Noncoherent Detection With Post-Detection Processing for digital FM Over Fading Channels

by

Abdulbaset Z. Atweri

M. Eng., 1995

Abstract

The thesis deals with the error rate performance of three binary narrow-band digital FM signals using Limiter-discriminator detection. The performance improvement due to several post-detection processing algorithms has been investigated. The signals considered are: minimum shift keying (MSK), Duobinary coded MSK and Gaussian filtered MSK (GMSK). Computer simulation was used extensively to study the receiver performance. The introduction of Inter-symbol Interference (ISI) through pre-modulation and pre-detection filtering severely degrades the receiver performance of the simple bit-by-bit detection algorithm. A K-state Viterbi equalizer, following an integrate and dump (I&D) post-detection filter, was used to reduce the effects of ISI and therefore improving the error rate performance of the receiver.

The BER performance was first investigated over an AWGN channel. It was shown that the I&D filter is more effective with 2-bit integration period. We consider two implementations of the Viterbi algorithm: A simple version based on the assumption that the noise is Gaussian at the output of the receiver, and a second complex version that takes the exact noise distribution. WE found that the second implementation outperforms the first one by 0.8 dB and 0.1 dB at error rates of 10^-2 and 10^-5 respectively. For the cases considered, the receiver provides the best performance when the 2-bit I&D filter is followed by a 4-state Viterbi equalizer. For the receiver pre-detection IF filter, a bandwidth-time product BT of 1 was approximately optimum.

The performance investigation of the receiver was then extended to the case of mobile fading channel using GMSK with B_tT product of 0.35 and h=0.5. The post-detection arrangements investigated are the 1-bit and 2-bit I&D filter followed by 4-state Viterbi equalizer. the performance of a receiver over a channel corrupted by co-channel interference and flat Rayleigh fading has been greatly improved compared to the conventional detection techniques. The performance of the receiver over a frequency-selective fading channel was also studied using a two-ray channel model. The effects of time dispersion on the eye pattern characteristics was first examined. Finally the error rate performance was studied. It was found that great improvements can be achieved when the equalizer is adapted to the channel conditions. With perfect adaptation the error floor can be reduced an order of magnitude compared to the fixed equalization case.

Back to the list of graduate students