Transcript
A Novel Wideband DPD Measurement Platform
Wideband Digital Pre-Distortion Measurement Platform for LTE/LTE-Advanced using Agilent SystemVue Jinbiao Xu, Agilent Technologies
Daren McClearnon, Agilent Technologies Oct, 2012
A Novel Wideband DPD Measurement
Digital Pre-Distortion (DPD): Problem Statement Agenda
1. Introduction and Problem Statement 2. Digital Pre-Distortion (DPD) Concepts 3. DPD verification with Agilent Hardware 4. DPD simulation with Agilent EDA Tools 5. Crest Factor Reduction (CFR) 6. PA Modeling 7. Summary
A Novel Wideband DPD Measurement 2
Digital Pre-Distortion (DPD): Problem Statement • Modern communication systems: • Signals have high peak-to-average power ratios (PAPR). • Must operate with high power-added efficiency (PAE).
• High PAPR is a consequence of high spectral efficiency • Multiple-Carrier Signals (MC GSM, MC WCDMA) • CDMA (WCDMA, CDMA2000) • OFDM (LTE, WiMAX)
• High PAE is achieved when the RF power amplifier (PA) is driven towards saturation
• Operation near saturation inherently results in higher signal distortion
A Novel Wideband DPD Measurement 3
DPD Problem Statement
Higher DC-RF Efficiency Increase Drive levels
Higher Peak Power Causes high distortion levels
Higher Spectral Efficiency “Back off” the drive levels
Conflicting requirements
How to handle signals with high PAPR, while driving the PA to operate with high PAE, while also having low signal distortion?
A Novel Wideband DPD Measurement 4
DPD Problem Statement High-efficiency PA design typically relies on: • Constraining the PA energy into a bandpass characteristic • Peaking the PA output energy by dynamically biasing the PA as a function of input power These techniques inherently result in PA nonlinearities with memory
Techniques to reduce distortion when operating at high efficiency: • PA linearization: Drive the PA closer to saturation, for a given level of distortion • Signal preconditioning: Reduce signal peaks without significant signal distortion
A Novel Wideband DPD Measurement 5
DPD Solution Approach CFR Higher DC-RF Efficiency Increase Drive levels
Higher Peak Power Causes high distortion levels
DPD
Higher Spectral Efficiency Higher throughput levels for subscribers
Solution: Preconditioning the signal (CFR) and correcting for the hardware (DPD) will both be discussed in this presentation
A Novel Wideband DPD Measurement 6
Agenda 1. Introduction and Problem Statement 2. Digital Pre-Distortion (DPD) Concepts
3. DPD verification with Agilent Hardware 4. DPD simulation with Agilent EDA Tools 5. Crest Factor Reduction (CFR) 6. PA Modeling 7. Summary
A Novel Wideband DPD Measurement 7
Digital Pre-distortion principles – compressing PA OUTPUT POWER
LINEAR GAIN
Psat Pdesired
PA, WITH GAIN COMPRESSION
Pactual
Pin
Pin needed to achieve Pdesired
INPUT POWER
A Novel Wideband DPD Measurement 8
Digital Pre-distortion principles – pre-expansion OUTPUT POWER
DPD GAIN EXPANSION
LINEAR GAIN
Psat PA, WITH GAIN COMPRESSION
+
LINEAR REGION
Maximum DPD REGION correctable power
INPUT POWER
A Novel Wideband DPD Measurement 9
Digital Pre-distortion principles – linearized result OUTPUT POWER
DPD GAIN EXPANSION
Psat
LINEARIZED DPD + PA PA, WITH GAIN COMPRESSION
+
LINEAR REGION
DPD REGION
Maximum correctable power
=
INPUT POWER
A Novel Wideband DPD Measurement 10
Linear Operation with time-varying envelope OUTPUT POWER
LINEAR GAIN
Psat
INPUT POWER
Peak-to-Avg Power Ratio (PAPR)
11
A Novel Wideband DPD Measurement
Nonlinear Operation – peaks are compressed
OUTPUT POWER
LINEAR GAIN
Psat
(compressed peaks)
INPUT POWER
CCDF (LTE)
Peak-to-Avg Power Ratio (PAPR)
12
A Novel Wideband DPD Measurement
DPD Pre-Expansion – peaks are exaggerated
OUTPUT POWER
LINEAR GAIN (expanded peaks)
Psat
INPUT POWER Possible Improvements
13
•
Compensate for artificially higher avg. signal power
•
Crest Factor Reduction (CFR)
A Novel Wideband DPD Measurement
DPD Net Result: Linear gain of complex-valued RF carrier envelope over a specific range of power levels RF Power Amplification
Baseband Digital Pre-Distortion OUTPUT POWER
LINEAR
LINEAR
DPD pre-expanded peaks
INPUT POWER
14
PA compresses peaks
INPUT POWER
A Novel Wideband DPD Measurement
What does a DPD look like? (Volterra Model) Volterra series pre-distorter can be described by Q
K
z ( n)
z k ( n)
where
Q
z k ( n)
m1 0
k 1
k
hk (m1 ,, mk ) mk 0
y(n ml ) l 1
Which is a 2-dimensional summation of power series & past time envelope responses Q
z (n) h0
Q
h1 (m1 ) y(n m1 ) m1 0
Q
h2 (m1 , m2 ) y(n m1 ) y(n m2 ) m1 0m2 0
A full Volterra produces a huge computational load. People usually simplify it into • Wiener model • Hammerstein model • Wiener-Hammerstein model • Memory polynomial model
15
A Novel Wideband DPD Measurement
DPD principles – Memory Polynomial Model If only diagonal terms are kept, Volterra reduces to “Memory polynomial” model.
Agilent uses the “Indirect Learning” algorithm to extract MP coefficients. As of SystemVue 2011.10, you can now add your own model, extraction algorithm, and even your own GUI. K
Q
z ( n)
a kq y(n q) y(n q)
k 1
k 1q 0
Where • •
K is Nonlinearity order Q is Memory length
L. Ding, G. T. Zhou, D. R. Morgan, Z. Ma, J. S. Kenney, J. Kim, and C. R. Giardina, “Memory polynomial predistorter based on the indirect learning architecture,” in Proc. of GLOBECOM, Taipei, Taiwan, 2002, vol. 1, pp. 967–971.
16
A Novel Wideband DPD Measurement
Wireless Transmitter Path – Where Is Your Product?
Env Tracking
BB PHY
CFR
DPD
DAC
Up convert
Adapt
ADC
Down convert
PA
Duplexer
• What is included with your actual product? • What IP do you have access to? Or, able to imitate? Able to modify? • What specification do you need to test against?
17
A Novel Wideband DPD Measurement
Agilent Measurement-based DPD Modeling Platform W1461 SystemVue W1918 LTE-A IP Library
BB TX PHY
Vector Signal Generator AWG ESG, MXG, PSG
Also: 3G, WLAN 60GHz, DVB, OFDM
W1716 DPD Step-byStep GUI
CFR
DPD model
DAC
Up convert
Generate Coefficients
ADC
Down convert
Throughput BER/FER ACPR EVM
PA
BB RX PHY
89600 VSA Optional Reference RX
18
Vector Signal Analyzer MXA, PXA, Modular
A Novel Wideband DPD Measurement
Measurement-Based DPD Modeling Flow Measured PA Input, Measured PA Output 1 Create DPD Stimulus
RF Input
2
3
4
5
19
RF Output
• DPD flow consists of 5 steps in SystemVue • Convergence improves with more iterations • 2-3 iterations are typical for real PAs
Get baseband complex waveforms of PA input and output Extract DPD Model (includes delay estimation and adjustment) Apply DPD Model, and Get DPD+PA Response Verify DPD Performance
A Novel Wideband DPD Measurement
Measurement-Based DPD Modeling Simplification: Calculated PA Input, Measured PA Output 1
•
Uses the Ideal BB stimulus waveform vs. measured PA output waveform to extract the DPD model.
•
• Advantages: - Single connection - PA remains “ON” - Easier to automate - Faster speed
•
Is typical of industry practice today
•
Linearizes the entire system, not just the PA
•
Provides very acceptable accuracy for quick Evaluation and MFG Test applications.
Create DPD Stimulus
RF Output
BB Input
2
3
4
5
20
Get baseband complex waveforms of PA input and output Extract DPD Model (includes delay estimation and adjustment) Apply DPD Model, and Get DPD+PA Response Verify DPD Performance
Assumptions: - Source flatness - Source linearity - No additional source signal conditioning
A Novel Wideband DPD Measurement
Simulation vs. Measurement DPD Extraction SIMULATION-BASED DPD ADS (predictive)
• ADS & GoldenGate Circuits as simulated RF DUTs - Complex loading, memory FX, dynamic behaviors • NVNA X-parameter measurement model, - Great for smaller solid-state devices
GG
CO-SIM, MODELS
CO-SIM, MODELS
X-parameters
MODEL
N5241,2 PNA-X
MEASUREMENT-BASED DPD
RF DUT
M9392A PXI VSA (>140MHz) or N9030A PXA (<140 MHz)
89600 VSA External Trigger
I,Q
RF Attenuator
M9330A AWG if > 100 MHz
21
N5182 MXG, or E8257D PSG as external modulator
RF DUT
A Novel Wideband DPD Measurement
Agilent Simulation-based DPD Modeling Platform W1461 SystemVue W1918 LTE-A IP Library
BB TX PHY
Agilent ADS Agilent GoldenGate RF circuit-level EDA software
Also: 3G, WLAN 60GHz, DVB, OFDM
W1716 DPD Step-byStep GUI
CFR
DPD model
DAC
Up convert
Generate Coefficients
ADC
Down convert
Throughput BER/FER ACPR EVM
PA
BB RX PHY
89600 VSA Optional Reference RX
22
A Novel Wideband DPD Measurement
Agenda 1. Introduction and Problem Statement 2. Digital Pre-Distortion (DPD) Concepts
3. DPD verification with Agilent Hardware 4. DPD simulation with Agilent EDA Tools 5. Crest Factor Reduction (CFR) 6. PA Modeling 7. Summary
23
A Novel Wideband DPD Measurement
Measurement-based DPD: Data capture 2 ways to transfer data from Instruments SystemVue Schematic sources grab live external waveforms at run-time into the DPD simulation 89600 VSA
gap
VSA_89600B_Source out
V1 {VSA_89600B_Source@Data Flow Models} VSATitle='Simulation output OutputType=Timed (Envelope/Real Baseband) VSATrace=B
COMMAND EXPERT
CommandExpertLink
C1 {CommandExpertLink@Data Flow Models}
Two methods to capture PA response data from vector signal analyzer (MXA/PXA, or PXI modular) 1) 89600 VSA software (convenient, but added cost) 2) Command Expert (free but requires customization)
24
A Novel Wideband DPD Measurement
DPD Measurement Automation: 2 Approaches Method 1 – Measure both PA Input and Output signals 1 Create DPD Stimulus
RF Input DC Power Analyzer
Adjust current to control switch
MXA / PXA
MXG
2
3
THRU DUT Power Splitter
• • • •
25
RF Output
Switch
4
Get baseband complex waveforms of PA input and output Extract DPD Model (includes delay estimation and adjustment) Apply DPD Model, and Get DPD+PA Response
Set the parameters in SystemVue 5 Verify DPD Performance Click “Go” in the script file. The DPD extraction process runs automatically. After DPD measurement, verify EVM, ACP vs Output Power.
A Novel Wideband DPD Measurement
DPD Measurement Automation: 2 Approaches Method 2 – Calculate PA Input, Measure PA Output 1 Create DPD Stimulus
RF Output
BB Input
2
3
MXG
MXA / PXA
Get baseband complex waveforms of PA input and output Extract DPD Model (includes delay estimation and adjustment)
DUT 4
Single connection allows automation, iterations Eliminates one measurement, physically faster Identical extraction algorithms, verification process
26
5
Apply DPD Model, and Get DPD+PA Response
Verify DPD Performance
A Novel Wideband DPD Measurement
Comparing Methods: BB Input vs. Measured RF 6-Carrier GSM
LTE-Advanced DL (20 MHz)
ACLR of DL 20 MHz System
ACLR
-2BW Lower
-1BW Lower
+1BW Upper
+2BW Upper
BB input
Raw PA output
54.06
35.33
35.68
53.58
Measured RF input
DPD+PA w/ BB input
55.05
50.15
52.28
54.59
DPD+PA w/ PA input
55.80
51.23
54.32
55.41
27
DPD+PA RESULTS
A Novel Wideband DPD Measurement
SystemVue DPD Modeling Flow for LTE/LTE-A Step 1. Create DPD stimulus waveform • •
Set LTE parameters such as BW, Resource Block allocation and others Choose between built-in LTE or LTE-Advanced waveform generation The download power and length of the waveform can also be set.
28
A Novel Wideband DPD Measurement
SystemVue DPD Modeling Flow for LTE/LTE-A Step 2. Capture PA response • •
SystemVue downloads directly to the MXG or M9330A AWG (source), and capture data back from PXA or M9392A (analyzer). Equipment parameters such as number of signal, trace assignment, and file name can be set.
THRU : Connect the MXG/AWG directly to the PXA/M9392A and click the “Capture Waveform” button. This is the true RF PA input.
DUT: Connect the MXG to the PA, connect the PA to the PXA/M9392A, and click the “Capture Waveform” button. The captured signal is the output of the PA DUT. The measured I/Q files are stored and used in following steps.
29
A Novel Wideband DPD Measurement
SystemVue DPD Modeling Flow for LTE/LTE-A Step 3. DPD Model Extraction •
PA AM-to-AM Characteristic
DPD model parameters such as number of training samples, memory order, and nonlinear order can be set.
DPD AM-to-AM Characteristic
A Novel Wideband DPD Measurement 30
SystemVue DPD Modeling Flow for LTE/LTE-A Step 4. Capture DPD+PA Response •
The signal is pre-distorted by the DPD model and re-downloaded into the MXG or AWG.
DPD+PA (measured RF output) PA input (original RF input)
Set the RF power DPD+PA AM-to-AM Characteristic
A Novel Wideband DPD Measurement 31
SystemVue DPD Modeling Flow for LTE/LTE-A Step 5. Verify DPD+PA response •
LTE performance for the DPD model used with the PA hardware is verified.
Spectrum, EVM and ACLR are calculated and plotted automatically
A Novel Wideband DPD Measurement 32
Accommodating Proprietary IP • • • •
Use your own extractor IP instead of Agilent’s Continue to enjoy an integrated environment Allows remote & distributed DPD teamwork Greater user control of algorithm details, IP security, performance, delivery date, quality, etc
Custom DPD Model Extraction (.m math language)
Custom Digital Pre-distorter (.m math language)
A Novel Wideband DPD Measurement 33
DPD of LTE-Advanced DL with Doherty PA (50W) Spectrum, ACLR and EVM results (5 MHz DL System) Raw PA output PA+DPD, after 1 iteration to extract DPD coefficients
ACLR (dB) ACLR
-2BW Lower
-1BW Lower
+1BW Upper
+2BW Upper
RF input (HW)
61.75
53.01
53.52
62.33
Raw PA output
50.25
31.98
31.56
48.19
DPD+PA output
57.96
49.00
48.63
58.57
EVM EVM (dB) Input signal
-23.44
Raw PA output
-21.33
DPD+PA output
-23.36
CFR was applied to this LTE-Advanced DL signal , with a maximum EVM target of 8%.
34
Vector Source：MXG Vector Analyzer: PXA
A Novel Wideband DPD Measurement
DPD of LTE-Advanced DL with LDMOS Doherty PA (200W) Spectrum, ACLR and EVM results (10 MHz DL System) Raw PA output PA+DPD, after 1 iteration to extract DPD coefficients
ACLR (dB) ACLR
-2BW Lower
-1BW Lower
+1BW Upper
+2BW Upper
BB input (sim)
58.67
49.63
49.17
58.01
Raw PA output
49.90
28.69
28.35
47.31
DPD+PA output
48.88
45.10
45.16
48.83
EVM EVM (%)
EVM (dB)
Simulation BB input
5.33
-24.46
Raw PA output
10.13
-19.89
DPD+PA output
5.52
-25.16
Vector Source：MXG Vector Analyzer: PXA
CFR was applied to this LTE-Advanced DL signal, with a maximum EVM target of 10% for 16-QAM.
35
A Novel Wideband DPD Measurement
DPD of LTE-Advanced DL with LDMOS Doherty PA (200W) Spectrum, ACLR and EVM results (20MHz DL System) ACLR (dB) ACLR
-2BW Lower
-1BW Lower
+1BW Upper
+2BW Upper
BB input (sim)
64.73
55.09
57.10
64.92
Raw PA output
51.01
30.69
30.04
49.50
DPD+PA output
50.31
45.16
45.56
51.40
Raw PA output PA+DPD, after 1 iteration to extract DPD coefficients
EVM EVM (%)
EVM (dB)
BB input signal (sim)
6.10
-24.28
Raw PA output
8.87
-21.04
DPD+PA output
6.88
-23.24
Vector Source：MXG Vector Analyzer: PXA
CFR was applied to this LTE-Advanced DL signal with a maximum EVM target of 10%,8% and 6% for QPSK, 16-QAM and 64-QAM, respectively.
36
A Novel Wideband DPD Measurement
LTE-A Results with 200W LDMOS Doherty PA Raw PA Output (DL 20MHz System)
37
A Novel Wideband DPD Measurement
LTE-A Results with 200W LDMOS Doherty PA DPD+PA Output (DL 20MHz System)
38
A Novel Wideband DPD Measurement
DPD of LTE-Advanced DL with LDMOS Doherty PA (200W) Results with (2x10MHz) Carrier Aggregation of 2 separate CC’s ACLR (dB) ACLR -2BW Lower
-1BW Lower
+1BW Upper
+2BW Upper
BB input (sim)
63.11
56.75
56.70
62.72
Raw PA output
50.58
30.80
30.22
49.06
DPD+PA output
51.74
45.75
45.73
51.18
Raw PA output PA+DPD, after 1 iteration to extract DPD coefficients
CC0 EVM (QPSK) EVM (%) EVM (dB) Baseband signal (sim)
0.21
-53.43
Raw PA output
3.03
-30.37
DPD+PA output
1.93
-34.28
CC1 EVM (16-QAM) EVM (%) EVM (dB)
39
Baseband signal (sim)
0.20
-54.11
Raw PA output
3.12
-30.11
DPD+PA output
1.93
-34.31
Vector Source：MXG Vector Analyzer: PXA
A Novel Wideband DPD Measurement
Multi-Standard Radio (MSR) into LDMOS Doherty PA (200W) 2 Carriers GSM
2 Carriers WCDMA
2 Carriers LTE
2 Carriers EDGE
Raw PA output PA+DPD
40
A Novel Wideband DPD Measurement
Wideband configurations: LTE-A 2x20MHz + 1x20MHz CA
Agilent M9330A AWG, M9392A VSA Source = M9330A AWG N5182 MXG Vector Analyzer= M9392A - 12bits ADC - up to 250MHz bandwidth PA output Spectrum (Blue) PA+DPD Spectrum (Red) PA input Spectrum (Green)
41
A Novel Wideband DPD Measurement
DPD of 802.11ac, using M9330A/M9392A (80MHz signal, with 3x oversampling = 240 MHz VSA BW)
42
A Novel Wideband DPD Measurement
Agenda 1. Introduction and Problem Statement 2. Digital Pre-Distortion (DPD) Concepts 3. DPD verification with Agilent Hardware 4. DPD simulation with Agilent EDA Tools 5. Crest Factor Reduction (CFR) 6. PA Modeling 7. Summary
43
A Novel Wideband DPD Measurement
DPD with Agilent EEsof EDA tools Predictive PA modeling and linearization Benefits of using RF Simulation for DPD • • • •
Predict the final DPD result, while Analog PA can still be changed De-risk module or wafer iteration, to save time and money Explore vendors, waveforms, statistical spreads, analog variables Validate system-level specifications with preliminary RF & BB
Trade offs: • Accuracy. Dynamic “circuit envelope” behavior depends on – the simulation engine (and any behavioral modeling) – the device-level transistor models, for traps, self-heating, mismatch
• Speed. – Real HW measurements >> faster than Simulations
Conclusion: it is still worth doing 44
A Novel Wideband DPD Measurement
Simulation-based, predictive DPD SystemVue co-simulation with circuit-level PA in ADS SystemVue
SystemVue
ADS Ptolemy
STIMULUS
RESPONSE
(circuit-system co-simulation)
CO-SIM
ADS reads data from SystemVue
CO-SIM
ADS circuit-level PA (circuit envelope simulation)
ADS sends data to SystemVue
ADS circuit-level PA，needs Circuit Envelop to co-simulate with SystemVue.
45
A Novel Wideband DPD Measurement
Simulation-based, predictive DPD SystemVue co-simulation with circuit-level PA in ADS Spect r um Anal yzer
Extract Capture PA input vs. output waveforms for DPD extraction
Spect r um Anal yzer
PAIn_spec Mode=TimeGate Start=0s SegmentTime=50μs
SystemVue
PAOut_spec Mode=TimeGate Start=0s SegmentTime=50μs
PA
R2 {ReadFile@Data Flow Models} File='Step1_BBData_Imag_Iter2.txt [Step1_BBData_Imag_FileName ] Periodic=YES
Fc
T
Im
123
Cx
Fc Env
Env
Im
PAInputData_Imag2 {Sink@Data Flow Models} StartStopOption=Samples SampleStart=0 SampleStop=99999 [NumOfCapturedSamples-1] DataFileName='Step2_PAOutputdata_Imag_Iter2.txt [filename_PAout_Q]
Cx
Re ADS Cosim R3 {RectToCx@Data Flow Models}
S1 {SetSampleRate@Data Flow Models} SampleRate=69.33e+6Hz [SamplingRate]
C1 {CxToEnv@Data Flow Models} Fc=2.505e+9Hz [FCarrier]
Re
A1 {ADSCosimBlockEnv@Data Flow Models} OutputFc=2.505e+9 [FCarrier] InputBlockSize=1000 [BlockSize] OutputBlockSize=1000 [BlockSize] InputID='SystemVueToADS OutputID='ADSToSystemVue
E1 {EnvToCx@Data Flow Models}
123
C7 {CxToRect@Data Flow Models}
PAInputData_Real2 {Sink@Data Flow Models} StartStopOption=Samples SampleStart=0 SampleStop=99999 [NumOfCapturedSamples-1] DataFileName='Step2_PAOutputdata_Real_Iter2.txt [filename_PAout_I]
R1 {ReadFile@Data Flow Models} File='Step1_BBData_Real_Iter2.txt [Step1_BBData_Real_FileName] Periodic=YES
123 Im
PAInputData_Imag1 {Sink@Data Flow Models} StartStopOption=Samples SampleStart=0 SampleStop=99999 [NumOfCapturedSamples-1] DataFileName='Step2_PAInputdata_Imag_Iter2.txt [filename_PAin_Q]
The UI to connect with ADS in 123 SystemVue, corresponding to the ADS schematic (Ptolemy co-sim with circuit-level design) in ADS.with digital ADS Generate PA output pre-distorter Re
C2 {CxToRect@Data Flow Models}
PAInputData_Real1 {Sink@Data Flow Models} StartStopOption=Samples SampleStart=0 SampleStop=99999 [NumOfCapturedSamples-1] DataFileName='Step2_PAInputdata_Real_Iter2.txt [filename_PAin_I]
R5 Im Fi l e =' Ste p 3 _ DPD_ Co e ffi c i e n ts _ Im a g _ Ite r2 .tx t [DPD_ Co e ffi c i e n ts _ Im a g _ Fi l e Name] Peri o d i c =YES Re
DPD
R6
Verify See linearized result, including DPD
46
SystemVue
Spect r um Anal yz r e
123 Afte rDPD M o d e =Ti m e Ga te Sta rt=0 s Seg m e n tTi m e =5 0 μs
R4 Fi l e =' Ste p 3 _ DPD_ Co e ffi c i e n ts _ Re a l _ Ite r2 .tx t [DPD_ Co e ffi c i e n ts _ Re a l _ Fi l e Na me] Peri o d i c =YES
DPD_ PAOu tp u tDa ta _ Im a g Sta rtSto p Op ti o n =Sam p l e s Sam p l e Sta rt=0 [Sta rtSam p l e ] Sam p l e Sto p =9 9 9 9 8 [Sto p Sam p l e - 1] Da ta Fi l e Na m e =' Ste p 4 _ DPD_ PAOu tp u td a ta _ Im a g _ Ite r2 .tx t [Ste p 4 _ DPD_ PAOu tp u t_ Im a g _ Fi l e Na m e ]
Im
Fc
T
DPD_Coe f
R8 Fi l e =' Ste p 1 _ BBDa ta _ Im a g _ Ite r1 .tx t [Ste p 1 _ BBDa ta _ Im a g _ Fi l e Na me]
Cx
DPD_O ut p t u DPD_Pr eDist or t er DPD_I npu t
Fc Env
Env
Im Cx
ADS Co s i m
Re
R1
G1 Ga i n =0 .9 7 8 [Powe rAl i g n m e n t]
D1 M e m o ry Ord e r=5 [M e m o ry Ord e r]
S4 Sam p l e Ra te =6 9 .3 3 e +6 Hz [Sam p l i n g Ra te]
No n l i n e a rOrd e r=9 [No n l i n e a rOrd e r] Nu m OfIn p u tSam p l e s =2 0 0 0 0 [Nu m OfIn p u tSam p l e s]
Fc Spect r um Anal yz r e
Cx
A1 Ou tp u tFc =2 .5 0 5 e +9 [FCa rri e r]
PA
In p u tBl o c k Si z e =1 0 0 0 [Bl o c k Size] Ou tp u tBl o c k Si z e =1 0 0 0 [Bl o c k Si ze] In p u tID=' Sy s te m Vue To ADS Ou tp u tID=' ADSTo Sy s te m Vue
R7 Fi l e =' Ste p 1 _ BBDa ta _ Re a l _ Ite r1 .tx t [Ste p 1 _ BBDa ta _ Re a l _ Fi l e Na m e]
T
C2 Fc =2 .5 0 5 GHz [FCa rri e r]
Re
Env
E2
C1
123 DPD_ PAOu tp u tDa ta _ Re a l Sta rtSto p Op ti o n =Sam p l e s Sam p l e Sta rt=0 [Sta rtSam p l e ] Sam p l e Sto p =9 9 9 9 8 [Sto p Sam p l e - 1] Da ta Fi l e Na m e =' Ste p 4 _ DPD_ PAOu tp u td a ta _ Re a l _ Ite r2 .tx t [Ste p 4 _ DPD_ PAOu tp u t_ Re a l _ Fi l e Na m e ]
S1 Sam p l e Ra te =6 9 .3 3 e +6 Hz [Sam p l i n g Ra te]
C3 Fc =2 .5 0 5 GHz [FCa rri e r]
Befo re DPD M o d e =Ti m e Ga te Sta rt=0 s Seg m e n tTi m e =5 0 μs
A Novel Wideband DPD Measurement
Simulation-based, predictive DPD SystemVue co-simulation with circuit-level PA in ADS
6-Carrier GSM Carrier Spacing: 4MHz Sampling Rate： 256 * 270.8333kHz =69.3333 MHz
40dB improvement after 2 iterations
PA input Spectrum (Green) PA output Spectrum (Blue) PA+DPD Spectrum (Red, first iteration)) PA+DPD Spectrum (Orange, Second iteration) 47
A Novel Wideband DPD Measurement
Simulation-based, predictive DPD SystemVue with native FCE model, extracted from GoldenGate Spect r um Anal yzer
Extract Capture PA input vs. output waveforms for DPD extraction
Spect r um Anal yzer
PAIn_spec Mode=TimeGate Start=0s SegmentTime=50μs
PAOut_spec Mode=TimeGate Start=0s SegmentTime=50μs
PA
R2 {ReadFile@Data Flow Models} File='Step1_BBData_Imag_Iter2.txt [Step1_BBData_Imag_FileName ] Periodic=YES
Fc
T
Im
123 Fc
Cx
Env
Env
Re
Im
PAInputData_Imag2 {Sink@Data Flow Models} StartStopOption=Samples SampleStart=0 SampleStop=99999 [NumOfCapturedSamples-1] DataFileName='Step2_PAOutputdata_Imag_Iter2.txt [filename_PAout_Q]
Cx
Fast Cir cuit Envelope
Re R3 {RectToCx@Data Flow Models}
S1 {SetSampleRate@Data Flow Models} SampleRate=34.67e+6Hz [SamplingRate]
C1 {CxToEnv@Data Flow Models} Fc=2.505e+9Hz [FCarrier]
F2 {FastCircuitEnvelope@Data Flow Models} File='SIM_PA_2p505GHz_m7dBm_level3_ampaccu...
E1 {EnvToCx@Data Flow Models}
PAInputData_Real2 {Sink@Data Flow Models} StartStopOption=Samples SampleStart=0 SampleStop=99999 [NumOfCapturedSamples-1] DataFileName='Step2_PAOutputdata_Real_Iter2.txt [filename_PAout_I]
123 R1 {ReadFile@Data Flow Models} File='Step1_BBData_Real_Iter2.txt [Step1_BBData_Real_FileName] Periodic=YES
Im
123
C7 {CxToRect@Data Flow Models}
PAInputData_Imag1 {Sink@Data Flow Models} StartStopOption=Samples SampleStart=0 SampleStop=99999 [NumOfCapturedSamples-1] DataFileName='Step2_PAInputdata_Imag_Iter2.txt [filename_PAin_Q]
Re
123 CMOS Handset PA C2 {CxToRect@Data Flow Models}
PAInputData_Real1 {Sink@Data Flow Models} StartStopOption=Samples SampleStart=0 SampleStop=99999 [NumOfCapturedSamples-1] DataFileName='Step2_PAInputdata_Real_Iter2.txt [filename_PAin_I]
Fast Circuit Envelope (FCE) model extracted from GoldenGate Generate PA output with digital pre-distorter (direct co-sim is also possible, but slower)
R5 Im Fi l e =' Ste p 3 _ DPD_ Co e ffi c i e n ts _ Im a g _ Ite r2 .tx t [DPD_ Co e ffi c i e n ts _ Im a g _ Fi leName] Pe ri o d i c =YES Re
Verify See linearized result, including DPD
48
PA
DPD
R6
Spect r um Anal yz r e
R4 Fi l e =' Ste p 3 _ DPD_ Co e ffi c i e n ts _ Re a l _ Ite r2 .tx t [DPD_ Co e ffi c i e n ts _ Re a l _ Fi l eName] Pe ri o d i c =YES
DPD_O ut p t u
Im
Cx
DPD_IDnp PD t _Pr eDist or t er u
Sta rtSto p Op ti o n =Sa m p l es Sa m p l e Sta rt=0 [Sta rtSa m p l e] Sa m p l e Sto p =9 9 9 9 8 [Sto p Sa m p l e- 1] Da ta Fi l e Na m e =' Ste p 4 _ DPD_ PAOu tp u td a ta _ Im a g _ Ite r2 .tx t [Ste p 4 _ DPD_ PAOu tp u t_ Im a g _ Fi l e Name]
Fc
T
DPD_Coe f
R8 Fi l e =' Ste p 1 _ BBDa ta _ Im a g _ Ite r1 .tx t [Ste p 1 _ BBDa ta _ Im a g _ Fi l e Name]
123 DPD_ PAOu tp u tDa ta _ Im a g
Afte rDPD M o d e =Ti m e Ga te Sta rt=0 s Se g m e n tTi m e =5 0μs
Fc Env
Fast Cir cuit Envelo e p
Env
Im Cx
Re
Re
R1
G1 Ga i n =0 .8 2 1 [Po we rAl i g n m e n t]
D1 M e m o ry Ord e r=5 [M e m o ry Ord e r]
S4 Sa m p l e Ra te =3 4 .6 7 e +6 Hz [Sa m p l i n g Ra te]
No n l i n e a rOrd e r=9 [No n l i n e a rOrd e r] Nu m OfIn p u tSa m p l e s =3 0 0 0 0 [Nu m OfIn p u tSa m p les]
Fc Spect r um Anal yz r e
Cx
F2 Fi l e =' SIM _ PA_ 2 p 5 0 5 GHz _ m 7 d Bm _ l e v e l 3 _ a m p a ccu.
E2
C1
123
R7 Fi l e =' Ste p 1 _ BBDa ta _ Re a l _ Ite r1 .tx t [Ste p 1 _ BBDa ta _ Re a l _ Fi l e Name]
T
C2 Fc =2 .5 0 5 GHz [FCa rri e r]
Env
DPD_ PAOu tp u tDa ta _ Re a l Sta rtSto p Op ti o n =Sa m p l es Sa m p l e Sta rt=0 [Sta rtSa m p l e] Sa m p l e Sto p =9 9 9 9 8 [Sto p Sa m p l e- 1] Da ta Fi l e Na m e =' Ste p 4 _ DPD_ PAOu tp u td a ta _ Re a l _ Ite r2 .tx t [Ste p 4 _ DPD_ PAOu tp u t_ Re a l _ Fi l e Na me]
Be fo re DPD S1 Sa m p l e Ra te =3 4 .6 7 e +6 Hz [Sa m p l i n g Ra te]
C3 Fc =2 .5 0 5 GHz [FCa rri e r]
M o d e =Ti m e Ga te Sta rt=0 s Se g m e n tTi m e =5 0μs
A Novel Wideband DPD Measurement
Simulation-based, predictive DPD SystemVue with native FCE model, extracted from GoldenGate
6-Carrier GSM Carrier Spacing: 600kHz
30dB improvement after 2 iterations
Sampling Rate： 128 * 270.8333kHz =34.6667 MHz
PA input Spectrum (Green) PA output Spectrum (Blue) PA+DPD Spectrum (Red, first iteration)) PA+DPD Spectrum (Orange, Second iteration)
49
A Novel Wideband DPD Measurement
Simulation-based, predictive DPD SystemVue with analog X-parameter model (100W PA)
Port_2 {*OUT} ZO=50Ω
MultiSource_3 {MultiSource} Source1=200 MHz at -10 dBm 1 3 VDC
X
2
SG1 {VDC} VDC=22V XP_1 {XPARAMS} File='.\100W_3Hz_3H_3Bias_PHD.mdf {*GND}
FILE
Analog X-parameter device is placed into a Spectrasys subnetwork (RF simulation domain)
50
A Novel Wideband DPD Measurement
Simulation-based, predictive DPD SystemVue with analog X-parameter model (100W PA) X-param PA Spect r um Anal yzer
Extract Capture PA input vs. output waveforms for DPD extraction
PAIn_spec Mode=TimeGate Start=0s SegmentTime=50μs
R2 {ReadFile@Data Flow Models} File='Step1_BBData_Imag_Iter1.txt [Step1_BBData_Imag_FileName ] Periodic=YES
PAOut_spec Mode=TimeGate Start=0s SegmentTime=50μs
Fc
T
Im
Spect r um Anal yzer
Cx
123
Fc
RF_Link
Env
Env
Im
PAInputData_Imag2 {Sink@Data Flow Models} StartStopOption=Samples SampleStart=0 SampleStop=99999 [NumOfCapturedSamples-1] DataFileName='Step2_PAOutputdata_Imag_Iter1.txt [filename_PAout_Q]
Cx
Re Re R3 {RectToCx@Data Flow Models}
S1 {SetSampleRate@Data Flow Models} SampleRate=34.67e+6Hz [SamplingRate]
C1 {CxToEnv@Data Flow Models} Fc=200e+6Hz [FCarrier]
Data1 {RF_Link@Data Flow Models} Schematic='Xparam_device FreqSweepSetup=Automatic EnableNoise=NO CalcPhaseNoise=NO
R1 {ReadFile@Data Flow Models} File='Step1_BBData_Real_Iter1.txt [Step1_BBData_Real_FileName] Periodic=YES
E1 {EnvToCx@Data Flow Models}
C7 {CxToRect@Data Flow Models}
123
PAInputData_Real2 {Sink@Data Flow Models} StartStopOption=Samples SampleStart=0 SampleStop=99999 [NumOfCapturedSamples-1] DataFileName='Step2_PAOutputdata_Real_Iter1.txt [filename_PAout_I]
123 Im
PAInputData_Imag1 {Sink@Data Flow Models} StartStopOption=Samples SampleStart=0 SampleStop=99999 [NumOfCapturedSamples-1] DataFileName='Step2_PAInputdata_Imag_Iter1.txt [filename_PAin_Q]
Re C2 {CxToRect@Data Flow Models}
123
PAInputData_Real1 {Sink@Data Flow Models} StartStopOption=Samples SampleStart=0 SampleStop=99999 [NumOfCapturedSamples-1] DataFileName='Step2_PAInputdata_Real_Iter1.txt [filename_PAin_I]
RF_Link
Brings RF networks (incl. X-parameter devices) up to the dataflow simulation Generate PA output with digital pre-distorter R5 Im Fi l e =' Ste p 3 _ DPD_ Co e ffi c i e n ts _ Im a g _ Ite r1 .tx t [DPD_ Co e ffi c i e n ts _ Im a g _ Fi leName] Peri o d i c =YES Re
Verify See linearized result, including DPD
DPD
R6
DPD_ PAOu tp u tDa ta _ Im a g
R4 Fi l e =' Ste p 3 _ DPD_ Co e ffi c i e n ts _ Re a l _ Ite r1 .tx t [DPD_ Co e ffi c i e n ts _ Re a l _ Fi leName] Peri o d i c =YES R8 Fi l e =' Ste p 1 _ BBDa ta _ Im a g _ Ite r1 .tx t [Ste p 1 _ BBDa ta _ Im a g _ Fi l e Name] DPD_O utp t u
Cx
DPDD _IPD np t_Pr eDist or t er u
Sta rtSto p Op ti o n =Sam p l es Sam p l e Sta rt=0 [Sta rtSam p l e] Sam p l e Sto p =9 9 9 9 8 [Sto p Sam p l e- 1] Da ta Fi l e Na m e =' Ste p 4 _ DPD_ PAOu tp u td a ta _ Im a g _ Ite r1 .tx t [Ste p 4 _ DPD_ PAOu tp u t_ Im a g _ Fi l e Name]
Fc
T
DPD_Coe f
Im
Fc Env
RF_Link
Env
Im Cx
Re
Re
R1
R7 Fi l e =' Ste p 1 _ BBDa ta _ Re a l _ Ite r1 .tx t [Ste p 1 _ BBDa ta _ Re a l _ Fi l e Name]
G1 Ga i n =1 .1 4 [Powe rAl i g n m e n t]
D1 S4 M e m o ry Ord e r=3 [M e m o ry Ord e r] Sam p l e Ra te =3 4 .6 7 e +6 Hz [Sam p l i n g Rate] No n l i n e a rOrd e r=1 1 [No n l i n e a rOrd e r] Nu m OfIn p u tSam p l e s =2 0 0 0 0 [Nu m OfIn p u tSam p les]
Fc
T
Spect r um Anal yz r e
Cx
S1 Sam p l e Ra te =3 4 .6 7 e +6 Hz [Sam p l i n g Rate]
51
123
Spect r um Anal yz r e
Afte rDPD M o d e =Ti m e Ga te Sta rt=0s Seg m e n tTi m e =5 0μs
Env
C3 Fc =0 .2 GHz [FCa rri e r]
Befo re DPD M o d e =Ti m e Ga te Sta rt=0s Seg m e n tTi m e =5 0μs
C2 Fc =0 .2 GHz [FCa rri e r]
Da ta 1 Sc h e m a ti c =' Xpa ra m _ d evice Fre q Swe e p Setu p =Auto m a tic Ena b l e No i s e =NO Ca l c Pha s e No i s e =NO
X-param PA
E2
C1
123 DPD_ PAOu tp u tDa ta _ Re a l Sta rtSto p Op ti o n =Sam p l es Sam p l e Sta rt=0 [Sta rtSam p l e] Sam p l e Sto p =9 9 9 9 8 [Sto p Sam p l e- 1] Da ta Fi l e Na m e =' Ste p 4 _ DPD_ PAOu tp u td a ta _ Re a l _ Ite r1 .tx t [Ste p 4 _ DPD_ PAOu tp u t_ Re a l _ Fi l e Name]
A Novel Wideband DPD Measurement
Simulation-based, predictive DPD SystemVue with analog X-parameter model (100W PA)
6-Carrier GSM Carrier Spacing: 600kHz
~40dB improvement (w/o memory effects)
Sampling Rate： 128 * 270.8333kHz =34.6667 MHz
FILE
PA input Spectrum (Green) PA output Spectrum (Blue) PA+DPD Spectrum (Red, first iteration)) PA+DPD Spectrum (Orange, Second iteration)
52
A Novel Wideband DPD Measurement
DPD Modeling Simplification: Automation UI Measurement-based
MXG
DUT
MXA / PXA
GG co-sim (or FCE model)
FastCircuitEnvelope
ADS co-sim
ADS Cosim
53
Both DPD extractions share the same UI: • Measurement-based • Simulation-based
Verification of simulation-based DPD Sweep power, re-extract DPD at each point, watch EVM, ACP EVM vs. Output Power
ACP vs. Output Power Lower/Upper ACLR w/o DPD
EVM w/o DPD
ACLR with DPD
EVM with DPD
Input waveform: • • • •
IEEE 802.11ac, 5 GHz WLAN No CFR (PAPR is 8.7dB) Bandwidth = 80MHz system 4x Oversampling rate=320 MHz
Output < 0 dBm
0 < Output < +16.5 dBm
DPD offers little benefit
DPD offers significant benefit
Device Under Test: •
54
WLAN “FCE” model extracted from Agilent GoldenGate RFIC simulator
A Novel Wideband DPD Measurement
Verification of simulation-based DPD Sweep power, constant DPD coefficients, watch EVM, ACP Question: “Do I need Adaptive DPD?”
Lower/Upper ACLR w/o DPD EVM w/o DPD EVM with DPD
Different (or fewer) DPD coefficients needed
55
Useful Range for this set of DPD coefficients
PA may be less correctable
Lower/ Upper ACLR with DPD
ACP may actually be worse out of range: turn DPD off.
ACP satisfies a spectral compliance mask
High DC-RF efficiency but poor ACP
A Novel Wideband DPD Measurement
Verification of simulation-based DPD Sweep power, re-extract at each point, see final Pout vs. Pin
Power Output With DPD
Linear Gain = 25.5dB
Using Crest Factor Reduction (CFR) to reduce the peaks, the average signal level can be increased farther to the right, resulting in higher DCRF Efficiency, and longer distance coverage
Signal with PAPR = 8.7dB must be backed-off, lower average power Signal with PAPR = 7.5dB can be driven to higher average power
56
A Novel Wideband DPD Measurement
Memory Polynomial vs. Volterra DPD models 802.11ac 80MHz, FCE PA Model Co-sim Memory Polynomial (21 coefficients)
ACPR
57
Lower
Upper
EVM (dB)
Original input
-56.19
-57.20
-47.16
PA Output (No DPD)
-36.66
-38.43
DPD+PA Iter1
-50.28
-49.95
DPD+PA Iter2
-53.39
-52.18
Volterra Series (24 coefficients)
Lower
Upper
EVM (dB)
Original input
-56.19
-57.20
-47.16
-29.88
PA Output (No DPD)
-36.68
-38.45
-29.90
-42.20
DPD+PA Iter1
-51.60
-49.79
-42.90
DPD+PA Iter2
-54.05
-54.29
-46.06
DPD+PA Iter3
-54.71
-55.26
-46.40
-44.41
ACPR
A Novel Wideband DPD Measurement
Verification after DPD model extraction Verifying Memory Order and Nonlinear Order in Memory Polynomial
EVM vs. Memory Order (@Nonlinear Order=7)
EVM and ACP are stable when memory order>=3.
ACPR vs. Memory Order (@Nonlinear Order=7)
Memory effect almost removed when memory order >=3.
EVM vs. Nonlinear Order (@Memory Order=3)
58
EVM and ACP are stable when nonlinear order>=7.
ACPR vs. Nonlinear Order (@Memory Order=3)
A Novel Wideband DPD Measurement
Verification after DPD model extraction A closer look at ACPR vs. Nonlinear Order (“how many terms do I need?”)
-39dB
Nonlinear=9 (@Memory=3) -56dB
Order=3
59
Order=11 A Novel Wideband DPD Measurement
Verification after DPD model extraction A closer look at ACPR vs. Memory Order (“how many terms do I need?”)
-43dB
-54dB Memory=3 (@Nonlinear=7) Memoryless
60
Order=5 A Novel Wideband DPD Measurement
Agenda 1. Introduction and Problem Statement 2. Digital Pre-Distortion (DPD) Concepts 3. DPD verification with Agilent Hardware 4. DPD simulation with Agilent EDA Tools 5. Crest Factor Reduction (CFR) 6. PA Modeling 7. Summary
A Novel Wideband DPD Measurement 61
Crest Factor Reduction (CFR) Concepts • Spectrally efficient wideband RF signals may have PAPR >13dB. • CFR preconditions the signal to reduce signal peaks without significant signal distortion • CFR allows the PA to operate more efficiently – it is not a linearization technique • CFR supplements DPD and improves DPD effectiveness • Without CFR and DPD, a basestation or handset PA must operate at significant back-off from saturated power to maintain linearity. The back-off reduces efficiency Benefits of CFR 1. PAs can operate closer to saturation, for improved efficiency (PAE). 2. Output signal still complies with spectral mask and EVM specifications
A Novel Wideband DPD Measurement 62
Crest Factor Reduction (CFR) Concepts
• •
If you can reduce the Peak-to-Average Power Ratio (PAPR) of the signal, then for a given value of Peak, you can raise the Average power (up & to the right, above) with no loss in signal quality. Thus, CFR enables higher PA efficiency by reducing the back-off, often by 6dB
A Novel Wideband DPD Measurement 63
CFR for LTE-Advanced Downlink OFDMA Controls EVM and band limits in the frequency domain. • Constrains constellation errors, to avoid bit errors. • Constrains the degradation on individual sub-carriers. Allows QPSK sub-carriers to be degraded more than 64 QAM sub-carriers. Does not degrade reference signals, P-SS and S-SS. Subcarriers of out-of band are set to NULL.
A Novel Wideband DPD Measurement 64
CFR for LTE-Advanced Downlink OFDMA • No side modifications for receiver • No out-of band spectral distortion (no spectral mask measurement pass/fail issue • EVM always meets specification •Good PAR reductions •No impact of timing and frequency and channel estimation of DL Q m {D A TA P ORT} D ata Ty pe=Integer B us =NO
0+0*j S C _S tatus {D A TAPORT} D ata Ty pe=Integer B us =NO
DC
MOD
V al ue=0 [0+0*j] D 2 {D P D _LTE _A _C FR _P os tP roc @ DPD Models} B andw i dth=B W 20 M H z [B andwidth]
256
O v ers am pl i ngO pti on=R ati o 4 [O v ers am pl i ngO ption] E V M _Thres hol d_Q P S K =0.12 [E V M _Thres hold_QPSK] E V M _Thres hol d_16Q A M =0.1 [E V M _Thres hol d_16QAM]
M1 M odul o=256
E V M _Thres hol d_64Q A M =0.06 [E V M _Thres hol d_64QAM] output {D A TA P ORT} D ata Ty pe=C om plex
Qm
M appi ngD ata {D A TA P ORT} D ata Ty pe=C om plex B us =NO
G2 G ai n=1
A
A3 B l oc k S i z es =1;600;6991;600 [[1,H al f_U s edC arriers ,D FT_z eros ,H al f_U s edCariers]
A 0+0*j
A2 B l oc k S i z es =600;600 [[H al f_U s edC arriers , H al f_U s edC arriers] z eros V al ue=0 [0+0*j]
D P D _R adi usClip C l i ppi ngThres hol d=16.5e-6 [C l i ppi ngThres hold]
DPD
B us =NO
SC_St at u s
LTE_A
out pu t
re f input
FFT
out pu t
DPD_Radi usClip
FFT
i fft1
fft
FFTS i z e=8192 [D FTS i ze] S i z e=8192 [D FTS i ze] D i rec ti on=Inv erse
FFTS i z e=8192 [D FTS i ze] S i z e=8192 [D FTS i ze] D i rec ti on=Forw ard
FreqS equenc e=0-pos -neg
FreqS equenc e=0-pos -neg
FFT
CFR_PostProc input
i fft2 FFTS i z e=8192 [D FTS i ze]
G1 G ai n=8192 [D FTS i ze]
S i z e=8192 [D FTS i ze] D i rec ti on=Inv erse FreqS equenc e=0-pos -neg
G3 G ai n=1
A Novel Wideband DPD Measurement 65
CFR of LTE-Advanced 20MHz Downlink QPSK modulation, CFR algorithm set to Max EVM = 10% Spectrums with and w/o CFR are same!
PAPR=9dB w/o CFR PAPR=6.8dB w CFR
A Novel Wideband DPD Measurement 66
CFR of LTE-Advanced 20MHz Downlink Algorithm EVM targets: QPSK < 10%, 16QAM < 8%, 64QAM < 6%
PAPR=8.9dB w/o CFR PAPR=7.2dB with CFR
Observed EVMs w/CFR
A Novel Wideband DPD Measurement 67
CFR of LTE-Advanced with Carrier Aggregation CFR Approach 1 • •
CFR performed separately on each Component Carrier (up to 20MHz BW) Component Carriers are then summed
CFR Approach 2 •
•
CFR is applied to the carrier-aggregated composite signal (up to 100MHz BW) Then each component carrier is re-filtered individually to remove out-of-band energy, and re-summed
A Novel Wideband DPD Measurement 68
CFR of LTE-Advanced with Carrier Aggregation Approach 1, 2x20MHz contiguous CA 1. 2.
Both CC0 and CC1 adopt 16-QAM and QPSK, respectively. CC1 magnitude threshold of polar clipping is a little larger than CC0 because QPSK modulation can tolerate larger EVM limit, according to EVM specification. Component Carrier 0 (CC0) HARQ _Bit s
Spect r um Anal yzer
LTE_A U E1_ChannelBit s
DL
11010
UE1_Dat a
UE1_M odSym bols
f r m _FD
B4 DataPattern=PN9
11010
CA_Spectrum Mode=TimeGate Start=0s SegmentTime=50μs
Src CFR
Cx
f r m _TD
LTE_DL_Src_CFR2 ShowSystemParameters=YES FrameMode=FDD Bandwidth=BW 20 MHz [Bandwidth] OversamplingOption=Ratio 4 [OversamplingOption] CyclicPrefix=Normal UEs_RevMode=0;0;0;0;0;0 [[0,0,0,0,0,0] CFREnable=YES ClippingThreshold=11.75e-6 [ClippingThreshold1] NumFrames=1
B1 DataPattern=PN9
Fc
T S1 SampleRate=122.9e+6Hz [SamplingRate]
Env
Fc=2.14e+9Hz [FCarrier1]
G1 GainUnit=voltage Gain=1
CCDF
Parameter
CCDF CC1_CCDF
CFREnable=YES
CCDF
Env
Fc Change CA_CCDF Start=0s Stop=50ms
HARQ _Bit s
CC0_CCDF1
LTE_A U E1_ChannelBit s
DL
11010
UE1_Dat a
UE1_M odSym bols
Src f r m _FD
CFR
B2 DataPattern=PN9
11010 B3 DataPattern=PN9
LTE_DL_Src_CFR1 ShowSystemParameters=YES FrameMode=FDD Bandwidth=BW 20 MHz [Bandwidth] OversamplingOption=Ratio 4 [OversamplingOption] CyclicPrefix=Normal UEs_RevMode=0;0;0;0;0;0 [[0,0,0,0,0,0] NumTxAnts=Tx1 CRS_NumAntPorts=CRS_Tx1 CFREnable=YES ClippingThreshold=13.05e-6 [ClippingThreshold2] NumFrames=1
Fc
T Cx
f r m _TD
S2 SampleRate=122.9e+6Hz [SamplingRate]
Env
C2 Fc=2.16e+9Hz [FCarrier2]
G2 GainUnit=voltage Gain=1
Component Carrier 1 (CC1)
A Novel Wideband DPD Measurement 69
CFR of LTE-Advanced with Carrier Aggregation Approach 1: 2x20MHz contiguous CA
CC0 PAPR =7.2 dB CC1 PAPR = 6.7dB 2x20MHz 2CC with CFR #1 PAPR = 8.2dB
EVM of PDSCH 16-QAM is 8.54% in CC0 and EVM of PDSCH QPSK is 11.11% in CC1. EVM values of P-SS, S-SS and RS <0.65% A Novel Wideband DPD Measurement
70
CFR of LTE-Advanced with Carrier Aggregation Approach 2: 2x20MHz contiguous CA 1. Both CC0 and CC1 adopt 16-QAM and QPSK, respectively. 2. Aggregate CC0 and CC1 first, then do polar clipping on the 40MHz bandwidth composite CA signal. 3. Each Component Carrier is filtered separately (20MHz each) 4. Combine the filtered CC0 and CC1 into one CA signal again. Filtering per each carrier
Component Carrier 0 (CC0) HARQ _B tsi
L TE_A
Spect r um Ana r e z ly
UE1_Channel tsB i
11010
DL UE1_Da t UE1_M odSy m sb l o
Src f r m_ D F
CFR B4 Da ta Patte rn=PN9
11010
Cx
L TE_DL _ Src _CFR2 Sho wSy s te m Para m eters=YES S1 Fra m e M o d e=FDD Ban d wi d th =BW 2 0 M Hz [Bandwidth] Sam p l e Ra te =1 2 2 .9 e +6 Hz [SamplingRate] Ov e rs a m p l i n g Op ti o n =Ra ti o 4 [Ov e rsamplingOption] Cy c l i c Pre fi x=Normal CFREna b l e=NO Cl i p p i n g Th re s h o l d =1 1 .1 5 e -6 [ClippingThreshold1]
Env
Fc Change
Fc =2 .1 4 e +9 Hz [FCarier1]
E3 Ou tp u tFc =2 .1 4 e +9 Hz [FCarier1] Ban d wi d th=0Hz Fc
Parameter CFREnable=NO
B1 Da ta Patte rn=PN9
CA_Spe ctrum M o d e =Ti meGate Sta rt=0s Seg m e n tTime=50μs
Fc
T
f r m_ D T
Fc Change
Env
Cx
inpu t out p t u DPD_RadiusClip
Fc
T Cx
Env
E2
DPD_ Ra d i u s Clip_1 S4 C4 Cl i p p i n g Th re s h old=0.14250 Sam p l e Ra te =1 2 2 .9 e +6 Hz [SamplingRate] Fc =2 .1 5 e +9 Hz [FCarier]
DL Src f r m_ D F
11010
E4 Ou tp u tFc =2 .1 6 e +9 Hz [FCarier2] Ban d wi d th=0Hz
UE1_Da t UE1_M odSy m sb l o
CFR
E5 Ou tp u tFc =2 .1 5 e +9 Hz [FCarier] Ban d wi d th=0Hz
Cl i p p i n g _CCDF Sta rt=0s Sto p =50ms
Spect r um Ana r e z ly
CCDF
UE1_Channel tsB i
B2 Da ta Patte rn=PN9
A3 Ou tp u tFc=Max
Fc Change
L TE_A
CCDF Fc Change Env
HARQ _B tsi
11010
Env
F1 FCe n te r=2 .1 4 e +9 Hz [FCarier1] Pas s Ban d wi d th=18e6Hz Pas s Ri p p le=0.1 Sto p Ban d wi d th=19e6Hz Sto p Ri p ple=80 M a x i m u m Order=2057
Fc
T
f r m_ D T
Cx
L TE_DL _ Src _CFR1 Sho wSy s te m Para m eters=YES S2 Fra m e M o d e=FDD Sam p l e Ra te =1 2 2 .9 e +6 Hz [SamplingRate] Ban d wi d th =BW 2 0 M Hz [Bandwidth] Ov e rs a m p l i n g Op ti o n =Ra ti o 4 [Ov e rsamplingOption] Cy c l i c Pre fi x=Normal CFREna b l e=NO Cl i p p i n g Th re s h o l d =1 2 .5 e -6 [Cl ippingThreshold2]
Env
C2 Fc =2 .1 6 e +9 Hz [FCarier2]
Component Carrier 1 (CC1)
CA_CCDF Sta rt=0s Sto p =50ms
F4 FCe n te r=2 .1 6 e +9 Hz [FCarier2] Pas s Ban d wi d th=18e6Hz Pas s Ri p p le=0.1 Sto p Ban d wi d th=19e6Hz Sto p Ri p ple=80 M a x i m u m Order=2057
CA_Spe c tru m _ Clipping M o d e =Ti meGate Sta rt=0s Seg m e n tTime=50μs
Polar clipping
B3 Da ta Patte rn=PN9
Combine carriers as CA signal A Novel Wideband DPD Measurement 71
CFR of LTE-Advanced with Carrier Aggregation Approach 2: 2x20MHz contiguous CA
2x20MHz 2CC w/o CFR PAPR = 9 dB 2x20MHz 2CC with CFR #2 PAPR = 7.4dB
EVM of PDSCH 16-QAM is 7.80% in CC0 and EVM of PDSCH QPSK is 7.82% in CC1. All EVM values of P-SS, S-SS and RS are about 7% A Novel Wideband DPD Measurement
72
Agenda 1. Introduction and Problem Statement
2. Digital Pre-Distortion (DPD) Concepts 3. DPD verification with Agilent Hardware 4. DPD simulation with Agilent EDA Tools 5. Crest Factor Reduction (CFR) 6. PA Modeling
7. Summary
A Novel Wideband DPD Measurement 73
PA Modeling with Memory Polynomial A nonlinear PA model with memory effects is a by-product of the DPD process
1. Create DPD Stimulus
It can be used as a transportable model for system-level simulations
Accuracy degrades with changes to Signal, RF carrier frequency, bandwidth, and power level.
2. Capture PA Response
3. Extract PA Model
4. Verify PA Response
A Novel Wideband DPD Measurement 74
PA Modeling with Memory Polynomial (MP) PA output
PA Model Extraction and Verification Im Re
R1 {ReadFile@Data Flow Models} File='Step2_PAOutputdata_imag.txtR3 {RectToCx@Data Flow Models}
NMSE
Sh i fte d _ PA_Out_Piece
Sc a l e d _ PA_In_Piece PA_ Out Sh i fte d _PA_Out DPD_ PACo e ffExtractor PA_In Sc a l e d_PA_In
R2 {ReadFile@Data Flow Models} File='Step2_PAOutputdata_real.txt
DPD_ Coef DPD_ Output DPD_ Pre Di s to rter DPD_ Input
In Test
123
NMSE
DPD_NMSE
In Ref PA_ Coef
D1 {DPD_PACoeffExtractor@DPD Models} ModelType=Memory Polynomial [ModelType] ModelIdentificationAlgorithm=LSE using QR [ModelIdentificationAlgorithm] MemoryOrder=3 [MemoryOrder] NonlinearOrder=9 [NonlinearOrder] NumOfInputSamples=61440 [NumOfInputSamples]
PA input R5 {ReadFile@Data Flow Models} File='Step2_PAInputdata_imag.txt
D3 {DPD_PreDistorter@DPD Models} D2 {DPD_NMSE@DPD Models} MemoryOrder=3 [MemoryOrder] NumOfInputSamples=61440 [NumOfInputSamples] NonlinearOrder=9 [NonlinearOrder] NumOfInputSamples=61440 [NumOfInputSamples]
123
Im Re Im
DPD_CoeffImag StartStopOption=Samples
R6 {RectToCx@Data Flow Models}
DPD Coefficients
Re R4 {ReadFile@Data Flow Models} File='Step2_PAInputdata_real.txt
NMSE StartStopOption=Samples
C1 {CxToRect@Data Flow Models}
123 DPD_CoeffReal StartStopOption=Samples
Real part of PA coefficients
Imaginary part of PA coefficients
0.39152231499472767 0.21240763496121581 -0.23274788419756332 -0.09408421097412327 -0.014555159450854303 -0.013485802116749272 -0.0018840630729567572 -0.0002707103038025503 -1.4961320837900501e-005 0.2518269309218949 -0.59748335589139812 0.47971434227815068 -0.20894251288419013-
0.627076629490954 -0.68351712455802727 0.59425895517848892 -0.021747841570654059 0.036844218259549213 0.012303248529776717 0.0020763571187302357 0.00028702357690964096 1.6024859831773016e-005 0.66865526723410285 -1.3558414032104396 1.0569299692538558 -0.40306476477955344
The complex Memory Polynomial coefficients are stored to an ASCII file for each: - Memory Order - Nonlinear Order
A Novel Wideband DPD Measurement 75
PA Modeling with Memory Polynomial Re-read coefficients from ascii file into MP model
PA Modeling Verification
R2 File='Step3_PA_Coefficients_Imag.txt Periodic=YES
Captured PA input waveform
Im Re R3
R1 File='Step3_PA_Coefficients_Real.txt Periodic=YES
Fc R5 File='Step2_PAInputdata_imag.txt Periodic=YES
Fc
T
Im
Fc
Spe c trum Analyzer
DPD_Coef DPD_O ut put
Cx
Env
Amplifier
Env
Cx
DPD_ Pre Di s to rter
Cx
Env
DPD_I nput
Re R6
S1 SampleRate=61.44e+6Hz [SamplingRate]
C1 Fc=2GHz
A1 GainUnit=dB Gain=6.5
E4
R4 File='Step2_PAInputdata_real.txt Periodic=YES
C4 Fc=2GHz
Spe c trum Analyzer
PA_Input Mode=TimeGate Start=0s SegmentTime=150μs
R8 File='Step2_PAOutputdata_imag.txt Periodic=YES
D3 MemoryOrder=3 NonlinearOrder=9 NumOfInputSamples=61440
PA_Output_SW Mode=TimeGate Start=0s SegmentTime=150μs
Fc
T
Im
PA Model by using memory polynomial coefficients
Spe c trum Analyzer
Cx
Env
Re R7
S2 SampleRate=61.44e+6Hz [SamplingRate]
C2 Fc=2GHz
PA_Output_HW Mode=TimeGate Start=0s SegmentTime=150μs
R9 File='Step2_PAOutputdata_real.txt Periodic=YES
Captured PA output waveform
A Novel Wideband DPD Measurement 76
PA Modeling with Memory Polynomial Distorted PA model results
A Novel Wideband DPD Measurement 77
Agenda 1. Introduction and Problem Statement 2. Digital Pre-Distortion (DPD) Concepts
3. DPD verification with Agilent Hardware 4. DPD simulation with Agilent EDA Tools 5. Crest Factor Reduction (CFR) 6. PA Modeling 7. Summary
A Novel Wideband DPD Measurement 78
Unified architecture, verification for Layer 1 Comms Augments general purpose tools, or, stands on its own Agilent SystemVue Cross-domain PHY modeling framework, for Model-Based Design
Baseband Algorithms
PHY IP
Dataflow Simulation
RF Simulators
Baseband Hardware Flows DSP/ASSP GPP/ARM Software FPGA/ASIC/SoC Software Hardware
RF Sys Architecture
RF Hardware Flows TEST
RFIC / MMIC Hardware SiP / Board Hardware
PHY system integration and verification Complete a working PHY using combinations of Software, RF/BB Hardware, Simulation, and Measurements
A Novel Wideband DPD Measurement 79
Agilent DPD Modeling Value for Enterprise System-level approach • • • • •
Open, standards-based modeling interfaces (.m, C++, HDL) for vendor and hardware neutrality Wireless standards IP leadership for confidence, coverage, interoperability and virtual system-level closed-loop BER/Throughput Modifiable DPD modeling IP, with quick results before implementation Excellence in RF modeling and Test Consistent flow: Same modeling approach for Simulation, R&D evaluation, Final Test
Enterprise connectivity for highest leverage • • • •
Test & Measurement leadership and software integration RF EDA flow leadership, with connectivity for predictive, preliminary results Reduces overall tool count, support, increases design re-use Connects islands of domain knowledge, tools, skills
Services for successful integration • • •
80
Local presence, worldwide Customization services and Training Aggressive product roadmap
Summary Modern communication systems with high spectral efficiency and wide bandwidths typically • Use signals with high peak-to-average power ratios (PAPR). • Operate RF PAs with high power added efficiency (PAE). • However, high PAPR results in driving the PA into higher distortion levels that requires PA back-off in drive level, which reduces PAE. The key problem is: How to handle signals with high PAPR, with the PA operating at high PAE, while maintaining low signal distortion?
• Digital Pre-Distortion (DPD) and Crest Factor Reduction (CFR) techniques together help overcome conflicting requirements. • SystemVue offers a practical DPD Design Flow usable with real PA hardware that includes integration with Agilent ESG/MXG and PXA/MXA instruments. A Novel Wideband DPD Measurement 81
Questions & Answers
A Novel Wideband DPD Measurement 82
“LTE-Advanced DPD using Agilent SystemVue” THANK YOU W1716 Digital Pre-Distortion Web - www.agilent.com/find/eesof-systemvue-dpd-builder App Note - http://cp.literature.agilent.com/litweb/pdf/5990-6534EN.pdf App Note - http://cp.literature.agilent.com/litweb/pdf/5990-7818EN.pdf App Note - http://cp.literature.agilent.com/litweb/pdf/5990-8883EN.pdf
SystemVue www.agilent.com/find/eesof-systemvue www.agilent.com/find/eesof-systemvue-videos www.agilent.com/find/eesof-systemvue-evaluation
Or, contact your regional Agilent resource www.agilent.com/find/eesof-contact
A Novel Wideband DPD Measurement 83
Appendixes
A Novel Wideband DPD Measurement 84
Summary: Agilent SystemVue For system architects and baseband algorithm developers • Improved productivity through model-based design & verification • Provides top-down System-Level cockpit for communications & defense • Unites Algorithm & Baseband with Agilent’s leadership in other domains, such as RF, Test, and Standards IP for • superior cross-domain effectiveness • earlier design maturity • higher performance, lower margins
A Novel Wideband DPD Measurement 85
Summary: SystemVue improves top-down System Design
with strengths in 4 key areas POLYMORPHIC BASEBAND ALGORITHMS & IP
Easily assemble Virtual PHYs
INDUSTRY-LEADING REFERENCE IP and APPS WiMAX DVB-S2 ZigBee OFDM
DPD RADAR MIMO Channel mmWave WPAN
ACCURATE RF & CHANNEL EFFECTS
HARDWARE & MEASUREMENT CONNECTIVITY
Quickly move Ideas to proven, real-world Hardware
A Novel Wideband DPD Measurement 86
Wideband configurations: LTE-A 2x20MHz Contiguous CA
Agilent M9330A AWG, M9392A VSA Source = M9330A AWG N5182 MXG Vector Analyzer= M9392A - 12bits ADC - up to 250MHz bandwidth PA output Spectrum (Blue) PA+DPD Spectrum (Red) PA input Spectrum (Green)
A Novel Wideband DPD Measurement 87
DPD of LTE-Advanced DL CA, using M9330A/M9392A 3x20MHz contiguous CCs, (60MHz signal BW)
A Novel Wideband DPD Measurement 88
DPD of LTE-Advanced DL CA, using M9330A/M9392A 2x20MHz + 20MHz non-contiguous CCs, (60MHz signal BW)
A Novel Wideband DPD Measurement 89
LTE-A Results with 200W LDMOS Doherty PA DPD+PA Output (10MHz System)
A Novel Wideband DPD Measurement 90
LTE-A Results with 200W LDMOS Doherty PA Raw PA Output (10MHz System)
A Novel Wideband DPD Measurement 91
LTE-A Results with 200W LDMOS Doherty PA DPD+PA Output (DL 20MHz System)
A Novel Wideband DPD Measurement 92
LTE-A Results with 200W LDMOS Doherty PA Raw PA Output (DL 20MHz System)
A Novel Wideband DPD Measurement 93
LTE-A Results with 200W LDMOS Doherty PA DPD+PA Output (DL Carrier Aggregation 2x10MHz System)
A Novel Wideband DPD Measurement 94
LTE-A Results with 200W LDMOS Doherty PA Raw PA Output (DL Carrier Aggregation 2x10MHz System)
A Novel Wideband DPD Measurement 95
