RF Design and Test Using

MATLAB and NI Tools

Tim Reeves – treeves@mathworks.com

Chen Chang - chen.chang@ni.com

© 2015 The MathWorks, Inc.

1
What are we going to talk about?

▪ How MATLAB and Simulink can be used in a wireless system design

workflow
▪ Wireless Scenario Simulation
▪ End-to-end Simulation of mmWave Communication Systems with Hybrid
Beamforming
▪ Developing Power Amplifier models and DPD algorithms in MATLAB
▪ Use of National Instruments PXI for PA characterization with DPD

2
Common Platform for 5G Development
Mobile and Connectivity Standards Unified Design and Simulation

Baseband RF MIMO &

PHY Front End Antennas

Deep Channels &

C-V2X
Learning Propagation

Prototyping and Testing Workflows

OTA Model-
Deploy to
Waveform Based
C/C++
Tx/Rx Design

3
What differentiates high data rate 5G systems from previous
wireless system iterations?
▪ High data rates (>1 Gbps) requires use of previously “under-used” (mmWave) frequency
bands

▪ mmWave requires MIMO architectures to achieve same performance as sub-6GHz

– Lower device power and high channel attenuation

▪ Antenna array, RF, and digital signal processing cannot be designed separately!
– Large communication bandwidth → digital signal processing is challenging
– High-throughput DSP → linearity requirements imposed over large bandwidth
– Wavelength ~ 1mm → small devices, many antennas packed in small areas

4
How is the presentation set up?
Link Level Modeling
Scenario Modeling TRANSMITTER

Digital
Baseband DAC PA
Front End
Channel

Digital PHY RF Front End Antenna

Digital
Baseband ADC LNA
Front End

RECEIVER

Hardware
5
What is the most basic way we can look at a wireless link?

Scenario Modeling
▪ Scenario Level Modeling
– RF propagation
– Multi-transmitter scenarios
– Coverage

6
What relevant items need to be included to analyze a realistic
5G coverage scenario?
• Multiple Transmitter Scenario for analyzing SINR

▪ Frequency = 4GHz
▪ Model 19 adjacent cells
▪ TX power = 44dBm
▪ Each cell has 3 sectors
▪ Antenna height = 25m

7
What are the different scenarios that can be analyzed?

▪ Select unique RF propagation scenarios ▪ Choose different antenna elements and

such as ‘Close-in’ and ‘Rain’ propagation array configurations to maximize
models. coverage.

8
What are the different use cases for Antenna Toolbox?

Antenna Element and Array Design Visualization and Analysis of 3rd party
Antenna Data

RF Propagation Visualization and Analysis

9
What type of fidelity do we want to add to a physical layer
model? Link Level Modeling
TRANSMITTER
▪ RF Front End
– Noise budget Baseband
Digital
DAC PA
Front End
– Gain Channel

Digital PHY RF Front End Antenna

– Non-linearity
Digital
– Tx linearization Baseband
Front End
ADC LNA

▪ Antennas RECEIVER

– Arrays
– Beamforming
– Propagation effects
– Loading

10
Why do link level modeling for a 5G mmWave system?

11
What needs to be included in a 5G system model to describe
typical operation?
▪ Include fidelity that comprises of array behavior, channel modeling, spatial
multiplexing and pre-coding and basic hybrid beamforming

System level design considerations

Design an array Add channel model Spatial multiplexing/ Hybrid beamforming

Precoding

12
What comprises the behavior between the Tx and Rx
antenna?
• Channel and RF propagation behavior

Signal Attenuation Wideband performance Scatter-rich propagation

13
What is Hybrid Beamforming?

RF RF

Baseband H Baseband

RF RF

Beamforming done in two stages:

– RF Beamforming (phase shifters in RF front ends)
– Digital Beamforming (digital filtering of baseband signal)

14
Why do you want to add RF (System-Level) models to your
PHY layer model?
▪ Design the architecture and define the specs of the RF components
▪ Integrate RF front ends with adaptive algorithms such as DPD, AGC, beamforming
▪ Test and debug the implementation of the transceiver before going in the lab
▪ Use models and measured data to gain insights in your design
▪ Provide a model of the RF transceiver to your colleagues and customers

15
Circuit Envelope to Trade-off Fidelity and Speed

Equivalent Baseband
Simulation speed

Spectrum
freq
Carrier

Circuit Envelope

Spectrum
freq
DC Carrier 1 Carrier 2
True Pass-Band

Spectrum
freq

Modeling fidelity
16
PA Linearization: Digital Pre Distortion (DPD) in Practice
Up-conversion
Baseband

DPD PA
RF Antenna loading
Adaptive coefficients
Timing

Down-conversion
PA characteristic
(actual)
Pout [dBm]

Compression

Memory

DPD characteristic

Pin [dBm] 17
PA Modeling Workflow

▪ Get I/Q (time domain, wideband) measurement data from your PA

▪ Fit the data with a memory polynomial (extract the coefficients) using MATLAB
▪ Verify the quality of the polynomial fitting (time, frequency)

Memory length →
Order →

18
What resources are available to characterize a PA Model?

MATLAB fitting procedure PA model coefficients

PA Data (White box)

PA model for circuit

envelope simulation

19
Why is static DPD modeling not enough for 5G systems?

▪ Circuit Envelope for fast RF simulation

▪ Low-power RF and analog components
– Up-conversion / down-conversion
– Antenna load
▪ Digital signal processing algorithm: DPD

20
Real-Life Example: AD9371 Transmitter + Observer

21
From Simulation to Implementation: HDL Code Generation

Automatically generate synthesizable HDL (Verilog / VHDL) code

▪ Make your model hardware “friendly”
▪ Estimate utilized resources
▪ Optimize model and generated code (speed, cost)
▪ Target FPGAs for rapid prototyping

22
How do we transition from software models to hardware?

▪ Implementing DPD in hardware

– Data streaming
– Prototype on hardware

Hardware
23
Connecting System-Level Models to Hardware for
Design and Verification

24
NI Front-End Module Test With DPD

▪ VST with 1 GHz instantaneous generation and analysis bandwidth

▪ Free NI-RFmx SpecAn with LUT, MPM, and GMP DPD models
▪ Free RFIC Test Software with DPD automation examples
PXI System
Front End Module
Front-End Module
▪ Generate reference waveform and Digital

1 acquire distorted waveform SMU

LNA
▪ Create predistortion model by VSA

2 comparing reference waveform to

distorted waveform
VST

3▪ Apply DPD to reference waveform using PA

predistortion model
Scope
4▪ Generate predistorted waveform and
Power PMIC ET
make measurements AWG
Modulator Power Supply

25
Traditional T&M Setup for MATLAB Based PA
Characterization with DPD Algorithm Running in MATLAB
▪ Familiar user experience for many engineers
▪ Slower measurement speed, Large physical footprint
▪ Expensive to upgrade or replace – even Software
▪ Difficult to synchronize for ET & DPD
▪ Tradeoffs between speed and accuracy

26
NI PXI Setup for MATLAB Based PA Characterization
with DPD & ET Algorithm Running in MATLAB
▪ Similar user experience as box-instruments
▪ Faster and FPGA-accelerated measurement speed, at
a fraction of the physical footprint
▪ Modularity for incremental upgrades
▪ Native synchronization technologies at sub nanosecond
accuracy
▪ R&D grade measurement accuracy with production test
speed

27
Enabling Integrated Semi PA Design & Validation Flow
Between LabVIEW & MATLAB
Design Validation
(MATLAB) (LabVIEW) Design V&V
(Sim-only) (T&M Only)
Stimuli Stimuli
Waveform LabVIEW
Generation MATLAB
RFmx
DPD MATLAB RFmx +
DPD DPD Algorithm (Custom) NanoSemi
DUT Sim Model Real
Waveform LabVIEW
MATLAB
DUT DUT Analysis RFmx
GUI LabVIEW
environment MATLAB
RFIC
Analysis Analysis

28
Enabling Integrated Semi PA Design & Validation Flow
Between LabVIEW & MATLAB
Design Validation
(MATLAB) (LabVIEW) Design V&V Design
(Sim-only) (T&M Only) (Integrated)
Stimuli Stimuli
Waveform LabVIEW
Generation MATLAB MATLAB
RFmx
DPD MATLAB RFmx + MATLAB
DPD DPD Algorithm (Custom) NanoSemi (Custom)
DUT Sim Model Real Real
Waveform LabVIEW
MATLAB MATLAB
DUT RFmx .NET API DUT Analysis RFmx
GUI LabVIEW
environment MATLAB MATLAB
RFIC
Analysis Analysis

29
Enabling Integrated Semi PA Design & Validation Flow
Between LabVIEW & MATLAB
Design Validation
(MATLAB) (LabVIEW) Design V&V Design V&V
(Sim-only) (T&M Only) (Integrated) (Integrated)
Stimuli Stimuli
Waveform LabVIEW LabVIEW
Generation MATLAB MATLAB
RFmx RFmx

LabVIEW MATLAB DPD MATLAB RFmx + MATLAB MATLAB

DPD Script Node DPD Algorithm (Custom) NanoSemi (Custom) (Custom)
DUT Sim Model Real Real Real
Waveform LabVIEW LabVIEW
MATLAB MATLAB
DUT DUT Analysis RFmx RFmx
GUI LabVIEW LabVIEW
environment MATLAB MATLAB
RFIC RFIC
Analysis Analysis

30
High-Power PA w/ DPD HW Demo Setup PXIe-1078 Chassis
PXIe-8840 Controller
PXIe-5840 VST
PXIe-4112 Power Supply

31
PA Design Engineer’s View in MATLAB

32
Validation Engineer’s View in LabVIEW

33
 Two Distinct Approaches to PA Characterization

Traditional Approach Platform-Based Approach

▪ Separate workflow for design and validation ▪ Integrated workflow for design and validation
▪ Different waveforms, PA models, analysis algorithm ▪ Same waveforms, PA models, analysis algorithm
▪ Expensive, large footprint, poor synchronization ▪ Modular, small footprint, sub-nanosecond
synchronization

34
35
 Qualcomm UK Uses MATLAB to Develop 5G RF NanoSemi Improves System Efficiency
Front-End Components and Algorithms for 5G and Other RF Products

Challenge Challenge
10x more waveform combinations in 5G than in LTE, Accelerate design and verification of RF power amplifier
making device validation much more complex and time- linearization algorithms used in 5G and Wi-Fi 6 devices
consuming
Solution
Solution Use MATLAB to characterize amplifier performance,
Use MATLAB to simulate hardware-accurate Tx and Rx develop predistortion and machine learning algorithms,
paths to predict system performance and optimize and automate standard-compliant test procedures
design parameters.
Results
Results ▪ Development time reduced by 50%
▪ Fully model RF transceiver and components ▪ Iterative verification process accelerated
▪ Securely release sensitive IP ▪ Early customer validation enabled
▪ Eliminate the cost of developing separate test suites
“With MATLAB, our team can deliver
“We use MATLAB models to
leading-edge IP faster, enabling our
optimize and verify the 5G RF
customers to increase bandwidth,
front end through all phases
push modulation rates higher, and
of development.”
reduce power consumption.”
Sean Lynch
Qualcomm 5G RF front end prototype
NanoSemi linearization IP development Nick Karter
Qualcomm UK, Ltd. and verification using MATLAB.
NanoSemi
36
Wrap up

▪ How MATLAB and Simulink can be used in a wireless system design

workflow
▪ Wireless Scenario Simulation
▪ End-to-end Simulation of mmWave Communication Systems with Hybrid
Beamforming
▪ Developing Power Amplifier models and DPD algorithms in MATLAB
▪ Use of National Instruments PXI for PA characterization with DPD

37
Learn More

▪ Where can you get more information about MathWorks tools for wireless
system modelling?

▪ MATLAB and Simulink for 5G Development

▪ White paper: RF PA and DPD linearization using MATLAB and Simulink

▪ White paper: Hybrid Beamforming for 5G Systems

38