Approved
Nonlinear pre-distortion using convex optimization
Yezi Huang ()
Start
2011-11-01
Presentation
2012-08-31
Location:
Finished:
2012-10-18
Master's thesis:
(Contact supervisor)
Abstract
In modern wireless communication, orthogonal frequency division multiplexing (OFDM) stands out as a popular due to its robustness to inter-symbol interference (ISI) caused by multi-path distortion. It is the multi-carrier scheme of OFDM system that enables high throughput and high spectral efficiency. However, those benefits are often dimmed by its inherent drawback of high peak–to–average–ratio (PAR). The PAR often leads to low power efficiency in the following power amplifier at the front–end of transmitter. Moreover, the non-linearity of the power amplifier distorting those peaks causes non-linear distortion of the signal at transmitter and bit–error–rate (BER) scheme degradation at the receiver. One way to remedy this problem is to use a powerful PAR reduction scheme before feeding the time–domain baseband signal to the power amplifier. Peak–clipping is a simple signal pre-distortion method to reduce PAR level, but it makes the whole system suffer from both in–band and out–of–band distortions. Many algorithms based on peak–clipping have been developed during the last two decades. Among those techniques, the active constellation extension (ACE) scheme was introduced as an effective way because it not only brought down the PAR level at transmitter, but also maintained the BER performance at receiver. The conventional ACE algorithm, however, is suboptimal in the sense of minimizing PAR but it has low complexity. Current development in computing and semi-conductor technology keeps increasing the admissible algorithm complexity. In this thesis, we propose an improved ACE–based PAR reducing solution by involving convex optimization. The proposed method is also extended to multiple antenna systems which utilize the spatial factor with a pre-processing step. Several factors that influence the performance of the proposed method are investigated. The complementary cumulative distribution function (CCDF) curve of PAR is used as a primary metric for evaluating PAR reduction level at transmitter. Furthermore, the BER performance at the receiver is measured in several system schemes. Performance comparisons are carried out comparing the non-distorted original signal, the peak-clipped signal, the conventional ACE-performed signal and the signal processed by our proposed method.
Supervisor: Thomas Magesacher (EIT)
Examiner: Stefan Höst (EIT)