrg) − 0.301 S (3.9)

where
S = Sg + Sn + Sp
Sn = 0.86 DN/(DN)L
Sp = 0.61 DNP/Cr2
P = Equivalent dynamic bearing load, lbf
N = Speed, rpm
Cr = Basic dynamic load capacity, lbf
D = Bore diameter, mm
(DN)L = Speed limit, rpm-mm
S = Half-life subtraction factor; for S = 1, the life drops 50%
Sg = Grease half-life subtraction factor; typically 0 for many greases
Sn = Speed half-life subtraction factor
Sp = Load half-life subtraction factor
Kt = Grease temperature factor = 2450 for acceleration factor of 1.5
for each 10°C
Tbrg = Bearing temperature, K
It should be noted that Equation (3.9) does not take into account of the effect of
grease quantity and may not cover all greases on the market, especially for those
newly developed greases which use synthetic oils.
REFERENCES
1. Fans, Chapter 3, ASHRAE 1983 Equipment Handbook, 1983.

2. Yeh, L. T., and Chu R. C., Thermal Management of Microelectronic Equip-
ment, ASME Press, New York, 2002.
3. Steinberg, D. S. Cooling Techniques for Electronic Equipment, Wiley, New
York, 1980.
4. Belevins, R. D., Applied Fluid Dynamics Handbook, Van Nostrand Reinhold,
New York, 1984.
5. Hill, T. B. and Hill, C. C., “Effects of Electronic Enclosure Layout on Fan
Performance,” 90-WA/EEP-6, ASME Winter Annual Meeting, Dallas, Texas,
November 25–30, 1990.
6. Maling G. C., and Yeager, D. M., “Minimizing Acoustical Noise in Electronic
Systems,” Electronics Cooling, Vol. 2, No. 2, May 1996.
7. Tian, X., “Cooling Fan Reliability: Failure Criteria, Accelerated Life Testing,
Modeling and Qualification,” Itherm 2006.
8. Kim, S. and Claassen, A., “How to Evaluate Fan Life,” Electronics Cooling,
Vol. 2, No. 2, 1996.
Chapter 4
Air Cooling and Heat Sinks
Simplicity and easy maintenance make direct air cooling a most attractive
approach in cooling of electronics. This approach includes both active and passive
methods. The former is related to forced convection while the latter represents
free or natural convection. The passive cooling has always been the preferred
choice. However, it is generally limited to the systems with very low heat dissipa-
tion because of poor heat transfer coefficients involved in the free convection.
Forced air cooling over printed circuit boards (PCBs) is often encountered in
electronic equipment. Various component size and spacing cause flow separation
over components and recirculation between components. The flow field is a fully
three-dimensional (3D) problem and extremely complex. Heat transfer involves
convection from components and the PCB to the air stream, conduction from
components to the PCB and heat spreading along the PCB, and also radiation
exchange among components and from components to surroundings.
4.1 NATURAL CONVECTION
The natural convection over a vertical plate is most simple and is also often encoun-
tered in the engineering applications. To enhance heat transfer, the fins are fre-
quently added to the surface that is also often referred to as the heat sink. Figure 4.1
shows a typical configuration of a continuous fin array. The flow field over a finned
heat sink is much complicated than the typical flow over a single plate or in parallel
plates because of the involvement of the third surface (fin base).
There are many studies of the natural convection from the multiple surfaces
such as parallel plates or finned heat sinks by numerical analysis or by experi-
ment. The finned heat sink as presented in Figure 4.1 consists of a number of
U-shaped channels. The corner regions formed by the base plate and the fins
result in a significant reduction in the heat transfer.
For vertically straight-fin heat sinks, several experimental data [1–3] are avail-
able. Among them, Izume and Nakamure [3] developed a mathematical relation-
ship describing heat transfer from the finned heat sink; however, their equation
does not hold in the limiting cases of very large or very small ratios of the channel
depth to channel width. To overcome this problem, Van De Pol and Tierney [4]
developed the following empirical equation applicable to any channel depth to
width ratios.
Nu r = (Ra * /ψ ){1 − exp[− ψ (0.5 / Ra*) 3/4]} (4.1)

F igure 4.1 A vertical straight fin heat sink.
where
Nu = hur/k,
r = 2LS/(2L + S),
Ra* = (r/H)GrrPr,
Grr = gb(r/m)2 (Tw – Ta) r3,
y = 24(1 – 0.483e–0.17/a)/{(1 + 0.5a)[1 + (1 – e–0.83a)(9.14a0.5eV*S – 0.61)]3,
a = S/L,
V = –11.8(1/in).
The above correlation is limited to the constant wall temperature condition

and is only applicable to the continuous straight fins. In addition, hu is the heat
transfer coefficient for a U-channel. It should be noted that Equation (4.1) can be
applied to various sizes of the heat sink. Radiation heat transfer must always be
included under the natural convection conditions. This is especially true for the
air at high altitudes where the effectiveness of natural convection is significantly
reduced due to reduction of the air density.
Yeh et al. [5] have performed a CFD analysis on the heat transfer from the vertical
finned heat sinks in the natural convection environments. The authors also com-
pared the results from the above equation with those of the CFD. A good agreement
between both sets of the results further validates the accuracy of Equation (4.1).
The equation is considered especially useful for optimizing the heat sink config-
urations because of its simplicity and ease for obtaining quick results. Yeh
Air Cooling and Heat Sinks • 33
et al. [5] further found that the radiation loss is relatively insensitive to fin spac-
ing if the spacing variation is small. Therefore, the natural convection alone can
be used to optimize fin spacing of the heat sinks under the natural convection
condition.
Yeh [6] utilized the above equation for optimizing component heat sinks. The
heat sink widths under consideration are 40 mm, 65 mm, and 90 mm and each
width with four different heat sink lengths (30 mm, 40 mm, 50 mm, and 60 mm)
in the flow direction. For a given heat sink width and length, analyses are per-
formed to determine the optimal fin spacing for the maximum natural convection
heat loss. The fin thickness for all configurations is 1.2 mm, and the heat sink is
made of anodized aluminum. The ambient is assumed at 50°C and the heat sink
base at 85°C. The former represents inlet air temperature over a printed circuit
board, and the latter corresponds to the component case temperature.
It should be noted that all other dimensions of the heat sink remain unchanged
while varying the fin spacing to achieve the maximum heat loss which is the prod-
uct of the heat transfer coefficient, heat transfer surface area, and the overall
surface efficiency. The summary results from Yeh [6] are shown in Figures 4.2,
4.3, and 4.4 for the heat sink widths of 40 mm, 65 mm, and 90 mm, respectively.
The optimal fin spacing is in theory a function of fin thickness, heat sink width,
and length. As shown in Figure 4.2 for the heat sink length of 30mm, its optimal
fin spacing is 4.34 mm. On the other hand, the optimal spacing for the heat sink
lengths of 40mm, 50mm and 60mm is 5.27 mm. Examining the results for the
heat sink length of 30mm in details reveal that the heat loss for the fin spaces of
4.34 mm and 5.27 mm are 4.047 and 4.039 Btu/hr, respectively. The difference in
the heat loss between these two fin spaces is so small that one can consider the
optimal fin spacing for the 30mm long heat sink to be 5.27 mm in order to reduce
manufacturing processes and part counts. Therefore, the optimal fin spacings
Figure 4.2 Natural convection loss for heat sink width of 40 mm.
for heat sink lengths of 30mm, 40mm, 50mm, and 60mm are 5.27mm, 5.18mm,
5.14mm for the heat sink width of 40mm, 60mm and 90mm, respectively. Fur-
thermore, for simplicity and practical application, the single fin spacing of 5.2
mm which is an average value of the above three optimal fin spaces (5.27mm,
5.18 and 5.14mm) is recommended for this family of heat sinks with length rang-
ing from 30mm to 60mm and width from 40mm to 90mm under the natural
convection environment.
4.2 FORCED CONVECTION

Simplicity, easy maintenance and low cost make direct air cooling a most attrac-
tive approach in cooling of microelectronics. In fact, almost all telecommunica-
tion equipment is cooled by air. Various types of heat sinks are often mounted to
the components in order to enhance convective heat transfer so that the compo-
nents can be kept below their respective temperature limits.
4.2.1 Heat Sink Thermal Performance

The presence of the fins on a given surface will affect the local heat transfer coeffi-
cient because the flow pattern has been changed. However, one should never just
look at the heat transfer coefficient alone in evaluating the thermal performance
of a heat sink. It should be noted that the fins generally will increase not only the
heat transfer but also the pressure drop. Any thermal design must always have a
good balance between the heat transfer and pressure drop. The thermal perfor-
mance (thermal resistance and pressure drop) of a heat sink can be illustrated as
shown in Figure 4.5.
The total thermal resistance of a finned heat sink consists of the average heat
sink resistance and the spreading resistance. The former corresponds to the con-
vective heat transfer from the fins, and the latter represents the heat conduction
spreading along the base of the heat sink. The total thermal resistance can be
expressed as follows:
Rt = Ro + Rc (4.2)
where
Ro = average heat sink resistance
Rc = heat sink spreading resistance
Figure 4.5 Typical thermal performance of a heat sink.

Average Heat Sink Thermal Resistance
For a uniform base temperature, the effectiveness of a heat sink is measured by

the total heat transfer rate which is defined as follows:
Q = ηo hAt ∆T (4.3)
where
Q = total heat transfer rate

ho = overall surface efficiency
= 1 – Af /At (1 – hf)
Af = fin surface area
At = total heat sink surface area for convection
hf = fin efficiency
h = heat transfer coefficient
DT = t emperature difference between average heat sink temperature and
fluid temperature
In addition, the average thermal resistance of the heat sink is also given by the
following equation:
Ro = Q/∆T = 1/ (ηo hAt ) (4.4)
The average thermal resistance of a heat sink is basically related to the fin
thermal performance, which is affected by the fin surface area, fin configuration,
fin material, and the mass flow rate of the fluid.
Base Spreading Thermal Resistance
The non-uniform temperature distribution along the heat sink base will signifi-
cantly reduce the thermal performance of a heat sink. Because of the increase in
the chip power in the recent years, the oversized heat sinks (larger than the com-
ponent foot print) are often needed in order to maintain the proper component
temperatures. Therefore, the spreading resistance of the heat sink base become a
critical issue for cooling of modern microelectronics.
In theory, the thermal spreading resistance can be eliminated if the base thick-
ness or thermal conductivity approaches infinite. The spreading resistance always
exists in the practical applications. Song et al. [7] and Lee et al. [8] provided a
closed form equation as given below to calculate the thermal spreading resistance.
Rc = [(Ap0.5 − As0.5 )/( kπ 0.5Ap0.5 As0.5 )] (4.5)
[λ kApRo + tanh( λt)]/[1 + λ kApRo + tanh( λt)]

and
λ = π1.5/Ap0.5 + 1/As0.5 (4.6)

where
As = footprint of heat source

Ap = footprint of heat sink base
t = thickness of heat sink base
k = thermal conductivity of heat sink base
Ro = average heat sink thermal resistance
The graphical presentation of Rc and Ro are illustrated in Figure 4.6. As indi-

cated in the figure, the maximum thermal spreading or constriction resistance
Rc, accounts for the local temperature rise over the average surface temperature
of the heat sink. It should be noted that the above correlation does not address the
shape of the heat source nor the heat sink base.
Later, Lee [9] extended the earlier work to examine the effects of the location
and the size of the heat sources. The effect of the heat source size on the heat sink
base temperature is shown graphically in Figure 4.7. As shown in the figure, the
maximum temperature at the center of the heat sink increases as the size of the
heat source decreases. As the temperature rises at the center, the temperature
along the edges of the heat sink decreases simultaneously. It should be noted
that the area-averaged surface temperature of the heat sink remains unchanged.
In other words, the area-averaged surface temperature of the heat sink depends
solely on the total heat load.
The effect of the location of heat source on the thermal resistance of the heat
sink is illustrated in Figure 4.8. The higher thermal spreading resistance in Case
B is due to the fact that heat spreading is limited to two directions while heat
spreading in Case A is in three direction as marked by the “arrows.” The thermal
spreading resistance is significantly increased for the oversized heat sinks.
Figure 4.6 Schematic view of local resistance or temperature varia-

tion of heat sink.
Figure 4.7 Normalized local temperature rise with various size

of heat sources.
Figure 4.8 Effect of heat source location on heat sink

thermal resistance.
4.2.2 Factors Affecting Heat Sink Thermal Performance
The following factors will have great impacts on the performance of the heat
sink.
Flow Bypass and Leakage

Figure 4.9 shows a typical straight fin heat sink attached to a component on the
board. Two important phenomena associated with the air-cooled heat sinks are
the flow bypass and the heat sink flow leakage as illustrated in Figures 4.10 and
4.11, respectively. The flow bypass rate equals to the difference between the chan-
nel and heat sink inlet flow rates, while the leakage flow rate equals the difference
between the heat sink inlet and exit flow rates. Both phenomena will reduce the
thermal performance of the heat sink.
No air flow bypass occurs if the cross-section of the test channel is equal to
that of the heat sink. However, the flow bypass always exists with air flowing over
F igure 4.9 Typical straight fin heat sink.
F igure 4.10 Flow bypass over heat sinks.
Figure 4.11 Heat sink flow leakage.
a printed circuit board. The main reason for the flow bypass is that the air will
flow through the least resistance path. The heat sink flow leakage also exits. One
way to eliminate the heat sink flow leakage is to add a cover over the heat sink
which flow is totally within a confined space.
Fin Configurations
The fins should have good aerodynamic characteristic to reduce the flow resis-
tance. Yeh [10a–10c] performed CFD analysis to evaluate the thermal performance
Figure 4.12 Various types of heat sinks.
of various types of heat sinks. Three different types of the heat sinks, including
extrusion fin, plain fin, and cell fins as shown in Figure 4.12 are under consider-
ation. The dimensions of the fins are given in Table 4.1.
The detailed information of a single fin for the cell fin heat sink is shown in
Figure 4.13. All cell fins are enclosed by the frame with the thickness of 0.5 mm.
The case without the flow bypass is first considered. In other words, the indi-
vidual heat sink is enclosed by the channel walls. The analysis is further extended
to the case with the flow bypass. The channel configurations for the heat sinks
with the flow bypass are illustrated in Figure 4.14.
The heat transfer coefficient and the thermal resistance of the heat sinks for
the case with the flow bypass are presented in Figures 4.15 and 4.16, respectively. The
reason for the lowest heat transfer coefficient for the cell fin hat sink is due to the
small spacing between fins which results in low air flow through the heat sink.
Table 4.1 Dimensions of Three Heat Sink Configurations
Extrusion fin (mm) Plain fin (mm) Cell fin (mm)

Length (L) 60.0 60.0 60.0
Width (W) 75.95 75.95 75.95
Height (H) 17.17 17.17 17.17
Base height (Hb) 2.95 2.95 2.95
Fin height (Hf) 14.22 14.22 14.22
Fin thickness (t) 1.3 0.2 0.1
Fin number 15 25 35
Figure 4.13 Configuration of single fin of cell fin heat sink.
17.61 mm
11 mm 11 mm 11 mm
Heat
Heat Sink
Sink Heat
Heat Sink
Sink Heat
Heat Sink
Sink
Components
Figure 4.14 Channel cross section with flow bypass.
Figure 4.15 Average heat transfer coefficient of heat sink without

flow bypass.
Figure 4.16 Average thermal resistance of heat sink without

flow bypass.
However, it has the largest heat transfer surface area among all heat sinks under
consideration. Therefore, the cell-fin heat sink has the lowest thermal resistance as
shown in Figure 4.16.
The heat sink heat transfer surface used to compute the heat transfer coefficient
is defined as the total surface area of the heat sink except the bottom of the heat
sink base overhanging over the components. The surface areas for the extrusion-
fin, the plain-fin, and the cell-fin are 46.74, 73.19, and 140.79 in2, respectively. The
cell-fin heat sink has the largest surface area for the convection, but the space
between the fins is very small that it limits its heat transfer coefficient.
The heat transfer coefficient and the heat sink thermal resistance with the flow
bypass are presented in Figures 4.17 and 4.18. Generally, the heat transfer coef-
ficients with the flow bypass are less than those without the flow bypass at the
same velocities because the less flow goes through the heat sinks.
The trends for the heat transfer coefficient and the thermal resistance are simi-
lar to those without the flow bypass with the exception of the low air velocity. As
can be seen from Figure 4.18, the thermal resistance of the cell-fin heat sink is across
over the plain fin and extrusion fin heat sinks for the velocities less than 2.5 and
1.5 m/sec, respectively. Because the heat transfer coefficient is so small for the cell-
fin heat sink at these velocities that the product of the heat transfer coefficient and
the surface area is also smaller than those of the other two heat sinks at the same
velocity.
The pin fin heat sink has been employed extensively in the direct air cooling
for microelectronics. The pin fin heat sink is better suite for the case where the air
flow is not parallel to the heat sink. In other words, there is no flow direction limi-
tation for the pin fin heat sink. The local heat transfer coefficient for a pin fin heat
sink is generally higher than that of the continuous fin heat sink. However, the
latter has a much larger surface area. Therefore, the overall thermal performance
Figure 4.17 Average heat transfer coefficient of heat sink with

flow bypass.
Figure 4.18 Average thermal resistance of heat sink with flow bypass.
of the continuous fin heat sink in general is better than that of a pin fin heat
sink. The comparison of thermal resistance between the pin heat sink made by
die-casting and the discrete continuous heat sink made by forging process is pre-
sented in Figure 4.19 [11]. It should be noted that the thermal conductivity of the
forged heat sink is higher than that of the die-casted heat sink.
Figure 4.19 Thermal resistance for pin fin and discrete continuous
heat sinks.
Mass Flow Rates
The heat transfer coefficient is a function of the mass flow rate (m = rVA) or
mass velocity (G = rV). Generally, the heat transfer coefficient varies with G0.5 and
G0.8 for the fully developed laminar and turbulent flow, respectively. Therefore,
the heat transfer coefficient will increase when the mass flow rate is increased.
However, on the other hand, the thermal resistance as defined in Equation (4.4)
decreases as the mass flow rate increases. In addition, the pressure drop which is
proportional to (rV 2) increases as the mass flow rate increases.
It should be noted that the mass flow rate should be utilized in calculating the
heat transfer coefficient or the pressure drop. It is a common mistake by adapting
the volumetric flow rate in evaluating the heat transfer or pressure drop. This is
especially true for the air cooling at high altitudes where the density is less than
that at the sea level because the air density is a function of altitudes. Omission
of density in thermal analysis may lead to the equipment overheating at higher
altitudes.
4.2.3 Heat Sink Thermal Enhancement
Once the factors affecting the thermal performance of a heat sink are fully
understood, thermal enhancements can be made accordingly. The thermal
improvements should focus at increasing the total heat transfer rate, not the heat
transfer coefficient alone.
Improved Fin Thermal Performance, Ro
As mentioned previously, the average heat sink resistance, Ro is generally related

to the fin thermal performance. Various optimizations, such as optimizations for
cost or weight/volume or heat transfer, can be made for a heat sink. However,
the present discussion is limited to the thermal optimization. The most common
one is to optimize the fin spacing or number of fins for the maximum heat trans-
fer rate. A minimum thermal resistance, Ro as defined in Equation (4.4) always
existed for a given heat sink as number of fins increasing over a wide range.
The heat transfer coefficient is a function of the mass flow rate. The mass
velocity over the individual fins decreases as the number of fins increases. There-
fore, heat transfer coefficient alone always decreases with increasing the number
of fins due to reduction of air flow over individual fins.
Enhanced Base Thermal Spreading, Rc
High thermal conductivity materials can be used to lower the thermal spreading
resistance at the heat sink base. Aluminum heat sink has been widely adapted in
the industry because of its light weight and low cost. However, its thermal con-
ductivity is much smaller than copper.
The following methods can be considered to reduce the base thermal spread-
ing resistance which in turns also lowers the total thermal resistance of the heat
sink.
1. Copper Embedded Heat Sinks

The oversize heat sinks are often required due to the increased component
power. Aluminum heat sinks are increasingly difficult to meet the thermal
requirements of high power components. The hybrid heat sink consisting
of aluminum fins and copper base can be employed to reduce the thermal
spreading resistance and also to lower the total thermal resistance of a
heat sink.
It is very difficult to measure or even compute the base thermal spread-
ing resistance alone. Therefore, the comparison is often made at the total
thermal resistance of the heat sink. McLean [11] has performed several
studies and the comparison between the aluminum and copper embed-
ded heat sinks are given in Table 4.2. The improvement is found to be
increased with the velocity.
2. Embedded Heat Pipe Base
The copper heat sink is sometimes not sufficient or too heavy to meet
thermal design requirements. Other effective methods such as the embed-
ded heat pipe or vapor chamber base heat sinks. The equivalent thermal
conductivity of a heat pipe is typically several times higher than copper.
Table 4.2 Thermal Performance of Aluminum and Copper Embedded Heat Sinks
AL heatsink
AL Heat sink Copperembedded
Copper embeddedheatheat
sink
sink
Dimension
Dimension 112 xx74
112 74xx30
30mm
mm 112 xx74
112 74xx30
30mm
mm
sinkbase
Heat sink base 7mm
7 mm 7mm
7 mm
Source
Heat source 37 xx 37
37 37mm
mm 37 xx 37
37 37mm
mm
Power
Power Dissipation
dissipation 80W
80 W 80W
80 W
Rt (℃/W)
Rt(℃/W) Rt((℃/W)
Rt ℃/W)
LFM
LFM Thermalresistance
Thermal resistance Thermal
Thermalresistance
resistance Improvement
Improvement
200
200 0.537
0.537 0.522
0.522 2.7%
2.7%
360
360 0.366
0.366 0.351
0.351 4.1%
4.1%
520
520 0.316
0.316 0.295
0.295 6.5%
6.5%
680
680 0.283
0.283 0.258
0.258 8.8%
8.8%
840
840 0.264
0.264 0.236
0.236 10.6%
10.6%
The base of a heat sink can be made of aluminum with the embedded
heat pipes. The comparisons between aluminum, embedded copper, and
embedded heat pipe heat sinks are presented in Table 4.3 [11].
3. Vapor Chamber Base
The vapor chamber is similar to the heat pipe which is a two-phase heat
transfer device. The vapor chamber offers two district advantages over the
heat pipes, and these are the direct contact to the heat source and uniform
spreading in all directions. The typical vapor chamber consists of the vapor
chamber, a (aluminum) fin stock, and an aluminum frame for mechanical
attachment. These three parts are soldered together as shown in Item A
in Figure 4.20 [12]. Another option is to directly embed the vapor cham-
bers into the base of the heat sink as given in Item B of Figure 4.20 [12].
Meyer [12] concludes that vapor chambers can offer thermal performance
improvement on the order of 10% to 30% over copper base and slightly bet-
ter than heat pipe base heat sinks. It should be noted that the vapor cham-
ber does include the wick structures as used in the heat pipes.
Recent development efforts are focusing at the new generation of the vapor
chambers with the base thickness of the heat sinks no more than 3 mm. The
thermal performance of such thin base vapor chamber heat sinks and the embed-
ded heat pipe heat sinks were studied by Glover et al. [13]. The test samples (heat
sinks with various types of constructions of vapor chambers) are given in Table 4.4.
The overall dimensions of the various heat sinks are 152.4 mm (width) ´ 101.6 mm
(length) (in flow direction) ´ 12.7 mm (total height). The tests were also con-
ducted at various orientations, including (a) horizontal with bottom heating (fins
Table 4.3 Comparisons of Aluminum, Copper Embedded, and Heat Pipe Embedded
Heat Sinks
AL heat sink Copper embedded Heat pipe embedded
heat
sink
Dimension 112 x 74 x 30 mm 112 x 74 x 30 mm 112 x 74 x 30 mm

Heat sink base 7 mm 7 mm 7 mm
Heat source 37 x 37 mm 37 x 37 mm 37 x 37 mm
Power dissipation 80 W 80 W 80 W
Rt (℃/W) Rt ℃/W)
( Rt (℃/W)
LFM Thermal resistance Thermal resistance Thermal resistance
200 0.537 0.522 0.516
360 0.366 0.351 0.335
520 0.316 0.295 0.271
680 0.283 0.258 0.237
840 0.264 0.236 0.217
Figure 4.20 Various types of vapor chamber heat sinks.
face up), horizontal with top heating (fins face down), side ways, and vertical with
air flow upwards. The heat sink thermal resistance which is defined as (Tb, max –
Ta, in)/Q, of heat sinks at various orientations is presented in Figure 4.21.
It is a well-known fact that the surface characteristics of a heater have great
effects on heat transfer in two-phase flow, especially in pool boiling. You, Rainey
and Ammerman [14] have developed and patented microporous coating surface
to enhance pool and flow boiling heat transfer. Based on this patent, a liquid
chamber heat sink is developed. The major differences between the vapor cham-
ber and liquid chamber are listed in Table 4.5.
4. Embedded Graphite Base

The thermal conductivity of the natural graphite is about 200 to 400 W/m °C.
However, the thermal conductivity can be increased over 1500 W/m °C after
Table 4.4 Tested Heat Sink Configurations
Base thickness
Manufacturers Weight (g) (mm) Descriptions
A-1 270 3 Extruded heat sink with 2 mm thick

embedded sintered power VC strip
in base
B-1 372 3 Wire-mesh VC base with Al zipper fins
C-1 318 3 VC base wick structure: heating side
sintered powder, cooling side-wire
mesh: Al zipper fins
C-2 246 3 Al base with 3 embedded heat pipes
in parallel, inserted 1” ´ 1” Cu block
in center of base, Al zipper fins
D-2 320 4 3 heat pipes cantilevered off Cu
center block base with Al stack fins
the special heat treatment [15]. Table 4.6 [15] lists the comparisons of TPG
(trademark of Momentive Performance Materials, Inc.) to other advanced
thermal materials. Due to relatively high planar thermal conductivity of
graphite, it has been considered as a thermal spreader. Therefore, graphite
can be embedded into the heat sink base to reduce the thermal spreading
resistance.
Figure 4.21 Measured heat sink thermal resistance.

Table 4.5 Tested Heat Sink Configurations
Vapor chamber Liquid chamber
1. Two phase based on evaporation Two phase based on nucleate boiling
2. Heat transfer by latent heat Heat transfer including latent heat and
microconvection
3. Liquid return by capillary force Liquid return due to condensation
(no wick needed)
4. Wick increasing thickness and weight Wickless reducing thickness and weight
Table 4.6 Comparison of Various Advance Thermal Materials
In-plane Specific
In-plane TC Thru-plane CTE Specific in-plane
Material (W/m – k) TC (W/m – k) (ppm/°C) gravity TC1
Aluminum 218 218 23 2.7 81

Copper 400 400 17 8.9 45
AlSiC-12 180 180 11 2.9 62
CuW 185 185 8.3 15.2 12
Carbon/ 400 40 –1.0 1.9 210
carbon
CVD 1100–1800 1100–1800 1–2 3.5 310–510
diamond
TPG 1500+ 10 –1 2.3 650
graphite
4.2.4 Materials and Manufacture Processes
The materials for the heat sinks are generally with high thermal conductivity
such as aluminum alloys (k < 240 W/m °C) and copper (pure copper k = 385
W/m °C). The former is much light and less expensive than the latter. The coef-
ficients of thermal expansion for aluminum and copper are about 24 and 16,
respectively.
The typical manufacturing processes include but are not limited to Stamp-
ing, Extrusion, Bonded/brazed, Die Casting, Forging, and Friction Stir Welding
(FSW). The extruded heat sink is most popular because it is inexpensive and
easy to make. However, the fin of this type heat sink is relatively thick which
results in the small surface area available for the convective heat transfer due
to limited number of fins on the heat sink. In addition, there is limitation for
the fin height-to-fin spacing. On the other hand, there is no limitation on the
fin height-to-fin spacing as well as the fin thickness for bonded fins. The typical
example of this type heat sink is the plain fins or folded fins bonded to a base
plate.
McLean [11] compares the heat sink thermal performance of various manu-
facturing processes. She first pointed out that the material thermal conductivity
as listed in Table 4.7 will be increased after annealing in the heat treatment. The
increased thermal conductivity is due to the improvement of the material crystal
structure. As given in the same table, the material thermal conductivity with the
forging process is better than those of the extrusion. The microstructures of the
parts for various manufacturing processes are summarized in Table 4.8.
Table 4.7 Material Thermal Conductivity Under Various Manufacturing Processes
W/o annealing
thermal With annealing
Raw Test conductivity thermal conductivity
material appearance (W/m × K) (W/m × K) Improvement
Copper 308.4 351.4 12.2%
With extrusion
thermal With forging
Raw Test conductivity thermal conductivity
Material appearance (W/m × K) (W/m × K) Improvement
AL 6063 177.9 203.3 14%
AL 1070 203.3 235.5 16%
Table 4.8 Microstructure of Parts (Made By Different Processes at 400´)

REFERENCES
1. Starner, K. E., and McManus, H. N., “An Experimental Investigation of Free

Convection Heat Transfer from Rectangular Fin Arrays”, J Heat Transfer 85,
1963.
2. Welling, J. R. and Wooldridge, C. R., “Free Convection Heat Transfer Coef-
ficients from Rectangular Vertical Fins”, J Heat Transfer 87, 1965.
3. Izume, K., and Nakamura, H., “Heat Transfer by Convection on Heated sur-
face with Parallel Fins”, Jap. Soc. Mech. Eng., 34, 1969.
4. Van De Pol, D. W., and Tierney, J. K., “Free Convection Nusselt Number for
Vertical U-Shaped Channels”, J Heat Transfer, 95, 1973.
5. Yeh, L. T., Yeh, Joseph and Chung, B. T. F., “Natural Convection from Finned
Heat Sinks”, IPack2007-33036, Vancouver, BC, Canada, July 8–12, 2007.
6. Yeh, L. T., “Optimizing Heat Sink on Chips for Natural Convection”, to be
published.
7. Lee, S., Song, S., Au, V., and Moran, K. P., “Closed Form Equation for Ther-
mal Constriction/Spreading Resistances with Variable Resistance Boundary
Condition”, Proceedings of the 1994 IEPS Conference, 1994.
8. Song, S., Lee, S., Au, V., and Moran, K. P., “Constriction/Spreading Resis-
tance Model for Electronic Packaging”, Proceedings of the 4th ASME/JSME
Thermal Engineering Joint Conference, Vol. 4, 1995.
9. Lee, S., “Calculating Spreading Resistance in Heat Sinks”, Electronics Cool-
ing, Vol. 4, No. 1, 1988.
10a. Yeh, L. T., “Thermal Characterization of Various Heat Sinks with/without
Flow By-Pass”, The 20th International Symposium on Transport Phenom-
ena, Victoria, BC, Canada, July 7–10, 2009.
10b. Yeh, L. T., “Thermal Performance Evaluations of Various Heat Sinks for Air
Cooling”, Itherm 2815, Itherm 2012, San Diego, CA May 30 – June 2, 2012.
10c. Yeh, L. T., “Thermal Performance of Various Heat Sinks”, IMPACT 2012,
Taipei, Taiwan, October 24–26, 2012.
11. McLean, J. H., “Comparison of Thermal Performance for Various Thermally
Enhanced Heat Sinks and Manufacturing Processes”, Workshop on Ther-
mal Management in Telecommunication Systems and Data Centers, Rich-
ardson, Texas, October 25–26, 2010.
12. Meyer, G. A., “Integrating Vapor Chambers into Thermal Solutions”, Elec-
tronics Cooling, Volume 16, No. 1, 2010.
13. Glover, G., Chen, Y., Luo, A., and Chu, H., “Thin Vapor Chamber Heat Sink
and Embedded Heat Pipe Heat Sink Performance Evaluations”, 25th IEEE
Semi-Therm Symposium, 2009.
14. You, S. M., Rainey, K. N., and Ammerman, C. N., “A New Microporous Sur-
face Coating for Enhancement of Pool and Flow Boiling Heat Transfer”,
Advance in Heat Transfer, Vol. 38, 2004.
15. Liu, X., “TPG and TC1050 Materials for Thermal Management of Electron-
ics”, Workshop on Thermal Management in Telecommunication Systems
and Data Centers, Richardson, Texas, October 25–26, 2010.
16. Yeh, L. T., and Chu, R. C., “Thermal Management of Microelectronic Equip-
ment”, ASME Press, 2002.
Chapter 5
System Thermal Analysis —

Rack/Cabinet (Part I)
The telecommunication equipment can be deplored in the indoor or the outdoor

environments, depending on its applications. Based on the functionality and per-
formance, the indoor systems include, but not limited to, the router, the opti-
cal network, the metro-ethernet, the access, and the core network. All of these
systems have the same basic mechanical configuration even though the system
application areas are totally different.
5.1 KEY ELEMENTS IN SYSTEMS
A typical telecommunication rack and the side view of a shelf are given in Figure 5.1.
As illustrated in the figure, the key elements include the card cage which hosts the
printed circuit boards, air filter, electric magnetic insulation (EMI) plates which pre-
vent electric magnetic waves from leaking into the ambient, and the fan trays in the
forced air cooling systems. In addition, the system also includes an air inlet section
and an outlet section.
5.1.1 EMI Plates

The EMI plates are generally installed on the top and the bottom of the card cage
to prevent the electric magnetic leakage from the microelectronics and the power
supplies to the environment. The EMI plate is a thin perforating plate (about
1 mm thick). For an effective electric magnetic insulation, the maximum percent-
age opening of the EMI plate is generally less than 65%.
The pressure drop can generally be expressed in the following form
DP = K(0.5rV 2) (5.1)
where K is the loss coefficient.

At the standard conditions, the pressure drop in an air flow system as pre-
sented in Equation (5.1) can further be reduced as follows:
DP = K(V/4005)2 (5.2)
Figure 5.1 A typical telecom rack and sketch of shelf.
where DP is in inches of water and V is ft/min.

The loss coefficient across over a thin perforating plate can be expressed as
follows [1]:
K (loss coefficient) = [0.707(1 – f )0.375 + (1 – f )]2/f 2 (5.3)
where f is the fraction of the total opening of a thin plate.

The loss coefficient versus the percentage of the opening is given in Figure 5.2.
An additional function of the EMI (perforation) plate is to make air flow more
uniform over the individual boards in the card cage due to additional flow
resistance.
Recently, the honeycomb structures have been used for the EMI purpose as
shown in Figure 5.3. The honeycomb structure has a much larger open than that
of the perforating thin plate. The fraction of the opening for a typical honeycomb
is about 0.9. However, in order to effectively reduce the electric magnetic leak-
age, the thickness of the honeycomb structure must be greater than 0.25 in. with
preferable of 0.5 in.
Another benefit of using the honeycomb structure is to straighten the air stream
as passing through the honeycomb structure. It should be noted that the pressure
drop over a honeycomb structure is much higher than the value solely based on
the percentage of the opening. This is due to the fact that the honeycomb structure
requires a much thicker structure to prevent the electric magnetic leakage.
System Thermal Analysis — Rack/Cabinet (Part I) • 55
Figure 5.2 Loss coefficient versus fraction of opening for a thin plate.
5.1.2 Air Filters
The air filter is needed to prevent the dust or any foreign particles flowing over
the microelectronics on the boards. Therefore, the filter should be placed in the
upstream of the card cage. Another reason for the air filter is to create a laminar
flow. The typical performance curve of a clean air filter, as shown in Figure 5.4a, is
Figure 5.3 Typical honeycomb structures.

Figure 5.4 a Performance curve of a clean air filter.
generated through a series of tests according to ASHRAE Standard 52.1-1992. For

convenience, Figure 5.4a is converted to Figure 5.4b which states that the pres-
sure loss across an air filter is a function of approaching air velocity. The pressure
drop over a dirty air filter could be increased several times larger than the value
of a clean one. The reason is that the majority of dusty or any foreign particles
are collected on the front surface of the conventional air filters which results in a
significant blockage to the air flow and leads to the high-pressure drop.
Figure 5.4b Pressure drop versus air flow of typical air filters.
The pressure drop across the air filter is relatively high and is a significant portion
of the total pressure drop in the system. As can be seen from the above figure, the
pressure drop is 0.07 in. of water at the air velocity of 300 ft/min which results in the
loss coefficient of the filter (K) is 12.5 according to Equation (5.2). On the contrary,
the loss coefficient for a typical board (not very high dense board) is probably less
than 6 or 7.
Consideration must be taken to reduce the pressure drop over the filter. Since
DP aV 2, Equation (5.1) can be rewritten as follows:
DP α(1/A)2 (5.4)
where A is the cross-section area of the filter.

Therefore, one way to reduce the pressure drop is to have as large air filter as
possible. The best location is to place the air filter under the card cage as shown
in Figure 5.1 because this position generally gives the largest surface area (over
the width and the depth of the rack). The air filter can be either placed in the
upstream or downstream of the fans. The former will prevent dust over the fans,
and the latter will make air flow more uniformly flow over the individual boards
due to the high flow resistance of the air filter.
5.1.3 Fans
The fan is the heart of any forced air cooling system. The fan propeller converts
torque from the motor to increase the static pressure cross the rotor of the fan
and to energize the air particles. The discussion here is focusing at the fan appli-
cation in the rack configuration.
Thermally, the location of the fans in a system is also an important factor
to consider. The fan located at the exit of a system is usually referred to as an
exhaust fan which draws air through the system. This type of installation reduces
the internal system pressure (to less than the ambient pressure) and possibly
draws dust or foreign particles into the system. The advantage of this method is
that no additional heat dissipated from the fans is added to the system. However,
the fans operate at higher air temperature at the exit and result in reduction of
the fan reliability as well as the fan life.
On the other hand, a blowing fan, which is located at the inlet or upstream
of the card cage in the system, creates a higher internal pressure and keeps dust
or any foreign particles out of the system. However, the inlet air is heated up
by the heat dissipation from the fan and thus the air temperature is increased
prior to entering the system or the card cage where the printed circuit boards are
located.
Typically, multiple fans which are arranged in parallel are placed in a fan tray.
Therefore, the acoustic noise of the system is significantly increased with the
number of fans. The accumulated delta noise level can approximately be deter-
mined by the following equation:
D(dBA) = 10 log (N) (5.5)

and, the total system acoustic noise is given as follows
(dBA)system = (dBA)sf + D(dBA) (5.6)
where N is the number of fans in the system and (dBA)sf is the acoustic noise
of a single fan.
5.2 SYSTEM LEVEL THERMAL ANALYSIS
The ultimate goal of the thermal analysis is to maintain the temperature of all
electronic components in the system below their required limit. The system level
thermal analysis is first performed to calculate system flow rate as well as the
flow rate to individual boards which is often also referred to as the slot flow rate.
The computed slot flow rate is then applied to the individual boards for the board
level thermal analysis which is to determine the temperature of the components
on the individual boards.
Since the detailed information of individual boards is generally not available
in the early design phase, one must assume some values of pressure drop for the
boards. This assumed value can be obtained from the existing similar boards or
just from the preliminary board layout. The assumption of this value is not very
critical for the purpose of the system level analysis. The reason is that the pres-
sure drop over the board is typically less than one third of the total system pres-
sure drop. Therefore, the effects of the errors in the estimated pressure drop over
the board on the system flow rate will be limited. In addition, the more accurate
estimated pressure drop over the board will be available once the detailed board
is developed.
No detail of the board layout is needed for the system level analysis. Each board
can be represented by a large component on the board with the board heat dissipa-
tion. The purpose of this assumption is to provide correct heat dissipation from
individual boards in order to compute air temperature rise over the boards and
also to show the component side of the boards. Furthermore, the assumed loss
coefficient over individual boards is then treated as the flow resistances at the inlet
and the exit of the boards. Another popular way to predict the air temperature rise
is to treat the power from the board as an internally volumetric heat generation in
the space, and the volumetric flow resistance is assumed over this space.
Figure 5.5 shows the simple sketch of a representative telecommunication
rack and its flow resistance over the system. The individual flow resistance is as
follows:
R1 = entrance
R2 = inlet section
R3 = 90° turn
R4 = air filter
R5 = card cage, including 2 EMI plates
R6 = 90° turn
R7 = exit section
R8 = exit
F igure 5.5 Sketch of typical rack and its flow resistance.
The flow resistance can be derived from Equation (5.1)
DP = K (0.5rV 2) = RQ2 (5.7)
where R is the flow resistance, and Q is the volumetric flow rate.

The intersection of the fan performance and the system resistance curve is the
system operation point, and the flow rate at the operation point is the average
system flow rate. In theory, the estimated system flow rate can be obtained by the
hand calculations. However, a CFD (computational fluid dynamics) analysis with
aids of the commercial software is typically employed in determining the system
flow rate as well as the flow rate to individual boards.
As stated previously, the system flow rate as well as the board flow rates can be
determined by the hand calculations instead of using the CFD. The hand calcula-
tion method is in general based on the flow resistance network. According to the
flow network modeling (FNM), the software analysis tool, “MacroFlow” [2] was
developed by Innovative Research for the system level flow analysis.
The CFD analysis is used to compute the velocity and pressure of each fluid
particle, while the FNM analysis is based on the lumped system and determines
the average value of the velocity and pressure of the system. The FNM method
is extremely efficient and fast for the system level flow analysis. For example, it
generally takes 3 to 5 days to build a system level CFD model of a typical telecom-
munication rack and may take several hours running the simulation to obtain the
converged results. On the other hand, it takes a couple of hours to develop FNM
model and less than 30 sec to obtain the solution. In most cases, the difference
between the FNM and the CFD results is generally less than 10%. The FNM is
very useful tool in the early phase of the system design, especially in defining the
system concepts. However, it cannot be adapted for the detailed thermal design
of the system.
5.3 PRACTICAL EXAMPLES

Example 5.1
A telecommunication rack consists of two identical and independent shelves as
shown in Figure 5.6 [3]. For clarity, the front door is not shown in the figure. Each
F igure 5.6 Telecommunication rack with two identical shelves.

F igure 5.7 Sketch of shelf.
rack consists of two identical shelves located on the top-to-bottom orientation. Each
shelf with overall dimensions of 25.45 in. (width) ´ 18.99 in. (depth) ´ 37.75 in.
(height) includes one fan tray with six fans, and three card cages with a total of 50
printed circuit boards (PCBs). The sketch of the shelf is given in Figure 5.7. Air at
40°C enters through the door from the front of the shelves, and then makes a 90° turn
upwards through PCBs, and finally turns another 90° to exit the system from the rear
of the shelf. The system level analysis is performed independently on each shelf.
The only difference between the upper and lower shelves in the rack is at the
opening of the front door. The shape and size of the opening (or air inlet) on the
front door are different. The lower shelf has a rectangular opening with the cross-
section area (90.73 in.2) identical to the inlet of the shelf. On the other hand, the
door opening at the upper shelf has an odd shape as shown in Figure 5.8 with the
cross-section area of 70.38 in2 which is smaller than the inlet of the shelf. The odd
shaped opening at the front is approximated by four different sizes of the rectan-
gular openings in the thermal model as presented in Figure 5.9. The total cross-
section area of these four rectangular openings is equal to that of the odd shaped
opening. The major components, such as the fan tray and three card cages inside
the shelf, are included in the detailed thermal mode as shown in Figure 5.9.
The bottom of each shelf is a fan tray with the height of 7.875 in. The fan tray
includes 6 EBM (DV6248TD) fans with each one consuming 90 W and generating
acoustic noise of 69 dBA. The outer diameter and the thickness of the fan are 6.75 in.
and 2 in., respectively. The fan operates at 48 V in the normal operation The fan loca-
tions relative to the mid-plane are shown in Figure 5.10.
The shelf consists of three card cages with the total power of 5626 W that are
located above the fan tray. The shelf is subdivided into the front and the rear sections
F igure 5.8 Odd-shaped opening at upper section of front door.
by a mid-plane. The front section of the shelf includes two identical card cages
with the total power of 2246 W, while there is only one card cage in the rear section
with the power of 3380 W. The mid-plane provides all electrical connection to the
boards inside the card cages. The air exit is located on the top of the card cages.
In addition, there are two perforate plates in the shelf for EMI (electrical magnetic
insulation) protection purpose. For simplicity, the air filter and the EMI plates are
modeled as a planar flow resistance with a zero thickness. The loss coefficients for
the EMI plates and the air filter are 1.7 and 14, respectively. The loss coefficient
F igure 5.9 System thermal model.

F igure 5.10 Fans locations.
of 1.7 corresponds to 63% opening on the thin perforation plate. In addition, the
ambient condition is assumed to be at 40°C and at the sea level.
The main purpose of the CFD analysis is to determine the air flow rate to indi-
vidual boards. The predicted air flow rates are then applied to the detailed board
level thermal analysis. The component junction temperatures and the pressure
drop across the boards are computed from the board level analysis of individual
boards. The total number of the cells in the current model is over 433,300. There
is no need to discuss the details of fundamental heat transfer theories involved
here because of using a commercial software tool.
The card cage sketch and the flow rate to individual boards for the upper front
cage are presented in Figure 5.11a. The GR-63 also requires the system to meet the
Figure 5.11a Card cage sketch and board flow rates for front card cage.
Figure 5.11b Card cage sketch and board flow rates for rear card cage.
thermal requirements under one fan failure condition. The flow rate for individual
boards for the case with Fan 3 failure is also included in the figure. The analysis
should be repeated for every fan failure condition so that the worst case can be iden-
tified. The board flow rates for the rear card cage are presented in Figure 5.11b.
As can be seen from the above figures, the flow rates to the front and the rear
card cages are 455 and 1020 CFM, respectively. The total flow rate for the upper
shelf is 1475 CFM with the system pressure drop of about 1.1 in. of water. It
should be noted that the operation point of the individual fans varies slightly
even for the identical fans. This is mainly due to the fan locations. Examining the
results in details also reveals that 69% of the total system flow goes to the rear
card cage which has 60% of the total system power.
Example 5.2
Another example is to employ a CFD analysis to examine the effects of the air
filter locations at the upstream and downstream of the fans on the flow distribu-
tion within the system [4]. A telecommunication shelf as shown in Figure 5.12
includes a card cage, a fan tray, an inlet, and an exit. The system includes four
EBM (DV6248TD) fans as adapted in Example 1. The overall internal dimensions
of the shelf are 21.2 in. (width) ´ 36.75 in. (height) ´ 19 in. (depth). There are
four compartments inside the card cage. Each of three identical compartments
contains an optical electronic module (OEM) and its motherboard. The fourth
compartment includes four small printed circuit boards.
The present discussion is focusing at the effects of the air filter location relative
to the fans on the system air flow rates. Parametric studies are performed to exam-
ine the effect of the location of the air filter on the flow distribution. Figure 5.13
shows the results for the case with four fans in operation. The difference in the flow
Figure 5.12 Sketch of shelf under consideration.
Figure 5.13 Air flow rates versus filter locations with four fans in
operation.
Table 5.1 Air Flow Rate for Four Fans in Operation with Air Filter
3 in. from Fan
Up-stream filter Down-stream filter

Compartment flow rate (CFM) flow rate (CFM)
OEM1 253 243
OEM2 208 248
OEM3 289 223
Boards 78 125
Total 828 839
rate between the cases with the air filter being located at the up- and the down-
stream of the fans is small (less than 11 ft3/min (CFM)) regardless the location of
the air filter.
For down stream air filter case, the total system flow rate increases as the dis-
tance from the fan pack increases. However, on the other hand, an optimum dis-
tance of 2 in. from the fan pack that generates the maximum system flow rate is
achieved for the up-stream air filter. The flow rate distribution to individual com-
partments with the air filter located at 3 in. from the fan is given in Table 5.1.
The results for the case with the one failure are presented in Figure 5.14. The
optimum distance for both cases (up- and down-stream air filters) is found to be
about 1 in. from the fan pack. Table 5.2 gives the flow rate to individual compart-
ments for one fan failure. The failed fan is located just below OEM1 and OEM2.
Therefore, little flow is supplied to these two modules when the air filter is located
upstream of the fans as given in Table 5.2. The table also indicates that the filter
located at the upstream of the fan provides uneven air flow to the compartments
when one fan fails. On the other hand, the case with the air filter placed in the
down stream of the fans gives a more uniform flow distribution to all compart-
ments as given in Table 5.2. This is because the latter pressurizes the plenum due
to a relatively high flow resistance of the air filter.
Figure 5.14 Air flow rates versus filter locations for one fan failure.
Table 5.2 Air Flow Rate for Three Fans in Operation with Air Filter 3 in. from Fan
Up-stream filter Down-stream filter

Compartment flow rate (CFM) flow rate (CFM)
OEM1 37 144
OEM2 45 126
OEM3 335 142
Boards 92 92
Total 509 504
Reversed flow (leakage 270 321
from bad fan)
The real concern is not at the total system flow rate but the flow rate to indi-
vidual modules. As can be seen from the above table, the flow rate to OEMs 1 and
2 for the case with the up-stream air filter is much smaller than those with the
downstream air filter. This may lead to an overheating of the components inside
OEMs 1 and 2. The flow leakage from the failed fan for the down-stream filter
is much larger than that of the upstream filter. This is because of the high flow
resistance of the filter.
It is a well-known fact that the flow rate generated by a fan will be decreased if
the fan is placed too close to any structures. The general rule is that the distance
between the fan and the blocking structure must be at least greater than the half
of the diameter of the fan. With the limited data available from this study, this
general rule seems to be also valid for the array of the fans.
To eliminate or reduce the flow leakage from a failed fan, a damper as shown
in Figure 5.15 may be added to each fan. However, the damper generally increases
the flow resistance and thus decreases the available air flow to the system. Another
concern about the damper is the possibility of failure. The fan may be blocked if
the damper fails to open on the good fans. The practice of adding the fan cover
F igure 5.15 Examples of fan damper/cover.

has been increasingly adapted in the industry, and the trends are expected to
grow in the future.
5.4 THERMAL DESIGN PROCESS
In order to insure the quality and the accuracy of the thermal analysis results,
one must have a good design process to follow. The recommended design process
presented below is summarized in the following design flow chart as presented
in Figure 5.16.
The flow chart below describes the complete thermal design process for a
telecommunication rack. The entire design process can be divided into 3 phases
which are summarized in the following steps:
1. Phase 1: System level design to determine the flow rate to individual

boards.
2. Phase 2: Board level design to compute the junction temperature of all
components.
(a) Using the computed the air flow rate to individual boards along with the
inlet air temperature to perform the detailed analysis for all boards.
(b) Design refinements are often required in order to meet the compo-
nent temperature limits. The most common refinement is to resize the
heat sink on the hot components. Others include the relocations of the
hot components or adding the baffles to guide the air flow to the hot
components or employing high thermal conductivity TIMs (thermal
interface materials).
Figure 5.16 Telecommunication rack thermal design process.

(c) The results from the board level analysis are the component junction
temperatures and the pressure drop (or the flow resistance) of the indi-
vidual boards.
3. Phase 3: Final design analysis to combine system and boards, including
3 detailed boards.
(a) The component temperature on a given board will be affected by the
presence of neighboring boards. Therefore, three detailed boards are
needed in the final analysis. The target board is sandwiched by two
detailed boards. The pressure drop or flow resistance calculated from
the board level analysis should be applied to the rest of other boards.
For the board located at both ends of the card cage, two boards (the tar-
get board and the neighbor board) will be needed in the final analysis.
(b) The component temperatures obtained for the target board are the
final results.
(c) The analysis should be repeated by rotating three-board group through-
out the entire system.
Theoretically, the final thermal model should include all detailed boards in the
system. Unfortunately, the thermal model in such case will be too large to run and
become unpractical. The adoption of the three-board group concept is to obtain
the accurate results within the reasonable time. In addition, the effects from the
boards located outside the three-board group are generally very limited.
As required by the GR-63, the thermal analysis must be performed for the
worst case of one fan failure condition. In theory, one must turn off the fan one
by one to determine the worst case of one fan failure case. However, most experi-
enced thermal engineers can quickly narrow down to a couple of possible worst
cases to be analyzed.
REFERENCES
1. Blevins, R. D., Applied Fluid Dynamics Handbook, Van Nostrand Reinhold,

New York, 1984.
2. Macroflow, a product of Innovative Research, Inc, Minnesota.
3. Yeh, L. T., “System Level Thermal Analysis of Large Telecommunication
Racks”, InterPack2003-35047, InterPack Conference, 2003.
4. Yeh, L. T., Chung, B. T. F., and Yeh, J., “A CFD Analysis of Telecommunica-
tion Racks, including Effects of Air Filter Locations”, InterPack2005-730008,
InterPACK Conference, 2005.
5. Yeh, L. T., and Chu, R. C., “Thermal Management of Microelectronic Equip-
ment: Heat Transfer Theory, Analysis Methods and Design Practices”, ASME
Press, New York, 2002.
Chapter 6
System Thermal Analysis —

Rack/Cabinet (Part II)
The system level thermal analysis and the complete thermal design process of a
telecommunication rack have been discussed in the previous chapter. Basically,
the system flow rate is determined by the fan operation point which is defined as
the intersection point of the fan performance curve and the system pressure drop
curve. The system pressure drop is the sum of the pressure drop over the inlet,
turns, air filter, EMI plates, boards, and exit.
This chapter will discuss the effects of the system configuration on the system
flow rates and various subjects related to the system thermal design.
6.1 EFFECTS OF SHELF CONFIGURATION ON SHELF FLOW
The effects of system height and/or depth on the system flow rates [1] will be dis-
cussed. As stated above, the system flow rate depends on the system pressure drop
(i.e., the system flow resistance). Figure 6.1 which was reproduced from Figure 5.5
shows the simple sketch of a representative telecommunication rack and its flow
resistance over the system. The individual flow resistance is as follows:
R1 = entrance
R2 = inlet section
R3 = 90 degree turn
R4 = air filter
R5 = card cage, including 2 EMI plates
R6 = 90 degree turn
R7 = exit section
R8 = exit and
Rt = total system flow resistance
= R1 + R2 + R3 + R4 + R5 + R6 + R7 + R8
The flow resistance can be expressed as follows
DP = K (0.5 ρ V 2) = R Q 2 (6.1)
where K is the loss coefficient, R (= 0.5 K r/A2) is the flow resistance, Q (= AV) is
the volumetric flow rate, A is the cross-section area, and V is the average velocity
across this cross-section area.
F igure 6.1 Sketch of typical rack and its flow resistance.
This study is to examine the effects of various parameters of the shelf configura-
tion on the system flow rate. The results predict the changes of the flow rate over the
value of the baseline configuration which is a known factor. The purpose of the analy-
sis is to develop a family of curves serving as the design guidelines, which can quickly
predict the effect of the changes of individual parameters on the system flow rate.
6.1.1 Effect of Shelf Height
Figure 6.2 illustrates the system configuration under consideration. Hi is the

height of the inlet and exit. It should be noted that the height of the inlet can be
different from that of the exit. For simplicity, both the inlet and exit assume to
have the same height. Additional assumptions are as follows:
1. All flow resistance from R1 through R8 of the baseline configuration

are known.
2. Except the height, all other dimensions of the shelf remain unchanged.
3. Flow resistance of other sections or parts of the system are unchanged.
From Equation (6.1), one obtains the following relationship for the inlet and
the exit
Rα (1/H 2 ) (6.2)
where H is the height of the inlet and the exit sections.

System Thermal Analysis — Rack/Cabinet (Part II) • 73
Figure 6.2 Sketch of shelf with various inlet/exit height.
Based on the above assumptions,
R /( R)b1 = (Hi/H )2 (6.3)
where (R)bl and Hi are the flow resistance and inlet/exit height of the baseline
configuration. Equation (6.3) states that increasing the inlet height (H > Hi) will
decrease the inlet flow resistance. The next step is to calculate the new total sys-
tem flow resistance, Rt based on the calculated flow resistance for the inlet/exit.
Once the new system flow resistance is known, the system new flow can readily
be determined by the following equation
Q/(Q)b1 = [(Rt)b1/ Rt] 0.5 (6.4)
The following example is to show how to compute the new system total flow
resistance and the flow rate:
Example 6.1: For H/Hi = 1.5,

From Equation (6.3), one obtains
R18 = 0.4444(R18)b1 (6.5)
If (R18)bl (= R1 + R8)bl = 0.1(Rt)bl for the baseline configuration, then the

new total system flow resistance for H/Hi = 1.5 is
Rt = 0.1 × R18 + 0.9(Rt) = 0.94444(Rt)bl (6.6)

and the new volumetric flow rate with H/Hi = 1.5 is
Q/(Q)bl = [(Rt)bl / Rt] 0.5 = [0.94444] 0.5 = 1.029 (6.7)
Equation (6.7) implies that the inlet/exit height increases by 50%, the system
flow rate increases by about 3% for (R18)bl = 0.1(Rt)bl if other parameters of the
system remain unchanged.
Following the above procedure, the family of the curves is developed as
shown in Figures 6.3 and 6.4 for the new flow resistance and the new flow rate,
respectively.
Figure 6.3 Inlet/exit height versus system flow resistance.
Figure 6.4 System flow rate ratios over inlet/exit height ratios.
6.1.2 Effect of Shelf Depth
Figure 6.5 will be used to examine the effect of the shelf depth on the system
(shelf) flow rate. The analysis is based on the following assumptions:
1. All flow resistance from R1 through R8 of the baseline configuration

are known.
2. Except the depth, all other dimensions of the shelf remain unchanged.
3. Flow resistance of other sections or parts of the system is unchanged.
From Equation (6.1), one obtains the following relationship for the card cage
plus with two EMI plates
R5 /(R5)bl = [(Ac)bl / Ac]2 (6.8)
where Ac (= W*D) is the cross-section area of the card cage normal to airflow
direction which is identical to the cross-section area of the air filter. W is the
width of the shelf and D is the depth of the shelf.
Figure 6.5 Sketch of shelf with various depth.

Equation (6.8) is further reduced to
Rα (1/D2 ) (6.9)
Similar to the above calculating procedure, the next step is to compute the new
system flow resistance and finally the new system flow rate. The effects of the shelf
depth on the flow resistance and the flow rate are given in Figures 6.6 and 6.7.
Figure 6.6 Shelf depth versus shelf flow resistance.
Figure 6.7 System flow rate ratios over shelf depth ratios.
6.1.3 Design Optimization Procedure
For convenience to users, the new family of curves is developed based on R1278
and R45. The R1278 represents the combined flow resistance of the entire sec-
tion of the inlet and the exit. Similarly, the R45 corresponds to the combined flow
resistance of the card cage (over printed circuit boards), two EMI plates and the
air filter. The new family curves are presented in Figures 6.8 and 6.9 for the effects
due to varying height and depth, respectively.
F igure 6.8 Effect of inlet/exit section height on system flow rate.
F igure 6.9 Effect of depth on system flow rate.

The design optimization procedure can be described as follows:
1. Performing CFD analysis or testing on the initial baseline configura-

tion to obtain the system flow rate, Q0 and all flow resistance (R1
through R8)bl. The obtained results will be the starting point in the
above figures.
2. Increasing the height of the inlet and/or exit section to a desired value
while keeping the shelf depth unchanged. The new system flow rate, Q1¢
(= Q1/Q0) (dimensionless) at the adjusted height can be computed from
Figure 6.8, and this new value will serve as the starting point of Figure 6.9.
3. Varying the depth to the desired value with the height of the inlet and exit
section fixed at the value of Step 2. The system flow rate, Q2¢ (= Q2/Q1)
(dimensionless) for the adjusted depth can be determined from Figure 6.9.
4. The system final dimensionless flow rate is the product of Q1 and Q2
and the dimensional flow rate is Q0 ´ Q1¢ ´ Q2¢.
It should be noted that the sequence of Steps 2 and 3 can be interchangeable.

The present analysis is to provide a quick result to evaluate the effects of the
parameters of the shelf configuration on the system flow rates. The analysis is
based on the assumption of the flow resistance for other sections of the system
remain unchanged. The guideline with the dimensionless form is developed for
general application.
The following example will show the calculation procedure about how to uti-
lize the above two figures to obtain the new system flow rate.
Example 6.2
Given: Baseline configuration gives Q0, R1278 = 0.5 Rt and R45 = 0.3 Rt
Find: the new total system flow rate when the inlet/exit height is increased to two
times of the baseline configuration and the depth is also increased to 1.5 times of
the baseline configuration.
Solution
Step 1 to compute Q1¢:
from Figure 6.8, we have Q1¢ = Q1/Q0 = 1.27
Step 2 to calculate Q2¢:
From Figure 6.9, we have Q2¢ = Q2/Q1 = 1.095
The total dimensionless system flow rate Q2 is 1.39 (= 1.27 ´ 1.095) Q0 which
is 39% higher than the flow rate at the base line configuration.
Example 6.3
A 19 in. telecommunication shelf with 12U height is under consideration. The
height of the inlet and exit is 1U. The system includes 14 slots and 12 axial fans.
The individual fan power consumption is 24.96 W. The results from the CFD
analysis on this baseline configuration indicate that (R18)bl = 0.5447(Rt)bl and
(R45)bl = 0.3(Rt)bl. Examine the effects on the system flow rate if the inlet and
exit height is increased to 1.5U, which results in the total height of the shelf of
13U (Figure 6.10).
For H/Hi = 1.5 and R18 = 0.5447 Rt

From Figure 6.3, we have Qnew/Qbaseline = 1.2.
The result indicates that the flow rate is increased by 1.2 times over the flow
rate at the baseline configuration when the height of the inlet and exit is increased
from 1 to 1.5U.
The potential opportunities due to the increased the system flow rates are as
follows:
1. Reducing the fan speeds to save the energy consumption and achieve
the system energy efficiency by keeping the system flow rate at the base-
line configuration.
2. Utilizing this increased the flow rate to reduce component tempera-
tures, which leads to higher system reliability. In addition, the results
also open the opportunity of using lower component temperature lim-
its of the commercial grade components, which are less expensive.
Figure 6.10 A telecommunication shelf under consideration.

To examine how much energy can be saved, one must understand the opera-
tion of the fan laws given below
Q = Φ ND3 (6.9a)
P = λρ N 3D5 (6.9b)
p = ψρ N 2D 2 (6.9c)
The variables involved in the above equations are fan size D, rotational speed
N, gas density r, volumetric flow rate Q, pressure p, power P, and the fan effi-
ciency h. Variables F, ψ, and λ are the constants for geometrically and dynami-
cally similar operation and are also referred to as the flow coefficient, pressure
coefficient, and power coefficient, respectively.
To maintain the same Qbaseline at H = 1.5Hi, the new fan speed and power con-
sumption can be computed according to the above fan laws as follows:
Nnew/Nbaseline = (1/1.2) = 0.83

Pnew/Pbaseline = (0.83)3 = 57.9%
Total energy saving = 12 ´ 24.96 ´ (1 – 0.579) = 126.1 W.

Total annual energy saving = 365 ´ 24 ´ 126.1/1000 = 1104.6 KW Hr.
For D/Di = 1.5 and R45 = 0.3 Rt:

The new configuration under consideration now is that the depth is increased by
50% while keeping H/Hi = 1.5.
From Figure 6.4, we have Q¢new/Qnew = 1.1
Q¢new/Qbaseline = (Q¢new/Qnew) ´ (Qnew/Qbaseline) = 1.1 ´ 1.2 = 1.32

N¢new/Nbaseline = (1/1.32) = 0.758
P¢new/Pbaseline = (0.758)3 = 0.435
Total system energy saving = 12 ´ 24.96 ´ (1 – 0.435) = 169.2 W

Total annual energy saving = 365 ´ 24 ´ 169.2/1000 = 1482.2 kW Hr
The above two examples clearly demonstrate the value of the family curves
presented in the above figures which allows obtaining quick solutions.
6.1.4 Accuracy of Results
The above family of curves generated [1] is based on the relationship between the
pressure drop and the flow resistance as described by Equation (6.1). The simpli-
fied analysis is also based on the assumptions listed in Section 6.1.2. The goal of
the present analysis is to provide quick solutions to examine the effects of shelf
height and/or the shelf depth on the system flow rates so that the system can be
optimized to achieve the energy efficiency.
The CFD analysis has been performed on several cases to evaluate the accu-
racy of the present analysis [1].
Case 1: R1278 = 0.596 Rt

The height of the inlet and exit of the system under consideration as shown in
Figure 6.11 is 1U for the baseline configuration.
Hot
Air
Boards EMI
(60%
Open)
Fans
Air
Filter
Cold
Air
F igure 6.11 System under consideration.

The detailed results from the CFD analysis are available for the baseline con-
figuration. Among them, the flow resistance for the inlet/exit section (R1278) is
0.596 Rt. The additional CFD analyses were also performed for the cases with
2U and 3U inlet/exit heights. The detailed CFD analysis results were given in
Tables 6.1 and 6.2. The former lists the pressure drop over individual sections of
the shelf, and the latter presents the system flow rate and pressure drop.
Table 6.1 CFD—Pressure Drop over Various Sections of Shelf
Configuration—1U Configuration—2U Configuration—3U

inlet/exit inlet/exit inlet/exit
(in of water) (in of water) (in of water)
Inlet/inlet section 0.3888 0.2164 0.142

+exit/exit section
System DP 0.6524 0.5808 0.5476
DP% 100% 89.0% 83.9%
R1278 for baseline configuration (Configuration – 1U)0.3888/0.6524 = 0.596.
Table 6.2 CFD—System Flow Rate and Pressure Drop
System Pressure System Air flow CFM ratio over

DP (in of water) rate (CFM) 1U baseline
Configuration—1U Inlet/exit 0.6524 287.2 1.0

(baseline)
Figure 6.12 Inlet/exit section height versus system flow rate.

Table 6.3 Comparison between CFD and Graphical Results
Height ratio Graphical flow rate ratio CFD flow rate ratio % Difference
1 (1U) 1.0 1.0 0.00

2 (2U) 1.345 1.349 −0.30
3 (3U) 1.458 1.495 −2.47
As can be seen from the above tables, the system flow resistance or system
pressure decreases and the system flow rate increases, respectively, when the
height of the inlet and exit increases.
Following the calculation procedure described in Section 6.1.1, an additional
curve for R1278 = 0.596 Rt is added to Figure 6.8, and the new charts are pre-
sented in Figure 6.12. The reason of creating the new curve for R1278 = 0.596 Rt is
to get an accurate result from the charts. For practical applications, the predicted
results can be obtained by the interpolation between curves for R1278 = 0.5 Rt
and R1278 = 0.7 Rt from Figure 6.8.
The graphical flow rate ratios over the height ratios from the above figure are
given in Table 6.3. Table 6.3 also includes the CFD results which are obtained from
Table 6.2. The graphical representation of Table 6.3 is presented in Figure 6.13.
Figure 6.13 Comparison between graphical and CFD results for

R1278 = 0.596 Rt.
Case 2: R1278 = 0.237 Rt

Another case with R1278 = 0.237 Rt was analyzed. Similar to Case 1, an addi-
tional curve for R1278 = 0.237 Rt is added to Figure 6.12 as shown in Figure 6.14
in order to get an accurate result.
Figure 6.14 Inlet/exit section height versus system flow rate.
The comparison between the CFD and graphical results are presented in Table
6.4 and Figure 6.15.
Graphical flow
Height ratio CFD results rate ratio CFD flow rate ratio % Difference
1 (1.4U) 410 CFM 1.0 1.0 0.00

1.47 (2U) 434.5 CFM 1.067 1.060 0.66
1.79 (2.5U) 440.9 CFM 1.093 1.075 1.67
Figure 6.15 Comparison between graphical and CFD results for

R1278 = 0.237 Rt.
Case 3: R5 = 0.44 Rt
This case is to study the effect of the shelf depth on the system flow rate. As before,
an additional curve for R5 = 0.44 Rt is added to the charts in Figure 6.7, and the
charts are shown in Figure 6.16. The comparison between the CFD and graphical
results is presented in Table 6.5 and Figure 6.17. Again, it should be noted that the
graphical results can be obtained by interpolation between curves for R5 = 0.3 Rt
and R5 = 0.5 Rt.
F igure 6.16 Shelf flow ratios versus shelf depth ratios.
It seems from Figure 6.17 that the difference between the CFD and graphi-
cal results increase as the shelf depth ratio increasing. This is possibly due to
the assumptions made in the analysis. One of the assumptions states that the
flow resistance of other sections or parts remains unchanged. This assumption
is no longer valid if the depth ratio is very large. In summary, it is seen from the
above three case studies that the graphical and the CFD results are in an excellent
agreement.
Graphical flow
Depth ratio CFD results rate ratio CFD flow rate ratio % Difference
1 (10 in.) 547.1 CFM 1.0 1.0 0.00

1.2 (12 in.) 598.1 CFM 1.075 1.093 −1.65
1.5 (15 in.) 623.8 CFM 1.150 1.140 0.88
2.0 (20 in.) 640.0 CFM 1.222 1.170 4.44
2.4 (24 in.) 644.4 CFM 1.254 1.178 6.45
Figure 6.17 Comparison between graphical and CFD results for R5 =

0.44 Rt.
6.2 VARIOUS SUBJECTS RELATED TO SYSTEM THERMAL

DESIGN
There is no “one size fits all” design approach. This section further discusses the
effects of various parameters on thermal design of telecom systems.
6.2.1 Effect of Series Fan Trays
It is a well-known fact that the fans in series will provide high available head (pres-
sure), while fans in parallel will give a high flow rate. The typical fan performance
curves for the fans in series and in parallel are illustrated in Figure 6.18 [2]. It
should also be noted that the flow rate is always increased for the fans in series.
However, the rate of increase decreases as the number of the fans increases.
It is of interest to also examine the effect of the fan trays in series. It should be
noted that the fans are arranged in parallel in the fan tray. Two cases, as shown
in Figure 6.19, are under consideration. The number of the fan trays is increased
from one to three in both cases. However, each shelf has one fan tray in Case B.
The flow rates from the CFD analysis are normalized with respect to one fan tray
case. The comparison between Cases A and B are presented in Figure 6.20. The
rate of increase for Case A is generally in agreement with the expectation. On the
other hand, the flow rate in Case B is only increased slightly as the number of fans
increases. The main reason is that the length of the flow path is also increased
with the number of the fan trays in Case B.
Figure 6.18 Performance curves of fans in series and parallel.
F igure 6.19 Fan trays in series.

Figure 6.20 Comparison of flow rate between Cases A and B.
6.2.2 Factors Affecting Component Temperature
The most critical factors affect the temperature of the components on the board
are the heat sink configuration and the airflow rate. However, both factors are too
complicated to give some sort of general design guidelines. Therefore, this sec-
tion examines other factors which are easily identified and may affect the compo-
nent temperature, while the heat sink configuration and the airflow rate remain
unchanged.
Thermal Interface Material (TIM)
The thermal interface material (TIM) is often inserted between the heat sink and
the component in order to enhance the heat transfer from the component to the
heat sink as illustrated in Figure 6.21. It is a well-known fact that the thermal resis-
tance always exists at the interface when two materials are brought together.
The total thermal resistance across TIM can be expressed as follows [2]:
Rt = Rth + Ri1 + Ri2 (6.10)
where
Rt = total thermal resistance

Rth = thermal resistance due to TIM thickness (=L/KA)
Ri1 = interface thermal resistance between heat sink and TIM
Ri2 = interface thermal resistance between component and TIM
Figure 6.21 Sketch of heat sink and component.
Rth is a function of material thermal conductivity, thickness, and void fraction

as well as its distribution, of TIM. Ri1 and Ri2, which can only be determined
experimentally, depend on bond process and surface characteristics and mate-
rial properties of Solids 1 and 2. It is difficult to compute Ri1 and Ri2 separately.
Therefore, Ri1 and Ri2 can be combined into a single value, and Equation (6.10)
can be rewritten as follows:
Rt = Rth (1 + C/Rth) (6.11)
where C represents the combined interface thermal resistance at two interfaces.
i. The C value becomes significant when Rth is small; this case implies
that thermal conductivity of TIM is very large or the thickness of TIM
is very small.
ii. The C value is negligibly small when Rth is large; this condition indi-
cates that thermal conductivity of TIM is very small or the thickness of
TIM is very large.
Equation (6.10) has been validated by the following experimental data pre-
sented in Figures 6.22 [2] and 6.23 [3].
The experimental data indicated that the combined interface resistance C
could vary from 10% to 80% of the theoretically calculated value (Rth) as epoxy
conductivity increased from 0.27 to 1.93 W/m °C. The following correlation was
developed by Yeh (2) to predict the total thermal resistance over an epoxy with
thickness ranging from 0.018 to 0.022mm:
Rt = 0.26(Rth)2 + 0.62Rth + 0.15 (6.12)
where the unit of the thermal resistance is cm2 °C/W.

Figure 6.23 shows the thermal resistance as a function of the applied pres-
sure. At low pressure <15 psi, the total interface thermal resistance can be more
than 50% of the thermal resistance of the material itself. The delicate leads and
Figure 6.22 Measured and calculated thermal resistance of an

epoxy joint.
Figure 6.23 Thermal pad joint thermal resistance versus pressure.

solder joints can be broken from an excessive pressure. The maximum pressure
for most heat sink clip is about 25 psi. However, often 15 psi is tolerated for most
components.
There are many types of TIMs, including, but not limiting to, thermal grease,
thermal paste, thermal gel, thermal epoxy and thermal pad. The characteristics
of thermal grease, thermal gel, or thermal paste are alike and can be lumped
into a single group and “thermal grease” will be used as an example for further
discussion:
i. Thermal Grease
–Only a small amount of thermal grease should be used, to just fill the void
spaces (cavity of rough surfaces) and still allow the high points of the
metal surfaces to touch each other.
–Too much grease may create voids within the heat compound and
increase the thermal resistance across the thickness of the grease.
–Care must be taken to insure no contamination or any foreign subjects
in grease.
–Thermal grease has a tendency to migrate in high temperature and to
evaporate in low pressure or vacuum environments. Once vaporized,
they may redeposit on adjoining surfaces or disappear altogether.
–It is generally very difficult to apply thermal grease uniformly over a
large area (> 3 × 3 in.).
ii. Thermal Epoxy

–Curing temperature and/or pressure may be important to epoxy bonding.
–Voids always exist within epoxy (% of voids depends on application
processes).
–Large voids and also void distribution may have a significant effect on
thermal resistance over epoxy.
iii. Thermal Pad

–Thermal pad is probably the only TIM that it can be used to address the
component height tolerance issue because of its compressibility.
–Thickness required for a thermal pad is generally larger than those of
thermal grease or epoxy.
–A large factor in reducing pad thermal resistance is the application of
pressure.
–Increasing pressure not only reduces the thermal contact resistance but
also can reduce the material thermal resistance by reducing the material
thickness.
–Increasing pad thermal conductivity generally will decrease material
compressibility.
Figure 6.24 [3] also shows the pressure limits for PQFP CPU and for most heat
sink clips. The delicate leads and solder joint can be broken with excessive pressure.
It is concluded that the thermal resistance at two interfaces of a joint can-
not be ignored in determining the total joint thermal resistance which is always
greater than Rth based on the material’s thickness alone.
F igure 6.24 Thermal resistance over various pad joints.
Thermal Conductivity of Printed Circuit Board
The board provides the structure support and electrical connections to all compo-
nents. However, it is seldom known that the board is the largest heat sink available
to the components. Increasing the board thermal conductivity increases the heat
conduction spreading along the board and reduces the component temperatures.
Therefore, the thermal conductivity of the board has a great effect on the compo-
nent temperature.
The thermal conductivity of the board depends on the construction and mate-
rials of the board. To improve the board thermal conductivity, one may consider
to employ polyimide in replacing of FR-4 in the board construction because the
conductivity of polyimide (k = 0.52 W/m °C) is about 50% higher than that of FR-4
(k = 0.35 W/m °C). The practical approach in the thermal analysis is to determine
the so-called equivalent thermal conductivity for the XY plane and the Z plane
which are function of the copper contents.
Baffles in Inlet/Exit Section
The baffle (often is also referred to as deflector) as shown in Figure 6.25 is used to
eliminate or minimize the flow recirculation zones in order to improve the system
flow. The effect of the baffle on the system flow in the forced air-cooled system
with fans is relatively limited. The main affect of the baffle is slight changes of the
flow distribution over the board which may have some effects on the component
temperatures. The baffles increase the airflow distribution to the rear section of
the board and therefore slightly decrease the temperature of those components
closer to the back plane. To be effective, the baffle must be added to the inlet and
exit section of the shelf simultaneously.
F igure 6.25 Sketch of shelf with baffles.
However, the effect of the baffle is much profound in the natural convec-
tion environment (no fan condition). In other words, the shelf flow rate will be
increased in the case without the baffles. This is due to the fact that the baffle
increases the system pressure drop. The available pressure head is pretty much
fixed in the natural convection conditions once the system height and the heat
load are defined.
Fillers in Card Cage

The understanding of the effect of the empty slots in the card cage on the ther-
mal analysis and the tests is extremely important. The detailed full scale thermal
model of the rack is sometimes too large for the practical applications. Simi-
larly, the fillers, which are often referred to as the dummy boards, are frequently
required during the system thermal tests to insure the proper flow distribution to
all slots (or boards).
The system under consideration is a low profile shelf as given in Figure 6.26. The
CFD analysis is performed for both cases; one with fully filled card cage (Case A)
and another with three-board card cage (Case B) as shown in Figure 6.27. The
F igure 6.26 System under consideration.
Figure 6.27 Top view of fully filled card cage and three-board
card cage.
target board is sandwiched by two neighbor boards. This arrangement is needed

to insure the accurate results for the target board which will be affected by the
neighbor boards. There is no noticeable difference in the predicted results for the
target board between Cases A and B (with empty slots in the shelf) in the normal
operation (all fans on) and one fan failed condition.
The results indicate that the target board is very insensitive to the operation
conditions in far away locations. In other words, the results of the target boards
are relatively independent of the performance of the other boards and fans
located at the distance from the boards under consideration. The main reason
is that the board thermal performance is generally governed by the fans directly
located under or above the boards under consideration. In this particular case,
the fans are located so closed to the card cage that the airflow distribution to the
boards is directly linked to the fans right below the board under consideration.
As required by GR-63, the components should also meet the respective tempera-
ture limits under the one fan failure. The effect of one fan failure is generally a
local phenomenon and only the board direct above or below the failed fan will
be affected.
The above finding is limited to the forced air cooling with the fans in the sys-
tem. The attention is now turned to the equipment without the fan. The compo-
nents on the board in such a system are subjected to the thermal radiation and
natural convection. The passive cooling is simple, cost effective and reliable. The
system under consideration as shown in Figure 6.28 is a 4U shelf, which includes
fifteen 2U boards. The shelf without the fan is cooled by natural convection. Ther-
mal radiation effect must always be included in the thermal analysis in a passive
cooling condition.
The purpose of the analysis is to examine the effects of the empty slots or the
fillers on the temperature of the components on the target board. The CFD analy-
sis is first performed on the fully heated card cage and the results are given in
Figure 6.29. The maximum component temperature for the target board is 122°C
and for two neighbor boards are 112°C and 114°C for Cards A and B, respectively.
These results are considered to be baseline values.
Figure 6.28 A passive cooled 4U shelf.

Figure 6.29 Results for fully heated boards in shelf.
The next step is to turn off the power to all the boards with exception of the
target board and Cards A and B. The results for the three-heated board in the fully
filled card cage are given in Figure 6.30. Due to radiation, heat is transferred from
the heated boards to the zero power boards which results in significant reduction
in the temperature of the components on the three powered boards.
The analysis is further extended to the case without the zero power fillers in
the card cage. In other words, all zero power cards in the above case are removed,
and the card cage only has three boards plus a lot of empty slots. The component
temperatures on the target board and Cards A and B are presented in Figure 6.31.
As can be seen from the figure, the component temperatures are reduced signifi-
cantly due to the empty slots. The reasons for the lower component temperatures
are because one side (opposite to the target board) of Cards A and B is basically
exposed to the ambient air induced by the natural convection and also radiation
to unheated structure in the surrounding.
Examining all results given in this section, the effects of fillers can be sum-
marized as follows:
1. Forced Air Cooled Systems

No noticeable difference in the component temperatures between the
cases of the fully loaded and partially filled card cages under the force air
cooling with the fans.
The main reason is that no mixing of airflow from fan exits due to the
limited space existing between the card cage and the fan tray. Furthermore,
the thermal performance of the individual boards is generally governed by
the performance of the fans directly located below or above.
2. Passive Cooled Systems
No significant difference in the component temperatures between the cases
with the zero power fillers and the empty slots under the natural convec-
tion conditions. However, both cases significantly reduce the component
Figure 6.30 Results for three-heated-board card cage with zero

power on fillers.
Figure 6.31 Results for three-heated-board card cage without fillers.

temperatures of the heated boards as compared with the fully loaded card
cage. Therefore, the thermal simulations and/or tests for the passive cool-
ing should be conducted under the actual operation conditions with the
full power on all boards in the card cage.
6.3 THERMAL ENHANCEMENT
To meet the component temperature limits, various methods can be considered

to reduce the temperature of the components on the board. The typical steps
which can be taken to improve the thermal performance of the telecommunica-
tion racks are described in the following section.
1. Board Level:
(i) Optimizing heat sinks and board layout
(ii) Enhancing board flow management (e.g., adding baffles to guide flow)
2. System Level:
(i) Optimizing system flow path for minimizing system DP in order to
increase flow rates
(ii) Utilizing perforating plates or baffles in inlet and/or exit section to
regulate the proper flow rates to the boards
(iii) Increasing the surface area of the air filter
(iv) Upgrading the fans without changing the size of the fan tray if possible
(v) Increasing the height of inlet and/or exit section of the system if
possible
Among all above factors, increasing the height of inlet/exit sections may be the
simplest way to do with most benefits as long as the total height of the rack
remains less than 7 ft which is the limited for North American market.
One critical thermal issue for the current high power and dense packed board
is the space available on the board. The large oversized heat sinks needed to
accommodate high power components further compound the thermal prob-
lems. The modularized board concept as presented in Figure 6.32 is developed to
address the above problems.
The unique feature of the modularized board is to add a thin metallic cover
which provides an extra large surface for heat transfer in air cooled systems or
interfaces with the cold plates in liquid cooled equipment. Other advantages of
the cover are as follows:
1. Protection for components during handling and/or shipping

2. Prevention from fire spreading between boards to meet GR-63 requirement
As noted previously, the board is the largest heat sink to all components on
the board. To utilize this large heat sink efficiently, it is recommended to place as
many low profile and low power chips on the bottom side of the board as possible
(up to 15% to 20% of total power of the board). This arrangement not only lowers
the surface heat flux on the top side of the board because of reduction of the num-
ber of the components but also provides additional space needed for mounting
Figure 6.32 Modularized board.
the heat sinks. The 6 to 8 mm gap on the bottom side of the board also give the
enough space for adding the low profile heat sink, including thin metal plates to
the components as needed.
As stated in the above section, the board cover can be utilized as the heat sink
to all components on the board if needed. Along this concept, a single large heat
sink can be adapted to cover multiple components as shown in Figure 6.33. The
current practice for attaching an oversize heat sink to the component is that the
heat sink is mechanically mounted to the board which requires punching holes
on the board. The holes on the board will greatly affect the board routing. Gener-
ally, four holes are needed to mount a heat sink to the component.
For a typical high power board, the board is populated with the components,
active and passive components. There is a limited space available for mounting
heat sinks. The single large heat sink will reduce the number of holes needed
on the board. It will require 12 heat sink holes on the board for three chips as
F igure 6.33 Large heat sink for multiple chips.
illustrated in Figure 6.33. However, the number of heat sink holes on the board
will be reduced to no more than eight if a single large heat sink as shown in the
figure is employed. In addition, the single large heat sink will also make chip
temperatures more uniform due to heat spreading along the base of the heat
sink.
6.4 THERMAL CONSIDERATION IN DATa CENTER OR

CENTRAL OFFICE
The inlet and/or exit location of the airflow for a telecommunication system gen-
erally does not affect the system thermal performance for stand alone equipment.
However, most telecommunication equipment is either deployed in a data center
or central office which hosts hundreds of similar equipment. The inlet and/or
exit location of the airflow will have some impacts on the system thermal perfor-
mance due to present of neighbor systems.
Rows of rows of the systems are placed in the data center as shown in Figure 6.34
which represents a typical data center layout. As can be seen, the air inlet is
located at the front side of the equipment facing to the cold aisle, and the hot air
exits from the rear of the system to hot aisles. CRAC stands for computer room
air conditioner. The system with the side inlet and exit is becoming unacceptable
in the today’s data centers because the hot exit air will significantly effects on the
thermal performance of the neighbor racks.
F igure 6.34 Equipment layout in data centers.
The cold air entrance at the top of the system should be avoided because the
some of hot exhaust air in the hot aisle will circulate to the front of the rack as
illustrated in Figure 6.35. Additional devices such as blanketing panels will be
needed to prevent the hot air from circulating back to the front of the rack if the
cold air inlet locates at the top of the racks.
F igure 6.35 Hot exit air circulation.

REFERENCES
1. Yeh, L. T., “Effects of System Height and/or Depth on System Flow Rates for
Telecommunication Racks”, to be published.
2. Yeh, L. T. and Chu, R. C., Thermal Management of Microelectronic Equip-
ment, ASME Press, 2002.
3. Latham, C. A., “Thermal Resistance of Interface Materials as a Function of
Pressure”, Electronic Cooling, Volume 2, No. 3, 1996.
Chapter 7
System Thermal
Analysis — Small Box
The typical telecommunication boxes for the indoor applications are small with
the height less than 4RU (7 in.). The “RU” stands for the rack unit which is 1.75
in. high. However, for simplicity, “U” is commonly used to replace “RU.” One of
the unique features of these typical boxes is the printed circuit boards being in
horizontal positions. As in the rack configurations, the cooling schemes of the
system can be either force air cooling with fans or passive cooling under natural
convection.
The challenges of the thermal design in the box configurations, especially for
low profile box are due to the limitation of the space available and the boards
being installed horizontally. The latter is especially critical to the natural convec-
tion environments.
7.1 ACTIVE COOLED BOXES
For a forced air cooling, the fans are typically placed on one side of the box. The
cold air enters from one side and hot air leaves at the opposite side of the box. The
design and operation are still subjected to GR-63 requirements.
As in any actively cooled indoor equipment, the air filter is generally required.
Air inlet and exit are through a perforating plate and the percentage of opening
is generally less than 65% in order to meet the EMI (electromagnetic interfer-
ence) requirements. In addition, the percentage of the opening must also meet
the structural requirements of the box.
Figure 7.1 shows a typical example of a 19 in. (width) box. The 19 in., which
is the most common one used in telecommunication industry, is a standardized
frame or enclosure for mounting telecom equipment. The 19 in. space includes the
mounting posts. The actual width of the 19 in. equipment is 17.5 in. (444.5 mm).
The depth of the box ranges from 220 to 300 mm. The height of the box can be
varied between 1 U and 4 U.
The fan selection is extremely important for the low profile box. Ultimately, the
fan performance must meet the system thermal requirements. The selection of the
fan is also limited by the available space inside the box. The size and the number
of printed circuit boards (PCBs) are system dependent. For example, eight small
PCBs as given in Figure 7.1 can be replaced by 4 large PCBs or any combination
F igure 7.1 Sketch of forced air cooling of box.
of small and large PCBs. Another design requirement is that the boards must be
able to be removed or plugged into the box freely. The minimum required airflow
rate (CFM) is in general about one fifth of the system power in watts. It should be
noted that the required airflow rates for a given system strongly depend on the
power and types of individual components as well as the board layout (component
placement).
The major thermal concern for the small box configuration is that the fans
are too close to the boards due to the limited space inside the box. Therefore, the
thickness of the fan becomes an issue. The current system employs 28-mm-thick
fans. If the system presented in figure is employed to the 4U shelf, the 80-mm fans
can be used for such cases.
The location of the fans is actually also depending on the system configura-
tion, especially the available space. Another example of force air cooled box is
presented in Figure 7.2. The system under consideration is a 1-U box with the
fans installed at the rear of the box. The box contains a large board with various
types of components on the board. The fans in the current system are installed in
the rear of the box which draws air from the vents at both side walls as shown in
Figure 7.2. Sometimes, the ambient air can also be drawn from the perforation at
the top of the box to enhance the thermal performance of the box.
In short, the placement of fans is system dependent and should be placed at
the least system resistance path. In addition, heat sinks are generally required for
most components, especially for those high power components.
Though the boards in the horizontal position are the most common cases in a low
profile box, the vertical boards as shown in Figure 7.3 have been adapted. However,
the height of the boards is limited to less than 2 U in order to make room for the
fan tray, and inlet and exit sections. The total height of the box would probably be
increased from 4 U to 5 U, as illustrated in Figure 7.4. In order to increase the board
surface area for mounting the components, the subboards or daughter boards are
often attached to the main boards. This design can support the system power up to
about 350 W.
System Thermal Analysis — Small Box • 105
Hot
Fan Fan
Air
Power
Supply
Air Air
Power
Supply
Components SFP if
existed
F igure 7.2 1 U box with forced air cooling.
7.2 PASSIVE COOLED BOXES
Heat transfer by natural (or free) convection has long been considered as one of
the most cost effective and reliable cooling methods. Natural convection with air
has many practical engineering applications and is of special interest to the cool-
ing of electronic equipment. The advantages of air cooling by natural convection
are simple, reliable, and cost-effective.
It should be noted that the fan tray in Figure 7.3 or Figure 7.4 can be removed,
and the system becomes pass cooled equipment. Under the natural convection
F igure 7.3 Vertical boards in low profile shelf.

Figure 7.4 Cross-section view of vertical boards in low profile shelf.
cooling, the system power as described in the above figures is reduced from 350 W
at forced air cooling to about 150W and the system height from 5U to 4U. Further-
more, the inlet and exit sections in the above figures can also be removed for the
case with or without fan tray under some limited conditions where sufficient spaces
above and below the box under consideration are available to minimize the effects
from the surroundings on the system.
The 1-U horizontal box under the natural convection is typically applied to
the low power systems in the past. With increasing speeds and functionalities,
the system power for a 1 U box has increased significantly in the recent years.
Because of the increasing demands on the system performance and reliability, the
thermal design is also becoming even more critical and important.
The overall dimensions of a typical 1 U box are 17.5 in. (444.5 mm) (width) ´
1.75 in. (44.5 mm) (height) with the depth ranging from 8.67 in. to 11.81 in. (220
to 300 mm). The box is generally mounted to an open telecommunication rack or
cabinet, as illustrated in Figure 7.5.
7.2.1 Market Survey
The summary of the survey of the current market for the passively cooled 1 U
horizontal box in telecommunication industry is given in Table 7.1. Table 7.1
Figure 7.5 1 U box in telecommunication rack.
includes the overall dimensions of the individual boxed. As can be seen from
Table 7.1, all products in the current market can only support the system power
around 50 W. The volumetric power density (W/L) is often used to evaluate the
thermal efficiency of the equipment in the natural convection as COP (coefficient
of performance) in the forced air cooled systems.
Passively cooled boxes are generally applied to 1 U or 2 U boxes which severely
limit the available space inside the box. In addition, the horizontal boards make
natural convection even worse. Most of products as mentioned in the above table
are in a sealed box environment with a few exceptions having some venting holes
on the top surface of the box. The heat generated from all components is first trans-
ferred to the enclosed walls and then heat is finally rejected from the walls to the
ambient by natural convection and radiation. To enhance the heat transfer, the
perforation can be made on the top cover and side walls as NEC Pasolink NEO
box. The opening on the top cover is generally limited to less than 60% in order to
meet the EMI requirements. The percentage of openings on the sidewalls should
be much smaller due to the structural requirements. Another way to enhance heat
transfer from components to the cover is often done by applying solid aluminum
blocks which are integrated part of the cover directly to components as Fujitsu
Flashwave 4020.
Table 7.1 Various Products of 1 U Box in Current Market
Products Overall dimensions Power Power density Ambient
(W ´ D ´ H) (mm) (W) (W/l) (°C)

Ericsson Mini-Link HC 483 ´ 280 ´ 44.45 45.0 7.49 −20–60
NEC Pasolink NEO 482 ´ 240 ´ 44.45 47.0 9.14 −5–50
Fujitsu Flashwave 4020 444.5 ´ 228.6 ´ 44.45 43.5 9.63 0–50
7.2.2 Case Studies
The system under consideration is given in Figure 7.6. The overall dimensions of
the box are 444.5 mm (W) ´ 246.4 mm (D) ´ 44.5 mm (H) and the total system power
is 53.13 W. The dimensions for the printed circuit board (PCB) are 378 mm (W) ´
190 mm (D) 1.6 mm (thickness). The component temperature limit which is
also the thermal design limits is generally at 125°C with exception for optical
components such as SFP/XFP.
In addition, the following assumptions are made to facilitate the thermal
analysis
a. The box is suspended in an open air space

b. The ambient air is at 55°C and at the sea level condition
c. The box is limited to passive cooling with natural convection and
radiation
d. The box is made from anodized aluminum with emissivity of 0.81
Item (a) actually implies that effects from other equipment located above and/
or below the current system under consideration are small and can be omitted.
This may not be always true in the actual installation and this issue of environ-
mental effects should be considered in the system design. Since no active cooling
is available, this 1 U horizontal box is subjected to the passive cooling, including
the natural convection and radiation. The analysis has been performed on vari-
ous configurations with aids of Flotherm software [1].
The ultimate goal of the thermal design is to maintain all component tempera-
tures below the respective temperature limits in order to achieve the desired sys-
tem reliability and performance. The heat is transferred to the ambient through
the box surfaces by the natural convection and radiation. The radiation must be
included in the case of the passive cooling condition.
F igure 7.6 Box internal layout.

F igure 7.7 Various heat sinks on board in box.
Instead of using small heat sinks for individual components which are the
general engineering practice, large heat sinks are employed in the present design.
Each heat sink covers a group of components with an exception of the 1/4 brick
power supply. Heat sinks are made from aluminum with thermal conductivity of
240 W/m °C and emissivity of surface coating at 0.95. The thickness of the fins
and the spacing between fins are 2 mm and 6.174 mm, respectively (Figure 7.7).
The temperature distribution over the top surface of the box is shown in Figure
7.8. The maximum temperature on the top of the box is 75.2°C. The component
Figure 7.8 Temperature distribution over top surface of box.

Figure 7.9 Component temperatures on board in box.
temperatures are given in Figure 7.9. All components have met the respective
temperature limits. To reduce the internal air temperature, a series of small holes
on the top surface of the box can be made so that the hot air can vent out the box.
To insure a good EMI protection, the opening on the perforate plate must be less
than 60%.
7.2.3 Other Applications
The above analysis, which is typical practice in the telecommunication industry,

is based on the assumption that the box is suspended in a large open air space. In
other words, there is no impact from the surrounding structures on system ther-
mal performance, which represents an ideal condition. It is of interest in examin-
ing the thermal performance of the box under the real world conditions.
Table Top Deployment
Though the box is typically mounted to a rack as shown in Figure 7.5; however, it
is also often to just place the box on a table. To simulate this condition, a steel plate
with the size of 920.6 mm ´ 399.3 mm ´ 1 mm is added to the thermal model. It
is assumed that the box is in perfect contact with the table top. In order to see the
maximum effect of the plate, the assumption of the zero thermal resistance at the
interface between the plate and box is made. Since the plate is much larger than
the bottom surface of the box, it results in both positive and negative impacts on
the box thermal performance. First, the plate acts as a heat sink to enhance the heat
transfer from the bottom surface of the box to the ambient. On the other hand, the
plate completely blocks the cold airflow from the space below to the box.
Figure 7.10 shows the temperature for the selective components for the case
with the box on a steel table top. The figure also includes the temperatures for the
cases with (1) stand alone box and (2) the box on a plastic table top. As can be seen
from the figure, the stand alone box has the best thermal performance in all three
cases under consideration. Due to low thermal conductivity of the plastic, it limits
the conduction spreading along the plastic table top which not only provides the
thermal insulation to the box but also blocks the cold airflow from the space below
the box. Therefore, it significantly increases the component temperatures inside
the box. Based on this analysis, it is recommended that the box be placed on metal
table in order to achieve a better thermal performance of the box.
The results from Figure 7.10 reveals that material thermal conductivity of the
table top have great effects on the component temperatures inside the box under the
passive cooling conditions. However, the material thermal conductivity of the table
top becomes irrelevant if the space between the box and table top exists as shown
in Figure 7.11. This space can easily be created by adding four or six support legs to
the bottom of the box. The results also indicate that the component temperatures
inside the box will decrease as the space between the box and table top increases and
approaches the results of the stand alone box if the distance from the box to the plate
is infinite.
Boxes Stack-Up
Because of small configuration, the box is sometimes placed in a cabinet or desk-

top. To maximize the available space, boxes are often stacked up to the top of
each other. The boxes stack-up do not have any effect thermally on the forced air
cooled boxes, but it will have severe impacts on the thermal performance of the
Figure 7.10 Component temperatures for boxes with/without plate.

Figure 7.11 Component temperatures versus space between box and

table top.
boxes under the natural convection environments. The heat transfer capability will
almost be reduced by a half for two boxes stack-up. Heat from the top surface of the
bottom box is blocked by the bottom surface of the top box. Similarly, heat from the
bottom surface of the top box is impeded by the top surface of the bottom box.
Following the discussion in the previous sections, the analysis is extended to
the cases with the stack-up boxes on the steel table top, and the results are given
in Figure 7.12. For two boxes stack up, the top box is superior to the bottom box
thermally. This is because the cold airflow is impeded by the large plate. In addi-
tion, there is radiation heat transfer from the top surface of the top box directly to
the surrounding cold environments. The similar trend (the top box is coolest) is
found in the case of the three boxes stack up on the steel table top. The results also
Figure 7.12 Component temperature in single and multiple boxes

stack-up on steel table top.
indicate that the middle box is the hottest one among the three boxes. In general,
the second box from the top one is hottest for the multiple boxes stack up.
The example presented here is to illustrate the expected trends. The actual
numerical values of the component temperatures depend on the system power
and configuration as well as component layout. In general, the component tem-
perature increases in an ascending order from the bottom box with an exception
of the top box.
The analysis is also extended to study the boxes stack up in an open space and
the results are given in Figure 7.13. The component temperatures at the bottom
box are about the same or slightly cooler than those of the top box for the cur-
rent box configuration. Again, the second box from the top box is hottest one in
multiple stack-up boxes.
In addition, the effects of the gap between the boxes on the component tem-
peratures are also examined Figure 7.14 shows the effects of the gap size between
boxes on the component temperatures. As should be expected, the component
temperature decreases as the gap size increasing. In theory, the component tem-
peratures in all boxes will approach to the values of the single box if the gap size
is very large (to the infinity).
Due to the cost of floor space in the data center and/or the central office, the
facility operators will typically place as many equipment (boxes) into a single
rack as possible. Therefore, the configurations presented in Figures 7.13 and 7.14
are much closer to the real world conditions and should be included in the ther-
mal design processes.
Figure 7.13 Component temperatures in single and three boxes stack-up.

Figure 7.14 Effects of gap size on component temperatures in boxes.
REFERENCES
1. Flotherm, a product of Mentor Graphic.

2. Yeh, L. T. and Chu, R. C., Thermal Management of Microelectronic Equipment,
ASME Press, 2002.
Chapter 8
Outdoor Electronic
Equipment
The outdoor electronic equipment includes a large variety of various types of sys-
tems. Among them are the outside plant (OSP) and the remote radio head (RRH)
for telecommunication applications. The OSP refers to all of the physical cabling
and supporting infrastructure such as conduit, cabinets, towers or poles, and any
associated hardware located between a demarcation point in a switching facility
and a demarcation point in another switching center or customer premises. On
the other hand, the RRH describes an operator radio control panel that connects
to a remote radio transceiver via electrical or wireless interface. From the meth-
ods of equipment mounting, the outdoor system can generally be classified into
two groups. One is the ground or vehicle mounted equipment and another is the
tower or pole mounted systems.
8.1 GROUND MOUNTED OUTDOOR ELECTRONIC

EQUIPMENT
For all air-cooled outdoor electronic equipment, heat generated from electronics
inside the enclosure is carried away by the air circulated internally. Heat is then
transferred to the external stream of the ambient air through an air-to-air heat
exchanger and is finally dumped into the ambient.
8.1.1 Thermal Designs of Outdoor Electronic Equipment
All outdoor equipments are subjected to extreme weather conditions, high humid-
ity, high ambient temperature and solar radiation. At the same time, they are
expected to operate with high reliability. A large variety of cooling techniques
are available in cooling of the outdoor electronics, including telecommunications
equipment. These include passive and active forced air cooling, the air condition-
ers or heat pumps, thermosyphons, and phase change materials (PCMs).
The following case studies are used as examples to describe thermal design
methodology for general outdoor electronic equipment:
All electronics are enclosed in a large sealed box [1] with the overall dimen-
sions of 66 in (length) ´ 36 in (width) ´ 28 in (depth). The system as shown in
Figure 8.1 includes 21 power supplies (116 W) and three racks (269.4 W) (also
Figure 8.1 Sketch of system under consideration.
referred to as card cages). The total system power is 385.4 W. Racks A, B, and
C contain 6, 18, and 15 printed circuit boards, respectively. The electronics and
the power supply are cooled by the internal air circulated inside the box. The hot
internal air is then cooled by the ambient air through an air-to-air heat exchang-
ers as illustrated in Figure 8.2. The internal air is first blown into the air duct by
three fans, and is then turned 90° horizontally flowing into the two separate heat
exchangers towards Rack A and B/C racks, respectively. After leaving the heat
exchangers, the internal air makes another 90° turn downward and enters Rack A
and Racks B/C in series, respectively. The air leaving A and C racks flow over the
floor from the front to the rear of the box, and then both air streams then com-
bine together, and finally flow sideways toward the internal fans. The hot internal
air is used to cool the power supply before it is recirculated into the internal air
duct by the fans. Three exhaust fans at the rear of the box draw the ambient air
from the front to the rear of the box.
The goal of the thermal design is to meet the following temperature limits:
a. Component junction temperature less than 105°C

b. Power supply base plate temperature less than 85°C
This outdoor equipment is designed for the worst high temperature environ-
ment described in Reference 2 as given in Figure 8.3. The calculated values given
in the figure are based on the equations developed by Yeh [2]. The figure gives a
daily cycle of the ambient air temperature and the direct solar flux on the vertical
and horizontal surfaces over 24 hours. The maximum air temperature is 120ºF
(49ºF) at 4 PM and the maximum solar incident is 355 Btu/hr-ft2 (1120 W/m2) at a
horizontal surface between 12 and 1 PM. The solar absorptivity and emissivity of
the surface of the box (equipment) are 0.66 and 0.89, respectively. The actual heat
absorption at the surface is the solar incidence time the solar absorptivity.
Outdoor Electronic Equipment • 117
Figure 8.2 Internal and external air flow paths.
Figure 8.3 Hot dry daily cycle of air temperature and solar flux.
A transient thermal analysis is performed on the system over a 24-hour period.

The internal air temperature histories at various locations, including at the inlet to
Racks A, B, and C are computed as presented in Figure 8.4 and the maximum air
temperature to the card cages is 60.2°C at 4 PM. The airflow rates to Rack A and
Racks B/C are 113.6 and 290.7 CFM, respectively. The calculated air temperatures
and the flow rates are then applied to individual racks to compute the junction
temperature of the components under the steady state condition. The results indi-
cate that all components and the power supply have met the respective tempera-
ture limits (105°C for microelectronics and 85°C for power supply base plate).
Since the outdoor equipment is exposed to the sun, the solar heat load should
always be included in the analysis. The present design approach represents the
most realistic conditions by performing the transient analysis first. The maximum
air temperature and the flow rate to the boards or box from the transient analysis
are then input to the steady state analysis of the boards or box. It should also be
noted that the double wall construction with ambient air flowing is adapted for
the top wall. The double wall will greatly reduce the solar heating to the electron-
ics inside the box.
The heat pipe heat exchanger as illustrated in Figure 8.5 can be considered to
replace the compact heat exchanger. The external and internal air streams are
also separated in the heat pipe heat exchanger.
Wankhede et al. [3] evaluated the cooling solutions for various outdoor elec-
tronics. Both CFD simulations and thermal tests were performed. Three varieties
of outside coating were examined, including white, black, and no coating (plain
aluminum finish). The solar absorptivity (a) and emissivity (ε) of three different
surface finishes are listed in Table 8.1. Among them, the white oil coating which
has the low solar absorptivity and the high surface emissivity makes it a very
favorable choice for cooling of the outdoor electronic enclosures.
F igure 8.4 Internal air temperature history.

F igure 8.5 Heat pipe heat exchanger.
One of configuration under consideration as shown in Figure 8.6 also employs

the heat pipe heat exchanger as described in Figure 8.5. The heat exchanger sys-
tem consists of two parallel heat pipes with the condenser ends located outside of
the enclosure. In addition, fins are bonded to the evaporator and the condenser
sections of the heat pipes to enhance the heat transfer.
The thermal design utilizes the double wall concept as mentioned in Figure 8.1
to minimize the effect of the solar heating. The schematic of this design concept is
illustrated in Figure 8.7. The maximum solar load is assumed to be 600 W/m2 for
Pune city in India during March which is one of the hottest months of the year.
The solar heating is applied to three adjacent surfaces, the top and two sides as
prescribed by GR-487 [4]. It should be noted that the CFD analysis performed is
limited to the steady state conditions. The study concludes that the effect of solar
heating on an outdoor enclosure is significant which can increase the internal air
temperature by 20%.
Table 8.1 Radiation Properties of Coating
Types of coating Solar absorptivity (a) Surface emissivity (e)
White oil 0.25 0.91

Black coating 0.88 0.88
Plain aluminum 0.08 0.09
Figure 8.6 Flow patterns inside enclosure.
F igure 8.7 Schematic of double wall enclosure.

8.1.2 Thermal Design of OutSide Plant (OSP)
The OSP has received considerable attention in the last few years. This is because
of significant increase in the wireless applications and also the great demands of
high speed videos and data transmission to the home and offices.
The typical OSP cabinet is shown in Figure 8.8. The OSP includes the wireless
and the land line applications; however, it has same basic thermal configuration
as illustrated in Figure 8.9.
Outside plants in the past were deployed in the remote locations. However,
with the growing consumer demand for higher data/video transmission speeds,
the outdoor cabinets are now frequently installed much closer to the users in resi-
dential areas. Therefore, the OSP must now operate much quitter. The acoustic
noise limit is reduced from 60 to 45 dBA per the newly updated version of GR487
[4]. In addition, the application of the solar heating specifies in GR-487 is differ-
ent from Example 8.1. The GR-487 describes that solar heating at 750 W/m2 is
applied to any three adjacent surfaces of the equipment (cabinet/enclosure) at the
steady state condition.
F igure 8.8 External view of typical OSP cabinet.

Figure 8.9 Basic thermal configurations in outdoor enclosure.
Conventional Heat Exchanger OSP
As shown in Figure 8.9, the OSP generally employs a compact heat exchanger to
transfer heat from the cabinet to the ambient. In order to overcome a relatively
high-pressure drop across the compact heat exchanger, more powerful fans are
required which lead to high acoustic noise and power consumption. The analysis
also requires a very large and fine meshed thermal model to represent the com-
pact heat exchanger which leads to a long computational time.
The attempt has been made by Trivedi et al. [5] to develop a simplified compact
model for the compact heat exchanger. The cabinet, including one heat exchanger
as shown in Figure 8.10 is under consideration. The methodology for development
a compact model is to replace the core of the heat exchanger by a volume flow
resistance and a volume of source resistance to simulate fluid flow and thermal
boundary conditions, respectively. The coefficients of the flow resistance and vol-
ume source resistance are obtained from the results of the detailed model. With
this simplified thermal model, the mesh count and the computational time have
been reduced by 45% and 63%, respectively. The results from all the simulations
indicate an excellent agreement between the compact model and the detailed
model with the error less than 10%. It should be noted that the solar heating is
not included in the analysis.
Mariam [6] provides more detailed information about the thermal design of
a typical OSP. The author also extends the analysis to the thermal design of the
battery compartment with the thermal electric coolers.
F igure 8.10 CAD model of cabinet.
Geo-Thermal Heat Exchanger OSP
To achieve energy efficiency of the equipment, a typical OSP cabinet as shown in

Figure 8.8 is placed on the top geothermal heat exchangers as presented in Figure 8.11
[7]. The conventional heat exchanger is replaced by the geothermal heat exchanger.
The cold air from the exit of the geothermal heat exchanger goes upwards
from the bottom of the left compartment in the OSP cabinet as shown in Figure
8.11. After flowing over all electronics and the power units, the hot air flows into
the right compartment through the duct on the top of the cabinet and is forced
down by small fans into the geothermal heat exchanger where the heat is trans-
ferred though the heat pipes to the soils. The cooled air once again goes into
the left compartment of the OSP cabinet to start the new flow cycle again. In
order to enhance the heat transfer, the fins are soldered to both condenser and
evaporator sections of the heat pipes. Nickel plating is recommended on external
surfaces of the heat pipes to prevent corrosion from contacting with soils. The
heat exchanger for the external air stream in Figure 8.9 is basically replaced by
U-shaped tubes plus the heat pipes under ground.
The soil temperature depends on the climate, including solar heating and the
depth of soil of the specific location. Since the ambient temperature and solar
F igure 8.11 Geothermal OSP.
heating are function of time, the soil temperature is also a function of time and
the depth from the ground surface. The soil temperature can be determined by
solving transient one-dimensional heat conduction in a semi-infinite solid with
time dependent thermal boundary condition at the ground surface. The soil tem-
perature profiles are illustrated in Figure 8.12. The left figure shows the transient
temperature profiles when the ground surface is heated at (T¥ + DT) where T¥ is
the soil initial temperature from the ground surface to the infinite depth region.
On the other hand, the right figure represents the soil temperature profile at a
given time with the ground surface at various temperatures, e.g., (T¥ + DT1), (T¥ +
DT2), and (T¥ + DT3).
The soil temperature decreases as the depth from the surface of the ground
increasing. There are two types of the soil temperature variations. One is due to
the daily cycle and another is from the seasonal variation. The soil temperature
below 5 ft is generally independent of the daily cycle of the ambient temperature
F igure 8.12 Typical soil temperature profiles.
and solar heating in the former case. On the other hand, the seasonal soil tem-
perature variation is within about 5°C at about 15 ft below the ground and there
is no change with seasons below 50 ft. The soil temperature below the ground
surface is typically colder than the ambient air in the summer. Therefore, geo-
thermal OSP is better suitable for the hot climate regions.
The geothermal cabinet is more energy efficient than the conventional one by
eliminating the conventional heat exchanger and also using the smaller fans to cir-
culate air through the underground-section. However, there are some inherit issues.
The disadvantages of the OSP with geothermal heat exchangers are as follows:
1. Complicated and costly installation procedures

a. A relatively large hole under the ground is required and the digging
may require special permits from the city, state, or federal govern-
ment agencies
b. The installation procedure must insure the soils in good contact with
the surfaces of the heat pipes for effective heat transfer to the soil
2. Service, repair, and maintenance issues
There is no possible or practical for repairs or services once the equip-

ment is installed under the ground. This is especially critical in the case of
natural disasters such as earthquakes.
Other issues are related to the system upgrades. Again, the geothermal heat
exchanger cannot be upgraded as needed in the future.
The conventional heat exchangers which are typical compact heat exchangers
in general require larger fans in order to overcome the pressure drop over the
heat exchanger. Large fans consume more power and also generate higher acous-
tic noise. The geothermal cabinets can achieve better energy efficiency and lower
acoustic noise but have inherit problems as described above. Therefore, the best
way to achieve energy efficiency for an outdoor enclosure is to adapt a natural
convection for the external stream through the heat exchangers in Figure 8.9.
However, due to the ineffective heat transfer of the natural convection, the system
power is somewhat limited.
Direct Air Cooled OSP
All previous discussions employ various types of the heat exchangers in the out-
door equipment to separate the internal and the external (ambient air) air streams.
This is because no direct contact is permitted between the electronic components
and the ambient air for most outdoor equipment. The main reason is to prevent the
electronics from contamination by the ambient air. The updated version of GR-487
allows the direct air cooling of the equipment as long as the special air filter which
can filter out the dust, sand, soot, pollen, and salt particles is employed.
The direct cooling eliminates the needs of using the heat exchanger. The advan-
tages of the direct air cooling as compared with the heat exchanger cooled OSP
are as follows:
a. Operates at the lower temperature due to low supply air temperature,

b. Eliminates the possible hot spots along the internal circulation air
path,
c. Increases system reliability and also operation life of electronics,
d. Reduces system complexity and the number of mechanical parts in
system,
e. Lower system power consumption resulting in higher COP.
Generally, the direct cooling will lower the internal air temperature of the cabi-
net by about 10°C as compared with that of the conventional heat exchanger OSP.
Gore’s ePTFE membrane filter [8] which is naturally waterproof is one of com-
mon used filters in the direct air cooling for the outdoor electronic equipment.
Comparison of dust protection among various types of the air filter is presented in
Figure 8.13 [8]. The results indicate the ePTFE filter is much superior to others.
Garcia and Cosley [9] adapt the double air filters design as given in Figure 8.14
for the outdoor enclosure. The primary one is the typical fibrous air filter for debris
and large particulates, and the secondary one is the hydrophobic filter such as
ePTFE filter for vapor transmission and small particulates. It is not necessary to use
F igure 8.13 Dust protection of various types of air filters.
F igure 8.14 Filtered vent thermal system concept.

double filters as described in the paper. It depends upon the environmental chal-
lenge. The ePTEF filter developed by Gore is designed with layered pre-filtration that
will protect from typical ambient dust and debris. If the environmental challenge is
more severe, then the use of a separate pre-filter will help to increase the life of the
hydrophobic filter.
The results from the field trial tests indicated that the pressure drop across the
primary filter increases with the time of exposure during the field trial and, on the
other hand, no significant increase in pressure drop over the secondary filter during
near a year trial test. The former implies that the primary filter functions properly.
This is because the accumulation of dust decreases the available free flow area and
increases the pressure drop across the filter. The latter reveals the benefit of the
primary filter in prolonging the exposure life of the secondary filter.
Natural Convection OSP
The natural convection has been utilized extensively to cool microelectronic

equipment because the passive cooled system is energy-efficient, reliable, and
Figure 8.15 External natural convection for OSP cabinet.

cost-effective. The natural convection cooled OSP which adopts the natural con-
vection for the external air stream is illustrated in Figure 8.15. As can be seen
from the figure, the hot side (internal air stream) is due to forced convection and
the cold side (external air stream from the ambient) is by natural convection. By
eliminating the fans in the external flow path, the equipment achieves greater
values of COP which has the higher energy efficiency and also lowers the acoustic
noise. It should be noted that the cabinet sometimes only has one side wall avail-
able for heat transfer between the equipment and the ambient. As always, the
radiation heat transfer must be included in the passive cooling. It is also often to
add a cover or shroud to protect the external fins which forms a counter-flow heat
exchanger as shown in Figure 8.16.
Applying the results from his earlier work (10), Yeh estimated the possible
cooling capacity of a natural convection OSP. The heat sink configuration and
the summary results are presented in Figure 8.17. The heat width is 56” and the
height ranges from 26”–60”. The heat sink base is assumed at constant tempera-
ture of 66°C which is an average value of the inlet and exit air temperature of
the heat sink as shown in Figure 8.17. The optimal fin spacing for the heat sink
under consideration is 0.513”. The heat loss from the heat sink which is based on
the correlations [2] for a straight fin array may over-predict over the CFD results
by 10%.
The use of the heat exchangers, including the geothermal heat exchangers,
as well as natural convection cooling is only capable of providing above ambi-
ent temperatures to the electronics inside an OSP. Typical temperature differ-
ence between the internal and the ambient air is about 20°C. The direct cooling
approach provides a near the ambient temperature to the electronics. By elimi-
nating the heat exchanger, the direct cooling reduces the size of the system ther-
mal control unit and the power consumption as well as the acoustic noise. An air
conditioner can be employed to provide the below ambient temperatures to the
Figure 8.16 Counter flow heat exchanger for OSP in passive cooling.
Figure 8.17 Estimated cooling capacity of natural convection OSP.
electronics. However, the size increase in the thermal system due to the air condi-
tioner unit results in significant increase in the power consumptions. Another way
to provide the below ambient temperatures is to utilize the thermoelectric coolers
which require a significant amount of additional (electrical) power input.
Recently, the fuel cell technology, which is one of the green technologies,
has received a great interest by the data center industry to improve the energy
efficiency. The application of the fuel cell technology has also been extended to
the OSP. The current application of this technology is limited to as the back up
power (i.e., back up to the batteries in the cabinet). The fuel cell OSP cabinet
is currently available in the Europe market. Further utilization of the fuel cell
technology to the OSP is expected to increase in the near future.
The summary of various types of cooling technologies employed by OSP is
listed in Table 8.2 where the comparisons are made for various types of the OSP
8.2 TOWER/POLE MOUNTED OUTDOOR ELECTRONIC

EQUIPMENT
Remote radio heads (RRHs) have become one of the most important subsystems
of today’s new distributed base stations. The remote radio head contains the base
station’s RF circuitry plus analog-to-digital/digital-to-analog converters and up/
down converters. RRHs also have operation and management processing capa-
bilities and a standardized optical interface to connect to the rest of the base sta-
tion. Figure 8.18 shows the tower mounted devices.
Table 8.2 Comparisons of Various Cooling Technologies for OSP
Thermal technology Advantages Disadvantages
1. Conventional heat 1. Separated internal and external air paths 1. Higher component temperatures due to higher
exchanger, Figure 8.9 2. No environmental contamination to inlet air temperature to electronics
electronics 2. High pressure drop over external stream heat
3. Applicable to wide range environments exchanger which requires large powerful fans
(region/weather)
2. Heat pipe heat 1. Separated internal and external air paths 1. Higher component temperatures due to higher
exchanger, Figure 8.5 2. No environmental contamination to inlet air temperature to electronics
electronics 2. Slightly low pressure drop over external stream
3. Applicable to wide range environments heat exchanger as compared with Case 1
(region/weather)
3. Geothermal heat 1. Separated internal and external air paths 1. Limited applicable environments (only in hot
exchanger, Figure 8.11 2. No environmental contamination to climatic regions)
electronics 2. Soil thermal properties varying with locations
3. Few fans needed for underground external air 3. Current design for power less than 1000W
flow path with small pressure drop 4. Complicate snd costly installation processes
4. Less power required and lower acoustic noise 5. Possible; additional governmental regulations
5. Inlet temperature to electronics less
dependent at ambient conditions
4. Direct air cooling, 1. Lower component temperature due to direct 1. Expensive air filter needed to prevent
Figure 8.13 cooling by ambient air environmental contamination on electronics
2. No eternal air path heat exchanger (less 2. Some limitations to applicable environmental
power required) (location/weather conditions)
5. External natural 1. No fan needed for external air path in heat 1. Limited system power due to less effective heat
convection cooling, exchanger transfer in natural convection
Figure 3.15 2. Less power required and lower acoustic noise
F igure 8.18 Picture of tower mounted devices.
8.2.1 Thermal Design of Remote Radio Head
The weight and volume are the key design factors for any tower or pole mounted
equipment in addition to the thermal requirement. A passive cooling, which is
simple, reliable, and cost-effective, is most often employed for these equipments. A
typical remote radio head (RRH) with a heat sink is shown in Figure 8.19. RRHs are
for the wireless applications and are typically mounted on the pole, tower or even
the top of the roof. As any outdoor equipment, the RRH is subjected to the solar
radiation heating and the design and operation are also governed by GR-487.
F igure 8.19 Typical remote radio head.
Natural convection from vertical finned heat sinks has been studied exten-
sively in the past. Among them, Yeh et al. [10] performed CFD analysis of the
finned heat sink under the natural convection environments. The authors first
optimize the fin spacing and also examine the thermal performance of various fin
configurations such as continuous fins, staggered fins, and in-line fins as shown
in Figure 8.20 and found that the thermal performance of the continuous finned
heat sink is the best. Yeh [11] further extends the analysis to study the effect of the
cover or shroud over the finned heat sink. The cover will enhance the natural con-
vection heat transfer. In addition, based on the limited results, there is no effect
of cover on the heat transfer if the distance between the heat sink fin tip and the
cover is greater than 2.5 times of the fin height.
The RRH typically employs the extruded finned heat sinks as shown in the
above figure in the past because of easy manufacturing and low cost. However,
the extrusion heat sink is bulky and heavy due to the thick base and fins which also
limit the available heat transfer surface area. The heat sink is typical made from
aluminum. The size of the extrusion heat sink will be significantly increased in
order to support the increasing system power. Since in most of cases, the heat load
is not uniformly distributed over the heat sink base, it would require a thick base to
spread heat to the entire heat sink.
F igure 8.20 Various types of finned heat sinks.
Skandakumaran et al. [12] adapt a thin, flat two-phase active heat spreader
(AHS) as shown in Figure 8.21 to improve the thermal performance of the tower
mounted RRH heat sinks. The dimensions of the heat sink base are 280 mm ´
540 mm. The active heat spreader (AHS) is basically a vapor chamber device. The
improvement is limited to the base plate of the power amplifier. The experiments
and simulations are performed using four different extrusion heat sink configura-
tions as presented in the figure. The thermal conductivity of an extruded aluminum
heat sink is 187 W/m °C. The heat sinks are tested at 75 W of thermal power under
the natural convection environment. The test and simulation results (maximum
Figure 8.21 Active heat spreader heat sinks.

Figure 8.22 Extruded heat sink thermal solutions.
temperatures) are given in Figure 8.22. The results reveal that the 4-mm base heat
sink exhibits the highest temperature while the 8mm base heat sink has the lowest
temperature. It appears that AHS device only improve the heat transfer by about
10% due to the small size of the device which limits the heat spreading capability.
Though copper has much higher thermal conductivity, the copper heat sinks
are also eliminated for consideration because of the weight factor. The con-
ventional heat pipes can be embedded into the base of heat sink to enhance
heat spreading. However, it may lead to significantly increase the thickness of
the heat sink base. In addition, a great number of heat pipes are required for a
large heat sink. Another option is to adopt embedded graphite in the heat sink
base. The thermal conductivities for the graphite are 1000+ W/m °C in the XY
plane and 20 W/m °C in the Z plane (thickness direction). The density of the
graphite is about the same value as the aluminum.
Liquid has been employed to cool the high power supercomputers recently.
Similarly, liquid can be considered for the RRH to deal with the increased sys-
tem power. Thermosyphon is often employed for the passive cooling conditions.
Thermosyphon includes two different cooling modes as illustrated in Figure 8.23.
One is the closed tube and another is closed-loop thermosyphones. The former
is a self-contained closed tube which is similar to the heat pipe, and the latter
is two-phase closed loop system. The important difference between a heat pipe
and a thermosyphon is at the mechanism of condensate return in these devices.
A heat pipe relies on the capillary action of an internal wick and a working fluid,
whereas a thermosyphon employs an external force such as gravity or centrifugal
force for condensate return to the evaporator. Since there is no wick involved, a
closed tube thermosyphon is often referred to a wickless heat pipe.
F igure 8.23 Basic thermosyphone models.
Similar to the heat pipes, the closed tube thermosyphons can be embedded
into the base of a finned heat sink. The cooling scheme of a closed-loop thermo-
syphon for RRH is illustrated in Figure 8.24. The section of the heated leg is a
liquid cold plate in a direct contact with the RRH power module, and the section
of the cooled leg is the base of the finned heat sink. The vapor flow is generated
due to the heating from the RRH electronics in the hot leg and vapor is then con-
densed to liquid by rejecting heat to the ambient through the finned heat sink by
natural convection and radiation. The flow circulation as shown in the figure is
because of the difference in fluid densities in the hot and the cold legs.
Since it is for outdoor applications, the freezing point of the fluid must be
below −40°C. The dielectric fluids such as Fluorinert liquids (FC-43, -75, -77, -87)
from 3M company and Ethylene Glycol/water mixture. The disadvantage of the
thermosyphon is that the thermal performance of the system will degrade signifi-
cantly as the inclined angle from the vertical plane increasing, and it practically
ceases to perform at the horizontal position.
The extruded heat sink is bulky and heavy because of thick fins which are typi-
cally greater than 1.5 mm. With the advances in the manufacturing technologies,
the plain fin heat sink can be made to replace the extruded heat sink in order to
reduce its weight. For example, for fin height of 2 in., the fin thickness and the
base can be reduced to 0.02 in. and 0.1 in., respectively. The weight for a plain fin
F igure 8.24 Passive cooling with closed-loop thermosyphon.
heat sink with the overall dimensions of 10.4 in. (width) ´ 12 in. (length in flow
direction) is only about 1 kg. This lightweight heat sink can be deplored at any
orientation, even at the horizontal position.
Another popular tower or pole mounted outdoor equipment is the ONT (optical
network terminal) as shown in Figure 8.25 which are subjected to the direct solar
heating and the ambient of 46°C. However, some of the ONTs are installed inside
of the garage of the home where the equipment is not directly exposed to the sun
but the air inside the garage is hotter than the ambient air due to the solar heat-
ing on the roof and the sidewalls of the building. Therefore, the equipment must be
designed for the minimum “ambient” temperature of 65°C under the natural convec-
tion conditions.
To reduce the cost, Corning plastic outdoor enclosure is frequently employed.
The basic thermal solution for an ONT can be illustrated in Figure 8.26. Gener-
ally, the electronic components are enclosed in a metal box to prevent EMI leak-
age to the environment and also from a direct contact between the ambient air
and the components. The system power for a fully functioned ONT could be up
to 25 W, and the overall dimensions of the envelope are approximately 300 mm
(height) ´ 240 mm (width) ´ 80 mm (depth). The solar absorptivity of the plastic
is about 0.44.
Solar heating is a major concern for any outdoor equipment. The solar ther-
mal loads in some cases are greater than the heat load from electronics inside
the equipment. Since the infrared (IR) energy emitted by the Sun is at much
shorter wavelengths than those emitted by a body near room temperature. This
Figure 8.25 A typical ONT with cover open.
Figure 8.26 Basic configuration of ONT.

distinct characteristic allows for the some of thermal control coatings with low
solar absorptivity and high emissivity being applied to the external surfaces of the
system for thermal management of solar radiation heating. This special coating
(AXJ-4020) with the solar absorptivity of 0.15 and the material emissivity of 0.9 is
currently available from AZ Technology.
REFERENCES
1. Yeh, L. T., “Transient Thermal Analysis of Outdoor Electronic Equipment”,

EEP-Vol. 23/HTD-Vol. 356, CAE/CAD and Thermal Management Issues in
Electronic Systems, ASME, 1997.
2. Yeh, L. T., and Chu, R. C., Thermal Management of Microelectronic Equip-
ment, ASME Press, NY, 2002.
3. Wankhede, M., Khaire, V., Goswami, A., and Mahajan, S. D., “Evaluation of
Cooling Solutions for Outdoor Electronics”, Volume 16, No. 3, Electronics
Cooling, 2011.
4. Telcordia Technologies Generic Requirements, GR-487-CORE, Issue 3, April,
2009.
5. Trivedi, A., Agonafer, D., Sivanandan, D., Hendrix, M., and Sahrapour, A.,
“Compact Modeling of A Telecommunication Cabient”, Proceedings of
IMECE2008, Boston, MA, October 31–November 6, 2008.
6. Mariam, F. A. I., Thermal Management of Outside Plant Telecommunication
Cabinet: Design and CFD Modeling Methodology, MS Thesis, The University
of Texas at Arlington, May, 2010.
7. Hong, Yuping, Shengqin, J., Liqian, Z., Qiao, C., Bianco, C., Griffa, G., and
Cucchietti, C., “Cooling System of Outdoor Cabinet using Underground Heat
Pipe”, INTELEC, San Diego, CA, 2008.
8. Braswell, M., and Poon, W., “High Reliability with Direct Air Cooling”, IWPC
Workshop, Chicago, IL, July 30, 2007.
9. Garcia, M. P., and Cosley, M. R., “Ambient Air Cooling of Electronics in An
Outdoor Environment”, 26th International Telecommunications Energy Con-
ference (INTELEC), September 19–23, 2004.
10. Yeh, L. T., Yeh, J., and Chung, B. T. F., “Natural Convection from Finned Heat
Sinks”, Proceedings of IPACK2007, Vancouver, British Columbia, Canada,
July 8–12, 2007.
11. Yeh, L. T., “Natural Convection from Finned Heat Sinks with/without Cover/
Shroud”, 19th International Symposium on Transport Phenomena, Reykja-
vik, Iceland, August 17th–21st 2008.
12. Skandakumaran, P., Khanikar, V., Smalc, M., Norley, J., and Reis, B., “Passive,
Lightweight Thermal Solutions for Remote Radio Head (RRH) Electronics”,
Itherm, 2008, 2008.
Chapter 9
Liquid Cooled Systems
In the cooling of electronic equipment, liquid cooling is frequently applied to high

power electronic equipment because of its high heat transfer capability. Liquid
cooling can further be divided into single- and two-phase flow systems. The lat-
ter involves phase change processes such as boiling or condensation that greatly
increases the heat transfer capability by utilization of the latent heat. In addition,
the phase change process takes place at the constant temperature.
Since there is no commercially available liquid cooled telecommunication
equipment presently, this chapter first focuses at the fundamental theories of
both single- and two-phase flow heat transfer, follows with the coolant selection,
and then briefly describes current liquid cooling technologies as well as their
applications in the electronics industry in general. Finally, system thermal design
will be discussed.
9.1 CONVECTION AND BOILING HEAT TRANSFER
Basic fundamentals of single phase convection and boiling heat transfer will be
discussed in this section.
9.1.1 Single-Phase Forced Convection

The heat transfer coefficient of a single-phase flow under forced convection is
a function of the Reynolds number (Re), Prandtl number (Pr), and the thermal
boundary conditions as well as the shape of the tube for fully developed lami-
nar flow or for fluids with Prandtl number less than one. The Prandtl number is
defined as Pr = υ/α, where υ is kinematic viscosity and α is thermal diffusivity of
the fluids. It is well known that the heat transfer coefficient increases with the
Reynolds number ( ρVL/µ), where ρ is density of fluid, V is velocity, L is the char-
acteristic length, and µ is viscosity of fluid.
The current cooling technologies used in the telecommunication industry are
still limited to the air cooling. The major difference between the air cooling and
the liquid cooling is due to the effects of the Prandtl number of the fluids. The
Nusselt number or heat transfer coefficient is in general as a function of Pr1/3.
For most of gases, the Prandtl number ranges from 0.7 to 1.0 which basically
eliminates the effect of the Prandtl number on the heat transfer coefficient. The
Prandtl number also has a great effect on the boundary layer development which
142 • THERMAL MANAGEMENT OF telecommunications EQUIPMENT
can be illustrated in Figure 9.1, where δ and δth are the thickness of the hydrody-
namic and thermal boundary layers, respectively. The heat and momentum are
diffused through the fluid at the same rate if the Prandtl is unit. If the Prandtl
number is greater than 1, the velocity profile will develop more rapidly than the
temperature profile, whereas the opposite trends are found in a fluid with Prandtl
number less than 1.
It should be noted that the temperature profile starts to develop at the point
where the heat transfer between the fluid and wall actually takes place. In other
words, due to heat conduction along the heating wall, the thermal boundary layer
will start at the upstream of the actually heated location. If the Prandtl number is
greater than 5, the velocity profile generally will develop so much faster than the
temperature profile that a solution based on a fully developed velocity profile will
apply quite accurately, even without accounting for the hydrodynamic starting
length. Such a condition frequently prevails in the dielectric fluids that are widely
used for electronic equipment because the Prandtl number for most dielectric flu-
ids is in general greater than 20 at the room temperatures. On the other hand, for
the case with Prandtl number approaches zero, the velocity profile never actually
develops while the temperature profile does develop. A uniform velocity profile
(slug flow) can be assumed over the entire thermal boundary layer.
In laminar flow or fluids with a Prandtl number less than 1, the heat transfer
coefficient depends not only on the thermal boundary conditions but also the
shape of cross section of the tube as shown in Table 9.1 [1] The two basic thermal
conditions are constant axial wall temperature (T condition) and constant heat
flux (H condition). For symmetrically heated noncircular tubes, two common
cases are:
1. H1 condition: constant axial wall heat flux with uniform peripheral

wall temperature (variable peripheral heat flux).
2. H2 condition: constant axial and peripheral wall heat fluxes (variable
peripheral wall temperature).
Figure 9.1 Prandtl number effect on boundary layers.

Liquid Cooled Systems • 143
Table 9.1 Nusselt Number and Friction Factor in Fully Developed Laminar Flow
Tube Shape f · Re NuT NuH1 NuH2
Rectangular Tube:
Aspect Ratio (a) = 2a/2b
2a: tube height; 2b: tube width
1.0 14.227 2.976 3.608 3.091

0.9 14.261 – 3.620 –
0.8 14.378 – 3.664 –
0.714 14.565 3.077 3.734 –
0.667 14.712 3.117 3.790 –
0.6 14.980 – 3.895 –
0.5 15.548 3.391 4.123 3.020
0.4 16.368 – 4.472 –
0.333 17.090 3.956 4.795 2.970
0.3 17.512 – 4.990 –
0.25 18.233 4.439 5.331 2.940
0.2 19.071 – 5.738 2.930
0.167 19.702 5.137 6.049 2.930
0.125 20.585 5.597 6.490 2.940
0.1 21.169 – 6.785 2.950
0.05 22.477 – 7.451 –
0 24.000 7.541 8.235 8.235
Circular tube 16.000 3.657 4.364 4.364
Triangle tube (equilateral) 13.333 2.470 3.110 –
It should be noted all above discussions are applicable for either liquid or gas
flow.
In addition to the heat transfer coefficient, the friction factor is also needed in
computing the pressure drop in a system. The general relation for pressure drop
in a duct can be expressed as follows:
DPt = DPf + DPs + DPa ± DPg (9.1)
where
DPt = total pressure drop,

DPf = pressure drop due to friction
= 4f(L/Dh)(G2/2rm gc),
DPs = pressure drop due to additional losses such as entrance, exit,
bends, fittings, valves, etc.
= åK(G2/2rm gc),
DPa = pressure drop due to flow acceleration (or momentum changes)
= (1/r2 − 1/r1) (G2/2gc),
DPg = pressure drop or gain due to gravity in a vertical flow

= rm gH/gc
“+” for vertical upward flow
“–” for vertical downward flow,
Dh = hydraulic diameter,
F = Fanning friction factor,
G = Mass velocity
= rV (V is flow velocity),
g = acceleration,
gc = gravitational constant
= 32.1739 lbm-ft/lbf-sec2 = 4.17 ´ 108 lbm-ft/lbf-hr2,
H = tube vertical height,
K = loss coefficient
9.1.2 Boiling Heat Transfer
From the viewpoint of thermodynamics, when liquid is heated at a constant pres-

sure, the temperature and volume of liquid will increase from A to B as shown
in Figure 9.2 [1]. Beyond point B, any additional heat added to the system is to
change liquid phase to vapor phase, and Point B is the saturated liquid at a given
pressure. In the region between points B and C, liquid and vapor phases coexist in
equilibrium. The amount of heat required to convert a unit of mass from liquid to
vapor is called the latent heat of vaporization, hfg (= hg − hf), where hg and hf are the
enthalpy of saturated vapor and liquid, respectively. In this region, the addition of
heat increases the vapor fraction and the volume of mixture, however, the tempera-
ture of the two-phase mixtures still remains at the saturated temperature. At point
C, all the liquid has vaporized and is referred to as the saturated vapor. Further
heating will result in a temperature rise of vapor (superheated vapor) along the
constant pressure line.
Figure 9.2 Liquid at constant pressure heating process.

Boiling is the process of evaporation associated with vapor bubbles in liquid.

The change of phase from liquid to vapor is caused by heat transfer or pressure
changes. The two basic types of boiling are pool boiling and flow boiling. The
former is the boiling on heated surfaces submerged in a pool of stationary liquid,
while the latter is the boiling on a surface in a flowing stream of liquid. Depend-
ing on the liquid temperature, boiling can also be classified into subcooled boil-
ing, in which the bulk fluid temperature is below the saturation temperature, and
saturated boiling, in which the bulk fluid temperature is uniformly at the satura-
tion temperature. The boiling curves for both pool and flow boiling are illustrated
in Figure 9.3, and their trends are very similar.
The heat transfer is so effective in the nucleate boiling region that the tempera-
ture difference between the heating wall and the liquid is very small, typically less
than 5°C. As the heat flux is increased continuously, the number of nucleation
sites on the heating surface increases until the maximum heat flux is reached at
point C. The maximum heat flux is often called the critical heat flux (CHF). At
point C, the vapor generation rate is so high that the surface is blanketed with a
vapor film which prevents liquid from contacting the surface.
For pool boiling, in Region CD as given in Figure 9.3, a portion of the heated
surface will undergo film boiling while others are under nucleate boiling, therefore,
the vapor film is unstable. A smooth transition from nucleate to film boiling can
only occur under temperature controlled conditions, which can be accomplished
by condensing vapors or by convection from high temperature fluids with high
mass flow rates. However, on the other hand, with electric heating, it is impossible
to operate in region CD. When the power is increased, the operating point shifts
rapidly from point C to C¢, which is at a much higher temperature. A physical
burnout of material could occur if point C¢ is above the melting point of the solid
surface. If the temperature of point C¢ is below the melting point, operations can
be maintained along the curve DE. The region DE is called film boiling because a
layer of vapor film continuously covers the surface and liquid does not contact the
surface. Because of high surface temperature in this region, radiation becomes
an important factor. The operating condition will change directly from point D to
nucleate boiling at point D¢ if the heat flux is reduced below point D. Then, power
again must be increased to point C before the process can shift to film boiling.
Figure 9.3 Boiling curves for pool and flow boiling.

9.2 COOLANT SELECTION
Selection of a coolant for a liquid cooled system is a very important design con-
sideration. A number of factors must be considered when selecting a coolant for
a given system. The first step is to establish the important system characteristics,
such as weight/volume limitation, power consumption requirement, and leak-
tightness. In general, a fluid may be very good in some respects and quite defi-
cient in others. Therefore, there is no single, ideal fluid that is good for all systems
and applications. Consequently, various fluid properties of the coolants should be
evaluated against the particular systems and applications under consideration.
The important properties that affect the selection of a coolant were discussed
in details by Knight [2] and these properties include toxicity, flammability, flash
point, vapor pressure, pour and freezing points, thermal decomposition, dielec-
tric strength, effects of moisture. One should never overlook the properties which
do not have any effects on the system thermal performance. For examples, the
toxicity and flammability are critical to the system safety and leakage. The fluid
electric strength and the effects of moisture will have great impacts on the system
electric insulation requirements.
Coolants must be chemically compatible with the materials, which they will
contact within the coolant loop, and from leakage. Also, particular attention must
be paid to the selection of seals, gaskets, and adhesives used in the cooling system.
Kelly [3] has rated the compatibility of a number of common compounds and
some of the general classes of fluids that might be used as coolants. For cooling of
electronics, the dielectric property of a liquid is also very important, especially for
a direct liquid cooling. The reader is encouraged to read both references.
The fluid properties that are directly related to thermal and hydrodynamic per-
formance of the coolant are specific heat (cp), thermal conductivity (k), viscosity
(m), and density (r). Specific heat is an indicator of the coolant’s ability to store
thermal energy within a given mass. Thus, it is desired that the fluid have a high
specific heat. Thermal conductivity defines the coolant’s ability to transfer heat
and should be high. Fluid viscosity increases the fluid resistance, which results in
a high system pressure drop and required high pumping power, and also reduces
the effectiveness of heat transfer, therefore, the viscosity of the fluid should be as
low as possible. The density of the coolant relates to the mass of the fluids. The
higher the density, the more mass is in a given volume. The density, unlike specific
heat and thermal conductivity, also has some negative attributes. Fluid weight
and pumping power increase with the density. Thus, it may be desirable to have
either a large or a small density, depending on the specific application.
The important properties affecting the boiling process for a given fluid are sur-
face tension (s) and latent heat of vaporization (hfg). A coolant with a low value of
surface tension will result in a small contact angle and wet the heat transfer sur-
face. This high wetting nature of the fluid deactivates many potential nucleation
sites on the surface that can lead to a significant surface temperature overshoot
at boiling incipience. Care must also be exercised because the liquids having a
low surface tension are more likely to cause leakage problems by sipping through
seals, gaskets, cracks, connectors, etc.
The coolants may be rated on the basis of heat transfer characteristics as they
affect mass flow rate, pumping power, volume, and weight. From the viewpoint of
thermo-hydraulic performance, fluids with a high heat transfer coefficient always
have a higher flow resistance that requires a higher pumping power to overcome
the higher system pressure drop. Depending on the system requirement, a care-
ful balance between the heat transfer coefficient and pressure drop is needed to
achieve an acceptable design for thermal control of the system.
For convenience, a figure of merit (FOM) is frequently used to compare the
effectiveness of various fluids. The FOM which is a dimensionless parameter rep-
resents the relationship among key properties or characteristics of a fluid under
prescribed flow conditions and is used to guide the selection and optimization of
fluids. FOMs based on heat transfer consideration only are derived from the gov-
erning thermal transport relations for single-phase (liquid) and two-phase (vapor
and liquid) flow, including forced and free convection; boiling incipience; and
critical heat flux for pool and flow boiling [4]. For example, for natural convec-
tion, the FOM can be defined as follows:
FOM = [br2cp k(1/n−1)/m]n (9.2)
where n is 0.25 for laminar flow and 0.33 for turbulent flow.
It is often used in the past to express the FOM to be function of the heat trans-
fer coefficient for the forced convection condition. For a fully developed laminar
forced convection, the heat transfer coefficient, h, only depends on the thermal
conductivity, k, and the channel hydraulic diameter, d, i.e., h µ k/d. Based on
this definition, the FOM does not include the specific heat of fluids in the case of
forced convection. Therefore, this definition for FOM appears to be incomplete.
Any good heat transfer fluids should possess the ability to transfer heat between
the wall and the fluid and should also have the ability to transport heat in the
fluid that is related to the specific heat of the fluid. Furthermore, in practical engi-
neering applications, it is beneficial to maximize heat transfer performance while
minimizing the required pump power to save the power consumption, system
weight and cost. As proposed by Yeh and Chu [1], the FOM for a single-phase flow
under forced convection be defined as follows:
FOM = cp h/P (9.3)
where P is the required pump power. If the pump power is not a consideration
factor, e.g., natural convection, let P = 1 in Equation (9.2).
For fully developed flow, Equation (9.3) can be simplified as follows
FOML a (kcpr/m) for laminar flow (9.4a)
and
FOMT α (k0.67cp1.33r1.05/m0.72) for turbulent flow. (9.4b)

Table 9.2 Coolant Figures of Merit at 40°C
Fluid Laminar Turbulent
Water 14.207 27.688

FC-43 0.155 0.633
FC-75 0.405 1.249
FC-78 0.666 1.751
Coolanol-25 0.250 0.980
Coolanol-45 0.071 0.397
Glycol/water 1.382 4.482
PAO(2CST) 0.222 0.950
Notes:
(1) Constant mass flow rate
(2) Based on English units: k (Btu/hr-ft-°F), cp (Btu/lbm-°F), r(lbm/ft3),
m (lbm/hr-ft)
Numerical values of FOM for various fluids at 104°F (40°C) are given in Table 9.2
[1]. The table indicates that water is the best coolant and is followed by ethylene
glycol/water (62.5% and 37.5% by weight). Since thermal properties of fluids are
a function of temperature, the temperature may greatly affect numerical values
of FOM. One needs to consider the operation temperature range of the system in
determining the FOM. The constants listed in Equations (9.4a) and (9.4b) may
vary with the correlations of the heat transfer and friction factor being employed
in Equations (9.3).
In general, a dielectric liquid which has a high dielectric strength to prevent
electrical breakdown is a preferred choice in cooling of electronics. Unfortunately,
the heat transfer capability of the dielectric fluids is much lower than that of the
water. The dielectric liquid is not normally required for the cases with indirect
liquid cooling. However, it is a must for the immersion cooling where liquid is in
a direct contact with electronics. Due to very low freezing point, typically below
−40°C, dielectric fluids should be employed in all outdoor electronic equipment.
Among dielectric fluids, fluorocarbon liquids (FC-series) are the most commonly
employed in cooling of electronics.
In summary, any good liquid coolant should have the following general
characteristics:
Physical Properties
a. High fluid density ( for heat transfer, low density from weight aspect)
b. Low pour or freezing point and low vapor pressure
c. High coefficient of thermal expansion (for free convection, low from
pressure aspect)
d. High specific heat and thermal conductivity
e. High flash point and high self-ignition temperature
Electrical Properties
a. Low dielectric constant

b. High dielectric strength
c. Low dissipation factor

d. High volume resistivity
Functional properties
a. High thermal stability

b. High oxidative stability
c. High hydrolytic stability
d. Good lubricity
e. Inertness to system materials (or compatibility)
f. Non-toxicity
g. Low cost
h. Low contamination (fouling) and good clean up characteristics
9.3 LIQUID COOLING TECHNOLOGIES
9.3.1 Direct Immersion Cooling
The simplest method of cooling is to immerse the electronic equipment in a pool

of liquid. Due to the direct contact between the liquid and electronics, a dielectric
fluid is a must. In other words, the liquid must have the dielectric strength needed
to provide electrical isolation between adjacent power/ground conductors and
the signal lines. In addition, the liquid must also be compatible with the materials
of all parts in the systems.
The thermal properties of all dielectric fluids are poor as compared with the
water. The basic characteristics of the dielectric fluids are as follows:
1. Excellent electrical properties, high dielectric strength, low dielectric

constant, and high electrical resistivity
2. High thermal stability
3. Low chemical reactivity
4. Poor thermal properties: low thermal conductivity, specific heat, and
heat of vaporization
5. Extremely low surface tension
6. High solubility of air
Liquid immersion remains a primary candidate for cooling of electronics,

especially for high power electronic systems. For low heat dissipation systems,
natural convection takes place in liquids. With increased power, the heat transfer
will shift from natural convection to pool boiling.
Parks and Bergles [5] performed a series of experiments to investigate the effect
of the heater size on the heat transfer coefficient in the case of natural convec-
tion. For flush heaters, the heat transfer coefficient increases as the heater width
decreases. The width effect is more pronounced in R113 than in water. The data
indicates that the heat transfer coefficient is more than 20% higher than those pre-
dicted by conventional correlations. This is attributed to leading edge effects.
Figure 9.4 Pool boiling data for a plain tube.
Marto and Lepere [6] conducted pool-boiling experiments with R-113 and
FC-72 on a plain copper tube with an outside diameter of 15.8 mm and the results
are given in Figure 9.4. As can be seen from the figure, a greater incipient boiling
superheat is required for R-113 than for FC-72. As stated above, the dielectric flu-
ids have extremely low surface tension which often results in a large temperature
overshoot in the case of pool boiling as shown in Figure 9.4. The large tempera-
ture overshoot is undesirable and unacceptable for electronic equipment. The
methods to eliminate or minimize the temperature overshoot at the incipience of
boiling are presented by Yeh and Chu [1].
The liquid-encapsulated module [7] as shown in Figure 9.5 was developed by
IBM. Heat generated by the immersed chips in perfluorinated liquid (FC-86) is
carried by natural convection to the walls and is then dumped into the ambient
Figure 9.5 IBM liquid encapsulated module (LEM).
through the external air-cooled heat sink or to a liquid cold plate. With the satu-
rated temperature of 62°C at the atmospheres pressure, the component junction
temperature is at 85°C. In addition, the author also discussed the nature of the
departure from natural convection (DNC) and departure from the nuclear boiling
(DNB) as illustrated in Figure 9.6. The former deals with the transition from the
Figure 9.6 Boiling curve with DNC and DNB.

natural convection to the nucleate boiling while the latter describes the transi-
tion of the heat transfer region from the nucleate boiling to the film boiling. The
DNC is generally associated with an excessive temperature over-shoot as shown
in Figure 9.4. However, on the other hand, the DNB often results in extremely
high wall temperature due to the poor heat transfer in the film boiling.
9.3.2 Indirect Liquid Cooling
Forced convection through a cold plate is frequently used to cool electronic

equipment. In most cases, flow is unidirectional from inlet to exit, which leads to
a significant temperature gradient in the cold plate. A counter flow arrangement
proposed by Yeh [8] as shown in Figure 9.7 may be applied to overcome the large
temperature gradient. Typical temperature profiles of the cold plate and fluid for
unidirectional flow and counter flow systems are illustrated in Figure 9.8. The
advantages of using counter flow over a unidirectional flow system are lowering
the cold plate wall temperature and the temperature gradient. Yeh [8] provides an
analytical solution given below for the cases where the wall heat dissipation and
the heat transfer coefficients are functions of the space.
qx = T2(x) − T1(x) = q0e−Bx + Q(x)e−Bx (9.5a)
q0 = T2(0) − T1(0) = {{[2h2(x)W/[h1(x) + h2(x)]}òLq²(x)dx
−{h1(x)h2(x)W/[h1(x) + h2(x)]}òLqxdx}/C2 (9.5b)
T1(x) − T1(0) = {h1(x)W/{C1[h1(x) + h2(x)]}}òx[2q²(x) + h2(x)qx]dx (9.5c)
Figure 9.7 Sketch of a counter flow system for electronic cooling.

Figure 9.8 Cold wall temperature distribution for unidirectional and

counter flow.
Tw (x) = [2q²(x) + h1(x)T1(x) + h2(x)T2(x)]/[h1(x) + h2(x)] (9.5d)
where
A = 2W[C2h1(x) + C1h2(x)]/{C1C2[h1(x) + h2(x)]},
B = Wh1(x)h2(x)(C2 − C1)/{C1C2[h1(x) + h2(x)]},
Q(x) = −òxAq¢(x¢)e–Bx¢dx¢,
Ci = micpi, i = 1, 2.
This approach was further extended to a multiple-channel counter flow system

[9, 10] and has been applied to solid state phased array radar [11].
The focus of this chapter is aiming at the mini or micro channel flow which
has practical applications for cooling of microelectronics. A liquid cooling of chip
package based on the microchannel technology as given in Figure 9.9 was first
proposed by Tuckerman [12, 13]. The integrated microchannels are directly at the
back of the silicon chip which greatly reduces the thermal path length from the
heat sources to the coolant. The height and width of individual cooling channels
are 300 μm and 50 μm, respectively. The power of 790 W is dissipated over the
1 cm ´ 1 cm chip. The experimental results with the water flow rate of 8.6 ml/sec
indicate that the temperature difference from the coolant inlet and maximum cold
plate temperature is 71°C and the calculated heat transfer coefficient is 11.13 W/
cm2 °C. Thermally, the microchannel cooling is very effective but its pressure drop
which is 31 psi (214 kPa) at the flow rate of 8.6 ml/sec is also too high because of
the small cooling channels to be acceptable for practical applications.
Yeh [14] performed similar tests with various types of cold plates. Figure 9.10
shows the picture of one of tested cold plates (the brass cold plate). The overall
F igure 9.9 Microchannel cooled silicon chip.
Figure 9.10 Microchannel cold plate and round heater.

dimensions of the cold plate are 1 in. ´ 1 in. ´ 0.064 in. The cold plate has 75 chan-
nels with the cross section of each channel is 0.01 in. (0.254 mm) by 0.006 in.
(0.1524 mm). A round foil heater with the cross section area of 0.15 in.2 is also
shown in the figure. Four thermocouples were instrumented on the backside of
the cold plate (opposite to the heater side).
Since no thermocouple is installed at the heater side of the cold plate, to deter-
mine the temperature on the heater side of the cold plate, a thermal model was set
up to simulate the test conditions. The measured and predicted temperatures at the
center of cooling section are presented in Figure 9.11. As can be seen from the figure,
an excellent agreement is found between the test and simulation results (average
about 1°C difference). The simulation also predicts the maximum cold plate tem-
perature on the heater side which is at the center of the heater. The predicted maxi-
mum cold plate temperature at this location with the flow rate of water at 33 ml/min
is 40°C above the inlet water temperature (21.8°) when the heater dissipates 50 W
with the equivalent heat flux of 334 W/ in.2 (518 W/cm2). It should be noted the flow
rates used in Tuckerman and Yeh tests are 8.6 ml/sec and 33 ml/min, respectively.
A thermal test was also performed on a swirl cold plate that has 30 trenches as
shown in Figure 9.12. The cross section of each trench is 0.01 in. by 0.016 in. Water
enters the center of the cold plate which impinges the heated side of the cold plate
and flows radically outwards along the channels to the outer ring as indicated in
Figure 9.12. The overall dimensions of this aluminum cold plate are same as the
above on (1.0 in. ´ 1.0 in. ´ 0.064 in.). The diameter of the nozzle which is located
at the center of the cold plate is 0.022 in. Again, a round heater (0.15 in.2) as one
given in Figure 9.10 was installed at the center of the cold plate.
The cold plate is tested at the vertical orientation as shown in Figure 9.12. The
water inlet is at 24°C. The measured cold plate temperatures on the heater side
F igure 9.11 Temperatures of microchannel cold plate.

F igure 9.12 Swirl cold plate.
at various water flow rates ranging from 33 ml/min to 73 ml/min are presented
in Figure 9.13. The open symbols (TC 11) represent the temperature at the cen-
ter of the cold plate. The hottest spot (TC12) represented by a closed symbol is
located near the center of upper half of the cold plate. For a flow rate of 33 ml/
min, the measured hottest spot is about 40°C above the water inlet temperature
when 50 W of heat are applied. This result is similar to that of the brass cold plate
discussed previously. However, the pressure drop in the swirl cold plate is much
lower than that of the brass cold plate. The heat transfer coefficient and the pres-
sure drop are correlated into the following equations, respectively.
For center location
Nud = 0.171 Red0.6 (9.6a)
For hottest spot location
Nud = 0.155 Red0.6 (9.6b)
For pressure drop
DP = −0.034 + 0.2417 ´ 10−3 Red + 0.1299 ´ 10−5 Red2 (9.7)
where Red is Reynolds number based on the nozzle diameter.

Figure 9.13 Thermal performance of swirl cold.
The microchannel coolers proposed by Tuckerman [12, 13] could not be fabri-
cated easily and also pressure drops were extremely high which were not suitable
for the practical application in the real world conditions. Recently, the advances
in manufacturing technologies, including the progress in high-rate, deep reactive
ion etching of silicon have made possible for the easy fabrication of silicon micro-
channel cooler at the reasonable cost [15]. In addition, the methods for reducing
the pressure drop have also been reported, including subdividing the flow into
multiple heat exchanger zones with shorter channel lengths and manifold designs
with large cross sectional area is presented in Figure 9.14. For flow rate at 1.01 l/
min, the overall thermal resistance and the pressure drop are 17.5 °C mm2 /W and
30 kPa, respectively.
F igure 9.14 IBM microchannel cooled silicon chip.
Yeh [16] conducted an experiment to study the heat transfer and flow friction
of offset fins as shown in Figure 9.15 with various liquids in narrow passage. The
fin stock used in cold plate slat for all heat transfer tests has 15 fins per inch.
All fins have a height of 0.04 in. (channel height), a thickness of 0.004 in., and a
length of 1.25 in. in the flow direction.
F igure 9.15 Thin rectangular offset fins.

Figure 9.16 Friction factor and Nusselt number for offset fins in nar-
row passage.
The summary results are presented in Figure 9.16 and are correlated into the
following equations:
f = C(x/D)−0.384(w/H)−0.092ReD−n (9.8)
and
Nu = 0.1 ReD0.74 Pr1/3 for ReD ³ 90 (9.9a)

Nu = 0.4 Re 0.43
D
Pr
1/3
for ReD < 90 (9.9b)
where C and n are 8.591 and 0.66 for water, respectively, and
Nu(hD/k) = Nusselt number

ReD(ρVD/μ) = Reynolds number based on hydraulic diameter D
Pr(μCp/k) = Prandtl number
Bland, Niggemann, and Parekh [17] developed a compact high intensity cooler
(CHIC) to cool 50 W/cm2 device for space applications. Therefore, the coolant
must not freeze in the system and should also have a low pressure drop to reduce
the pumping power. In addition, the cooler must also be lightweight and com-
pact. The geometry (cross-section) of this heat exchanger cannot be larger than
1 cm ´ 1 cm square area. The primary coolant in the system is Freon 11.
The cooler consists of a stack of thin copper orifice plates and spacers which
are bonded together and arranged to provide liquid jet impingement heat transfer
on successive plates as shown in Figure 9.17. There are cutouts on the spacers
in the region of the orifices. The hole pattern on alternate orifice plates is offset
by half the hole pitch so that jets from orifices are targeted at locations midway
Figure 9.17 CHIC heat exchanger design concept.
between holes on the next orifice plate. This stack is bonded to the base plate
which directs the liquid to the drain channels and provides the means by which
the heat exchanger is attached to the heat source or device to be cooled.
Figure 9.17 shows the fluid entering through an inlet tube into an inlet ple-
num where it separates into three streams and passes through orifices in the top
plate. The liquid jets impinge on the next orifice plate to remove heat conducted
through the spacer plates from the heat sources. Heat is first conducted to the
base plate of the cooler and then is conducted through contacting areas between
the orifice plates and the space plates and is finally transferred to the coolant
from the orifice plates. Heat transfer path is shown by the dashed-line arrows,
while the flow path is represented by the solid arrows.
The liquid from impinging jets then flows transversely into the nearby orifices
and impinges on the next orifice plate again. The process is continued until the
jets impinge on the base plate of the drain channels. The coolant then flows later-
ally through the narrow gap inside the outer shell into the drainage slot and out
through the fluid outlet tube.
Thermal tests on the wafer thin coolers [18] were performed for the heat flux
ranging from 5 to 125 W/cm2. The primary application of these liquid cooled
devices is to remove heat from compact gallium arsenide (GaAs) diode wafers
used in laser communications as shown in Figure 9.18. The tested wafer thin cool-
ers are a double pass microchannel cooler, two types of single pass microchannel
coolers, and two versions of jet impingement coolers such as CHIC mentioned
above. In a typical application, GaAs diode wafers are fabricated including 40 to
400 individual diode emitters. For maximum efficiency, all diodes must operate
below or near 25°C with the maximum temperature gradient less than 1°C.
The coolers are designed to support high heat fluxes over 100 W/cm2 with mini-
mal temperature gradients at the heating surface. The thin coolers have a heat
F igure 9.18 Laser diode/cooler stack.
input area of 1 cm2 and the thickness ranging from 1 to 1.8 mm with water as cool-
ant. One single pass cold plate was made of beryllium oxide and other four coolers
were made of copper. Typical results for heat flux at 125 W/cm2 with coolant flow
rate of 10 kg/hr indicated that the heat transfer coefficients and maximum surface
temperature gradient were 8 W/cm2 °C and 0.7°C for double pass microchannel,
6.5 W/cm2 °C and 6.8°C for the copper single pass microchannel, 3.9 W/cm2 °C and
6.1 °C for the beryllium oxide single pass microchannel, 3.5 W/cm2 °C and 2.5°C for
the first impingement cooler and 5.3 W/cm2 °C and 2.4°C for the second impinge-
ment cooler, respectively.
9.4 INDUSTRY TRENDS IN LIQUID COOLED ELECTRONIC

EQUIPMENT
Since most of telecommunication systems with exception of some outdoor equip-

ment are cooled by air, it is of interest in examining the current industry trends in
liquid cooled electronic equipment, especially in computer industry. In addition to
direct immersion cooling, two general schemes in liquid cooling of chip package
are direct internal cooling of silicon die and external cooling of the package through
the cold plate. The former includes the microchannel cooling as described by Tuck-
erman and Pease [12], Tuckerman [13], and Colgan [15] where the cooling channels
are integral part of the silicon die. On the other hand, the latter is to replace the con-
ventional heat sinks used in the air cooled systems by a liquid cold plate which is in
contact with the component case (or lid). The flow pattern in both the internal and
external cooling methods includes liquid through channels and jet impingement.
Water has long been used in cooling of computers. One of most noticed liquid
cooled computers in the earlier dates was IBM Thermal Conduction Module (TCM)
Figure 9.19 IBM liquid cooled thermal conduction module.
as shown in Figure 9.19 [19]. IBM TCM is an indirect liquid cooling scheme. Heat
generated from microelectronics (chips) is conducted through spring loaded pis-
tons to the cold plates on the top. To reduce the thermal contact resistance, air
is replaced by helium to fill up the space in the individual piston because helium
has a higher value of thermal conductivity. The construction of TCM is extremely
complicated.
To modify the thermal path between the chip/package and the water cooled cold
plate, Andros and Shay [20] proposed utilizing a miniature flexible heat pipe to con-
nect the chips and the cold plate as illustrated in Figure 9.20. Similar modification
as shown in Figure 9.21 was also made by Clark and Metreaud [21]. The latter also
Figure 9.20 Liquid cooling of bare chip with flexible heat pipe.
Figure 9.21 Liquid cooling of bare chip with flexible heat pipe
extension into liquid flow.
eliminates the contact thermal resistance at the interface between the cold plate and
the flexible heat pipe by extending the flexible heat pipe directly into the liquid flow.
Liquid jet impingement was employed for cooling a large scale computer
(Fujitsu M-780) as given in Figure 9.22 [22]. As shown in Figure 9.22, Rcond is
the thermal resistance due to conduction from the chip junction to FTC (flexible
thermal conductor) heat transfer plate surface, and Rconv is the thermal resistance
F igure 9.22 Fujitsu FACOM M-780 computer.

due to convection (water jet impingement) from the FTC heat transfer plate to the
coolant. The total thermal resistance is estimated to be 2.4°C/W with chip power
of 9.5 W (336 chips on board). The coolant is supplied by the coolant distribution
and control unit (CDUC) as shown in the left side of the figure.
The NEC supercomputer SX-2 as illustrated in Figure 9.23 [23] employs a high-
density multi-chip package and a liquid (water) cooling module (LCM) structure
which consists of a heat transfer block (HTB), a cold plate and 36 studs placed in
machined holes in the HTB to insure good contacts with LSI chip carrier mounted
on a multi-layer substrate. The cooling unit as shown on the left of the figure sup-
plies the cooling water to the cold plate of LCM. The average thermal resistance
is measured at 3.74°C/W with the maximum system power of 3000 W.
Yeh [11] employed dielectric liquid (coolanol-25) to cool RF modules with a
large cold plate as presented in Figure 9.24. Heat generated from RF modules was
transferred by heat pipes to the aluminum cold plate consisting of 53 extruded
channels with fluid flowing in opposite direction in alternating channels. Each
heat pipe serves a pair of RF modules. The total system power is 7.4 kW. The cool-
ant is supplied to the cold plate at seven gallons per minute (GPM) with pressure
of 70 PSIG and a maximum temperature of 46°C.
Once coolant enters supply manifolds, the fluid flows in opposite directions
through the adjacent fluid channels and is collected in the discharge manifolds as
it leaves the system. The counter flow scheme is selected to reduce the maximum
cold plate temperature and also to minimize the temperature difference of over
the cold plate. Each fluid channel which is made by the extrusion process includes
several subchannels in order to enhance the liquid convection heat transfer of the
cold plate by increasing the heat transfer surface. As presented in Figure 9.24,
Figure 9.23 NEC SX-2 cooling system.

Figure 9.24 Liquid cooled clod plate for RF systems.
F igure 9.25 IBM Power 575 system.

various subchannels in each case (from top, Case A to bottom, Case D) are inves-
tigated to meet the following requirements : (1) thermal performance, (2) weight,
(3) producibility, (4) structural integrity and (5) cost. The initial design of 10 rect-
angular subchannels (Case A) is not acceptable because the sharp, square sub-
channels are not feasible in the extrusion process and Cases B and C are rejected
because the wall thickness is too thin.
The hybrid cooling over the boards has recently received considerable attention. In
other words, the high power components on the board are cooled by the liquid while
the rest components are still cooled by the air. Figures 9.25 [24] and 9.26 [25] present
some examples of this cooling scheme. The former corresponds to IBM Power 575
system while the latter is Fujitsu GS 8900. The issue of possible condensation inside
the rack must carefully be considered for the system with hybrid cooling.
The possible fan failure and also the acoustic noise issue may still exist in the
hybrid cooling scheme. In addition, the liquid loop (tubes) may become obstacle
to the certain air flow to the components. Therefore, the active liquid cooling is
the best approach to achieve the energy efficiency both at equipment level and
data center and also to support the much higher system power while eliminating
the acoustic noise from air cooling.
F igure 9.26 Fujitsu High End Server (GS8900).

Figure 9.27 IBM Zurich liquid cooled board.
IBM Research Zurich Lab in conjunction with ETH developed a warm water
(60°C) cooled supercomputer (Aquasar) using microchannel cold plate for cooling
the components as shown in Figure 9.27 [26]. The detailed cooling scheme of
individual cold plate developed by Escher [27, 28] is given in Figure 9.28. For a
flow rate of 1.3 l/min, the total thermal resistance between the maximum heater
temperature and fluid inlet temperature is 0.09 cm2 °C/W with a pressure drop of
0.22 bar on a 2 ´ 2 cm2 chip.
Careful planning of the fluid network must be made to insure the proper flow
rate to individual boards and also to individual components on given boards to
avoid overheating. The former is related to the system flow distribution and the
later represents the flow distribution on the board level.
Figure 9.28 Scheme of a microchannel heat sink.

9.5 SYSTEM THERMAL DESIGN
While the major effort during the design of any liquid cooled electronic systems
are at the cooling aspects, attention must also be given to the practical design
aspects which are reflected in the maintainability of the equipment. To achieve
ease of maintenance and repair, system designs should be simple, and provide
easy access to the internal parts. Detailed maintainability studies are required
for any equipment based on selected cooling concepts. Factors requiring consid-
eration include: (1) a minimum of external connections for the cooling circuitry,
(2) quick dis-connect devices, (3) enclosures which are easy to disassemble and/or
remove from the working electronic equipment, (4) mounting devices from which
the electrical components are easily removed, and which are easily installed and
removed from the chassis, and (5) seals for use with pressurized, immersion and
vaporization cooling concepts. In addition, the system should not be required to
avoid drain the coolant during the service.
The degree of maintainability achieved of given equipment depends not only
on the simplicity of its design and accessibility of its components but also on the
reliability of the system/components. Therefore, the advanced cooling concepts
are more complex and less accessible than those using air cooled systems, this
does not necessarily imply that such equipment requires more maintenance. The
increase in reliability expected from the liquid cooling should reduce the frequency
of required repairs which result in the greater availability of the equipment for
use despite the fact that it may take a longer time for a repair as compared with
the air cooled systems.
9.5.1 Direct Liquid Cooling
As in air cooling, liquid cooling can be either direct or indirect cooling. Dielectric
liquids must be utilized in the direct cooling because liquids are in a direct con-
tact with electronic components. Direct cooling can employ natural convection/
pool boiling or forced convection/flow boiling mode of heat transfer with either
single- or two-phase flow as shown in Figure 9.29A and 9.29B [29]. The electronic
components are immersed in a pool of liquid in the former case and the liquid is
forcefully circulated over electronic components in Figure 9.29B.
In the simplest case as shown in Case A of Figure 9.29A, an electronic equip-
ment is immersed in a sealed contained. Heat generated by electronics is removed
by liquid through natural convection and is further carried away to contact with
the walls of the container and is finally cooled by transferring heat to outside of
the container. The flow circulation is caused by the density difference between the
cold and hot fluids. This process is often referred to as a natural circulation. In
addition, the heat sink as given in Cases B and C is actually a condenser to con-
dense the vapor back to liquid. It should be noted for liquid filled containers as
given in Figure 9.29A, account must be taken of the volumetric change due to its
temperature rise.
If forced convection immersion cooling is used, the heat removed from the equip-
ment is rejected to the secondary fluid by means of a suitable heat exchanger. The
higher heat transfer coefficient associated with the forced convection significantly
improves the system heat removal capacity. However, such forced convection
F igure 9.29A Passive direct liquid cooling.
F igure 9.29B Active direct liquid cooling.

system requires an additional pump and heat exchanger which increases the over-
all power consumption, weight and cost. An accumulator is often added to the
system to allow for liquid expansion and minimize vapor lock.
For forced convection and/or flow boiling as presented in Figure 9.29B, the
flow pattern is parallel in Case A while is in series in Case B. The parallel flow
arrangement is preferred and it has many advantages over the case with a series
flow scheme. The major advantage of the parallel flow system is to provide mini-
mum pressure drop and maximum distribution of the cold liquid. The system
pressure drop is extremely small for the parallel flow arrangement as compared
to the series flow system because of small flow rate (fraction of the total flow)
and short flow length of each parallel flow path. In addition, the temperature
difference among all boards is relatively small. On the other hand, the fluid is
continuously heated in the series flow arrangement which results in a large the
temperature difference between the inlet and exit. It should be noted that the
pressure drop through each of the parallel path is essentially the same. In order
to provide sufficient coolant to each flow path, the flow rate to each path (boards)
must be proportional to the heat load of individual paths (boards). Therefore, the
flow rate to individual paths must be properly regulated.
The nucleate boiling will eventually take place in either natural or forced con-
vection liquid immersion if the system power is sufficiently high. The nucleate
boiling is the most effective heat transfer mechanism. In addition to extremely
high heat transfer coefficients in this region, the boiling conditions are charac-
terized by relatively small temperature differences between the heating surfaces
and the coolant, typically less than 5°C. Therefore, one should take advantage of
this efficient heat transfer mechanism in the design of liquid cooling systems for
electronic equipment. One major consideration in selection of the coolant for this
type system is the boiling point of the fluid. The boiling temperature of a fluid is
a function of the saturated pressure of the system which must be lower than the
required temperature limits of electronic components used in the equipment. The
desirable boiling point of given fluids for the system under consideration can be
estimated by the following equation
Tbp = Tj – Q* ´ qj−c − DTsat (9.10)
where
Tbp = liquid boiling point

Tj = component junction temperature limit
Q* = component power
qj−c = component junction-to-case thermal resistance
DTsa = difference between component case and liquid saturated
temperatures
9.5.2 Indirect Liquid Cooling
In an indirect cooling, the coolant does not make a direct contact with the elec-
tronic components. Heat is removed from electronics by various heat transfer
modes to a cooling device such as cold plate, and heat is further carried away to
another heat exchanger where the hot liquid is cooled and is returned to the cold
plate. The liquid cooling process is repeated again and again. A condenser will be
required if the boiling takes place in the cold plate. The thermal resistance at the
interface between the cold plate and components is extremely critical in any indi-
rect liquid cooling system and care must be made to minimize this interface ther-
mal resistance. Indirect liquid cooling has certain advantages over immersion or
direct liquid cooling. Among them are easier accessibility for maintenance, less
possibility of fouling equipment, less concerns of handling of coolants, and great
flexibility on coolant selections where, for example, dielectric fluids are no longer
required and high heat transfer fluids such as water can be used.
A single fluid heat exchanger designed for removing heat from any heat sources
is called a “cold plate.” The cold plate is a special case of a heat exchanger which
involves two fluids in the system. The cold plate can be employed in either air or
liquid cooling systems. Cold plate cooling concepts take advantage of the large
number of efficient heat exchanger configurations, e.g., various types of fins. As
in any heat transfer systems, the key factor in thermal design of a cold plate is to
maximize heat transfer rate and to minimize the pumping power.
9.5.3 General Design Considerations
Though liquid cooling has long been employed in the computer industry for a
long time, no liquid cooled indoor telecommunication equipment is commercially
available at the present time. Water is the most common fluid used in the high
power computers because of its high heat transfer capability. However, water can
not be used in a direct immersion cooling and nor for outdoor applications. The
former is due to its poor dielectric properties and the later is because of its high
freezing point. Water also must be properly treated prior to being used in any
cooling systems. Most chemicals used in water treatment are proprietary formu-
las. It is very important to select a compatible pair of liquid coolant and wetted
materials throughout the entire liquid loop. For material compatibility, copper
must be used for the entire liquid loop if water is adopted as the coolant. On the
other hand, aluminum can be employed for all dielectric fluids, including Glycol.
De-ionized water which is often used as coolant is compatible with stainless steel.
All above three types of materials (copper, aluminum, and stainless steel) are
compatible with Glycois.
All cooling methods discussed previously only deal with liquid cooling of indi-
vidual components or boards (subsystems). Generally, coolant first goes into an
inlet or supply manifold once it enters the system (equipment) and then flows
through inlet connecting tubes to the cold plates associated with individual boards
and finally liquid is collected at the exit or discharge manifold through exit con-
necting tubes prior to leaving the system. Since the system (equipment) always
consists of a number of boards (or subsystems) whose power may vary, one must
make sure proper flow rate be distributed to individual boards. The flow rate to
individual boards should be properly regulated in proportion to the power of the
boards. Sometimes, a large flow rate will still be needed even for a low power
board if this board includes some critical components such as optical modules
Figure 9.30 Redundant liquid cooling system.
Figure 9.31 Saturation pressure versus temperatures for water and

FC-72.
Figure 9.32 Saturation pressure versus temperatures for Freon-12

and ammonia.
with low component temperature limits. The simplest way to have a proper flow
distribution to individual boards is to place various sizes of the orifices in the inlet
connecting tubes or exit connecting tubes.
In addition, the equipment must also be designed with redundant cooling
loops to insure the system reliability though it makes system more complicated.
One of examples of the redundant cooling loops is presented by Yeh [11]. The
sketch of the redundant cooling system for the cold plate as shown in Figure 9.24
is given in Figure 9.30. The system has two independent and separate flow paths
through the entire cold plate.
The above discussion of indirect liquid cooling is limited to the single-phase
liquid cooling. Two-phase flow cooling has much higher thermal efficiency than
that of single-phase liquid cooling. However, the two-phase flow system is much
more complicated which requires a condenser so that the vapor can be condensed
for reuse. For water cooled systems, including all current supercomputers, water
must be boiled at around 60°C in order to maintain the chip junction temperature
below 85°C. The boiling point of water at atmospheric pressure (14.696 psia) is
100°C. Therefore, system pressure must be adjusted to below the atmospheric pres-
sure at about 2.9 psia as given in Figure 9.31 by pulling the vacuum. Figure 9.31
shows the saturation pressure as a function of the saturation temperature for the
water and the FC-72. As can be seen from the figure, the boiling point for FC-72 at
the atmospheric pressure is at about 61°C which is much lower than that of water
(100°C). Similar curve for refrigeration fluids such as Freon-12 and Ammonia is
given in Figure 9.32.
Figure 9.33 Estimated flow rates required of dielectric fluids for

given system power.
Generally, the effects of coolant properties on heat transfer and friction fac-
tor is minimum because of the temperature difference between the inlet and
exit of a cold plate is not large enough to be concerned. However, it may become
critical in the microchannel cooling systems. This is due to the fact that very
little flow is used in such systems which result in large temperature gradient of
the fluid between the inlet and the exit. The microchannel liquid cooling tech-
nologies are generally employed in cooling of high power systems and it typi-
cally has a large pressure drop because of extremely small channel cross section
even operating at small flow rates. With these factors, high system power and
high pressure drop, the liquid coolant could be boiled near or at the exit of the
cooling systems.
For any forced convection system, the pressure drop is as important as the
heat transfer. A high-pressure drop system will require a large pump which not
only consumes more power but also is heavy and costly. High heat transfer always
accompanies by the high pressure drop. A good system design must have a good
balance between the heat transfer and the pressure drop. Therefore, the goal of
system thermal design is to maximize heat transfer and to minimize pressure
drop. The best way to achieve the goal is to adopt efficient cooling schemes which
will require the least flow rate. For dielectric fluid systems, the required system
flow is proposed and is given in Figure 9.33. This curve is generated based on
the limited data by Yeh [11] and the past experience of the authors. The actual
flow rate required is totally depending on the system thermal requirements and
the selective cooling methods. Generally, the higher flow rates are needed for
higher liquid inlet temperatures. The proposed curve is intended to be used as a
general guideline in determining estimated flow rates required for the given sys-
tem power. The much less flow rates will be needed if the coolant is water because
of its high heat transfer capacity. To limit power consumption of the pump, it is
recommended that the system pressure drop should exceed 20 psia under the
supplied flow rate and the coolant inlet temperature.
REFERENCES
1. Yeh, L. T. and Chu, R. C., “Thermal Management of Microelectronic Equip-

ment”, ASME Press, 2002.
2. Knight, A. F., “Choice of Fluids for Cooling Electronic Equipment”, Electro-
Technology, June, 1963.
3. Kelly, D. J., “Resistance of Materials to Hydraulic Fluids”, Machine Design,
January, 21, 1972.
4. Lee, T. Y., Saylor, J., Simon, T., Tong, W., Wu, P. S., and Bar-Cohen, A., “Fluid
Selection and Property Effects in Single- and Two-phase Immersion Cooling”,
Proceedings of Intersociety Conference on Thermal Phenomena in Electronic
Components, May, 1988.
5. Park, K. A., and Bergles, A. E., “Natural Convection Heat Transfer Charac-
teristics of Simulated Chips”, Heat Transfer in Electronic Equipment, ASME
HTD-48, 1985.
6. Marto, P. J. and Lepere, V. J., “Pool boiling Heat Transfer from Enhanced
Surfaces to Dielectric Fluids”, J. of Heat Transfer, Vol. 104, 1982.
7. Oktay, S., “Departure from Natural Convection (DNC) in Low-Temperature
Boiling Heat Transfer Encountered in Cooling Micro-Electronic LSI Devices”,
Proceedings of 7th International Heat Transfer Conference, 1982.
8. Yeh, L. T., “Analytical Solutions for a Counterflow Heat Exchanger with Space
Dependent Wall Heat Dissipation,” Heat Transfer in Electronics Equipment,
ASME HTD-48, 1985.
9. Yeh, L. T., and Gingrich, W. K., “Numerical Solutions for a Multiple-Channel
Counterflow Heat Exchanger with Space Dependent Wall Heat Dissipations”,
Proceedings of the 8th Int. Heat Transfer Conference, 6, San Francisco, CA,
1986.
10. Gingrich, W. K., Yeh, L. T., and Fuhr, T. D., “Transient Temperature Distribution
for a Multiple-Channel Counterflow Heat Exchanger with Space Dependent Wall
Heat Dissipations”, ASME Winter Annual Meeting, San Francisco, CA, 1989.
11. Yeh, L. T., “Thermal Design of a Multiple-Channel Bidirectional-Flow Cold Plate
for Solid-State Phased Array Radars”, Proceeding International Symposium on
Cooling Technology for Electronic Equipment, Honolulu, HI, March, 1987.
12. Tuckerman, D. B. and Pease, F. F., “High Performance Heat Sinking for VLSI”,
IEEE Electron Devices Letter, EDL-2, 1981.
13. Tuckerman, D. B., “Heat Transfer Microstructures for Integrated Circuits”,
Doctoral Thesis, Stanford University, 1984.
14. Yeh, L. T., Unpublished data, 1989.
15. Colgan, E. V., Bezama, R. J., Gaynes, M., and Marston, K. C., “A Practical
Implementation of Silicon Microchannel Coolers”, Electronics Cooling, Vol-
ume 13, Number 4, 2007.
16. Yeh, L. T., “An Experimental Study of Offset Fins in a Narrow Channel”, Pro-
ceedings of 1st International Conference Aerospace Heat Exchanger Technol-
ogy, Palo Alto, CA, 1993.
17. Bland, T. J., Niggemann, R. E., and Parekh, M. B., “A Compact High Intensity
Cooler (CHIC)”, 13th Intersociety Conference on Environmental System, July
11−13, 1983.
18. Grote, M. G., Hendron, R. E., Kipp, H. W., Lapinski, J. R., “Test Results of
Wafer Thin Coolers at Heat Fluxes from 5 to 125 W/cm2 °C”, 18th Intersoci-
ety Conference on Environmental Systems, San Francisco, CA, July 11–13,
1988.
19. Oktay, S. and Kammerer, H. C., “A Conduction Cooled Module for High Per-
formance LSI Devices”, IBM J Research Development, 26 (1), 1982.
20. Andros, F. E., and Shay, R. J. E., “Micro Helix Thermo Capsule”, US patent
No. 4,313,492, February, 1982.
21. Clard, B. T. and Metreaud, C. G., “Cooling device for Multilayer Ceramic
Modules”, IBM Technical Disclosure, 20 (5), 1977.
22. Yamamoto, H., Udagawa, Y., Suzuki, M., “Cooling System for FACOM M-780
Large-Scale Computer”, Proceeding International Symposium on Cooling
Technology for Electronic Equipment, Honolulu, HI, March, 1987.
23. Mizuno, T., Okano, M., Matsuo, Y., and Watari, T., “Cooling Technology for The
NEC SX Supercomputer”, Proceeding International Symposium on Cooling
Technology for Electronic Equipment, Honolulu, HI, March, 1987.
24. Ellsworth, M. J., Jr., Campbell, L. A., Simons, R. E., Iyengar, M. K., Schmidt,
R. R., Chu, R. C., “The Evolution of Water Cooling for IBM Large Server Sys-
tems : Back to Future”, Itherm, 2008.
25. Wei, J., “Hybrid Cooling for Fujitsu Large Computer Systems”, 2010 IEEE
SPJW, Kyoto, Japan, 2010.
26. Wrunschwiler, T., Meijer, G. I., Paredes, S., Escher, W., and Michel, B., “Direct
Waste Heat Utilization from Liquid-Cooled Supercomputer”, Proceedings of
the 14th International Heat Transfer Conference, Washington, DC, August
8−13, 2010.
27. Escher, W., Michel, B., and Poulikakos, D., “Efficiency of Optimized Bifurcat-
ing Tree-like and Parallel Microchannel Networks in the Cooling of Electron-
ics”, Intl. J. of Heat and Mass Transf. 52, 2009.
28. Escher, W., Wrunschwiler, T., Michel, B., and Poulikakos, D., “Experimental
Investigation of an Ultra-thin Manifold Microchannel Heat Sink for Liquid-
Cooled Chips”, Intl. J. of Heat and Mass Transf., 2010 (current under review).
29. Helenbrook, R. G., and Anthony, F. M., “Development of Liquid Cooling Tech-
niques for Advanced Airborne Electronic Equipment”, Ball Aerospace Com-
pany, AFFDL-TR-71-129, March, 1972.
Chapter 10
Thermal Management
and Energy Efficiency
The network traffic in telecommunication industry has grown very rapidly every
year since its inception. As projected, the network traffic demand will reach tens
or hundreds of Tb/s in a couple of years. In response to the projected growth, new
design and architecture are needed in order to face the power-density challenges
in the next generation telecommunication networks. As pointed out by Vukovic
[1], there are two aspects of these issues. One is the network architecture and
another is telecommunication equipment design.
The main goal of the architecture is to minimize the total power per network
throughput. To achieve this goal, one can employ the so called all-optical net-
work (AON) which keeps the transfer of bits in the optical domain. The AON can
develop a “less power hungry” network with the added benefit of being able to
cope successfully with the expected network growth and cost reduction require-
ments. The second issue deals with the telecommunication equipment design.
The key for the successful design is to explore techniques and technologies for
reducing power at the source by integrating state-of-the-art industry advances
in high-density packaging and high speed interconnects. The key challenge of
high-density and high speed equipment is the thermal management which must
include a balanced combination of materials, technologies and cost.
The cost of thermal control unit is typically a very small percentage of the total
system cost of any electronic equipment. For example, the cooling cost ranges
from less than 1% for the basic PCs and to 3–5% for large servers and approaching
10% of largest super computers, especially for the liquid cooled systems. Figure
10.1 shows the cooling cost versus power [1A]. As expected, the higher system
power, the more complicated and expansive the thermal solutions will be.
This chapter is to serve as the summary of the book which describe the current
industry trends of thermal management in the wide range of telecommunication
systems and the data centers.
10.1 POWER DENSITY OF TELECOMMUNICATION

EQUIPMENT
Currently, telecommunication equipment is facing numerous challenges. The

major one is the increased power density. Architects of the next-generation core
F igure 10.1 Cooling cost versus heat dissipation.
network are demanding exponentially higher system throughputs to keep up with

network traffic growth. With projected node data transfer rates from 40–100 Gb/
sec currently to 200–400 Gb/sec in the next couple of years, and extrapolating
from current equipment, future nodes would consume and dissipate up to 100s
of kilowatts of power.
Because of the advances in circuit and component technologies, electric circuits
have become more efficient and thus heat dissipation from individual transistors
has also become less. However, miniaturization of circuits greatly decreases the
size of individual devices and increases the number of such devices that can be
Figure 10.2 Maximum chip power trends in telecommunication

applications.
Thermal Management and Energy Efficiency • 179
Figure 10.3 Maximum chip heat flux in telecommunication

applications.
integrated on a single chip. The net result is that the chip heat dissipation has
significantly increased as illustrated in Figures 10.2 and 10.3, respectively [2].
The trend reflected in the projected growth in the heat flux based upon the
footprint area of the equipment is shown in Figure 10.4 [3]. As can be seen from
the figure, the system power for the core network is the highest among all prod-
ucts presented in the figure.
Figure 10.4 Thermal loading at facility level for various products.

10.2 CRITICAL THERMAL ISSUES
The equipment power consumption is continuously increased at a fast pace.

Since the current telecommunication systems are still focusing at the air cooling,
large high performance fans are becoming a must for the high power systems in
order to provide the necessary airflow rates. In addition to the system thermal
issues, two major concerns about these large fans are the power consumption
and the acoustic noise of the fans. The former is related to the equipment energy
efficiency, while the latter is strictly regulated by the government and industry
standards such as NEBS in USA and North America, and the ETSI in Europe.
The increased system power results in a significant increase in the power
consumption and the cost of powering and cooling of the equipment as well as
the operation cost of the data centers in the last few years. For example, AT&T
and Verizon have reportedly spent over 1 billion and 500 million US dollars,
respectively to operate their data centers throughout the United States in 2006.
The cost of electricity has steadily increased in the last few years. The operation
cost will be even higher due to the expansion of the date centers. In short, the total
operation cost could easily be doubled in the next few years. The similar trends
are also found in the worldwide base. The operation cost is the major concern for
the data centers. Therefore, the energy efficiency becomes the top priority for the
equipment manufacturers and also the data center operators.
10.2.1 Equipment Energy Efficiency
The COP (coefficient of performance) is often adopted in the telecommunication

industry to measure the thermal or energy efficiency of the equipment. The COP
is defined as the ratio of the total system power, including power from electronics,
power supply/converters, and fans to the total power consumption of fans. There-
fore, the higher power consumption of the fan, the lower system COP will be. In
general, the COP of any equipment under the worst-case condition with fans at
full speed should be greater 10.
10.2.2 Equipment Acoustic Noise
The equipment for the indoor applications must meet the GR-63 requirements of
NBES [4] for the USA and North American markets. For the acoustic noise lim-
its, the GR-63 states that the acoustic noise of any equipment must be less than
78 dBA (sound power) at the ambient of 27°C. On the other hand, ETSI states
that the acoustic noise (sound power) should be limited to 72 dBA or less at the
ambient of 23°C.
The high acoustic noise is a byproduct of high power systems associated with
the fans. The acoustic noise can be reduced by either active or passive solution or
combination of both. The passive solution is achieved by adding acoustic noise
reduction (absorption) materials at the proper locations throughout the system or
installing the acoustic door for the racks. Since the characteristics of the acoustic
noise are very similar to that generated by a microphone-speaker, it is possible
to create an identical sound wave with 180° phase difference to cancel the fan
acoustic noise. Based on this finding, the ANC (active noise cancellation) device
which generally includes several electronics on a small printed circuit board,
the speaker and the microphone is developed. An ANC device is typically to be
installed at the downstream of the fans.
10.3 THERMAL MANAGEMENT AND ROADMAP
Air cooling has long been considered and adapted for cooling of electronics. This
is especially true for the consumer electronics used in homes and offices. The
main reason is that the air cooling is relatively simple, cost-effective, and safe. Air
cooling is also a main cooling method for all current telecommunications equip-
ment. However, due to poor thermal properties of the air, there is a system power
limitation for the air cooling.
The thermal technology map for various power systems with different cooling
schemes is developed and presented in Figure 10.5 [5]. The horizontal axis shows
the estimated heat transfer coefficients. The left vertical axis in the figure gives
the system power while the right vertical axis represents the system complexity
and cost. Generally, the cooling system becomes more complicated and costly
when the system power increases. It should be noted that Figure 10.5 should only
be used as a general guideline. It is believed that the upper limit of the forced air
cooled rack is about 25,000 W. Any system power greater than the above value,
F igure 10.5 Thermal technology map.

liquid cooling may be considered. For practical engineering applications, the

forced air cooling is not recommended for any system over 20KW even it works
thermally because of acoustic noise and the COP issues.
The overall thermal management challenges in the next generation telecom-
munication equipment are as follows:
1. Heat removal from high density and high heat flux packages within
reducing volume with low pressure drop.
2. Hot spot thermal management in packaging level
3. Heat spreading from chip packages (small area) to heat sinks (large
area)
4. Thermal integration with EMI for high speed, low voltage packages
5. Integration acoustic noise management into system thermal solutions
6. Precise temperature control for optical/photonic packaging
7. New interface materials with high thermal conductivity and match-
ing CTE
8. Advanced analysis tools for thermal modeling
9. Improvement on PCB thermal performance, i.e, increasing overall
thermal conductivities
10. Integration energy efficiency into system thermal design
One must also consider the liquid cooling technologies for the high power tele-
communication systems, especially for the system power over 25KW in order to
meet thermal and acoustic noise requirements as well as achieving system energy
efficiency. Specific thermal management methods in chip and system levels will
be discussed later.
10.3.1 Chip Level Thermal Management
The thermal issues in the chip packaging are due to the increase in chip power
and the local hot spots. The primary goals of thermal management are to mini-
mize the component junction temperatures and the temperature gradients of the
chip packages. The detailed thermal paths within a chip package are given in Fig-
ure 10.6, and the explored view of a simplified sketch of the chip packaging and
the heat sink are illustrated in Figure 10.7. The thermal resistance, qjc (junction
to case) which includes a combination of a series and parallel thermal resistance
from the junction (heat source) to the case should be as low as possible. Among
all thermal resistance, the TIM1 (thermal interface material) is the most critical
factor because it controls the heat transfer from the junctions to the case.
The possible methods of reducing the DT across the TIM1 can be described as
follows:
a. Thin film thermoelectric device (by Nextreme)

b. Indigo-1 Phase change metallic interface with K = 20–40 W/m °K (by
Enerdyne Solution)
c. Reactive bonding of multilayered foils composed of thousands of nano-
scale layers (by Reactive Nano Technologies, Inc.)
Figure 10.6 Thermal paths in chip package with heat sink.
Figure 10.7 Sketch of chip packaging and heat sink.

d. Any materials with high thermal conductivity, e.g., Indium with K =

86 W/m °K (by Indium Corporation).
Similarly, the following methods can be adapted to lower the DT of the TIM2
a. Indigo-1 phase change metallic interface (by Enerdyne Solution)

b. Thermal grease with high thermal conductivity
c. Thin metallic foils
Generally, heat is generated from metalized areas of the silicon dies inside a
package and is transferred to the component case (lid) and also to the printed
circuit board. Heat conduction is the primary heat transfer mode inside a com-
ponent. Cooling by conduction often requires the transfer of heat across various
materials and interfaces that may be laminated or bonded together. Therefore,
materials used in the packaging should have high thermal conductivity for heat
conduction. In addition, the coefficient of thermal expansion (CTE) of materials
is as important as the thermal conductivity. This is due to the fact that any mis-
matching of CTE often results in separation of the joints. Therefore, materials
should not only have high thermal conductivity but also have matching CTE.
Because of increasing in component power, several metal-matrix composite
materials such as Al/Gr (aluminum/graphite), Cu/Gr (copper/graphite) that have
thermal conductivity as high as aluminum or copper, but with the coefficient of
thermal expansion close to silicon or gallium arsenide have been developed. The
specific value of the CTE can also be obtained by adjusting the percentage of indi-
vidual elements in the composite. The thermal conductivity and CTE of various
materials is presented in Figure 10.8 [6].
Due to the increased power requirement, the vertical integration of the circuit
die, which is also referred to as “3D Packaging,” has received considerable atten-
tion in the recent years. Thermally, the vertically stack chip packages are even
more demanding than the traditional 2D chips where heat is generally conducted
through the multiple interfaces and layers of the different materials to the case
(lid) and to the board in the traditional 2D package. The major challenge in the
thermal design of the 3D chip package is how to remove heat from interior layers
of the dies. For these high power 3D chips, the traditional air cooling techniques
are not possible because of extremely poor thermal properties of the air. It will
require unrealistically high airflow rates for effective air cooling. The microchan-
nel water cooling on the integrated circuits was first proposed by Tuckerman and
Pease [7, 8]. The backside heat removal using the microchannel cooling technol-
ogy has been shown effective for the single layer die. The interlayer integrated
water cooling proposed by Alfieri et al. [9] as shown in Figure 10.9 is a very prom-
ising approach.
One of special type components which are unique to telecommunication systems
is the small form-factor pluggable (SFP) transceiver. The SFP as shown in Figure
10.10 is a component used for both telecommunications and data communications
applications. Generally, a SFP is inserted into a steel cage with perforation on the
top surface as illustrated in Figure 10.10. The cage is soldered to the printed circuit
board. The overall dimensions of a SFP are 8.5 mm (H) ´ 13.4 mm (W) ´ 56.5 mm
Figure 10.8 Thermal conductivity and CTE of various materials.
(L) and its heat dissipation typically ranges from 0.7 to 1.2 watts with the module
temperature limit at 70°C. Because the SFP must be freely removed or plugged into
its cage, a small air gap (about 0.25 mm) exists on all sides between the SFP mod-
ule and the cage which results in a very high thermal resistance. With typical inlet
or ambient air temperature of 50 to 55°C. It is sometimes difficult to maintain the
SFP module case temperature at 70°C, especially under natural convection envi-
ronment. Therefore, a heat sink is often needed (Figure 10.11).
Another product from SFP family is XFP which is larger than SFP in size. The
overall dimensions of a typical XFP are 8.5 mm (H) ´ 18.35 mm (width) ´ 78.0
F igure 10.9 3D IC package (IBM Zurich Lab).

F igure 10.10 SFP module/cage on board.
mm (L). The XFP power can be up to 3.5 W with the temperature limit of 70°C.
Actually, the power of some of the newly developed SFP can reach up to 3.5 W too
and makes thermal design ever more challenge and difficulty. Because of the air
gap (about 0.25 mm) between the module and cage along with the low allowable
temperature limit (70°C), SFP or XFP becomes one of most critical components,
especially for 3.5 W components on the board thermally. Lee and Lodhia [10]
performed thermal analysis and testing on 1.5 W XFP at air various velocities
ranging from 1.016, 1.778, and 2.54 m/sec. However, their work was mainly aim-
ing at the development of thermal models.
To overcome the high thermal resistance of the air gap, a large rectangular hole
is cut on the top surface of the cage so that the heat sink base can be in contact
with the SFP module as shown in Figure 10.12. The bottom piece with trapezoid
shape which is an integrated part of the heat sink base will make a direct contact
with the SFP module. Therefore, the thermal resistance between the heat sink
F igure 10.11 Thermal models for XFP.

Figure 10.12 Heat sink and SFP/XFP cage.
and the module can be reduced. The compounded thermal problems are due
to a series of SFP/XFP in a row on a given board as shown in Figure 10.13. The
number of modules in a row may be more than 20. Because of preheating of air
stream, the modules at far downstream are very difficult to meet the temperature
limit of 70°C. Large heat sink which serves multiple components by a single heat
sink as shown in Figure 10.12 is recommended in order to reduce the maximum
temperature of the modules at the downstream of the airflow path. For module
power greater than 2.5 W, especially at 3.5 W, it is sometimes impossible for these
modules at the downstream of the airflow path to meet the temperature limit of
70°C with the inlet air or ambient air over 50°C. Therefore, an industry grade SFP/
XFP whose temperature limit is at 85°C is required. In addition, thermoelectric
coolers can also be employed to resolve the thermal issues of these components.
However, thermoelectric cooler would require additional electric power.
10.3.2 System Level Thermal Management
Among the many system level challenges faced in telecommunication equipment

manufacturers, the power density is the most critical one thermally. Extrapolat-
ing from current systems, future nodes would dissipate hundreds of kilowatts for
power and the power trends in the projected growth are expected to be continu-
ously increased as previously presented in Figure 10.4. The power reduction strat-
egies and high-density thermal management become two of the critical elements
for the future telecommunication equipment. The thermal technology map, as
shown in Figure 10.5, can be used to select the proper cooling schemes.
Figure 10.13 Example of series of SFP modules on board.
Air cooling has been and is still the key element for the thermal management
of the telecommunication equipment. To achieve the efficient thermal manage-
ment schemes in air cooling, the following factors should be considered
1. Optimization of heat sinks on components

2. Thermal interface materials (TIMs) with high thermal conductivity
3. Good airflow management with lower system pressure drop and better
flow distribution over the boards
The purpose of using a heat sink is to enhance the heat transfer from the
component to the air stream. The typical extrusion heat sinks may not be suit-
able for the component power greater than 35 W. The main reason is the lim-
ited available heat transfer surface. This is due to the fact that the extrusion fin
heat sinks are bulky and heavy with thick fins. The other types of fin structures
such as the plain fins may be considered. Another issue to deal with is the base
spreading resistance of the oversize heat sinks. To reduce the thermal spreading
resistance, the common methods are to increase the lateral heat conduction at
the heat sink base. The approaches include, but are not limited to, the embedded
graphite spreader (or any high thermal conductivity materials), embedded heat
pipes, vapor chamber, or liquid chamber base.
The flow management is another important factor for achieving the effective
thermal management scheme. The objective of the flow management is to obtain
proper flow distribution over the boards and also to minimize the system flow
resistance or pressure drop in order to increase the system flow. The baffles are
also often employed to improve the flow distribution over the board or to guide
flow to where it is needed. In addition, system design must have a good balance
between the desired system thermal performance and the pressure drop. In order
to reduce the acoustic noise and also to achieve system energy efficiency, i.e., high
COP, the following guideline for the system pressure drop versus flow rates as
presented in Figure 10.14 is recommended for air cooled systems.
Liquid has often been used for cooling of the high power systems. The greatest
advantage of liquid cooling over the air cooling is its cooling capability. Though
the liquid cooling has long been employed in the high-performance computers,
the current practice of using liquid (i.e., water) at the equipment in the data cen-
ters is aiming at the reduction of the air temperature to the data centers. IBM
System 360 Model 91 in 1967 was the first IBM computer to use water-cooling
in any form. The water-to-air heat exchangers as shown in Figure 10.15 [11] are
installed at the door and are inserted between the card cages when the door is
closed. The purpose of the water-to-air heat exchangers is to reduce the inlet air
temperature to the upper card cages (boards) in downstream of the airflow. As
can be seen from the figure, the DT air is reduced by more than 20°C with the heat
exchangers in the rack.
On the other hand, the IBM RDHX (rear door heat exchanger) [11] which is a
water-to-air heat exchanger is attached to the exits of the server rack as illustrated
in Figure 10.16. The overall dimensions are 25 in. (W) ´ 5.6 in. (D) ´ 76.4 in. (H).
The main purpose of the RDHX is to lower the temperature of the exit air from
the servers to the date center so that the energy consumption of the air-conditioning
Figure 10.14 Recommended system maximum pressure drop versus

flow rates.
Figure 10.15 IBM water–air heat exchangers in cabinets.
F igure 10.16 IBM rear door heat exchanger.

units in the data center can be reduced. It should be noted that the RDHX does
not affect the thermal performance of the equipment itself.
The hybrid cooling of the supercomputers [11, 12] has recently received con
siderable attention. In other words, the high-power components on the board
are cooled by the liquid, while the rest components are still cooled by the air. The
issue of possible condensation inside the rack must carefully be considered. The
possible fan failure and also the acoustic noise issue may still exist in the hybrid
cooling scheme. In addition, the liquid loop (tubes) may become an obstacle of
the certain airflow to the components. Recently, IBM Zurich Research Lab in
conjunction with ETH developed a warm water (60°C) cooled supercomputer
(Aquasar) [13, 14] using microchannel cold plate for cooling the components.
The active liquid cooling (total liquid cooling) is the best approach to achieve the
energy efficiency both at equipment level and data center and also to support the
much higher system power while eliminating the acoustic noise from air cool-
ing. The actively liquid cooled rack has following advantages over the current air
cooled rack.
• Liquid cooling can support much higher system power

• Liquid cooled rack makes equipment much more reliable
– By eliminating fans, equipment reliability is significantly increased
– Effective liquid cooling lowers component temperatures
• Liquid cooled rack is much compact and small
– By eliminating fan tray, air filter, air inlet and exit spaces
– Slot size (spacing between boards) is small
• Equipment is almost maintenance free (no need to replace air filter
and fans)
– Operation cost is also reduced by eliminating fans and air filters
• Liquid cooled rack can totally eliminate acoustic noise from fans and
high speed airflow
• Liquid cooled rack is independent of ambient conditions (e.g., room air
temperature) as long as chilled water is available
• Eliminating hot air from equipment will reduce data center tempera-
ture and decrease energy consumption in data centers
10.4 THERMAL DESIGN OF LIQUID COOLED TELECOM

EQUIPMENT
The liquid cooling has not been adapted commercially in any telecommunication
equipment. However, due to the increased system power and also the concern about
the energy efficiency, the liquid cooling has received considerable attention in tele-
communications industry recently. Alcatel-Lucent [15] has developed a sealed cabi-
net using dielectric mist for cooling as illustrated in Figure 10.17.
The large droplets of dielectric mist from the atomizer are injected into the
upstream of airflow entering the printed circuit boards. The liquid mist is heated
by the electronics on the board and is evaporated along the air stream. In other
Figure 10.17 Alcatel-lucent enhanced cooling in a sealed cabinet.
words, heat generated from electronic components is to vaporize the dielectric

mist instead of heating up the air stream. The mixture of air and the dielectric
vapor flows over an array of heat pipes and the dielectric vapor is then condensed
and the liquid is fallen into the liquid collector and finally is pumped into the
atomizer for reuse.
Cisco also employed the hybrid cooling for the high power systems which
are the modified 7609 Series [16]. The back side of chassis, including the heat
exchangers, tubing and the pumps is shown Figure 10.18. The system under con-
sideration is a standard 19 in. rack which consists of 9 boards with the board
pitch of 1.8 in. The maximum system power per shelf is about 10 KW with the
maximum nominal power per board of 1000 W.
Both systems from Cisco and Alcatel-Lucent are still in the very preliminary
phase in the laboratory. Some of other telecommunication equipment manufac-
turers have also started working on the liquid cooled systems.
The chilled water is readily available in most of the data centers. The sketch of
the active liquid cooling can be depicted as illustrated in Figures 10.19 and 10.20
[5]. Figure 10.19 shows a general liquid cooling loop in the equipment. The heat
exchanger as shown in the figure can be the cooling towers or condensers of the
data center. The cooling concept of the active liquid cooled rack is given in Figure
10.20. The liquid enters the inlet manifold which provides proper flow rate to the
individual boards and leaves the equipment through exit manifold. The chilled
water in Figure 10.20 is directly taken from the data center. It should be noted
that the pump may not be needed in this case if water from the date center has
F igure 10.18 Cisco liquid cooled rack : back of chassis exposed.
Figure 10.19 General liquid cooling loop in equipment.

Figure 10.20 Active liquid cooled rack with facility water.
sufficiently high pressure. The flow rate to individual boards must be regulated in
proportion to the power of the boards.
Figure 10.20 shows the liquid cooled equipment with the facility chilled
water. On the other hand, Figure 10.21 [5] illustrates the liquid cooled equip-
ment with the self supplied liquids. In this case, an additional heat exchanger is
required in order to cool the hot liquid from the equipment to be reused again.
This heat exchanger can be a liquid-to-liquid heat exchanger or an air cooled
condenser.
Figure 10.21 can also be applied to the outdoor systems wherein dielectric liq-
uids must be used in the outdoor applications. This is due to the fact that the out-
door equipment is typically required to operate at the temperature below −40°C
or even −50°C. In general, a dielectric liquid which has a high dielectric strength
to prevent electrical breakdown is a preferred choice in cooling of electronics.
Unfortunately, the heat transfer capacity of the dielectric fluids is much lower
than that of the water. The dielectric liquid is not normally required for the cases
with indirect liquid cooling. However, it is a must for the immersion cooling
where liquid is in direct contact with electronics.
The board assembly, including board and liquid network must be able freely
to be removed or plugged into the equipment. To do so, each board requires at
least two quick disconnects for the liquid loop. A good quick disconnect which is
extremely expansive is needed to avoid any liquid leakage after the board assembly
is unplugged. Figure 10.22 gives a sketch to illustrate the connection of the board
assembly to the system. It should be noted while the board is plugged into the back
plane for the electrical connection and, at the same time, the liquid loop must be
Figure 10.21 Active liquid cooled rack with self supplied coolant.
Figure 10.22 Sketch of board assembly connection to system.

connected to the system too. The fluid connections between liquid network and the
liquid inlet and exit manifolds in the system are made through the quick disconnect
fittings. The figure, which shows one cold plate for all components on the board,
is just for illustration only. Due to various heights of the components on the board,
multiple cold plates may be required.
Water is the most common fluid used in the high power computers because of
its high heat transfer capability. However, water cannot be used in a direct immer-
sion cooling nor for outdoor applications. The former is due to its poor dielectric
properties and the later is because of its high freezing temperature. When water
is used, it also must be properly treated for anti-corrosive and anti-fungal. For
material compatibility, copper or stainless steel must be used for the entire liq-
uid loop if water is adopted as the coolant. On the other hand, aluminum can
be employed for the dielectric fluids. However, stainless steel is recommended if
de-ionized water is the coolant. All above three types of materials (copper, alumi-
num, and stainless steel) are compatible with Glycois.
10.5 ENERGY EFFICIENCY
As stated previously, the equipment power consumption is continuously increased

at a fast pace. In addition, the cost of powering and cooling in the equipment at
the data centers has also increased significantly in the last few years. Therefore,
the energy efficiency at the equipment level and the data center become a top
priority in the industry.
Figure 10.23 [17] illustrates the power consumption in a typical data center.
Among them, the power consumption for the IT and cooling equipment are
45.4% and 30%, respectively. Another example for the 1 U servers, the ratio of
the 3-year site cost to the 1 U server cost is given in Figure 10.24 [18]. As can be
seen from the figure, the 3-year cost of powering and cooling servers is about
1.5 times of the cost of the servers in 2006, and this factor is projected to be over
22 times at the worst case by 2012.
In addition, the power consumption of equipment has a great impact on the
power required in a data center which is directly related to the operation cost.
Based on the estimate by Verizon [19], 1 W saved on the equipment level will
save a total of 2.41 W in their respective the data center as shown in Figure 10.25.
Therefore, it is extremely important for the equipment makers to do a better job
on energy efficiency of their products. It should be noted that this cascade factor
varies slightly with the individual data centers.
The following steps [5] should be considered to achieve the energy efficiency
at the equipment level:
1. Using low power consumption components, especially for power

supply.
2. Utilizing industry grade components whenever possible. The industry
grade parts are more expensive but are more reliable. In addition, the
industry grade component can withstand higher temperature that in
turn requires less airflow.
Figure 10.23 Typical data center energy breakdown.
Figure 10.24 Ratio of 3-year site cost to 1 U server costs.
3. Employing efficient thermal control schemes. The fans can operate at

low speeds because the less flow is required with effective heat transfer
methods.
4. Designing equipment with power management features. With this fea-
ture, some of boards can shift to standby mode in the conditions of low
traffics or demands.
Figure 10.25 Cascade effect of energy savings from equipment to

data center.
5. Adapting passive cooling methods whenever possible. The passive cool-

ing is simple, reliable, cost effective and energy efficient but is limited
to the low power systems due to poor thermal properties of air.
6. Adapting liquid cooling, especially the actively liquid cooled
equipment.
In addition, one should minimize the total power per network throughput by
transferring bits in the optical domain from the network architectural point of
view. This so-called all-optical network is a less power hungry network.
10.6 ALTERNATIVE ENERGY
With significant increase in the energy consumption and cost in the past few
years, it becomes necessary to explore various types of alternative energies to
replace or to supplement the energy consumptions in the equipment as well as
in the data centers. Among them, the thermoelectric power generation and the
fuel cell technology power generation have been received considerable attention
recently.
10.6.1 Thermoelectric Power Generations
Two major classes of application of thermoelectric devices are heat pumps and
power generators. In the former, electrical energy is supplied to transport thermal
energy from one location to other locations, while in the latter, thermal energy is
converted into electrical energy. The power generation from the waste heat of the
electronic systems is the subject area to be discussed here.
The basic theory of the thermoelectric device is due to the Seebeck effect which
states an electromotive force (emf) or voltage will be generated if two junctions
of two different materials are held at different temperatures. This emf is propor-
tional to the temperature difference between two junctions as follows
dV = α12dT (10.1)
and
α12 = α1 − α 2 (10.2)
where ai is Seebeck coefficient for material i with unit of V/°C.

For convenience, the figure of merit, Z which is a measurement of the useful-
ness of a material in a thermoelectric device is defined as follows
Z = α 2/(σ K) (10.3)
where σ is the electrical resistivity (ohm-cm), and K is the thermal conductivity

(W/m °C).
The sketch of a thermoelectric power generation is given in Figure 10.26 [20].
Generally, the higher temperature difference between the hot and cold sides, the
higher thermal efficiency will be. The thermal efficiency is limited by the material
development. Two critical factors that dictate power output in a thermoelectric
power generator are as follows:
1. The amount of heat flux that can be successfully transferred through

the module, and
2. the temperature difference between the hot and cold side.
Due to relatively small air temperature difference, typically less than 20°C,
between the inlet and exit in telecommunication equipment, the thermal efficiency
F igure 10.26 Basic sketch of thermoelectric power generator.

of the thermoelectric power generator is extremely small. The thermal efficiency

with DT of 20°C is estimated to be about 2% in 2010 and approaching 5% by
2013.
Figure 10.27 [20] shows one of possible schemes that the waste heat from
equipment at data center can be converted to the electricity by the thermoelectric
power generators. Even with very small thermal efficiency, it is still worthy to
try to utilize the waste heat from telecom or computing systems for the energy
generation in the data center because an extremely high amount of heat is gener-
ated. The following example is to demonstrate the possibility of using waste heat
recovery to generate electric power in order to supplement the power consump-
tion in the data center.
Example 10.1
Assume 5% thermal efficiency with DT of 20°C for a 20 KW telecommunication
equipment. As stated previously in Figure 10.25, 1 W saved in telecom equipment
will save 2.41 W in the data center.
Solution
Savings on equipment level for 20 KW system:
Annual energy saving = 20 KW ´ 0.05 ´ (24 ´ 365) = 8760 KW-HR
Annual cost saving = 8760 KW-HR ´ $0.11 = $963.6
Savings on data center for 20 KW system:
Annual energy saving = 2.41 ´ 8760 KW-HR = 21111.6 KW-HR
Annual cost saving = 21111.6 KW-HR ´ $0.11 = $2322.3
The above calculations are based on the average residential rate of $0.11/KW-HR.
Figure 10.27 Possible locations of thermoelectric power generators

in data center.
10.6.2 Fuel Cell Technology for Power Generation
The fuel cell is an electrochemical conversion device which was first discovered by
Christian Friedrich Sconbein in 1838. It produces electricity from fuel (on the anode
side) and an oxidant (on the cathode side) which react in the presence of an electro-
lyte. The reactants flow into the cell and the reaction products flow out of it.
The basic principal of the fuel cell technology is as follows:
H2 + 0.5 O2 ® Electricity + Heat + Water
Due to electrochemical actions, the hydrogen fuel and oxygen produce electric-
ity and heat with the byproduct of water. The hydrogen is the primary fuel source
for the fuel cell. The process of fuel reforming allows for the extraction of hydro-
gen from many sources such as natural gas and propane or any other hydrogen-
containing fuels. A typical fuel cell can produce a voltage up to 1V at the full rated
load. Therefore, in order to deliver the desired amount of energy, the fuel cells
can be combined in series or in parallel circuits, where series yield higher voltage
and parallel allows a stronger current to be drawn. Such a design by combining
individual fuel cells is called a fuel cell stack. To produce electrical power for a
working application requires more than just one fuel cell stack. A fuel cell system
may include fuel cell processing, thermal management, water management, power
conditioning, electric grid connection, and energy storage modules.
F igure 10.28 Direct application of fuel cell energy to OSP operation.

Fuel cell systems offer the potential for clean, reliable energy generation and
have been employed to supplement the energy consumption in the data centers.
Fuel cells have recently been adapted as the back energy to or even to replace the
batteries in the outside plants (OSPs). Plug Power’s [21] developed GeneCore 5 KW
products to be used as an emergency backup power over 72 hours or more for
the cell tower customers. Furthermore, Clear Edge has developed several fuel cell
systems to power home and/or business buildings by utilized available natural gas.
With OSP moving closer to the residential area where natural gas is readily avail-
able, direct application of fuel cell systems to power OSP as shown in Figure 10.28
becomes feasible.
10.7 FUTURE RESEARCH NEEDS
The continued development of new and improved thermal management tech-

nologies will require the combined efforts of industry based development and
university based research with a focus at practical applications. Extensive heat
transfer, thermo-fluid, and thermo-mechanical research are needed to define new
opportunities and to improve predictability and reliability.
The specific research needs are recommended as follows:
1. Thermal spreaders
– Inexpensive, light weight, high thermal conductivity, anisotropic mate-
rials (possibly composites) offering a closer thermal expansion match
to the CTE of Silicon (Si) or Gallium Arsenide (GaAs) or other die
materials.
– Techniques for achieving improved thermal spreading within a chip to
alleviate hot spots due to localized high heat fluxes.
– High capacity small heat pipes, including micro heat pipes for embed-
ding to heat sink base or high performance vapor chambers in order to
reducing base thermal spreading resistance.
2. Thermal Interface Materials (TIMs)
– Thermal pads, greases, pastes, epoxies, and elastomers load with high
thermal conductivity nanoparticles.
a. Overall thermal conductivity of TIMs should be greater than
15 W/m °K.
b. Compressibility of thermal pad must be greater than 50%.
– Innovative new interface materials, including natural graphite base
materials.
– Carbon nanotube (CNT) base technology for TIM applications and
bumps to substrates as shown in Figure 10.29 for Fujitsu 100 W class
amplifiers in the mobile base stations.
– Novel techniques/materials to minimize interfacial stresses and thermal
resistance.
– Correlations and analytical relations to predict fatigue life of bonded
interfaces.
– Void free processes for thermal pastes and epoxies bonding.
F igure 10.29 Carbon nanotubes bump on electrodes and substrate.
3. Air cooling
– High performance, low power consumption, and low acoustic noise fans
or blowers.
– High efficiency analysis tools
a. Models and correlations to accurately predict heat transfer in tran-
sition and perturbed low Reynolds number flow over packages and
in heat sink passages.
– Heat sink design and optimization for minimizing heat sink thermal
resistance subject to mass and volume constraints.
– Advanced manufacturing techniques for light weight metal or composite
material heat sinks.
– Novel miniature synthetic jet fans for enhancing hot spot cooling.
– Novel 3D microstructure air filters to enhance product life.
a. Current air filters collecting dusts and dirt on the front surface
which leads to much shorter life.
– Special subjects
a. Large heat sink for group of components
As illustrated in Figure 10.13, a large heat sink is utilized to serve a
group of components. This design concept has recently received consid-
erable interests, especially on the high power boards. An oversize heat
sink which is generally required for any high power component requires
4 mounting screws attaching to the board. The heat sink mounting is
becoming a major issue due to the following reasons:
i. Due to increased component power density, the number of com-
ponents required heat sink is significantly increased. On the
other hand, the available space for heat sink mounting is limited
because of a large number of components on board.
ii. Holes on board required for heat sink mounting will make board
wire routing extremely complicated and difficult.
Therefore, the concept of using a large single heat sink to serve a

group of components becomes necessary for high power boards in order
to reduce the number of screws needed for mounting heat sinks.
b. Synthetic jets
Using synthetic jets for cooling of electronics has received a great deal
of interest in recent years. A synthetic jet flow is a type of flow which is
made up of the surrounding fluid. Synthetic jet flow can be developed
in a number of ways such as an electromagnetic driver, a piezoelectric
driver, or even a mechanical driver, e.g., piston. Impinging jets are a well
established technique for achieving high local convective heat trans-
fer rates. Therefore, synthetic jets are employed to enhance local heat
transfer coefficient. It should be noted that the flow rate of a synthetic
jet which is generated from the surrounding fluid is relatively small. The
heat transfer enhancement is due to high velocity jet action.
The discussion here will be focused at the synthetic jets generated by an actua-
tor and a piezoelectric fan.
i. Piezoelectric fan synthetic jets

Piezoelectric fans are vibrating beams, disks or plates whose vibratory
motion is actuated using a piezoelectric material. The basic theory is that
an ultra low power alternating currents input to the piezoelectric element
forces oscillations throughout the beam which can be used to induce flow in
surrounding fluid and provide heat transfer enhancement. As these devices
(piezoelectric fans) typically run at resonance, significant vibration ampli-
tudes are achieved with small amount of input power. In addition, these
devices produce little or no audible noise. The pictures of a typical piezo-
electric fan and its application to a heat sink are shown in Figure 10.30.
As described by Kimber and Garimella [22, 23], thermal performance
of a piezoelectric fan may be affected by a total of five variable parameters
such as frequency, amplitude, width, length and gap of piezoelectric fan.
Key physical insights and results from their work include the following:
– The maximum performance for a given fan can be described in terms
of the vibration frequency and oscillation amplitude only.
– The frequency is found to be somewhat more influential than the
amplitude in determining the largest attainable heat transfer rates.
– The width, length, and frequency, each exhibit the same influence on
the value of optimum gap and the exponential decay rate from the
peak performance.
The comparison of the heat transfer coefficients of piezoelectric fan and those
other types of heat transfer is presented in Figure 10.31 [24].
Experimental investigation of the thermal performance of piezoelectric fans is
presented by Acikalin et al. [25]. Prototypes of the fans were built and tested to
assess their feasibility and cooling performance and determine optimal locations
for the fans. An enclosure the size of a cellular phone and commercially available
laptop computer were used to demonstrate the cooling feasibility of the fans.
Piezoelectric fans were found to offer significantly localized cooling, exceeding
Figure 10.30 Piezoelectric fan and its application to heat sink.
Figure 10.31 Thermal performance comparison with piezoelectric fan.

enhancements in convective heat transfer coefficient of 100% while exhibit-

ing low power consumption, minimal noise and small dimensions. Additional
work by the authors is given by Acikalin and Garimella [26]. The heat transfer
enhancement by piezoelectric fans is limited to the localized cooling. Therefore,
the piezoelectric fans have never intended to replace the conventional fans. The
best applications of such fans are in the low power systems, especially in the low
profile boxes under the natural convection environments. Acikahn, Sauciuc and
Garimella [27] studied the piezoelectric fan for low-form-factor electronics cool-
ing. The schematic of two experimental orientations is shown in Figure 10.32. The
cases with the best thermal performance in each orientation are given in Figure
10.33. As can be seen from the figure, the thermal performance of the horizontal
Figure 10.32 Schematic of two experimental orientations.
Figure 10.33 Thermal performance of best cases of horizontal and

vertical orientations.
Figure 10.34 Thermal performance comparison between piezoelec-

tric and small axial fans.
orientation is slightly better than that of the vertical orientation. For a piezoac-
tuator power input of 31 mW, the calculated thermal resistances are 6.5°C/W and
7.8°C/W for the horizontal and vertical orientations, respectively. The comparison
of the thermal performance of the best piezoelectric fan and two commercially
available small axial fans is presented in Figure 10.34. As indicated in the figure,
the piezoelectric fans not only have lower thermal resistance but also consume
less power as compared to small axial fans. In summary, the piezoelectric fans
have been employed as means to enhance local convective heat transfer while
requiring only small amounts of power with lower acoustic noise.
ii. Electromagnetic actuator synthetic jets

Another way to create the synthetic jets is by a synthetic jet actuator. The
synthetic jet actuator as shown in Figure 10.35 [28] consists of a cylindrical
cavity with a sharp-edged orifice and a flexible diaphragm which is located
on the opposite of the orifice. The bottom of the orifice extends outwards
to provide a confined outflow condition. A synthetic jet comprises a train
of vortices formed by successive ejection and suction of fluid across an ori-
fice. The synthetic jet which is generated by pulse action of the diaphragm
is impinging at the heated surface with the distance of H. When the jet
is impinging onto a heated surface, the highly pulsatile nature promotes
entrainment, mixing, and breaking up the thermal boundary layers that
lead to a significant increase in the local heat transfer coefficient.
Figure 10.36 illustrates the flow pattern for a synthetic jet actuator
operating in open ambient environment. The basic operation principle for
Figure 10.35 Schematic diagram of impinging synthetic jet.
Figure 10.36 Synthetic jet actuator operation in open ambient

environment.
a synthetic jet actuator is that a jet is generated due to the pulse action
of the diaphragm. The surrounding air is sucked into the device from the
orifice when the jet leaves the orifice with a train of vortices. The process is
repeated again and again. The device itself creates no net mass flux.
The thermal performance of various cooling schemes, including a pair
of synthetic jets and piezoelectric fan is given in Figure 10.36 [24]. As can
be seen from the figure, the thermal performance of the synthetic jet is
slightly better than that of the piezoelectric fan (Figure 10.37).
When a single impinging synthetic jet cools a heated surface as shown
in Figure 10.35, a forced cross-flow is required to supply fresh cooling
medium. However, Smith and Glezer [29] investigated the flow field of
pair of adjacent synthetic jets. Persoons et al. [30] applies this vectoring
effect in a pair of synthetic jets, thereby, combing (i) high local heat trans-
fer rates and (ii) the cross-flow required for effective cooling. The purpose
of their paper is to determine the optimal conditions for convective cool-
ing as a function of jet-to-surface spacing and phase difference between
the jets. Figure 10.38 shows the schematic of the test rig by Persoons et
al. [30].
Figure 10.37 Thermal performance comparisons with synthetic jet.

Figure 10.38 Schematic of test rig for a pair of synthetic jets.
In summary, both piezoelectric fan and synthetic jet actuator operate on

the principle that jet generated periodically breaks up the thermal bound-
ary layer on a heated surface and also induces significant entrainment on
the surrounding fluid. However, the main difference is that piezoelectric
fans are simpler, relying only on the vibrating motion of a cantilever beam
to propel fluids towards a heated surface. One the other hand, a synthetic
jet actuator generally requires three basic components: (1) an orifice to
form vortices that make up the synthetic jet flow, (2) a cavity connected to
the orifice, and (3) an oscillating diaphragm that drives the fluid motion. It
should also be noted that both devices are commercially available.
4. Liquid cooling
– Compact, cost-effective and reliable liquid cooling solutions
– High performance and reliable pumps and condensers if needed
– Development or improvement on thermal properties of new or existing
dielectric fluids
a. New technologies, including surface treatment to reduce tempera-
ture overshoot in pool boiling (direct immersion cooling) of dielectric
fluids
– Enhancement on microchannel cooling technologies to create low cost
and reliable cold plates
a. Mature microchannel cooling technology a must for 3D packaging
b. Integrated microchannels cooled silicon chips
5. Other advanced cooling technologies
– High performance, high capability micro heat pipes
a. Embedded micro heat pipes for enhancing heat sink base plate
spreading and also reducing base thickness
– High performance, high capacity long heat pipes or large vapor
chambers
– High COP thermoelectric devices
a. New thermoelectric materials and fabrication methods for perfor-
mance improvement
b. Thin film thermoelectric for TIM1 applications in device packages
– Technologies for refrigeration cooling
a. Highly reliable miniaturized parts such as compressors, condens-
ers and evaporators
b. Low cost, low noise refrigerators using solid state, vapor compres-
sion, or absorption cycles
10.8 SUMMARY AND CONCLUSION
With increasing power on both component and system levels along with the
demands of the reliability and performance, the thermal management becomes
even more critical. The high airflow rates needed by the increased system power
require the large high speed fans which not only consume more power but also
increase acoustic noise. Power reduction strategies and high power thermal man-
agement become two of the critical elements for design of the future telecommu-
nication equipment.
Air cooling has long been used for cooling of electronics, especially for the con-
sumer electronics used in home and offices. It is simple, economical, and safe. Air
cooling is also most commonly used for telecommunications equipment. While
larger high speed fans may generate enough airflow to adequately cool the high
power equipment, however, the air cooling is not recommended for any system
over 20 KW because of the concerns about the fan power consumption which is
directly related to the system COP and operation cost, and the excessive acoustic
noise which often exceeds the required limits.
The operation cost of the data centers is skyrocketed in the recent years
and it is also expected to be even higher in the future. The energy efficiency
becomes the top priority for the data center operators and the equipment man-
ufacturers. The methods to achieve the energy efficiency on the equipment
have been discussed in the previous sections. Among them, the active liquid
cooling is most attractive. The active liquid cooling not only can support even
higher system power but also can eliminate the acoustic noise. And above all,
the active liquid cooled rack is not only reliable but also energy efficient by
eliminating all fans and the air filter which are required in the case of air cool-
ing. It is believed that the active liquid cooling will be the choice of the future
thermal management for the core telecommunication equipment. It is further
expected the liquid cooled telecom equipment will be commercially available
in the next few years.
While industry trends toward the high power systems, one should never ignore
the value of the small box configuration equipment which can often be found in
office, home and even in the corners of the hallways of the residential buildings.
From reliability and cost point views, passive thermal control should be adapted
whenever possible. Efforts should also be considered to convert the fan-cooled
small box configuration equipment to passive cooled system to achieve equipment
energy efficiency for reduction in the operation cost and also to increase equip-
ment reliability. The piezoelectric fans and synthetic jets discussed previously will
have great benefits to low power and low profile boxes due to the enhancement on
local heat transfer over hot spots.
Due to high-energy cost, alternative energies should be considered as the sup-
plemental energy to reduce the operation cost of the equipment and the data cen-
ters. The thermoelectric power generation and the fuel cell technology appear to
be feasible. Both of them provide clean energy with little or no pollution. Further
researches to improve the thermal efficiency and the cost reduction will still be
needed.
As discussed in previous sections, 1 W saved at the equipment will lead to a
saving of more than 2.41 W in the data center. This clearly demonstrates that the
energy efficiency at the system level has a great impact on the energy consumption
and operation cost in the data center. With the cost of energy continuous increase,
the energy efficiency becomes a must for the equipment manufacturers and the
data center operators. Therefore, the energy efficiency should always be integrated
into the system thermal designs both at the equipment level and the data center.
The first few papers in dealing with the thermal management and energy effi-
ciency challenges in telecommunication systems and data centers was recently
published by Garimella et al. [31, 32]. Though the paper is mainly focusing at the
high-level discussions on the importance of the energy efficiency of telecommu-
nication equipment and data centers; however, it does include a vast of data and
information useful to the thermal engineers in these areas.
REFERENCES
1. Vukovic, A., “Power Density Challenges of Next Generation Telecommunica-

tion Networks”, ElectronicsCooling, Vol. 9, No. 1, February, 2003.
1A. Gunther, Binns, Carmean, and Hall, “Managing the Impact of Increasing
Microprocessor Power Consumption”, Intel Technology Journal, Vol. 5,
No. 1, February 2001.
2. “Datacom Equipment Power Trends and Cooling Applications”, American
Society of Heating, Refrigerating and Air-Conditioning Engineers, Inc, 2005.
3. Kenneth G. Brill, “2000–2010 Product Heat Density Trends Chart”, White
Paper, The Uptime Institute, 2006.
4. Telcordia Technologies Generic Requirements, GR-63-CORE, Issue 3, March,
2006.
5. Yeh, L. T.,“Thermal Management and Energy Efficiency of Telecommunication
Equipment”, Presented at 2nd Workshop on Thermal Management in Telecom-
munication Systems and Data Centers, Santa Clara, CA, April 25–26, 2012.
6. Zweben, C., “High Performance Thermal Management Materials”, Advanced
Packaging, February, 2006.
7. Tuckerman, D. B. and Pease, F. F., “High Performance Heat Sinking for VLSI”,
IEEE Electron Devices Letter, EDL-2, 1981.
8. Tuckerman, D. B., “Heat Transfer Microstructures for Integrated Circuits”,

Doctoral Thesis, Stanford University, 1984.
9. Alfieri, F., Tiwari, M. K., Zinovik, I., Poulikakos, D., Brunschwiler, T., and
Michel, B., “3D Integrated Water Cooling of A Composite Multilayer Stack of
Chips”, Proceedings of 14th International Heat Transfer Conference, Wash-
ington, DC, USA, August 8–13, 2010.
10. Lee, T. and Lodhia, A., “Component and Board-Level Thermal Modeling
Techniques for XFP Optical Transceivers”, IMAPS Advanced Technology
Workshop on Thermal Management for High Performance Computing and
Telcom/Wireless Application, October 24–26, 2002.
11. Ellsworth, M. J., Jr., Campbell, L. A., Simons, R. E., Iyengar, M. K., Schmidt,
R. R., Chu, R. C., The Evolution of Water Cooling for IBM Large Server Sys-
tems: Back to Future, Itherm, 2008.
12. Jei Wei, “Hybrid Cooling for Fujitsu Large Computer Systems”, 2010 IEEE
SPJW, Kyoto, Japan, 2010.
13. Escher, W., Wrunschwiler, T., Michel, B., and Poulikakos, D., “Experimental
Investigation of an Ultra-thin Manifold Microchannel Heat Sink for Liquid-
Cooled Chips”, Intl. J. of Heat and Mass Transf., 2010 (current under review).
14. Escher, W., Michel, B., and Poulikakos, D., “Efficiency of Optimized Bifurcat-
ing Tree-like and Parellel Michrochannel Networks in the Cooling of Elec-
tronics”, Intl. J. of Heat and Mass Transf. 52, 2009.
15. Bahadur, V., Hodes, M., Lyons, A., Krishnan, S., and Garimella, S. V.,
“Enhanced Cooling in a Sealed Cabinet using an Evaporating-Condensing
Dielectric Mist”, Proceedings of ITHERM, Orlando, FL, 2008.
16. Glover, G., “The Next Generation Router System Cooling”, MS Thesis, Cali-
fornia Polytechnic State University, 2009.
17. Schmidt, R. R., “Recent Advances and Future Challenges in Data Center”,
7th International Business & Technology Summit, Natick, MA, August 22–23,
2007.
18. Brill, K. G., “Data Center Energy Efficiency and Productivity”, White Paper,
The Uptime Institute, 2007.
19. Graff, C., “Verizon’s Thermal Management Program for Network Equipment
Design”, 2009.
20. Fisher, T., “Waste Heat Recovery from Electronics”, CTRC Project review
meeting, Purdue University, 2010.
21. Parsons, M. J. and Josefik, N. M., “Accelerating Production Readiness using
Lean Product Development”, Proceedings of ASME 2009 7th International
Fuel Cell Science, Engineering and Technology Conference, Newport Beach,
CA, June 8–10, 2009.
22. Kimber, M. L. and Garimella, S. V., “Measurement and Prediction of the Cool-
ing Characteristics of a Generalized Vibrating Piezoelectric Fan”, Interna-
tional Journal of Heat and Mass Transfer, Vol. 52, 4470–4478, 2009.
23. Kimber, M. L. and Garimella, S. V., “Cooling Performance of Arrays of Vibrat-
ing Cantilevers”, Journal of Heat Transfer, Vol. 131, 111401, 2009.
24. Persoons, T., “Advanced Air Cooling Techniques for Thermal Management
of Electronics, with a Focus on Synthetic Jets”, 2nd Workshop on thermal
Management in Telecommunication Systems and Data Center, Santa Clara,
CA, April 25–26, 2012.
25. Acikalin, T., Wait, S. M., Garimella, S. V., and Raman, A., “Experimental
Investigation of the Thermal Performance of Piezoelectrically Actuated Fans”,
Heat Transfer Engineering, Vol. 25, No. 1, 2004.
26. Acikalin, T. and Garimella, S. V., “Analysis and Prediction of the Thermal
Performance of Piezoelectrically Actuated Fans”, Heat Transfer Engineering,
Vol. 30, No. 6, 2009.
27. Acikalin, T., Sauciuc, I., and Garimella, S. V., “Piezoelectric Actuators for Low-
Form-Factor Electronics Cooling”, IPACK2005-73288, 2005.
28. Persoons, T., McGuinn, A., and Murray, D. B., “ A General Correlation for the
Stagnation Point Nusselt Number of an Axisymmetric Impinging Synthetic
Jet”, International Journal of Heat and Mass Transfer, Vol. 54, 2011.
29. Smith, B. L., and Glezer, A., “Vectoring of Adjacent Synthetic Jets”, AIAA Vol. 43
No. 10, 2005.
30. Persoons, T., O’Donvan, T. S., and Murray, D. B., “Heat Transfer in Adjacent
Interacting Impinging Synthetic Jets”, Paper No. HT2009-88440, ASME Sum-
mer Heat Transfer Conference, San Francisco, CA, July 19–23, 2009.
31. Garimella, S. V., Yeh, L. T., and Persoons, T., “Thermal Management Chal-
lenges in Telecommunication Systems and Data Centers”, IEEE Transac-
tions on Components, Packaging and Manufacturing Technology, Vol. 2, No. 8,
2012.
32. Garimella, S. V., Persoons, T., Weibe, J., and Yeh, L. T., “Technological Drivers in
Data Centers and Telecom Systems: Multiscale Thermal, Electrical, and Energy
Management”, Applied Energy, Vol. 107, July 2013.
1
2
3
4
5
6 Index
7
8
9
10
11
Page numbers followed by f and t indicate figures and tables, respectively.
12
13
A thermoelectric power generations,
14
Acoustic noise, 26–27, 57. See also 198–200, 199f, 200f
15
Fan Axial fans
16
equipment, 180–181 propeller, 21
17
Active cooled boxes, 103–105, 104f, tube, 21
18
105f. See also Small box (system vane, 21
19
thermal analysis)
20
Active heat spreader (AHS), 134, B
21
134f Baffles in inlet/exit section, 92–93, 93f.
22
AHS. See Active heat spreader (AHS) See also Rack/cabinet (system
23
Air cooling, 181, 188, 203 thermal analysis)
24
Air cooling and heat sinks Base spreading thermal resistance, 36,
25
forced convection, 34 37f, 38f
26
heat sink thermal enhancement, Base thermal spreading, enhanced
27
44–49, 46t, 47f, 47t, 48t, 49t (Rc), 45–49, 46t, 47f, 47t, 48f,
28
heat sink thermal performance, 48t, 49t. See also Heat sink
29
35–44, 35f, 37f, 38f–44f, 40t thermal enhancement
30
materials/manufacture processes, Board assembly connection, 195f
31
49–50, 50t Board thermal conductivities, 18f
32
natural convection, 31–34, 32f, 33f, Boiling
33
34f curves, 145f
34
Air filters, 55–57, 56f. See also Rack/ defined, 145
35
cabinet (system thermal Boiling heat transfer, 144–145, 144f,
36
analysis) 145f
37
dust protection, 127f critical heat flux (CHF), 145
38
performance curve of, 56f departure from natural convection
39
pressure drop versus air flow, 56f (DNC), 151, 151f
40
Air flow paths (internal/external), departure from nuclear boiling
41
117f (DNB), 151, 151f
42
Air flow rates, 65f, 66f, 66t, 67t, 104 film boiling, 145
43
Air handling systems. See Fan flow boiling, 170
44
All-optical network (AON), 177, 198 nucleate boiling, 170
45
Alternative energy. See also Energy pool boiling, 145
46
efficiency saturated boiling, 145
47
fuel cell technology, 201–202, 201f subcooled cooling, 145
48
49
ASME_THERMAL_MANAGEMENT_Index.indd 218 Manila Typesetting Company 09/12/2013 08:50AM

Index • 219
Booser grease life equation, 28 Computational fluid dynamics (CFD) 1

Boundary layers analysis, 59–60, 63, 81, 82t, 83t, 2
hydrodynamic, 142, 142f 84t, 85t, 86f 3
thermal, 142, 142f thermal model, 13f 4
Boxes stack-up, 112–114, 112f, 113f, Continuous fins, 133, 134f 5
114f. See also Small box (system Controlled collapse chip connection 6
thermal analysis) (C4), 5, 6f 7
Box internal layout, 108f Controlled collapse chip connection 8
carrier (C5), 11 9
C Convection and boiling heat transfer. 10
CAD model of cabinet, 123f See Liquid cooled systems 11
Carbon nanotube (CNT) base Conventional heat exchanger 12
technology, 202, 203f advantages/disadvantages, 131t 13
Card cages, 61, 63f, 64f OSP, 122, 122f, 123f. See also 14
fillers in, 93–98, 94f, 95f, 96f, 97f Outside plant (OSP) 15
Cascade effect of energy savings, 198f Coolant distribution and control unit 16
Centrifugal fans, 21. See also Fan (CDUC), 164 17
CFD analysis. See Computational fluid Coolant selection, 146–149, 148t. See 18
dynamics (CFD) analysis also Liquid cooled systems 19
Chip level thermal management, Cooling cost, 178f 20
182–187, 183f, 185f, 186f. See Cooling technologies for OSP 21
also Energy efficiency (comparisons), 131t 22
Chip packaging, 5–6, 6f. See also Counter flow heat exchanger, 129f 23
Component Counter flow system for electronic 24
Closed-loop thermosyphons, 135, cooling, 152f 25
137f Critical heat flux (CHF), 145 26
Coating 27
outside, 118 D 28
radiation properties of, 119t Data center/central office, 100–101, 29
Coefficient of performance (COP), 180 101f 30
Coefficient of thermal expansion Data center energy breakdown, 197f 31
(CTE), 6, 9f, 184, 185f Departure from natural convection 32
Cold plate, 171 (DNC), 151, 151f 33
Cold wall temperature distribution, Departure from nuclear boiling 34
153f (DNB), 151, 151f 35
Compact high intensity cooler (CHIC), Design optimization procedure, 77–80, 36
159, 160f 77f, 79f. See also Rack/cabinet 37
Component. See also Printed circuit (system thermal analysis) 38
board (PCB) Dielectric fluids, 136, 148, 194, 196 39
advanced packaging technology, thermal properties of, 149 40
11–14, 12f, 13f Dielectric material, 16 41
chip packaging, 5–6, 6f Direct air cooled OSP, 126–128, 127f. 42
thermal management, 7–11, 7f, 9f, See also Outside plant (OSP) 43
10f Direct air cooling, 21 44
Component temperatures, 112f advantages/disadvantages, 131t 45
on board in box, 110f Direct immersion cooling, 149–152, 46
for boxes with/without plate, 111f 150f, 151f. See also Liquid 47
gap size, effects of, 114f cooled systems 48
49

220 • THERMAL MANAGEMENT OF MICROELECTRONIC EQUIPMENT
1 Direct liquid cooling, 168–170, 169f External natural convection cooling,

2 Down-stream filter flow rate, 66t, 67t 128f
3 Dust protection (air filters), 127f advantages/disadvantages, 131t
4 Extruded heat sink thermal solutions,
5 E 135f
6 Electrical resistivity, 199
7 Electric magnetic insulation (EMI) F
8 plates, 53–54, 55f. See also Fan
9 Rack/cabinet (system thermal acoustic noise, 26–27
10 analysis) axial
11 honeycomb structure, 54, 55f propeller, 21
12 Electromagnetic actuator synthetic tube, 21
13 jets, 207–211, 208f, 209f, 210f. vane, 21
14 See also Synthetic jets centrifugal, 21
15 EMI plates. See Electric magnetic damper/cover, 67f
16 insulation (EMI) plates defined, 21
17 Energy efficiency, 177, 178f, 196–198, failure
18 197f, 198f catastrophic, 27
19 alternative energy causes of, 28, 28f
20 fuel cell technology, 201–202, 201f software, 27
21 thermoelectric power generations, inlet/exit conditions, 23–25, 24f, 25f
22 198–200, 199f, 200f laws, 22–23
23 liquid cooled telecom equipment, performance, 22, 22f
24 thermal design of, 191–196, reliability, 27–29
25 192f, 193f, 194f, 195f and system pressure, 26
26 power density of telecommunication tested, characteristics of, 25t
27 equipment, 177–179, 178f, 179f tray, 61, 62, 105
28 research needs, 202–211, 203f, Figure of merit (FOM), defined, 147
29 205f–210f Fillers in card cage, 93–98, 94f, 95f,
30 roadmap, thermal management and, 96f, 97f. See also Rack/cabinet
31 181–182, 181f (system thermal analysis)
32 chip level thermal management, Film boiling, 145
33 182–187, 183f, 185f, 186f Filtered vent thermal system concept,
34 system level thermal 127f
35 management, 187–191, 188f, Fin configurations, 39–43, 40f, 40t,
36 189f, 190f 41f, 42f, 43f, 44f. See also Heat
37 thermal issues sink thermal performance
38 equipment acoustic noise, Fin thermal performance, improved
39 180–181 (Ro), 45
40 equipment energy efficiency, 180 Flip chip packaging, 5
41 EPTEF filter, 128 Flow boiling, 169
42 Equipment Flow bypass and leakage, 38–39, 39f.
43 acoustic noise, 180–181 See also Heat sink thermal
44 energy efficiency, 180 performance
45 Equivalent thermal conductivity. See Flow management, 188
46 Printed circuit board (PCB) Flow network modeling (FNM), 59.
47 Ethylene Glycol/water mixture, 136 See also Rack/cabinet (system
48 Evaporation, 145 thermal analysis)
49

Index • 221
Flow resistance, 58–59, 59f, 71, 72f, Heat flow, 17 1

74f, 122. See also Rack/cabinet Heat flux, 145, 179f 2
(system thermal analysis) Heat pipe heat exchanger, 118, 119, 3
Fluorinert liquids, 136 119f 4
Fluorocarbon liquids (FC-series), 148 advantages/disadvantages, 131t 5
Forced air cooling, 96, 104f, 105f Heat sink, 7, 7f, 8, 129, 133 6
Forced convection, 152, 169. See Air AHS, 134f 7
cooling and heat sinks on board in box, 109f 8
Forced convection, single-phase, chip packaging and, 183f 9
141–144, 142f, 143t. See also and component, 89f 10
Liquid cooled systems design, 203 11
Fuel cell technology, 201–202, 201f. extruded, 135f, 136 12
See also Alternative energy finned, 134f 13
Fujitsu FACOM M-780 computer, 163 microchannel, 167f 14
Fujitsu High End Server (GS8900), for multiple chips, 100f 15
166f piezoelectric fan and, 205f 16
and SFP/XFP cage, 187f 17
G thermal enhancement, 44. See also 18
Gallium arsenide (GaAs) chip, 8 Air cooling and heat sinks 19
Gallium arsenide (GaAs) diode wafers, Rc (enhanced base thermal 20
160 spreading), 45–49, 46t, 47f, 47t, 21
Gallons per minute (GPM), 164 48f, 48t, 49t 22
Gamma function, 28 Ro (improved fin thermal 23
Geothermal heat exchanger performance), 45 24
advantages/disadvantages, 131t thermal resistance, average, 35–36 25
Geothermal heat exchanger OSP, use of, 188 26
123–126, 124f, 125f. See also Heat sink thermal performance, 35, 27
Outside plant (OSP) 35f. See also Air cooling and 28
GR-63, 2, 180 heat sinks 29
GR-487, 2, 121, 132 average heat sink thermal 30
Ground mounted outdoor electronic resistance, 35–36 31
equipment. See also Outdoor base spreading thermal resistance, 32
electronic equipment; Outside 36, 37f, 38f 33
plant (OSP) factors affecting 34
outside plant (OSP), thermal design fin configurations, 39–43, 40f, 40t, 35
of, 115–120, 116f, 117f, 118f, 41f, 42f, 43f, 44f 36
119f, 119t, 120f flow bypass and leakage, 38–39, 37
thermal designs, 115–120, 116f, 39f 38
117f, 118f, 119f, 119t, 120f mass flow rates, 44 39
Heat spreader, 11 40
H Heat transfer, 8, 129, 160, 170 41
Heat absorption, 116 Heat transfer block (HTB), 164 42
Heat dissipation, 149, 152, 178f Heat transfer coefficient, 141, 143, 43
Heat exchanger 147, 152, 204, 206 44
CHIC, 160f Hot exit air circulation, 101f 45
conventional, 122, 122f, 123f Hybrid cooling, 192 46
geo-thermal, 123–126, 124f, 125f Hydrogen fuel, 201 47
use of, 129 Hydrophobic filter, 126 48
49

1 I Liquid cooled telecom equipment,

2 IBM liquid cooled thermal conduction thermal design of, 191–196,
3 module, 162f 192f, 193f, 194f, 195f. See also
4 IBM liquid encapsulated module, 151f Energy efficiency
5 IBM microchannel cooled silicon chip, Liquid cooling, 210
6 158f of bare chip, 162f, 163f
7 IBM Power 575 system, 165f Liquid (water) cooling module (LCM),
8 IBM Zurich liquid cooled board, 167f 164
9 Indirect liquid cooling. See also Liquid Liquid encapsulated module (LEM),
10 cooled systems 151f
11 cooling technologies, 152–161, Liquid jet impingement, 163, 164
12 152f–161f Loss coefficient, 54, 55f, 71
13 system thermal design, 170–171
14 Inlet/exit height, 74f, 77f M
15 In-line fins, 133, 134f MacroFlow (software analysis tool), 59
16 Internal air temperature history, 118f Mass flow rates, 44. See also Heat sink
17 thermal performance
18 K Mean time to failure (MTTF), 28
19 Kinematic viscosity, 141 Microchannel cold plate
20 and round heater, 154f
21 L temperatures of, 155f
22 Laminar flow, 142, 143t Microchannel heat sink, 167f
23 Laser diode/cooler stack, 161 Microchannel technology, 153, 154f
24 Leakage, flow bypass and, 38–39, Microchannel water cooling, 12
25 39f. See also Heat sink thermal Microstructure, 50, 50t
26 performance Modularized board, 98, 99f
27 Liquid cooled rack, 191, 193f, 194f,
28 195f N
29 Liquid cooled systems Natural convection, 31–34, 32f, 33f,
30 convection and boiling heat transfer 34f, 105, 106, 133. See also Air
31 boiling heat transfer, 144–145, cooling and heat sinks
32 144f, 145f estimated cooling capacity of, 130f
33 single-phase forced convection, external, 128f
34 141–144, 142f, 143t OSP, 128–130, 128f, 129f, 130f. See
35 coolant selection, 146–149, 148t also Outside plant (OSP)
36 cooling technologies NEC SX-2 cooling system, 164f
37 direct immersion cooling, 149– Normal thermal conductivity, 17–18,
38 152, 150f, 151f 18t. See also Printed circuit
39 indirect liquid cooling, 152–161, board (PCB)
40 152f–161f Nucleate boiling, 170
41 industry trends, 161–167, 162f–167f Nusselt number, 141, 143t, 159f
42 system thermal design, 168–175
43 direct liquid cooling, 168–170, O
44 169f Optical electronic module (OEM), 64
45 indirect liquid cooling, 170–171 Optical network terminal (ONT), 137
46 redundant liquid cooling system, configuration of, 138f
47 171, 172f with cover open, 138f
48 saturation pressure, 172f, 173f OSP. See Outside plant (OSP)
49

Index • 223
Outdoor electronic equipment equivalent thermal conductivity 1

ground mounted normal thermal conductivity, 2
outside plant (OSP), thermal 17–18, 18t 3
design of, 121–130, 121f–130f, planar thermal conductivity, 4
131t 15–16, 15t, 16f 5
thermal designs, 115–120, 116f, thermal conductivity of, 92 6
117f, 118f, 119f, 119t, 120f thermal consideration, 19 7
tower/pole mounted, 130, 132f Printed wiring board (PWB). See 8
remote radio head (RRH), Printed circuit board (PCB) 9
thermal design of, 132–139, Propeller fans, 21. See also Fan 10
133f–138f PTH, thermal balls and, 11 11
Outside plant (OSP), 121, 121f 12
cabinet, 2 R 13
conventional heat exchanger, 122, Rack/cabinet (system thermal analysis) 14
122f, 123f accuracy of results, 80–85, 81f, 82f, 15
direct air cooled, 126–128, 127f 82t, 83f, 83t, 84f, 84t, 85f, 85t, 86f 16
geo-thermal heat exchanger, air filters, 55–57, 56f 17
123–126, 124f, 125f baffles in inlet/exit section, 92–93, 93f 18
natural convection, 128–130, 128f, CFD analysis, 59–60 19
129f, 130f component temperature, factors 20
Overmolding, 11 affecting 21
thermal conductivity of PCB, 92 22
P thermal interface material (TIM), 23
Packaging technology 87–92, 89f, 90f, 92f 24
advanced, 11–14, 12f, 13f. See also data center/central office, 100–101, 25
Component 101f 26
chip, 5–6, 6f EMI plates, 53–54, 55f 27
Passive cooled boxes. See Small box examples, 60–68, 60f, 61f, 62f, 63f, 28
(system thermal analysis) 64f, 65f, 66t, 67f, 67t 29
Passive cooled systems, 98 fans, 57–58 30
Passive cooling methods, 198 fillers in card cage, 93–98, 94f, 95f, 31
Perforation, 54 96f, 97f 32
Phase change materials (PCM), 115 flow network modeling (FNM), 59 33
Piezoelectric fan synthetic jets, flow resistance, 58–59, 59f 34
204–207, 205f, 206f, 207f. See series fan trays, effect of, 86, 87f, 35
also Synthetic jets 88f, 89f 36
Planar thermal conductivity, 15–16, shelf configuration on shelf flow, 37
15t, 16f. See also Printed circuit 71–72, 72f 38
board (PCB) design optimization procedure, 39
Plenum depth ratio, 25f 77–80, 77f, 79f 40
Power consumption, 196 shelf depth, effect of, 75–76, 75f, 76f 41
Prandtl number, 141, 142, 142f shelf height, effect of, 72–74, 73f, 42
Pressure drop, 22, 26, 53, 57, 82t, 143, 74f 43
174 system level thermal analysis, 58–60, 44
recommended system maximum, 59f 45
189f telecom rack 46
Printed circuit board (PCB), 2, 61, 103 and flow resistance, 58–59, 59f 47
construction, 14–15, 14f and shelf, 54f 48
49

1 thermal design process, 68–69, 68f case studies, 108–110, 108f, 109f,
2 thermal enhancement, 98–100, 99f, 110f
3 100f market survey, 106–107, 107t
4 Rack unit (RU or U), 103 table top deployment, 110, 111f,
5 Rc (enhanced base thermal 112f
6 spreading), 45–49, 46t, 47f, 47t, Small form-factor pluggable (SFP)
7 48f, 48t, 49t transceiver, 184, 185, 186f,
8 Rear door heat exchanger (RDHX), 188f
9 189, 190f Soil temperature, 123, 124, 125f
10 Redundant liquid cooling system, 171, Solar heating, 121, 137
11 172f. See also Liquid cooled Solar radiation, 115, 132
12 systems absorptivity, 116, 118, 119t
13 Remote radio head (RRH), 2, 3f, 130, emissivity, 116, 118, 119t
14 132, 133f solar flux, 116, 117f
15 thermal design of, 132–139, Solder bumping, 5, 6f
16 133f–138f Sound power, 26
17 Reynolds number, 141, 156 Sound pressure, 26
18 Roadmap, thermal management and. Specific heat, 146
19 See Energy efficiency Stack-up boxes, 112–114, 112f, 113f,
20 Router, 1 114f
21 Staggered fins, 133, 134f
22 S Static efficiency, 22
23 Saturation pressure, 172f, 173f. See Supercomputer, 191
24 also Liquid cooled systems Surface tension, 146
25 Sconbein, C.F., 201 Swirl cold plate, 156, 156f
26 Seebeck effect, 199 thermal performance of, 157f
27 Series fan trays, 86, 87f, 88f, 89f. Switch, 1–2
28 See also Rack/cabinet (system Synthetic jets
29 thermal analysis) electromagnetic actuator, 207–211,
30 Shear stress, 6 208f, 209f, 210f
31 Shelf, 61f. See also Rack/cabinet piezoelectric fan, 204–207, 205f,
32 (system thermal analysis) 206f, 207f
33 depth, effect of, 75–76, 75f, 76f System level thermal management,
34 fully heated boards in, 96f 187–191, 188f, 189f, 190f. See
35 height, effect of, 72–74, 73f, 74f also Energy efficiency
36 passive cooled, 95f System pressure
37 Short circuit, 14 fan and, 26. See also Fan
38 Single-phase forced convection, loss, defined, 26
39 141–144, 142f, 143t. See also System thermal design. See Liquid
40 Liquid cooled systems cooled systems
41 Small box (system thermal analysis) System thermal model, 62f
42 active cooled boxes, 103–105, 104f,
43 105f T
44 passive cooled boxes, 105–106, 105f, Table top deployment, 110, 111f, 112f.
45 106f, 107f See also Small box (system
46 boxes stack-up, 112–114, 112f, thermal analysis)
47 113f, 114f Tape automated bonding (TAB), 5
48
49

Index • 225
Telecommunication, defined, 1 Thermal enhancement, 98–100, 99f, 1

Telecommunication cabinet, 2f 100f. See also Rack/cabinet 2
Telecommunication equipment (system thermal analysis) 3
power density of, 177–179, 178f, Thermal epoxy, 91 4
179f. See also Energy efficiency Thermal grease, 91 5
router, 1 Thermal interface material (TIM), 6
switch, 1–2 7, 8, 68, 87–92, 89f, 90f, 92f, 7
Telecommunication rack, 107f 182, 202. See also Rack/cabinet 8
and flow resistance, 58–59, 59f. (system thermal analysis) 9
See also Rack/cabinet (system Thermal loading, 179f 10
thermal analysis) Thermal management, 7–11, 7f, 9f, 11
practical examples, 60, 60f 10f. See also Component 12
thermal design process, 68f Thermal model, 13, 13f 13
thermal performance, improving, Thermal pad, 91 14
98 Thermal pad joint thermal resistance, 90f 15
Telecommunication shelf, 79f Thermal performance 16
Temperature distribution, 109f, 110 of aluminum and copper embedded 17
Temperature gradient, 152 heat sinks, 46t 18
Tested heat sink configurations, 48f, metrics, 8 19
49f Thermal resistance, 87, 88, 91. See also 20
Thermal boundary conditions, 141, Air cooling and heat sinks 21
142 base spreading, 36, 37f, 38f 22
Thermal conduction module (TCM), of epoxy joint, 90f 23
161, 162 heat sink average, 35–36 24
Thermal conductivity, 8, 9f, 10, 146. measured, 48 25
See also Printed circuit board over pad joints, 92f 26
(PCB) versus pressure, 90f 27
and CTE, 185f Thermal spreaders, 202 28
material, 50t Thermal technology map, 181, 181f, 187 29
normal, 17–18, 18t Thermal tests, 155, 160 30
of PCB, 92. See also Rack/cabinet Thermocouples, 155 31
(system thermal analysis) Thermoelectric devices, 199, 211 32
planar, 15–16, 15t, 16f Thermoelectric power generation, 33
Thermal design 198–200, 199, 199f, 200f. See 34
goal of, 109 also Alternative energy 35
of liquid cooled telecom equipment, Thermosyphon, 135, 136, 136f 36
191–196, 192f–195f closed-loop, 135, 137f 37
of outdoor electronic equipment, Thin film thermoelectric device, 182 38
115–120, 116f–120f, 119t Thin rectangular offset fins, 158f 39
of outside plant (OSP), 121–130, 3D chip packaging, 11, 12f 40
121f–130f, 131t 3D packaging, 184, 185f 41
of remote radio head (RRH), Tower mounted devices, 132f 42
132–139, 133f–138f Tower/pole mounted outdoor 43
Thermal design process, 68–69, 68f. electronic equipment, 130, 44
See also Rack/cabinet (system 132–139, 132f, 133f–138f 45
thermal analysis) Tube axial fans, 21. See also Fan 46
Thermal diffusivity, 141 Two-phase flow cooling, 173 47
48
49

1 U W
2 Underfiller materials, 10 Water, 171, 196
3 Up-stream filter flow rate, 66t, 67t Water–air heat exchangers, 190f
4
5 V X
6 Vane axial fans, 21. See also Fan XFP, thermal models for, 185, 186,
7 Vapor chamber heat sinks, types of, 47f 186f
8 Volumetric power density, 107
9
10
11
12
13
14
15
16
17
18
19
20
21
22
23
24
25
26
27
28
29
30
31
32
33
34
35
36
37
38
39
40
41
42
43
44
45
46
47
48
49

ABOUT THE AUTHORS
Lian-Tuu Yeh, Ph.D. and PE

ASME Fellow
Lian-Tuu Yeh received his BS in mechanical engineering from National Cheng

Kung University, Taiwan, in 1967 and his M.S. and Ph. D in mechanical engi-
neering from the University of Akron, Akron, Ohio in 1972 and 1976, respec-
tively. Dr. Yeh has been practicing in thermal sciences for more than 35 years
over diverse industries that include major companies such as Babcock & Wil-
cox, Texas Instruments, Lockheed-Martin, Boeing, Chiaro Networks, Fujitsu Net-
works and Futurewei Technologies in energy/power, electronics, aerospace, space
and telecommunications industries. His work has been concentrated in thermal
management of wide range of electronic systems. Dr. Yeh has made steady and
significant contributions in the development of advanced cooling technologies for
high-power electronic systems. His experience includes various types of liquid- or
air-cooled electronic equipment, and above all, he has more than 10 years direct
experience in thermal management of telecommunication systems. Dr. Yeh has
been a member of Thermal Technology working group of INEMI (International/
National Electronics Manufacturing Initiate) since its inception in 2002 and he
has been serving as section lead for NETCOM since 2008. As an adjunct professor,
Dr. Yeh taught the course of thermal analysis and design of electronic equipment
for 2 years (1994, 1995) at the Department of Aerospace and Mechanical Engi-
neering at the University of Texas at Arlington.
Dr. Yeh has been an active participant in and session, as well as conference
organizer, of national and international heat transfer conferences. He has served
as a technical reviewer for various heat transfer journals and conferences, and also
for US National Science Foundation in reviewing proposals for research grants.
Dr. Yeh is a practicing engineer and is also researcher. He has published more than
62 technical papers in the field of heat transfer and a textbook entitled “Thermal
Management of Microelectronic Equipment” by ASME Press in 2003. Dr. Yeh is also
a co-author for the book chapter, “Thermal Management” of “flip chip advanced
technologies” edited by Professor C.P. Wong and published by Springer Publishing
in 2013. In addition, he also received one US patent for liquid cooling of electronic
equipment.
Dr. Yeh was elected to a Fellow of the American Society of Mechanical Engi-
neering (ASME) in 1990. He has served and is still an active member of the ASME
K-16 Committee on Heat Transfer in Electronic Equipment and the former mem-
ber of American Institute of Aeronautics and Astronautics (AIAA) Thermophysics
Technical Committee. He has been listed in Who’s Who in the South and South-
west (1988–1989) and Who’s Who in the world (1994–1995). Dr. Yeh is also a
registered professional engineer in the state of Texas.
Late Dr. Richard C. Chu

IBM Fellow Emeritus
Member, National Academy of Engineering, USA
Academician, Academia Sinica, ROC
Dr. Richard C. Chu is an IBM Fellow Emeritus in the Systems and Technology
Group in Poughkeepsie, NY. He is an internationally recognized authority in cool-
ing technology for high-performance computers. For more than four decades, he
has played a key role in developing the world’s most advanced computer cooling
solutions.
Joining IBM in 1960 as a thermal engineer, he worked on the development of
the IBM System/360. His 1965 invention of a multi-level air–liquid hybrid cooling
design was pivotal to IBM’s successful introduction of its System/360 Model 91.
The introduction of this system marked the beginning of the water-cooling era
which lasted throughout the years of bipolar technology.
Throughout his career, he has been a prolific technical innovator receiving
60 IBM invention achievement awards for over 200 patents and 150 invention
publications. As one of the inventors of the cooling scheme for the IBM Thermal
Conduction Module, he received an IBM Outstanding Innovation Award and a
Corporate Award. This scheme and his modular cold plate cooling system con-
cept formed the basis for the cooling design of IBM’s largest computers for over
15 years. Variations of this cooling concept were adapted by other major main-
frame computer manufacturers during this time period.
About the Authors • 227
The importance of his many contributions to IBM and the industry led to his
appointment as an IBM Fellow in 1983, the company’s highest technical honor.
He still continues in this capacity today, exploring cooling technology for future
computers. In 1996, he initiated an effort to study the feasibility of using lower tem-
perature cooling to enhance CMOS performance. He co-invented a dual-channel
evaporator cold-plate enabling modular refrigeration cooling (MRC), which has
been used on most IBM’s CMOS-based high performance computers since 1997.
The importance of his contributions extends well beyond IBM. He has been a
strong promoter of a close working relationship between industry and academia.
For 25 years, he was the principal individual responsible for IBM’s sponsorship
of electronic cooling research at 12 universities. He is widely known for his active
participation in ASME and IEEE conferences, presenting papers and often serv-
ing as a keynote speaker. He is a co-author of two books on electronic cooling.
He also chaired a National Electronic Manufacturing Initiative (NEMI) technical
working group in 2000 and 2002 to produce a Thermal Management Roadmap
chapter for the NEMI Technology Roadmap.
He has been widely recognized for his contributions, receiving the ASME
Heat Transfer Memorial Award, ITherm Memorial Award, SEMI-THERM Sig-
nificant Contributor Award, InterPACK Conference Achievement Award, Chinese
American Academic and Professional Society Achievement Award and the Chi-
nese American Corporate Achievement Award from the Organization of Chinese
Americans. In addition to being a member and past president of the IBM Acad-
emy of Technology, he is an ASME Fellow and a Fellow of the American Asso-
ciation for the Advancement of Science. He is the recipient of the Distinguished
Alumni Award from both of his alma maters, the National Cheng-Kung University
and Purdue University. He has also received honorary doctorate degrees from
Purdue University and the American University of the Caribbean. Most recently
he was honored as an Asian American Engineer of the year.

Thermal Management of Telecommunications Equipment

Uploaded by

Copyright:

Available Formats

Thermal Management of Telecommunications Equipment

Uploaded by

Document Information

Copyright

Available Formats

Share this document

Share or Embed Document

Sharing Options

Did you find this document useful?

Is this content inappropriate?

Copyright:

Available Formats

Thermal Management of Telecommunications Equipment

Uploaded by

Copyright:

Available Formats

What is the book about?

What is the book about?

What are some of Richard C. Chu's contributions?

What are some of Richard C. Chu's contributions?

THERMAL MANAGEMENT

ASME Press Book Series on Electronic Packaging

ASME shall not be responsible for statements or opinions advanced in papers

Requests for special permission or bulk reproduction should be addressed to the

Library of Congress Cataloging-in-Publication Data

Yeh, L.-T. (Lian-Tuu), 1944-

Chapter 2 Component and Printed Circuit Board 5

Chapter 3 Fans and Air Handling Systems 21

3.4 FAN AND SYSTEM PRESSURE�������������������������������������������������������� 26

Chapter 4 Air Cooling and Heat Sinks 31

Chapter 5 System Thermal Analysis—Rack/Cabinet 53

Chapter 6 System Thermal Analysis—Rack/Cabinet 71

6.2 VARIOUS SUBJECTS RELATED TO SYSTEM THERMAL DESIGN���� 86

Chapter 7 System Thermal Analysis—Small Box 103

Chapter 8 Outdoor Electronic Equipment 115

Chapter 9 Liquid Cooled Systems 141

9.2 COOLANT SELECTION����������������������������������������������������������������� 146

Chapter 10 Thermal Management and

About the Authors 225

1.1 Typical telecommunication cabinet and box configuration........................2

4.9 Typical straight fin heat sink.......................................................................39

6.16 Shelf flow ratios versus shelf depth ratios......................................................... 85

8.10 CAD model of cabinet.......................................................................................123

9.29A Passive direct liquid cooling concepts.............................................................169

10.35 Schematic diagram of impinging synthetic jet...............................................208

Miniaturization and convergence are key drivers in microelectronics and tele-

Dereje Agonafer, ASME & AAAS Fellow

The need to develop reliable microelectronic devices capable of operating at high

Telecommunication, which is defined as the use of information transmission over

t­ elecommunications network points-of-termination (POTs) to provide

A core network is the central part of a telecommunication network that pro-

Figure 1.1 Typical telecommunication cabinet and box configuration.

Figure 1.2 Typical outside plant and remote radio head.

1. Telcordia Technologies Generic Requirements, GR-63-CORE, Issue 3, March,

Component and Printed

2.1.1 Chip Packaging Technology

1. Higher I/O density

Figure 2.1 Typical flip chip package with heat sink.

Figure 2.2 Common solder bumping technologies.

F igure 2.3 Thermal representation of flip chip and heat sink.

Figure 2.4 Thermal conductivity and CTE of various materials.

θ j - a = (Tj,max - Ta) / P (2.1)

θ j - b = (Tj,max - Tpcb) / P (2.2)

θ j - c = (Tj,max - Tc) / P (2.3)

a. Effect of C4 and underfill thermal conductivity

F igure 2.5 Cross-section of package in wind tunnel.

a. Effect of C4 and underfill thermal conductivity

Figure 2.6 Schematic diagram of various thermal enhancements.

thermal conductivity smaller than 3 W/m °C. 0.6 to 10 W/m °C is the

2.1.3 Advanced Packaging Technology

F igure 2.7 Heat conduction spreading in die.

F igure 2.8 IBM 3D packages.

F igure 2.9 Cross-section of CFD thermal model.

2.2 PRINTED CIRCUIT BOARD

A typical printed circuit board consists of a multilayer of dielectric material and

F igure 2.11 Typical construction of a PCB.

Chapter 2 Component and Printed Circuit Board 5

Chapter 3 Fans and Air Handling Systems 21

3.4 FAN AND SYSTEM PRESSURE�� 26

Chapter 4 Air Cooling and Heat Sinks 31

Chapter 5 System Thermal Analysis—Rack/Cabinet 53

Chapter 6 System Thermal Analysis—Rack/Cabinet 71

6.2 VARIOUS SUBJECTS RELATED TO SYSTEM THERMAL DESIGN�� 86

Chapter 7 System Thermal Analysis—Small Box 103

Chapter 8 Outdoor Electronic Equipment 115

Chapter 9 Liquid Cooled Systems 141

9.2 COOLANT SELECTION�� 146

Chapter 10 Thermal Management and

t elecommunications network points-of-termination (POTs) to provide