Vent Holes at Spandrel PDF
Vent Holes at Spandrel PDF
Vent Holes at Spandrel PDF
ABSTRACT
A novel solution to overcome problems associated with the use of excessive window areas and
shallow floor plates for daylighting buildings is to use active sunlight tracking systems to deliver
controlled sunlight to areas that are inaccessible by conventional methods. This technology is
commercially available to designers with minimal limitations to building orientation and form.
The technology tracks the sun at the top of the building with an array of motorized mirrors and
redirects a stationary beam of sunlight to concentrating optics located in sunshades and
spandrel panels on each floor of the building. The concentrated beam of sunlight from the
spandrel is distributed through the building using hybrid light pipe technology and fixtures.
This paper addresses the challenges of integrating an optical system into the building envelope,
including limitations on insulation thickness, requirement for clear glass to the exterior and
interior of the spandrel, and controlling dust and moisture within the spandrel to maintain the
effectiveness of the optics. By means of analysis, testing, and validation, this system is
compared with conventional curtain wall spandrels for controlling heat, air, and moisture.
Keywords: daylighting, building envelope, curtain wall, desiccant, thermal analysis
INTRODUCTION
According to the United States Department of Energy, in 2008 commercial buildings
consumed 18.4% of the total primary energy demand. Out of this, nearly 26% of the energy
was used for lighting (D&R International 2011). While slowly decreasing, there is still a
significant amount of energy consumed in commercial buildings for lighting. Using daylighting
as a solution has been shown to significantly increase lighting quality and occupant comfort,
however, traditional daylighting methods have several limitations (Boyce 2003). These factors
provide an opportunity to develop and manufacture new daylighting systems and methods that
compete with conventional methods.
This paper introduces a non-traditional daylighting system; one that includes active tracking of
the sun, concentration of sunlight, and distribution through hybrid light pipe fixtures deep within
the buildings core. The system is designed primarily for office buildings because offices are
often occupied during the daytime and have large floor areas that are not fully accessible to
daylight. While striving for improved energy efficiency and occupant comfort in buildings, new
technologies can often introduce new challenges for designers. Detailed design review,
1
testing, and monitoring is required to reduce the uncertainty in the long-term performance of
new products for the construction industry and facilitate earlier adoption.
This paper focuses on how the challenges of integrating this new technology into common
curtain wall systems were met through design, testing, verification, and long term monitoring.
The challenges include physical integration within curtain wall spandrels, meeting thermal
performance requirements, and controlling heat, air, and moisture transfer in the spandrel.
FIGURE 1: Active daylighting system showing the integrated components that collect and
distribute sunlight deep into multi-story buildings. Image courtesy: SunCentral Inc.
2
Due to the increased depth occupied by the mirrors in the OSP, the depth of the panel is
substantially larger than IGUs. This results in significantly more air volume than an IGU,
requiring venting of the air space to prevent large temperature-induced pressure
changes.
With venting, moisture and dirt from the exterior air must be controlled as they are
detrimental to the functionality of the optical components. The OSP also requires a clear
front pane of glass to allow sunlight to enter so it cannot be opacified similar to common
spandrel glass.
The space required for the optical elements results in limited space for thermal
insulation. High performance thin insulation is required for this application to provide
insulation levels that are comparable to common spandrel sections.
The OSP is designed to meet these requirements, make adoption simpler, and help comply with
energy codes and standards like ASHRAE 90.1. Most glazing spandrels with insulated
backpans are opaque elements that generally do not meet the prescriptive insulation values or
effective U-values for cold climates. A common approach is to add continuous insulation5
inboard of metal-pans in order to keep below the maximum prescriptive thermal transmittance
requirements for steel-framed walls. However, this approach often yields spandrel section Uvalues that do not meet the requirements and can even result in an unacceptable risk of
condensation for some applications (Morrison Hershfield 2011).
Since the prescriptive path is generally not used for buildings utilizing spandrel panels, the
thermal performance of the OSP is assessed to meet or surpass the thermal performance of
spandrel panels used in common practice.
The insulation is generally continuous except at floor slabs, anchors, and at the transition to the vision areas
The OSP maintains the air and moisture barrier continuity to the mullions via silicone sealant
gaskets and glazing tape similar to common four sided structural silicone glazing. The metal
backpan provides the same level of floor separation as common practice. Insulation is also
provided inside the enclosure secured to the backpan.
FIGURE 3: Optical panel integrated within standard spandrel area of curtain wall
To integrate the OSP with a standard 25 mm (1 in.) glazing pocket similar to common IGUs, a zframe extrusion is used to provide a front flange for a glazing spacer and front pane of glass,
and a back flange that supports the backpan. This allows for roughly 115 mm (4.5 in.) of space
between the backpan and the front glass for optics and insulation in the common 6 mullion
systems targeted by the manufacturer. The output window at the back of the spandrel is sealed
with a clear pane of glass, allowing transmission of concentrated sunlight into the light pipe. The
light pipe fixture is aligned to this output window and distributes the concentrated sunlight deep
into the building (See Figure 4).
FIGURE 4: OSP integration with into curtain wall at slab (mirrors not shown)
Thermal Performance
The thermal performance of the OSP is compared to common practice for both thermal
transmittance and risk of condensation. An objective of the analysis is to demonstrate that the
thermal performance is not compromised by this design approach and is at least as adequate
as common practice.
Since the optical elements within the OSP occupy approximately 90 mm (3.5 in.) of the depth,
this results in only 25 mm (1 in.) of space available for insulation. Considering that common
spandrels fully insulated with semi-rigid insulation require roughly 100 mm (4 in.) to reach a
nominal resistance of R-16.8, a highly insulating material is needed in addition to methods that
5
mitigate thermal bridging around the perimeter of the OSP. Vacuum Insulated Panels (VIPs)
were deemed appropriate for insulating the backpan due to their highly insulating properties (R40/25 mm). Aerogel blanket (R-10/25 mm) was selected for the perimeter of the z-frame as it
can easily form and route around the z-frame and other optical elements in the OSP, while still
providing a very low thermal conductivity. Aerogel blankets can also be used to insulate the
backpan instead of VIPs if the thermal transmittance requirements can be met (see Figure 6).
Modeling procedures and software were extensively validated as part of ASHRAE Research Project 1365-RP
Thermal Performance of Envelope Details for Mid & High-Rise Buildings.
Nominal
Spandrel 1D
Insulation
hrft2oF/BTU
(m2K/W)
R-value
hrft2oF/BTU
(m2K/W)
U-value
BTU/hrft2oF
(W/m2K)
8.4
(1.48)
6.27
(1.11)
0.159
(0.905)
16.8
(2.96)
7.50
(1.32)
0.133
(0.757)
10.0
(1.76)
6.43
(1.13)
0.156
(0.883)
40.0
(7.04)
8.11
(1.43)
0.123
(0.700)
For the common unitized curtain wall system modeled, with R-16.8 insulation at the backpan,
the effective R-value of the spandrel area is reduced to R-7.5 due to the thermal bridging
resulting from the highly conductive curtain wall and backpan framing.
Different material configurations for the OSP insulation were considered to provide design
options for the manufacturer. Case 3 in Table 1 has Aerogel at the backpan and at the
perimeter of the Z-frame, whereas Case 4 utilizes a VIP panel at the backpan and Aerogel at
the perimeter. Since the VIP is four times less conductive for the same thickness compared to
Aerogel, it provides more insulation but also has some limitations on cost and form factor
compared to Aerogel. As can be seen in Table 1, the overall R-value for Case 3 surpasses
Case 1, and Case 4 surpasses Case 2 (baseline). Therefore, it can be shown that either Case 3
or Case 4 can be used for insulating the OSP depending on the performance requirements of
the building where the system will be installed. However, Case 4 will outperform most common
spandrels in overall R-value.
Through a design optimization process using 3D thermal modeling, interesting results were
attained. Extending the Aerogel to the perimeter of the glass to fill the air gap between the zframe flange and the glass by the foam spacer improved the overall thermal resistance of the
OSP by nearly R-2. This demonstrates that detailing of insulation can drastically improve the
effectiveness of insulation within the spandrel.
The thermal gradient image in Figure 7 identifies areas with higher heat flow with wide bands of
single colors and areas of high thermal resistance with narrow bands of different colors spaced
closely together. The simulated color chart provides a temperature index with 0 (blue) as the
outdoor temperature and 1 (red) as the indoor temperature. As can be seen by the thermal
gradient lines at the center of Figure 7, a source of thermal bridging for the OSP is through the
small glass output window at the bottom of the backpan where there is no backpan insulation.
For the design of the OSP, optical constraints limit the options available for improving the
thermal performance in this area, such as using an IGU instead of a single pane of glass to
provide added resistance for the output window.
7
Output Window
Light Pipe (for reference)
When comparing the thermal gradients of the traditional spandrel system (Figure 8) and the
OSP (Figure 9), the OSP maintains similar interior frame and slab temperatures compared to
the traditional system. The temperature at the output window is slightly reduced relative to the
rest of the backpan and poses a slightly higher risk for condensation. However, by conducting
heat from the interior, the light pipe connected to the output window will keep this area of the
panel warmer.
(1)
Where
P = Pressure, Pa
V = Volume, m3
T = Temperature, K
C, t, e = Constants
Assuming that the pressure is kept constant, this formula can be rearranged to find the
displaced volume between two temperature conditions:
(2)
A change in temperature of even 60C, which is within the expected service range of spandrel
panels exposed to sunlight, results in a pressure increase of approximately 20 kPa (3 psi) on
the glass panel. This exceeds most allowable design stresses for glazing. Therefore, a means
to quickly displace roughly 1/5th of the entire air volume is required to prevent a pressure
differential when temperature changes occur.
Some shadow boxes use breather tubes or weep holes to vent the air space to the exterior
environment as the exterior air generally contains a lower moisture load than the indoor
environment in most climates. However, traditional shadow boxes relying on a single pane of
glass at the front of the panel have historically been prone to condensation issues and visible
streaking on the clear front glass. This is an undesirable effect for the OSP as it impacts
aesthetics and results in moisture damage to the optics.
Improved shadow box designs (See Figure 10) employ an insulated glazed unit (IGU) so that
the inside lite (within the spandrel) is warmer than the outside lite. This allows for temperature
regulation and reduces the likelihood that the temperature of the inner pane of glass will dip
below the dew point of the vented airspace.
FIGURE 10: Improved shadow box design with IGU. The vented air cavity is located between
the IGU and the aluminum shadow box.
Although using an IGU is a feasible approach for shadow boxes, the OSP design only allows for
a single pane of glass to be used due to the limited depth available for optical components and
the need to maximize lighting performance. Therefore, this presents a high risk of condensation
for the OSP, especially in conditions when the temperature of the glass is below the
temperature of the exterior air due to night-sky-radiation.
To make any condensation invisible to the exterior, traditional vented spandrel panels utilizing
one pane of glass are generally opacified and rely on weep holes to allow the condensate to
drain out of the system. This approach is not an option for the OSP since the front pane of glass
has to be clear to allow sunlight to enter the system.
To eliminate the risk of condensation for the OSP, desiccant is required to adsorb the moisture
in the air and as a result, maintain a low dew point. This is a similar approach to that used to
keep IGUs dry, but unlike IGUs, due to the temperature-induced breathing cycles of the OSP,
moist air is introduced into the enclosure any time the temperature changes, requiring high
levels of desiccant to maintain a dry environment.
The challenge is determining the vapor load that the OSP experiences over the product life and
as a result, the amount of desiccant required. The goal is to provide pressure equalization, while
minimizing free vapor diffusion through the breather tube into the OSP. A prototype was built to
size the breather tube and test assumptions for the amount of desiccant required.
equalization, the OSP must allow a maximum flow of roughly 2L/min. These assumptions are
validated in the Testing and Validation section of this paper.
A pressure gauge was used to measure pressure decay as a function of time for different tube
sizes and air filter configurations. The experiment was conducted with tube sizes that were
either 3 mm diameter (0.125 in.) or 6 mm diameter (0.25 in.) with and without air filters. The
data collected can be seen in Figure 11.
FIGURE 12: Flow rate and pressure graph used for sizing of breather tube
The OSP is therefore designed with a breather tube that provides pressure equalization and
butyl/silicone perimeter seals similar to common IGUs to control moisture and vapor ingress into
the enclosure (Figure 13).
11
Method A: a simplified calculation method that assumes the OSP will encounter one
large breathing cycle per day as the air heats up and cools, and relying on known
climatic data to determine average vapor loads of the exterior air.
Method B: experimental technique using a scale prototype with humidity sensors and
flow gauges to create a model for expected vapor load (addressed further in the Testing
and Validation section of this paper).
The simplified calculation method (Method A) was utilized before the experimental methods and
relied on some basic assumptions. Since the largest temperature fluctuations occur on sunny
days when the sun heats the OSP past the temperature of the exterior air, it is assumed that a
maximum daily temperature fluctuation for the air inside the OSP will be about 60C (ie. a
gradual temperature swing from 20C to 80C). This assumption is based on experience and is
validated in the Testing and Validation section using Methods B and C. Due to the thermal lag
effect created by heat stored in the aluminum framing, glass, and optical components, it is also
assumed that minimal flow is experienced when small exterior temperature fluctuations occur.
Therefore, it can be approximated that the enclosure will undergo, on average, one breathing
cycle per day heating up when the sun comes out and cooling down at night time. Based on
this assumption and using Equation 1, the daily airflow exchange is calculated for a regular
sunny day. Since the OSP has roughly 120 L of air, and the temperature fluctuations induce a
22% volume change, roughly 26 L of air will be displaced per day. These assumptions are
12
conservative in general and the validation of these assumptions is presented in the Testing and
Validation section of this paper.
To determine the amount of water vapor that has to be adsorbed by the desiccant with each
breathing cycle, historical data for various climates was used to find the average vapor density
of air throughout the year. For example, in Vancouver, the average air vapor density is around
10 g/m3 for the summer and 5 g/m3 for the winter (see Figure 14). These densities vary by
climatic regions in North America, so the upper end of the density (10 g/m3) was used as a
conservative number that could be applied for a full year, and in other regions using Method A.
FIGURE 14: Typical seasonal hygrothermal data for Vancouver, British Columbia
By using the breathing cycle volume per day (26 L) and the vapor density of the air, the amount
of water vapor that the desiccant must adsorb with each breathing cycle can be determined.
The desiccant selected is a Type 4A molecular sieve desiccant due to its ability to adsorb a
large quantity of vapor in relation to its weight, and its common use in IGUs.
The adsorption capacity of the molecular sieve desiccant is roughly 24 wt%. However, once the
desiccant reaches this level, the rate of adsorption is significantly lower and its ability to adsorb
additional moisture is reduced. This means that the drying capacity and speed at which the air
can be dried is reduced, increasing the likelihood of condensation. To take a conservative
approach, the adsorption rate vs. equilibrium water capacity of the desiccant provided by the
supplier was assessed and it was determined that the desiccant would still perform as desired
at roughly 15 wt% capacity. Therefore, for every 1 g of water vapor adsorbed, roughly 6 g of
desiccant is needed. And so, for every breathing cycle, 0.26 g of water is absorbed and this
requires around 1.8 g of desiccant per day to keep the OSP dry. Based on the conservative
calculations using Method A, to design a system that maintains a dry environment for at least 20
years, a 120L enclosure requires roughly 13 kg (28.5 lb) of desiccant.
Due to the high water retention of molecular sieve desiccant, it was assumed that the desiccant
will not desorb any moisture that it adsorbs over time.
13
(3)
where
DV = Mass diffusivity for air (binary mixture), mm2/h
P = Pressure, 101.3 kPa
T = Temperature, 293K
For this scenario, DV is equal to 25.3 mm2/h. The vapor load on the OSP due to free flow
diffusion can be calculated using:
(4)
where
= Mass flow rate of vapor, kg/s
DV = Mass diffusivity for binary mixtures, mm2/h
MB = Relative Molecular Weight
Acs = Cross Sectional Area, m2
Ru = Universal gas constant
pb0 = Vapor pressure at one end of tube, kPa
pbL = Vapor at the end of tube of length, kPa
L = Length of tube, m
T = Temperature, 293K
Using a psychometric chart to determine vapor pressure, with the OSP constantly dried to
around 5% RH by the desiccant the vapor pressure differential (pb0 - pbL) between the exterior
air relative and the OSP for the summer in Vancouver is approximately 1.5 kPa. This can be
assumed as a conservative average value throughout the year.
Therefore, using Equation 4 for a breather tube 0.3 m long, the vapor load (0.0025 g H2O/day)
due to diffusion is quite negligible relative to the vapor intake during regular breathing cycles.
This demonstrates that vapor diffusion through the breather tube with the diameter and length
selected introduces a small amount of moisture into the enclosure and does not have a major
impact on the desiccant required in the OSP. In addition, due to the dynamic temperatures
experienced by the OSP, there are few scenarios where the airflow between the enclosure and
the exterior environment will be stagnant.
14
FIGURE 15: Prototype photos E1 at bottom left, E2 at top left and the desiccant and scale in E2
(right image). Condensation visible on the enclosure was on the outside of the glass.
The exterior humidity and temperatures were monitored to determine the vapor load on the
enclosure with each breathing cycle. Interior air temperatures were monitored to calculate the
flow in and out of the enclosure to validate the previous assumptions for flow rates (Method A)
and to compare to measured flow rates using flow gauges (Method B).
FIGURE 16: Photo of prototype enclosure E2 with desiccant. Components labeled: (1) breather
tube (2) R/H sensor (3) air filter (4) desiccant scale (5) desiccant (6) aluminum frame (7) passthrough sensor connectors (8) data collection unit (9) top and bottom glass.
15
The scale had to be positioned inside the enclosure to prevent having to take the desiccant out
of the enclosure for measurements. This introduced some challenges monitoring the weight of
the desiccant because of the environmental exposure of the scale. Due to the drop in relative
humidity and the temperature fluctuations of the enclosure, the scale was damaged. The scale
could not be repaired or replaced without having to open the enclosure and restart the
experiment altogether. Even though the scale showed insignificant mass gain even 1 year after
the prototype was deployed, the readings are not reliable to use in absolute terms. Although the
desiccant could be removed and weighed with another scale, gaining access to the enclosure
would permanently disrupt the experiment. This was not done in order to keep the experiment
running as part of long term monitoring.
Since only a small fraction of the anticipated desiccant weight for an enclosure of this volume
was used in the prototype, by monitoring the conditions inside the enclosure over time, a
desiccant lifetime can be extrapolated for the design of the OSP.
Approximately 2800 grams (6.2 lb) of desiccant was used in the enclosure with 120 L of air
volume. Data collected shows that 18 months into the experiment, the relative humidity of the
enclosure continues to be kept near 0% RH (within an error margin of 5-10% based on the
sensor manufacturers specifications). This shows that even with just 1/5th of the desiccant mass
required calculated using Method A, the desiccant has not been saturated. From the data
collected, it can be determined that an OSP with a desiccant weight of 13 kg (28.5 lb) and this
air volume would last at least 8 years in a climate like Vancouvers (1.5 years already proven x 5
times the desiccant mass). However, since the desiccant is not yet saturated a much longer
lifetime can be anticipated and the other validation methods can be relied on to demonstrate
this. The desiccant in the prototype will be continuously monitored to further validate the vapor
loads and determine if the air inside the enclosure is being dried.
Flow rates and anticipated vapor intake was validated using experimental data. Since the flow
data for Enclosure #1 and Enclosure #2 were very similar, only data from E2 (containing
desiccant) is used for validation. Figure 17 below shows the data collected for July 24 and 25 of
2013. July 24th was a partly sunny day as can be seen in the solar radiation data for the day. As
a result of passing clouds, there were continuous fluctuations in the enclosure air temperature,
but it is evident that the heat stored in the glass and framing inside the enclosure helps mediate
the temperature changes because the fluctuations in air temperature are not as drastic as the
changes in solar radiation. Measured flow rates also show changes in the flow direction that
correspond with the changing temperature of the air inside the enclosure. July 25th was sunny,
resulting in mostly outward flows in the early part of the day and inward flows for the later part of
the day as the enclosure air cooled (see Figure 17).
The maximum flow rate was approximately 20 L/hour (0.33L/min), which is far below the
calculated maximum of 2L/min based on the previous assumption of a maximum change in
temperature of 40C in 10 minutes. This shows that even on sunny days, the temperature
spikes induced by periods of clouds/sun are mitigated by the heat stored in the enclosure,
keeping the temperature of the air in the enclosure more constant (see Figure 19).
Of interest is also the difference between the exterior air temperature at the hottest (roughly
30C) and the air temperature inside the enclosure (nearly 65C), which shows a significant
increase in air temperature due to solar radiation. When installed in a spandrel section of a
building, the OSP will likely experience less temperature swings than the fully exposed
prototypes due to thermal regulation provided by the building.
16
Experimental data collected for the flow rates between the enclosure and exterior is compared
to calculated data based on the interior air temperature. The data shows that the measured flow
rates are significantly lower than the calculated rates using Equation 2, which assume that a
temperature change will be directly proportional to a volume change. However, the data follows
the same flow pattern (see Figure 19).
FIGURE 19: Measured flow rates with flow sensors vs. calculated flow using temperature
By summarizing the data for July and December in Table 2, we can compare the three different
methods of determining desiccant spent. Method A uses the min/max temperatures and an
average vapor density of the air. Method B relies on using the measured flows from the flow
sensors and adding up the flows for the entire day. Method C uses the measured temperatures
from the interior air and Equation 1 to convert to flows. Both Methods B and C use measured
relative humidity to determine vapor loads. As can be seen, using Method C results in nearly
twice the amount of desiccant spent for each day relative to Method B. Using Method A is less
conservative than Method C and results in a slightly lower desiccant weight spent. For Method
A, if we were to use vapor densities of 10 g/m3 of air as per our initial conservative approach, it
would result in a slight underestimate for the summer and, as expected, a very conservative
estimate for the winter.
Table 2: Measured and Calculated Flows
December 4th
(sunny)
Method
Description
December 5th
(partly sunny)
July 24
(party sunny)
July 25
(sunny)
Flow In Flow Desiccant Flow In Flow Desiccant Flow In Flow Desiccant Flow In Flow Desiccant
(L)
Out (L) Spent (g)
(L)
Out (L) Spent (g)
(L)
Out (L) Spent (g)
(L)
Out (L) Spent (g)
Calculated using
min/max approach
11.4
11.4
0.23
9.7
9.7
0.19
18.3
18.3
1.47
18.3
18.3
1.46
8.3
8.0
0.15
7.5
7.0
0.12
15.6
14.9
1.25
12.0
11.0
0.96
Calculated using
13.2
measured temperature
12.8
0.26
12.7
11.9
0.21
34.2
33.9
2.82
22.4
22.6
1.85
Notes:
1) For desiccant spent, mass is calculated based on 15% wt H2O adsorption capacity.
2) For method A, 3 g/m^3 of air is used as the vapor density and for the summer 12 g/m^3 of air is used.
2) For methods B, real-time measured flow is used.
3) For method C, the measured humidity of the air and measured temperature is used.
18
19
more than 1.5 m x 1.5 m (5 ft x 5 ft), this is likely not a big concern for introducing added vapor
into the enclosure.
SUMMARY
Through detailed design, testing, and validation, a new daylighting technology was integrated
with common curtain wall construction methods. Challenges of mechanically integrating the
OSP into curtain wall were met by applying standard glazing practices and approaching the
design similar to an IGU module. Even though the optics introduced space constraints within the
OSP, to meet thermal performance goals, detailed analysis and optimization was performed to
demonstrate that highly insulating materials can be used to provide adequate insulation for the
product. Analysis also showed that using Aerogel and careful detailing can reduce thermal
bridging and greatly improves the overall thermal effectiveness of the OSP.
To address the complex science behind heat, air and moisture movement through the building
envelope, several calculations were performed and validated using measured data from
prototypes. The experiments show that conservative calculations can be reliable for determining
the amount of desiccant required for the OSP.
This paper demonstrates that although integrating new technologies within the building
envelope introduce new challenges, solutions can be implemented when a methodical approach
is used that includes analysis, testing, monitoring, and validation. This work also presents an
opportunity for evaluating similar technologies in shadow boxes that encounter condensation
issues or for other building components that are highly visible and require sealing.
20
REFERENCES
ASHRAE. 2010. ASHRAE 90.1 Energy Standard for Buildings Except Low-Rise Residential
Buildings, Atlanta, GA: American Society of Heating, Refrigerating and Air-Conditioning
Engineers Inc.
ASHRAE. 2009. Handbook of Fundamentals. Atlanta, GA: American Society of Heating,
Refrigerating and Air-Conditioning Engineers Inc.
D&R International. 2011. 2010 Building Energy Data Book, Silver Spring, Maryland: United
States Department of Energy.
P. Boyce, C. Hunter, O. Howlett. 2003. The Benefits of Daylight through Windows. Troy, New
York: Lighting Research Center Rensselaer Polytechnic Institute.
L. Edwards, P. Torcelli. 2002. A Literature Review of the Effects of Natural Light on Building
Occupants. Golden, Colorado: National Renewable Energy Laboratory.
Morrison Hershfield Ltd. 2011. ASHRAE 1365-RP Thermal Performance of Building Envelope
Construction Details for Mid- and High-Rise Buildings, Atlanta, GA: American Society of
Heating, Refrigerating and Air-Conditioning Engineers Inc.
Morrison Hershfield Ltd. 2014. Building Envelope Thermal Bridging Guide Analysis,
Applications, and Insights, Vancouver, BC.
http://www.bchydro.com/powersmart/business/programs/newconstruction.html?WT.mc_id=rd_oldhiperfbuildingprogramWT.mc_id=rd_construction
W. Knowles, B. Hubbs, P. Kernan, G. Finch. 2010. A New Paradigm for the Design of
Sustainable Buildings. BEST 2 Conference Energy Efficiency EE3.
21