Advanced Architectures for Predistortion Linearization of RF Power Amplifiers

Transcription

1 Advanced Architectures for Predistortion Linearization of RF Power Amplifiers, Youngcheol Park, and Wangmyong Woo chool of Electrical and omputer Engineering Georgia Institute of echnology Presented at he IEEE opical Workshop on Power Amplifiers La Jolla, A

2 Acknowledgement his work was supported in part by Danam UA, an Jose, A. and by he Yamacraw Design enter, an economic development project funded by the tate of Georgia. hanks also to irenza Microdevices, and Ericsson UA for donating the power amplifiers used in this study. 2

3 Outline Introduction to Predistortion LU Updates using ub-sampling Receivers RF Envelope Predistortion ummary and onclusions 3

4 Predistortion oncept his back-off value is (Ideal limiter saturation point-papr of input signal) PA response after predistortion algorithm PA Response his back-off value is determined to give no significant nonlinear region for peak input power Input operating range with Pre-D PAPR Input operating range without Pre-D Average power Peak power 4

5 orrection echniques for ellular Base tations orrection echnologies orrection apability* orrection Bandwidth Relative ost Feed Forward db > 100 MHz High Envelope Feedback db < 5 MHz Med Analog Pre-Distortion 5-10 db > 25 MHz Low Adaptive Pre-D db > 50 MHz Med * IMD orrection based on 8-one ontinuous Random Phase 5

6 Predistortion Linearization Predistortion echnology Bandwidth Relative IMD orrection omments Open Loop Analog Pre-D High Low imple Implementation Adaptive Analog Pre-D (Work function) Moderate Moderate No I/Q stream required Digital Baseband Pre-D Moderate Good Depends on DP computational capability 6

7 Baseband Pre-Distortion I/Q tream Input DP Pre-D DA Error ignal PA RF Output Σ AD Adaptive Filter to Minimize ME AD/DA Requirements: 20 MHz BW 100 MHz with 5 th order IMD 3X over sample 300 Msps (on both I and Q) DP peed: > MA/sec 0.18 µm MO: ~1.5 nw/ma ~1W D power LU update must also be performed 7

8 rends in Analog-to-Digital onverter echnology Adapted from R. H. Walden, Performance rends for Analog-to-Digital onverters, IEEE ommunications Magazine, February 1999, pp

9 Pre-D Results with Low Power Amplifier (0.5W lass-ab GaAs HFE) Without Pre-D With Pre-D 9

10 Disadvantages of Baseband Pre-D ampling Requirements DP speed (power) Digital I/Q input stream required LU update must be performed in background 10

11 Outline Introduction to Predistortion LU Updates using ub-sampling Receivers RF Envelope Predistortion ummary and onclusions 11

12 ampling in Predistorters A predistorter samples signals at the input and output of a power amplifier to identify its AM-AM, AM-PM characteristics. Input ignal f Predistorter PA Output ignal f RF Downconverter RF Downconverter ample & Hold ircuit ample & Hold ircuit A/D onverter A/D onverter DP 12

13 Input Nyquist Rate Input Nyquist rate is when f IF +BW/2=f s -f IF BW/2 Again, this is equivalent to the second figure in terms of aliasing herefore, this is 33% of the output Nyquist Rate, considering 3 rd order IMD. 0 f IF f s - f IF f s 0 f s 13

14

15 Output Nyquist Rate With predistortion systems, the analog-to-digital conversion at the output of a PA is traditionally done with above Output- Nyquist rate in order to avoid the aliasing at the output spectrum. 0 f L f IF f H f s - f IF f H,i = f L,i f s 15

16 ub-sampling fs=10mhz f L,i =1.17 fs=5.01mhz (67% overlapped) f L,i =6.2 0 f L f IF =2 f H= f s =10 f H,i 0 f L f H,i f s =5.01 f IF =2 BPF f s fs<3.55 (More than 133% overlapped) f L f IF =2 f H 16

17 ampling Requirements for Nonlinear ystem Identification he goal of the sampling system is the pre-d system is not to reconstruct signals, but to identify the nonlinear distortion. herefore, it is not necessary to do Nyquist rate sampling. According to the Generalized ampling heorem*, a nonlinear system may be accurately identified if the output signal is sampled at the input Nyquist rate. *John simbinos, and Kenneth V. Lever, ampling Frequency Requirements for Identification and ompensation of Nonlinear ystems, in Proc. IAP, vol. III, 1994, pp

18 Basic Operation of the Generalized ampling heorem x(t) Nonlinear Mapping f - ( ) y(t) = f(x(t)) y(t) ampler y(t Nonlinear k ) g(y(t k )) t Mapping k =k s g( ) = f --1 ( ) LPF x(t) 18

19 ub-ampling Architecture AD imulation 19

20 ub-ampling Architecture AD imulations (cont.) 1.0 mag(h2) Input ignal H2 H Aliased ignal mag(h1) Mega_Hertz Figure 1: imulated AM-AM haracteristic when f s = 10 MHz (includes 3 rd order distortions). Figure 2: Frequency pectrum when f s = 10 MHz (includes 3 rd order distortions) Input ignal 0 Aliased ignals mag(h2) H2 H Figure 3: imulated AM-AM mag(h1) haracteristic when f s = 5 MHz (70% of output spectrum is overlapped) Figure 4: Frequency pectrum Mega_Hertz when f s = 5 MHz (70% of output spectrum is overlapped). 20

21 ub-sampling Architecture est Bed /H ircuit HP89410 H2 IF Filter LPF Power Amplifier Lo Pulse Generator Err 880MHz H1 HP4432 P 21

22 Measurement Results Vout [V] Vout [V] Vin [V] Vin [V] Figure 6: Measured AM-AM haracteristic of HF-0189 when f s = 10 MHz (includes 3 rd order distortions). Figure 7: Measured AM-AM haracteristic of HF-0189 when f s = 5.3 MHz (67% of output spectrum is overlapped). 22

23 Indirect Learning Predistortion Architecture Basic Algorithm : y( n) = f ( z( n)), e( n) = z( n) g( y( n)), wheref( ) is the nonlinear transfer function of PA, and g( )is an estimate of f ( ) calculated by method of Least quares -1 23

24 Predistortion Results APR Improvement [db] Overlapped pectrum of Output ignal [%] APR Improvement Using ubsampling Predistortion Architecture (Negative values in x-axis indicates sampling above Nyquist rate of the output signal) 24

25 Predistortion Results (cont.) irenza HF-0589: 2W GaAs HFE PA Figure: measurements over P in variations. 25

26 Predistortion Results (cont.) Motorola MRF W i LDMO 26

27 Impact of Memory Effects Memoryless PA PA with trong Memory Effects 27

28 Impact of Memory Effects (cont.) 28

29 Outline Introduction to Predistortion LU Updates using ub-sampling Receivers RF Envelope Predistortion ummary and onclusions 29

30 RF Envelope Predistortion Predistortion is performed directly on the modulated RF carrier using a high-speed vector modulator VMOD is driven by a look-up table (LU) that is indexed by the instantaneous input power level Kusunoki, et al., demonstrated 6-7 db APR improvement without adaptive feedback* *.Kusunoki, et al., Power Amplifier Module with Digital Adaptive Predistortion for ellular Phone, 2002 IEEE M- Int. Microwave ymp. Dig., pp , eattle, WA, June 4-6,

31 Adaptive RF Envelope Pre-Distortion est Bed 31

32 FPGA LU Implementation (Xilinx Vertex-II) LVD: Low Voltage Differential ignaling UAR: Universal Asynchronous Receiver ransmitter DM: Digital lock Manager 32

33 RF Envelope Pre-Distortion AD imulations dbm Mega_Hertz (a) 2x ampling (b) w/o Pre-D (c) 4x ampling (d) ource APR Improvement, db nbits AD for DP AD for LU DA for VMOD 33

34 RF Envelope Pre-D est Bed τ LU 34

35 ignal Integrity AD-LU-DA Path Agilent 33120A F/AWG (15MHz) 10MHz ine ignal P R-232 D 3054 DPO (500MHz, 5G/s) LU out 10bit/ 100MHz 12bit/ 100MHz I Q AD EV Board [AD9214] (105MHz) LU Board DA EV Board [AD9765] (125MHz) LU in Agilent 1692A 68ch LA Linear function loaded into LU estbed diagram 35

36 ignal Integrity AD-LU-DA Path (cont.) (3-a) 10MHz input, No filtering (3-b) 10MHz input, Filtering (fc=20mhz) (3-c) Glitch due to the crosstalk of data lines and overflow 36

37 Outline Introduction to Predistortion LU Updates using ub-sampling Receivers RF Envelope Predistortion ummary and onclusions 37

38 ummary and onclusions ub-sampling architecture developed to reduce sampling and processing requirements ample rate 2 x BW of input signal Memory effects may increase this RF Envelope pre-d is an alternative to baseband digital pre-d No digital I/Q stream required Lower power, high BW 38