“4g for everyone” - keysightliterature.cdn.keysight.com/litweb/pdf/5990-6742en.pdf · live...

“4G For Everyone”Extended RF Performance with Digital Pre-Distortion

Live webinar presented June 17, 2010http://www.agilent.com/find/eesof-systemvue

Copyright © Agilent Technologies 20101

Dr. Jinbiao XuAgilent TechnologiesMaster System Engineer

John BaprawskiRF/MW System Consultant

1. Introduction and Problem Statement

2. Digital Pre-Distortion (DPD) Concepts

3. Example using DPD

4. Crest Factor Reduction (CFR) Concepts

5. Example using CFR

6. DPD Hardware Implementation

7. Using X-Parameters with PA and DPD

8. Practical Considerations

9. Summary

Agenda

Copyright Agilent Technologies 2010SystemVue DPD webcast

Jinbiao XU

2 6/17/2010

3

• Improved productivity through model-based design

• Provides top-down ESL cockpit for Radar design

• Unites Algorithm & Baseband with Agilent leadership in RF, Test, and Radar IP for

• superior cross-domain effectiveness• earlier design maturity• higher performance, lower margins

About Agilent SystemVueFor system architects and baseband algorithm developers

3

SystemVue DPD webcastJinbiao XUCopyright Agilent Technologies 2010

6/17/2010

SystemVue – The shortest path from imagination to hardware for communications & defense

Easily assemble Virtual PHYs

Quickly move Ideas to proven, real-world HardwareWiMAX

DVB-S2ZigBeeOFDM

DPDRADARMIMO ChannelmmWave WPAN

HARDWARE & MEASUREMENTS

REFERENCE IP & APPS

BASEBANDALGORITHMS & IP

ACCURATE RF & CHANNEL EFFECTS

SystemVue DPD webcastJinbiao XUCopyright Agilent Technologies 2010

6/17/2010

• 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.

Problem Statement


Jinbiao XU5 6/17/2010

Conflicting requirements

Problem Statement


Jinbiao XU6

How to handle signals with high PAPR, while driving the PA to operate with high PAE, while also having low signal distortion?

High PAE

Signals with high

PAPR

HighSpectral

Efficiency

Higher Drive levels

High Distortion

levels

“Back off” the drive

levels

6/17/2010

• High-efficiency PA design typically relies on :• Constraining the PA energy into a bandpass characteristic• Peaking the PA energy by dynamically biasing the PA versus input

powerThese techniques inherently result in PA nonlinearities with memory

• Techniques used to reduce distortion when operating at high efficiency:• PA linearization – enables driving the PA closer to saturation for a

given level of distortion• Signal preconditioning – Reduces signal peaks without significant

signal distortion

Problem Statement



Solution Approach


Jinbiao XU8

Solution: Preconditioning the signal (CFR) and correcting for the hardware (DPD) will both be discussed in this presentation

High PAE

Signals with high

PAPR

HighSpectral

Efficiency

Higher Drive levels

High Distortion

levels

Higher throughput levels for

subscribers

CFR

DPD

6/17/2010

Two key power amplifier characteristics• AM-to-AM

• Change in output power gain (in dB) versus input power,relative to small signal gain

• AM-to-PM• Change in output phase (in degrees) versus input power,

relative to small signal conditions

Without memory effects• There is a one-to-one relationship between the AM-to-AM and

AM-to-PM values vs. input powerWith memory effects

• There are multiple AM-to-AM, AM-to-PM values associated with input power due to the PA memory and the previous signal values

Definitions



• Today’s emerging 4G signals use spectrally efficient formats with high PAPR and require PAs with high PAE

• The highest performance modes of these standards are sensitive to PA memory effects

AM-AM characteristics for a PAWith and without memory effects

SystemVue DPD webcastJinbiao XU

10

AM-to-AM for PA with memoryAM-to-AM for PA without memory

Copyright Agilent Technologies 2010

6/17/2010









9. Summary

Agenda



• DPD enables operating the PA in its high PAE region near saturation without significant signal distortion.

• The DPD approach discussed here addresses PAs with memory • Realistic training signals are used• Actual PA output is measured• The DPD network pre-distorts the PA input signal so that

combined [DPD+PA] behaves more linearly

• Based on Agilent Technologies Measurement Research Labs• OFDM MIMO research drove the need for DPD development• Agilent Multi-Carrier Power Amplifier (MCPA) test system

• Custom set-ups for high power basestation PAs

Digital Pre-Distortion (DPD) Concepts



1. Characterize DUT PA using DUT input and output waveforms

2. Calculate complex memory coefficients using QR and SVD algorithms

3. Apply memory polynomial coefficients to construct memory polynomial pre-distorter

4. Pre-distorted signal cancels nonlinear distortion in PA (including major memory effects)

Steps to Achieving Digital Pre-distortion (DPD)



Theory of Operation – Desired Pout

• Uncorrected PA: Pout vs. Pin curve saturates at Psat (in dBm)

• DPD+PA: Pout-pd vs Pin curve is linear, up to a higher limit,where Max(Pin) is the maximum correctible input power

Linear Output

Input Power Pin Pin-pd

Output Power

Pout Pout-pd

Desired Output Saturation

Psat

Max Correctable Pin



Theory of Operation – Operating Region

θo-pd = k θout

Input Power

Output Phase

Pi Pi-pd

Desired Output Linear Output



Memory Polynomial Algorithm• As the signal (such as 3GPP LTE) bandwidth gets wider, power amplifiers

begin to exhibit memory effects. Memoryless (LUT) pre-distortion can achieve only very limited linearization performance.

• Volterra series is a general nonlinear model with memory. It is unattractive for practical applications because of its large number of coefficients.

• Memory polynomial reduces Volterra’s model complexity. It is interpreted as a special case of a generalized Hammerstein model. Its equation is as follows:

K is Nonlinearity order and Q is Memory order


Jinbiao XU16

∑∑= =

−−−=K

k

Q

q

kkq qnyqnyanz

1 0

1)()()(

6/17/2010

Signal Training to derive the Memory Polynomial

1. Pre-distorter training: Nonlinear coefficients are extracted from the PA input and PA output waveforms (ie – on real physical behavior)

2. Copy of PA : The DPD model accurately captures the nonlinearity with memory effects


Jinbiao XU17

zUUUa HH 1)(ˆ −=

[ ]TNzzzz )1(,),1(),0( −= [ ]KQQK uuuuU ,,,,,, 1010 =

[ ]Tkqkqkqkq Nuuuu )1(,),1(),0( −=

1)()()(−−−

=k

kq Gqny

Gqnynu

Memory Polynomial Coefficients

[ ]TKQQK aaaaa ˆ,,ˆ,,ˆ,,ˆˆ 1010 =

6/17/2010









9. Summary

Agenda



5-Step Measurement-Based DPD Modeling Flow

1. Create DPD Stimulus

2. Capture PA Response

3. Extract DPD Model

4. Capture DPD+PA Response

5. Verify DPD+PA Response



DPD Hardware Verification Platform

10MHz Reference

10MHz Reference

External Trigger

External Trigger

Attenuator (20dB)

1. Instrument setup to capture PA input signal

2. Instrument setup to capture PA output signal

Agilent E4438C Agilent E4440A

Agilent E4440AAgilent E4438C


Jinbiao XU20

Thru measurement

6/17/2010

SystemVue DPD Hardware Flow for LTE

The download power and length of the waveform can also be set.


Jinbiao XU21

Step 1. Create DPD stimulus waveform• LTE parameters such as bandwidth, Resource Block allocation and others

can be set

6/17/2010

Connect the ESG directly to the PXA and click the “Capture Waveform” button. The captured signal is the input of the PA.

Connect the ESG to the PA, connect the PA to the PXA, 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.


Jinbiao XU22

Step 2. Capture PA response• SystemVue interfaces directly to the ESG (source) and PXA (analyzer).• Instrument parameters such as number of signal, trace assignment and file

name can be set


6/17/2010

DPD AM-to-AM Characteristic


Jinbiao XU23


Step 3. DPD Model Extraction• DPD model parameters such as number

of training samples, memory order and nonlinear order can be set

PA AM-to-AM Characteristic

6/17/2010

Set the RF power


Jinbiao XU24


Step 4. Capture DPD+PA Response• The signal is predistorted by the DPD model and downloaded into the ESG.

DPD+PA output (blue) and original raw signal (green) is displayed

6/17/2010


Step 5. Verify DPD+PA response• LTE performance for the DPD model used with the PA hardware is verified

Spectrum, EVM andACLR are calculatedand plotted automatically



• Spectrum:

• ACLR (dB): DPD-PA restores performance ~to the original ACLR.

• EVM: DPD-PA output signal has same (-26.8 dB) as input signal

LTE Results for PA with Memory


Jinbiao XU26

ACLR -2BW Lower

-1BW Lower

+1BW Upper

+2BW Upper EVM (dB)

Raw signal 127.35 103.75 103.66 127.30 -26.80PA output 51.08 38.90 38.96 51.15 -24.97DPD-PA output 84.65 80.07 79.96 84.62 -26.78

RED - PA outputBLUE - DPD+PA outputGREEN – Original raw signal

6/17/2010

GSM Results with 4-Carriers for PA with Memory4-Carrier GSM,30 MHz Bandwidth

No DPD

DPD+PANonlinearOrder=9.

DPD+PA NonlinearOrder=11.











9. Summary

Agenda



• 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 PA must operate at significant

back-off from saturated power to maintain linearity. The back-off reduces efficiency

Benefits of CFR1. PAs can operate closer to saturation, for improved efficiency (PAE).2. Output signal still complies with spectral mask and EVM

specifications

Crest Factor Reduction (CFR) Concepts



Crest Factor Reduction (CFR) Concepts

If you can reduce the Peak-to-Average Ratio of the signal, then for a given amplitude 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



Crest Factor Reduction for Multiple-Carrier Signals• Multiple-Carrier Signals (such as GSM, WCDMA, WiMAX) already have high

PAPR. • In the future, they will also include multiple waveforms (ie - LTE with 3G WCDMA).• Therefore CFR will increase in importance for Multi-Carrier PA (MCPA)

linearization.


Jinbiao XU31

CFR algorithm for multiple carrier signals• PW (Peak Windowing)-CFR• NS (Noise-Shaping) -CFR• PI (Pulse Injection)-CFR• PC (Peak Cancellation)-CFR

6/17/2010

CFR for 3GPP LTE DL OFDM Signal• 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.• All control channels (PDCCH, PBCH, PCFICH and PHICH) adopts

QPSK threshold.











9. Summary

Agenda



CFR Clipping CFR Filtering

WCDMA Example with 4-Carriers



WCDMA Example with 4-CarriersSimulation results


Jinbiao XU35

With CFR

With CFR

PAPR With CFR PAPR No CFR5.89dB 9.49dB

6/17/2010

LTE Downlink • 10MHz BW• Sampling Rate=61.44MHz• QPSK• EVM threshold < 10%

LTE Example with Crest Factor Reduction


Jinbiao XU36

CFR Design with Clip and Filter

PAPR With CFR PAPR No CFR6.867dB 9.05dB

CCDF with CFR

CFR causes little ACLR degradation

6/17/2010









9. Summary

Agenda



Real-Time SystemVue DPD

SystemVue for DPD Investigation / Design• Integrated algorithm design, validation and real-

time implementation solution• HDL, C++ code generation available• Shortens time to market for DPD designs

FPGA implementation challenges• High FPGA computational resource requirement• High numerical range for model coefficients

• 10-20 ~ 1020

• High requirement on accuracy, # bits• High computation load

• 10s~100s GFLOPS• Additional implementation complexity if

accommodating different PAs

Real-time DPD IP

FPGA(Real-time DPD)

Baseband signal

RF card

DPD hardware board

DPD algorithm design

DPD real-time implementation

DPD verification



Real-Time SystemVue DPD IPGeneric FPGA core with customized

parameters to shorten time to market and R&D cost

• Variable memory order & nonlinear order to cover different PAs

• Variable FPGA footprint depending on model orders

• Accurate real-time implementation needed to achieve the same performance as simulated DPD

• About 20dB ACLR improvement depending on the PA

-60 -50 -40 -30 -20 -10 0 10-15

-10

-5

0

5

10

Pout [dBm]

Pin

/Pou

t [dB

]

DPD gain improvement

SystemVueFPGA

No DPD (red),DPD in SystemVue (green)Real-Time DPD FPGA (blue)


Jinbiao XU39

DPD AM-to-AM SystemVue (blue) FPGA (red)

6/17/2010









9. Summary

Agenda



X-Parameters for System Design


Jinbiao XU41

• Accurate: X-parameters are the mathematically correct superset of S-parameters

• Versatile: Applicable to both small and large signal conditions for linear and nonlinear components vs. Power, Freq, Bias, Zload, etc

• Design flow closure: Brings finished RF designs back up to the architecture level, and makes the IP safely transportable

Measure and view X-parameters with the Agilent NVNA and ADS

6/17/2010

Using X-Parameters with SystemVue DPD


Jinbiao XU42

• X-parameters (Version 2 and earlier) accurately models frequency-domain nonlinearities, but not nonlinearity memory effects

• Research to add nonlinearity memory effects to X-parameters is currently in development at Agilent (keynote IMS 2009 paper)

• SystemVue DPD can work directly on the current generation of X-parameters and P2D type models• Set the DPD model memory order = 0• Gives quick estimate of achievable system performance

6/17/2010

EVM (dB)Frame 1

EVM (dB)Frame 2

PA input -26.016 -26.016Raw PA output -25.810 -25.807DPD-PA output -26.013 -26.013

LTE DPD Simulation with X-parameter PA (1W)Spectrum, ACLR and EVM results

Copyright Agilent Technologies 2010


43

EVM (dB)-2BW Lower

-1BW Lower

+1BW Upper

+2BW Upper

PA input 71.01 64.34 63.43 70.84Raw PA output 65.39 49.06 49.28 65.33DPD-PA output 67.47 62.84 62.20 67.34

ACLR (dB)

• Low power PA (1 W)• X-parameters were measured with

Agilent NVNA• Same SystemVue DPD flow used for

these LTE spectrums, EVM and ACLR results.


6/17/2010









9. Summary

Agenda



• DPD and CFR are more likely to be needed when

• The PAE must be maximized

• The PA must be operated at high power

• The nonlinearity has memory

• The analog part is used with wideband or multiband signals

• DPD and CFR are needed for modern systems that operate with high RF spectral efficiency

• Multiple-Carrier Signals (Multiple-Carrier GSM, Multiple-Carrier WCDMA)

• CDMA (WCDMA, CDMA2000)

• OFDM (LTE, WiMAX)

Practical Considerations









7.. Using X-Parameters with PA and DPD


9. Summary

Agenda



• 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 these 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.

Summary



Questions & Answers

48Copyright Agilent Technologies 2010


6/17/2010

“4G For Everyone” Now you can use DPD, too!

49

THANK YOUFor more information

www.agilent.com/find/eesof-systemvue www.agilent.com/find/eesof-systemvue-videoswww.agilent.com/find/eesof-systemvue-evaluationwww.agilent.com/find/ims2010-microapps

Or, contact your regional Agilent resource• www.agilent.com/find/eesof-contact


Jinbiao XU6/17/2010

Appendix



LTE Results with Doherty PA (30W)AM-AM characteristics


Jinbiao XU51

AM-AM30W Doherty PA

AM-AMDPD correction network

6/17/2010

LTE Results with Doherty PA (30W)Spectrum, ACLR and EVM results


Jinbiao XU52

EVM (dB)

-2BW Lower

-1BW Lower

+1BW Upper

+2BW Upper

PA input 51.61 50.99 50.01 51.82PA output 43.59 27.64 27.42 42.26

DPD-PA output 50.15 40.43 40.08 47.78

BLUE – Raw PA outputRED – DPD+PA output

ACLR (dB)

6/17/2010

1. Lei Ding, Zhou G.T., Morgan D.R., Zhengxiang Ma, Kenney J.S., Jaehyeong Kim, Giardina C.R., “A robust digital baseband predistorter constructed using memory polynomials”, Communications, IEEE Transactions on, Jan. 2004, Volume: 52, Issue:1, page 159-165.

2. Lei Ding, “Digital Predistortion of Power Amplifiers for Wireless Applications”, PhD Thesis, March 2004.3. Roland Sperlich, “Adaptive Power Amplifier Linearization by Digital Pre-Distortion with Narrowband Feedback using

Genetic Algorithms”, PhD Thesis, 2005.4. Helaoui, M. Boumaiza, S. Ghazel, A. Ghannouchi, F.M., “Power and efficiency enhancement of 3G multicarrier

amplifiers using digital signal processing with experimental validation”, Microwave Theory and Techniques, IEEE Transactions on, June 2006, Volume: 54, Issue: 4, Part 1, page 1396-1404.

5. H. A.Suraweera, K. R. Panta, M. Feramez and J. Armstrong, “OFDM peak-to-average power reduction scheme with spectral masking,” Proc. Symp. on Communication Systems, Networks and Digital Signal Processing, pp.164-167, July 2004.

6. Zhao, Chunming; Baxley, Robert J.; Zhou, G. Tong; Boppana, Deepak; Kenney, J. Stevenson, “Constrained Clipping for Crest Factor Reduction in Multiple-user OFDM”, Radio and Wireless Symposium, 2007 IEEE Volume , Issue , 9-11 Jan. 2007 Page(s):341- 344.

7. Olli Vaananen, “Digital Modulators with Crest Factor Reduction Techniques”, PhD Thesis, 2006 8. Boumaiza, et a, “On the RF/DSP Design for Efficiency of OFDM Transmitters” , IEEE Transactions on

Microwave Theory and Techniques, Vol. 53, No. 7, July 2005, pp 2355-2361.9. Boumaiza, Slim, “Advanced Memory Polynomial Linearization Techniques,” IMS2009 Workshop WMC

(Boston, MA), June 2009.10. Amplifier Pre-Distortion Linearization and Modeling Using X-Parameters, Agilent EEsof EDA

References



http://ieeexplore.ieee.org/xpl/RecentIssue.jsp?punumber=26�



For more information aboutAgilent EEsof EDA, visit:

www.agilent.com/find/eesof

For more information on Agilent Technologies’products, applications or services, pleasecontact your local Agilent office. Thecomplete list is available at:

www.agilent.com/find/contactus

Contact Agilent at:

AmericasCanada (877) 894-4414Brazil (11) 4197 3500Mexico 01800 5064 800United States (800) 829-4444

Asia PacificAustralia 1 800 629 485China 800 810 0189Hong Kong 800 938 693India 1 800 112 929Japan 0120 (421) 345Korea 080 769 0800Malaysia 1 800 888 848Singapore 1 800 375 8100Taiwan 0800 047 866Thailand 1 800 226 008

Europe & Middle East

Austria 01 36027 71571Belgium 32 (0) 2 404 93 40Denmark 45 70 13 1515Finland 358 (0) 10 855 2100France 0825 010 700*

*0.125 €/minuteGermany 07031 464 6333Ireland 1890 924 204Israel 972-3-9288-504/544Italy 39 02 92 60 8484Netherlands 31 (0) 20 547 2111Spain 34 (91) 631 3300Sweden 0200-88 22 55Switzerland 0800 80 53 53United Kingdom 44 (0) 118 9276201Other European Countries:www.agilent.com/find/contactus

Product specifications and descriptions in this document subject to change without notice.

© Agilent Technologies, Inc. 2010Printed in USA, October 25, 20105990-6742EN

www.agilent.comwww.agilent.com/find/eesof-systemvue

http://www.agilent.com/find/eesof�

http://www.agilent.com/find/contactus�

http://www.agilent.com/find/contactus�

http://www.agilent.com/�

http://www.agilent.com/find/eesof-systemvue�



“4g for everyone” - keysightliterature.cdn.keysight.com/litweb/pdf/5990-6742en.pdf · live...

Documents