IEEE NJ Coast Seminar, 12/01/2008 Walid Ahmed et. al. Performance Analysis of a Novel Real-Time Closed-Loop Technique for Non-Linearity Correction of Power Amplifiers Walid K. M. Ahmed, Senior Member IEEE Qing Li, Member IEEE Ajit Reddy, Senior Member IEEE This work has been done while the authors were with Tyco Electronics
27
Embed
IEEE NJ Coast Seminar, 12/01/2008Walid Ahmed et. al. Performance Analysis of a Novel Real- Time Closed-Loop Technique for Non- Linearity Correction of.
This document is posted to help you gain knowledge. Please leave a comment to let me know what you think about it! Share it to your friends and learn new things together.
Transcript
IEEE NJ Coast Seminar, 12/01/2008Walid Ahmed et. al.
Performance Analysis of a Novel Real-Time Closed-Loop Technique for Non-
Linearity Correction of Power Amplifiers
Walid K. M. Ahmed, Senior Member IEEE
Qing Li, Member IEEE
Ajit Reddy, Senior Member IEEEThis work has been done while the authors were with Tyco Electronics
IEEE NJ Coast Seminar, 10/28/2008Walid K. M. Ahmed
Outline
General Closed-Loop Block Diagram. Motivation: The EDGE/GSM Challenge! Predictive Algorithm. DTx System Model and PA Closed-
Loop Correction. Results.
IEEE NJ Coast Seminar, 10/28/2008Walid K. M. Ahmed
Closed-Loop Approach: General
Generic Pictorial Illustration of a Block Diagram for the Proposed Technique
Target/Reference
Signal
Non-LinearDevice
Detection/Measurement
Module
Pre-DistortionModule
IEEE NJ Coast Seminar, 10/28/2008Walid K. M. Ahmed
Motivation: The EDGE/GSM Challenge
Highly efficient PA implies severe non-linearity. Hence, PA characteristics need to be corrected for, or linearized, in baseband, which involves pre-distortion of the baseband signal prior to modulating the PA. Otherwise, acceptable spectral and waveform quality performance can not be achieved for the transmitted signal.
The severe PA non-linearity is also expected to change versus many parameters, such as temperature, PA power level setting (PA bias), PA load (VSWR), PA part-to-part tolerance and aging.
Open-loop (factory calibration) of such a non-linear PA behavior is difficult since it is not possible to store such a large number of PA correction table versus so many parameters.
On-chip non-real-time closed-loop solutions, e.g., calibrate the PA before every call or when turning the phone on and use the generated tables for the entire transmission duration, may not work if the PA is expected to change its AMAM/AMPM behavior during transmission due to, for example, load and VSWR changes.
IEEE NJ Coast Seminar, 10/28/2008Walid K. M. Ahmed
Motivation: The EDGE/GSM Challenge Hence, frequent calibration is needed, which implies
the need for an adaptive closed-loop solution. Two challenges come with such an approach:
When to calibrate? Standards usually do not allocate off-air time for such an operation to be done!
How to maintain signal quality while calibrating and transmitting at the same time?
IEEE NJ Coast Seminar, 10/28/2008Walid K. M. Ahmed
Motivation: The EDGE/GSM Challenge
In EDGE/GSM, for example, the switching-transient requirements are very stringent. It is not possible to let the PA run free of correction (i.e., without linearization) at any point in time. It is also not possible to have inaccurate correction tables applied to the PA. Moreover, a high-efficient PA will have a high level of non-linearity that will change from one EDGE/GSM burst to the next due to (at least) load/VSWR changes. Hence, pre-stored (or pre-loaded) correction tables will not suffice.
The question then becomes: How and when do we calibrate the highly-non-linear PA in an EDGE/GSM application EVERY BURST?!!
IEEE NJ Coast Seminar, 10/28/2008Walid K. M. Ahmed
Motivation: The EDGE/GSM Challenge
The answer to the question of WHEN is attained by calibrating the PA on the EDGE/GSM burst ramp-up.
In order to do that, the power ramp-up profile must be designed in a special way to ensure all amplitude levels are covered, yet a smooth rising ramp is achieved in order not to violate the tough EDGE/GSM switching-transients requirements. A simple, novel, robust and arbitrarily tunable ramp-up
design technique has been proposed by Ahmed (see Walid K. M. Ahmed, “Method and Apparatus for Signal Power Ramp-Up in a Communication Transmitter”. Patent pending, U.S. Patent Serial #
11/385,212).
IEEE NJ Coast Seminar, 10/28/2008Walid K. M. Ahmed
Motivation: The EDGE/GSM Challenge
Augm
ente
d Sym
bols
for R
ampin
g up
Tail Symbols of a Normal Burst
IEEE NJ Coast Seminar, 10/28/2008Walid K. M. Ahmed
Motivation: The EDGE/GSM Challenge
The answer to the question of HOW is the focus of this presentation.
IEEE NJ Coast Seminar, 10/28/2008Walid K. M. Ahmed
Motivation: The EDGE/GSM Challenge
...The chicken and egg problem: A major challenge in calibrating on the ramp is that while collecting the PA
response data, we are also transmitting! Hence, we must have correction data available to ensure quality transmission…but didn’t I just say we don’t even have correction yet since we are still calibrating!!
Moreover, the attempt at performing such an on-the-fly correction/linearization of the PA while ramping up will suffer from memory effects, that is, the collected measurements of the PA AM/AM and AM/PM non-linearity lag behind the current bias point of the PA due to the inevitable forward and feedback loop path delay. Hence, resulting-in inaccurate correction (or inversion of the PA non-linearity), which clearly leads to un-acceptable performance that fails the stringent requirements of the EDGE/GSM specifications.
Accordingly, one must use a mechanism that can break into the future in some sense and predict what the next non-linearity point on the PA curve would look like based on previous on-the-fly measurements!
A novel algorithm has been proposed by Ahmed and Li (see Walid K. M. Ahmed and Qing Li, “Method and Apparatus for a Non-Linear Feedback Control System”. Patent pending)
IEEE NJ Coast Seminar, 10/28/2008Walid K. M. Ahmed
Motivation: The EDGE/GSM Challenge
An Example of EDGE time-burst/time-mask/ST graph
IEEE NJ Coast Seminar, 10/28/2008Walid K. M. Ahmed
Motivation: The EDGE/GSM Challenge
ST performance without AM/PM correction during ramp-up
IEEE NJ Coast Seminar, 10/28/2008Walid K. M. Ahmed
Predictive Algorithm
PA Output
(y)Actually
measured
PA Input
(x)
PA Output
Actually
measured
Required Point
PA Input
( )yfx 1-=
Begin Prediction
Path Delay Elapsed?
Path Delay Elapsed?
Extrapolate Corrected PA Input
Extrapolate Corrected PA Input
Apply PA Input Uncorrected, or from Preloads
Apply PA Input Uncorrected, or from Preloads
Prediction Time Over?Prediction
Time Over?
Stop and Use Measured Data as Correction LUT for Modulation
Stop and Use Measured Data as Correction LUT for Modulation
Simplified Pictorial Illustration of the Non-Linear Predictive Algorithm
IEEE NJ Coast Seminar, 10/28/2008Walid K. M. Ahmed
Predictive Algorithm
Xi-1 Xi
dX = Xi– Xi-1
dy = yi – yi-1yi-1
yi
Xi
Yidy/dx
If for any xi in set X, there is a yi in set Y following the functionyi = f(xi) (i = 1, 2, …n),
then we can find Yi = f(Xi) = f(xi) + f ’(xi)*(Xi – xi) + 1/(2!) f ’’(xi)*(Xi – xi)2 + …
If (Xi – xi) is small enough, we can simplify it to linear (1st order) extrapolating:Yi = f(Xi) f(xi) + f ’(xi) * (Xi – xi) = yi + dy/dx * (Xi – xi)
yn
xi yi
yi = f(xi)x1
x2
xn y1 y2
Xi Yi
IEEE NJ Coast Seminar, 10/28/2008Walid K. M. Ahmed
DTx System Model
AmplitudeCorrection
+ PhaseModulator
AFE
Antenna
PA
Low Band
High Band
PA
PhaseCorrection
AmplitudeEstimation
PhaseEstimation
Detector
Multi DAC
Sine/CosineGeneration
Phase
Amplitude AmplitudeCorrection
+ PhaseModulator
AFE
Antenna
PA
Low Band
High Band
PA
PhaseCorrection
AmplitudeEstimation
PhaseEstimation
Detector
Multi DAC
Sine/CosineGeneration
Phase
AmplitudeA “differential-phase” down-converter is used, i.e., RF input and RF output of the PA are mixed to save an extra un-modulated LO and to obtain AMAM while “zoom” into AMPM (see Walid K. M. Ahmed and Dale Douglas, “Multi-Mode Selectable Modulation Architecture Calibration and Power Control Apparatus, System, and Method for Radio Frequency Power Amplifier”. Patent pending, U.S. Patent Serial # 11/347,455).
IEEE NJ Coast Seminar, 10/28/2008Walid K. M. Ahmed
PA Closed-Loop Correction
EDGE AMSignal
EDGE PMSignal
Extrapolation/PredictionAlgorithm
(Used duringRamp-up onlyuntil full LUTs
are built.Then LUTs are
used for modulationand Ramp-down)
AMPath
PMPath
PA
AMAMand
AMPMDetection
CorrectedAM
CorrectedPM
IEEE NJ Coast Seminar, 10/28/2008Walid K. M. Ahmed
Performance Results
Amplitude Predictor
Phase Predictor
Numerical noise is filtered by a 3rd order Bessel filter
IEEE NJ Coast Seminar, 10/28/2008Walid K. M. Ahmed
Performance Results
DTx EDGE/GSM PA Characteristic Examples:Low Band
High Band
AMAM AMPM
AMAM AMPM0 100 200 300 400 500 600
0
5
10
15
--> States
-->
Vol
ts
0 100 200 300 400 500 600-40
-20
0
20
40
60
80
100
120
--> States
-->
Deg
rees
0 100 200 300 400 500 6000
2
4
6
8
10
12
14
16
--> States
-->
Vol
ts
0 100 200 300 400 500 6000
50
100
150
200
250
--> States
-->
Deg
rees
Low Band
High Band
AMAM AMPM
AMAM AMPM0 100 200 300 400 500 600
0
5
10
15
--> States
-->
Vol
ts
0 100 200 300 400 500 600-40
-20
0
20
40
60
80
100
120
--> States
-->
Deg
rees
0 100 200 300 400 500 6000
2
4
6
8
10
12
14
16
--> States
-->
Vol
ts
0 100 200 300 400 500 6000
50
100
150
200
250
--> States
-->
Deg
rees
IEEE NJ Coast Seminar, 10/28/2008Walid K. M. Ahmed
Performance Results EDGE/GSM Switched Transient and ACPR Specification
Center Frequency (kHz)
Bandwidth (kHz)
ACPR Limit (dBm)
ST Limit
(dBm)
400 30 -54 -23
600 30 -60 -26
1200 30 -60 -32
1800 100 -65 -36
IEEE NJ Coast Seminar, 10/28/2008Walid K. M. Ahmed
Performance Results
GSM Low Band Performance:
0 5 10 15 20 25 30 35 400
5
10
15
20
25
30
35
40
45
50Switched Transient Margin at Antenna
Power Setting in (dBm) -->
Mar
gin
in (
dB)
-->
ST400
ST600ST1200
ST1800
0 5 10 15 20 25 30 35 40-110
-105
-100
-95
-90
-85
-80
-75
-70ACPR
Power Setting in (dBm) -->
AC
PR
in (
dBc)
-->
ACPR400
ACPR600
ACPR1200
ACPR1800
0 5 10 15 20 25 30 35 40-5
0
5
10
15
20
25
30
35Peak, Average Power
Power Setting in (dBm) -->
Out
put
Pow
er in
(dB
m)
-->
Peak Power
Average Power
0 5 10 15 20 25 30 35 40-10
0
10
20
30
40
50
60PAE
Power Setting in (dBm) -->
PA
E in
(%
) --
>
PA
ANT
0 5 10 15 20 25 30 35 400
5
10
15
20
25
30
35
40
45
50Switched Transient Margin at Antenna
Power Setting in (dBm) -->
Mar
gin
in (
dB)
-->
ST400
ST600ST1200
ST1800
0 5 10 15 20 25 30 35 40-110
-105
-100
-95
-90
-85
-80
-75
-70ACPR
Power Setting in (dBm) -->
AC
PR
in (
dBc)
-->
ACPR400
ACPR600
ACPR1200
ACPR1800
0 5 10 15 20 25 30 35 40-5
0
5
10
15
20
25
30
35Peak, Average Power
Power Setting in (dBm) -->
Out
put
Pow
er in
(dB
m)
-->
Peak Power
Average Power
0 5 10 15 20 25 30 35 40-10
0
10
20
30
40
50
60PAE
Power Setting in (dBm) -->
PA
E in
(%
) --
>
PA
ANT
IEEE NJ Coast Seminar, 10/28/2008Walid K. M. Ahmed
Performance Results
GSM High Band Performance:
0 5 10 15 20 25 30 350
5
10
15
20
25
30
35
40
45
50Switched Transient Margin at Antenna
Power Setting in (dBm) -->
Mar
gin
in (
dB)
-->
ST400
ST600ST1200
ST1800
0 5 10 15 20 25 30 35-110
-105
-100
-95
-90
-85
-80
-75
-70ACPR
Power Setting in (dBm) -->
AC
PR
in (
dBc
) --
>
ACPR400
ACPR600
ACPR1200
ACPR1800
0 5 10 15 20 25 30 35-5
0
5
10
15
20
25
30
35Peak, Average Power
Power Setting in (dBm) -->
Ou
tpu
t P
ow
er i
n (d
Bm
) --
>
Peak Power
Average Power
0 5 10 15 20 25 30 35-10
0
10
20
30
40
50
60
70PAE
Power Setting in (dBm) -->
PA
E in
(%
) --
>
PA
ANT
0 5 10 15 20 25 30 350
5
10
15
20
25
30
35
40
45
50Switched Transient Margin at Antenna
Power Setting in (dBm) -->
Mar
gin
in (
dB)
-->
ST400
ST600ST1200
ST1800
0 5 10 15 20 25 30 35-110
-105
-100
-95
-90
-85
-80
-75
-70ACPR
Power Setting in (dBm) -->
AC
PR
in (
dBc
) --
>
ACPR400
ACPR600
ACPR1200
ACPR1800
0 5 10 15 20 25 30 35-5
0
5
10
15
20
25
30
35Peak, Average Power
Power Setting in (dBm) -->
Ou
tpu
t P
ow
er i
n (d
Bm
) --
>
Peak Power
Average Power
0 5 10 15 20 25 30 35-10
0
10
20
30
40
50
60
70PAE
Power Setting in (dBm) -->
PA
E in
(%
) --
>
PA
ANT
IEEE NJ Coast Seminar, 10/28/2008Walid K. M. Ahmed
Performance Results
EDGE Low Band Performance:
0 5 10 15 20 25 30 350
5
10
15
20
25
30
35
40
45
50Switched Transient Margin at Antenna
Power Setting in (dBm) -->
Mar
gin
in
(dB
) --
>
ST400
ST600ST1200
ST1800
0 5 10 15 20 25 30 35-90
-85
-80
-75
-70
-65
-60
-55ACPR
Power Setting in (dBm) -->
AC
PR
in
(dB
c) -
->
ACPR400
ACPR600
ACPR1200
ACPR1800
0 5 10 15 20 25 30 35-5
0
5
10
15
20
25
30
35Peak, Average Power
Power Setting in (dBm) -->
Out
put
Pow
er
in (
dB
m)
-->
Peak Power
Average Power
0 5 10 15 20 25 30 35-10
0
10
20
30
40
50PAE
Power Setting in (dBm) -->
PA
E in
(%
) --
>
PA
ANT
0 5 10 15 20 25 30 350.54
0.56
0.58
0.6
0.62
0.64
0.66
0.68
0.7
0.72
0.74EVM
Power Setting in (dBm) -->
EV
M in
(%
) --
>
0 5 10 15 20 25 30 350
5
10
15
20
25
30
35
40
45
50Switched Transient Margin at Antenna
Power Setting in (dBm) -->
Mar
gin
in
(dB
) --
>
ST400
ST600ST1200
ST1800
0 5 10 15 20 25 30 35-90
-85
-80
-75
-70
-65
-60
-55ACPR
Power Setting in (dBm) -->
AC
PR
in
(dB
c) -
->
ACPR400
ACPR600
ACPR1200
ACPR1800
0 5 10 15 20 25 30 35-5
0
5
10
15
20
25
30
35Peak, Average Power
Power Setting in (dBm) -->
Out
put
Pow
er
in (
dB
m)
-->
Peak Power
Average Power
0 5 10 15 20 25 30 35-10
0
10
20
30
40
50PAE
Power Setting in (dBm) -->
PA
E in
(%
) --
>
PA
ANT
0 5 10 15 20 25 30 350.54
0.56
0.58
0.6
0.62
0.64
0.66
0.68
0.7
0.72
0.74EVM
Power Setting in (dBm) -->
EV
M in
(%
) --
>
IEEE NJ Coast Seminar, 10/28/2008Walid K. M. Ahmed
Performance Results
EDGE High Band Performance:
0 5 10 15 20 25 30 350
5
10
15
20
25
30
35
40
45Switched Transient Margin at Antenna
Power Setting in (dBm) -->
Ma
rgin
in (
dB)
-->
ST400
ST600ST1200
ST1800
0 5 10 15 20 25 30 35-90
-85
-80
-75
-70
-65
-60
-55
-50ACPR
Power Setting in (dBm) -->
AC
PR
in (
dBc
) --
>
ACPR400
ACPR600
ACPR1200
ACPR1800
0 5 10 15 20 25 30 35-10
-5
0
5
10
15
20
25
30
35Peak, Average Power
Power Setting in (dBm) -->
Ou
tput
Po
we
r in
(dB
m)
-->
Peak Power
Average Power
0 5 10 15 20 25 30 35-10
0
10
20
30
40
50PAE
Power Setting in (dBm) -->
PA
E i
n (%
) --
>
PA
ANT
0 5 10 15 20 25 30 350.5
0.6
0.7
0.8
0.9
1
1.1
1.2
1.3
1.4
1.5EVM
Power Setting in (dBm) -->
EV
M i
n (
%)
-->
0 5 10 15 20 25 30 350
5
10
15
20
25
30
35
40
45Switched Transient Margin at Antenna
Power Setting in (dBm) -->
Ma
rgin
in (
dB)
-->
ST400
ST600ST1200
ST1800
0 5 10 15 20 25 30 35-90
-85
-80
-75
-70
-65
-60
-55
-50ACPR
Power Setting in (dBm) -->
AC
PR
in (
dBc
) --
>
ACPR400
ACPR600
ACPR1200
ACPR1800
0 5 10 15 20 25 30 35-10
-5
0
5
10
15
20
25
30
35Peak, Average Power
Power Setting in (dBm) -->
Ou
tput
Po
we
r in
(dB
m)
-->
Peak Power
Average Power
0 5 10 15 20 25 30 35-10
0
10
20
30
40
50PAE
Power Setting in (dBm) -->
PA
E i
n (%
) --
>
PA
ANT
0 5 10 15 20 25 30 350.5
0.6
0.7
0.8
0.9
1
1.1
1.2
1.3
1.4
1.5EVM
Power Setting in (dBm) -->
EV
M i
n (
%)
-->
IEEE NJ Coast Seminar, 10/28/2008Walid K. M. Ahmed
Power pk: PA:33.9dBm Ant:32.1dBmPower Avg: PA:30.8dBm Ant:29dBmMod. PAE PA:45.6% Ant:30.0%Avg Current 750mA
EVM 0.649%
IEEE NJ Coast Seminar, 10/28/2008Walid K. M. Ahmed
Conclusions
A novel predictive real-time closed-loop non-linearity correction technique has been proposed and tested.
Despite the severe non-linearity of the PA, the predictive algorithm has been able to achieve healthy performance margins compared to the stringent limits set by the EDGE/GSM standard.
The results indicate that the predictor based closed loop approach is very promising compared to other linearization techniques.
It offers a simple, yet robust, efficient and intelligent solution that is not-only independent of modulation/transmission protocol constraints, but is also flexible enough to handle a wide range of PA non-linearity levels, in order to make the PA correction reliable and robust against variations due to temperature, process, power and aging.
It should be noted that this technique can be applied to any PA modulation architecture (e.g., polar or IQ).
IEEE NJ Coast Seminar, 10/28/2008Walid K. M. Ahmed
Conclusions, continued
Also, this technique can be applied during actual modulation/data-transmission periods (i.e., not necessarily on a ramp-up time such as in EDGE/GSM) if the performance specifications allow for that. This is possible in interference-averaging modulation schemes such as CDMA technologies, where switching transients are not of critical importance. Hence, we can afford longer prediction periods in order to allow for sometime during modulation such that the signal covers all possible modulation signal levels (to collect a complete correction table). Clearly, in such a scenario, we can design the predictor to predict in both directions (i.e., forward and backwards prediction).
This technique can also be enhanced to perform PA compression prediction. Hence, in addition to correcting for the PA non-linearity on-the-fly, the algorithm can also predict when the PA gets into compression and prevents such an event to avoid future signal clipping if the PA is driven hard into compression (see Walid K. M. Ahmed, “Algorithm, Method and Apparatus for Real-Time Adaptive Compression-Control in Power Amplifiers”. Patent