A Robust and Low Sampling Rate Digital Predistortion Algorithm for Broadband
PA Modeling and Predistortion
Gongzhe Su, Wenhua Chen, Silong Zhang
Department of Electronic Engineering
Tsinghua University
Beijing, China
e-mail: chenwh@tsinghua.edu.cn
Fadhel M Ghannouchi
Department of Electronic and Computer Engineering
University of Calgary
Calgary, Canada
e-mail: fghannou@ucalgary.ca
Abstract—A robust and low sampling rate digital predistortion
(DPD) algorithm for broadband power amplifiers (PAs) is
proposed in this paper. To alleviate the burden on the
sampling rate, an approximated model extracted by band-
limited signals is employed to approach the broadband
behavior. Due to the assistance of the low pass filter (LPF) in
the feedback loop and the adaptive basis function scheme, the
sampling rate can be reduced drastically and the algorithm is
more robust. Experimental results show that more than 21 dB
adjacent channel power ratio (ACPR) improvement is
achieved for 60MHz 3-carrier LTE signal application only with
92.16Msps sampling rate, and the cavity RF filter in the
feedback loop is not necessary any more.
Keywords-Broadband Modeling, Digital Predistortion,
Memory Polynomial, Power Amplifiers.
I. INTRODUCTION
Digital predistortion is widely recognized as an effective
and robust method among all the linearization techniques.
Based on behavioral modeling, the predistorter can easily
predict the PA output and compensate for the distortions [1].
To ensure the accuracy of behavioral models, at least 5x
bandwidth expansion is required for the feedback signal path
[2]. In the last decade, the signal bandwidth of wireless
systems increases to 100MHz. Therefore, the DPD system
requires 500MHz bandwidth in feedback path and more than
500Msps sampling rate analog to digital converters (ADCs),
which is not affordable and practically impossible.
Recently, several approaches were proposed to solve this
problem. In [2], a band-limited Volterra series based DPD
was presented by Yu et al., which uses indirect learning
architecture and bandwidth-constrained least square (LS)
method [3] for model extraction. With the assistance of
cavity filters, ADC sampling rate is reduced and in-band
distortion is compensated perfectly. But the intermodulation
distortion (IMD) out of the filter band cannot be controlled
without the cavity filter at the PA output. In [4], Ying Liu
proposed a band-limited memory polynomial (BLMP)
model, and LS method was used to extract the forward
memory polynomial (MP) model with reduced sampling
rate. Based on this, a significant improvement is achieved.
However, this method may meet instability problem as the
number of model coefficients increase. Therefore, the
memory effects of PAs may not be well compensated.
DPD
Foward Model
Extraction
Inverse Model
Extraction
Coupler
DAC
ADC
Antenna
u(n) x(n)
y
F
(n)
y(n)
Anti-alias
LPF
-
DUT
DUT Model LPF Model
+
Digital Signal Processing
e(n)
z(n)
Down-
converter
Analog Circuits
r(n)
y
M
(n)
Fig. 1. System structure for proposed DPD algorithm
In this paper, a robust and low sampling rate DPD
algorithm is proposed for broadband PA model extraction.
An approximated forward model is extracted through
iteration with adaptive basis function scheme. Then, the
DPD predistorter can be obtained by model inversion.
Therefore, with an ultra low sampling rate in feedback
signals, a significant improvement of output frequency
regeneration in wideband spectrum can be achieved and the
output cavity filter can be removed.
II. DIGITAL PREDISTORTION FOR BROADBAND
APPLICATIONS
A. System Structure
Fig. 1 shows the system structure for proposed DPD
algorithm. The block in red dot lines shows the forward PA
model extraction procedure, which uses an iterative method
with low sampling rate feedback signals based on the
adaptive basis function scheme. During each iteration, the
most significant basis functions are dynamically selected.
The obtained forward model is an approximated behavioral
model of PA. Based on this model, it is possible to estimate
the inverse model of PA which is marked as blue in Fig. 1. In
this approach, no microwave cavity filter at PA output is
necessary, only the baseband LPF is needed for anti-alias
filtering and harmonic suppression after down converting.
978-1-4799-460 / $31.00 ©2014 IEEE