1201
1-4244-1176-9/07/$25.00 ©2007 IEEE.
Digital Predistorter Architecture with Small Signal
Gain Control for Highly Nonlinear RF Power
Amplifiers
Oualid Hammi, Slim Boumaiza,
and Fadhel M. Ghannouchi
iRadio Lab., Dept. of Electrical and Computer Engineering
Schulich School of Engineering, University of Calgary
Calgary, AB, Canada
ohammi@ucalgary.ca
Bill Vassilakis
Powerwave Technologies, Inc.
Santa Ana, CA, USA
Abstract—In this paper, a digital predistorter (DPD)
architecture is proposed for the linearization of highly
nonlinear RF power amplifiers. This digital predistorter
architecture uses a complexity reduced and a computationally
efficient procedure to synthesis the predistortion function.
Unlike conventional digital predistortion architectures that
require more than one characterization to get a perfect match
between the PA’s nonlinearity and that of the DPD, the
proposed architecture uses a single characterization and
iteratively optimizes the predistortion function performance by
controlling the predistorter’s small signal gain. Experimental
validation carried on a highly nonlinear RF power amplifier
demonstrates the ability of the predistorter’s small signal gain
control to improve the linearity performance.
I. INTRODUCTION
RF power amplifiers deployed in current wireless
communications base stations are required to meet stringent
linearity requirements while achieving the highest possible
power efficiency. The linearity constraint, that consists in
meeting the spectrum emission mask requirements, is due to
the nature of the transmitted signals which have non constant
envelops. In addition, the efficiency of the power amplifier
needs to be maximized since it will dominate the running
costs of the base stations in terms of DC power consumption
and power dissipation. This calls for linearity versus
efficiency trade-off. Such trade-off is very critical in the case
of modern communication signals used in third generation
(3G), 3G and beyond (3G+) and WiMAX systems. In fact,
these communication standards employ advanced code
division multiple access (CDMA) and orthogonal frequency
division multiplexing (OFDM) techniques that result in time
domain signals having high peak to average power ratio
(PAPR). Accordingly, the use of brute force power
amplifiers requires considerable back-off levels and will lead
to poor power efficiency.
Currently, linearization techniques are being considered
to extend the linear region of continuously driven power
amplifiers. For a given linearity level, this reduces the
required back-off and increases the achievable power
efficiency [1]-[3]. Among the linearization techniques that
have been reported in the literature, predistortion, and
especially digital predistortion, is the most suitable for base
station power amplifiers applications. In fact, feedforward
systems are complex in nature and usually lead to low
overall power efficiency due to the linear amplifier used in
the distortion cancellation loop. In addition, feedback
systems are very narrowband and are unable to handle the
signal bandwidths used with base station power amplifiers
and especially multi-carriers power amplifiers (MCPA). In
contrast, predistortion technique presents inherent ease of
implementation, low cost and good linearity versus
efficiency trade-off in comparison with feedforward and
feedback based linearizers. Furthermore, the use of digital
predistortion, that has the ability to accurately synthesize the
AM/AM and AM/PM compensation functions, makes
possible the migration from mildly nonlinear power
amplifiers to highly nonlinear power amplifiers. This further
boosts the overall power efficiency of the linearized power
amplifier.
The performance of digital predistortion systems
fundamentally relies on the perfect match between the
synthesized predistortion functions and the actual nonlinear
characteristics of the power amplifier. Since the behavior of
the power amplifier depends on the statistics and bandwidth
of the input signal (continuous wave (CW), two tones,
CDMA, etc,…) [4] as well as its average power [5]-[6], it is
important to keep the same signal type and especially the
same average power at the input of the power amplifier
between the characterization and the linearization steps. The
use of the instantaneous input and output waveforms based
characterization technique reported in [4] ensures the first
condition relative to the nature of the driving signal. For
This work was supported by the Alberta Informatics Circle of
Research Excellence (iCORE), the Natural Sciences and Engineering
Research Council of Canada (NSERC), and Canada Research Chairs
(CRC).