Topics in nonlinear signal processing and system identification

1
Topics in nonlinear signal processing and system
identification
Adaptive Volterra Filter for Pre-distorsion of a
RF High Power Amplifier
2
Source of information

• J. Tsimbinos Identification and Compensation of
  Non-Linear Distorsion, 1995.
• Chapter 5 Identification and Compensation of
  Nonlinearities With Memory Using the Volterra and
  Wiener Models.

3
Introduction

• Non-linearities in RF HPA (Radio Frequency High
  Power Amplifier) is a large problem, since it can
  cause Amplitude/Phase distortion and can also
  cause ISI (Inter-Symbol-Interference).
• This can create difficulties in Digital Radio
  Systems, especially for multicarrier techniques
  like WCDMA.

4
RF High Power Amplifier
Transmitter The D/A converter, the up-converter
and the amplifier all contain Non-Linear
elements which will distort the signal. In this
presentation we will look on the HPA.
5
RF High Power Amplifier

• Example of a RF HPA characteristics (Data from a
  TWTA)

6
RF High Power Amplifier

• The Non-Linearities of the HPA will distort the
  signal we want to transmit. A 16-QAM signal is
  shown .

Before HPA
After HPA
7
RF High Power Amplifier
The real part of the 16-QAM signal shown before
HPA (blue) and after (red).
8
Pre-distorter
By pre-distorting the 16-QAM signal with the
inverse of the Non-Linear system we will get the
original signal out from the non-linear HPA.
H
F
x(n)
u(n)
y(n)
9
Pre-distorter
Before pre-distorter
After pre-distorter
After HPA
10
Pre-distorter
The real part of the 16-QAM signal shown before
pre-distorter (blue) and after HPA (red).
11
Adaptive pre-distorter
When e(n) approaches zero, d(n) will approach
s(n).
s(n)
e(n)
12
Adaptive pre-distorter

• The filters w are adaptive Volterra filters.
• The Non-linear system can be described using a
  Volterra series

13
Adaptive pre-distorter

• In the adaptive pre-distorter we are using a
  truncated Volterra filter (second order)
• An LMS-type of algorithm is used for updating the
  Volterra coefficients.

14
Adaptive pre-distorter
HPA output without pre-distorsion
Before pre-distorter
Error signal of the adaptive algorithm
HPA output with adaptive pre-distorter
15
Conclusion

• Adaptive Volterra algorithm is fast and can
  handle signals with large amplitude and phase
  shifts like a WCDMA signal.
• It is very difficult to get a stable algorithm
  and it is very sensitive to choice of stepsize
  for the adaptive algorithm.