Channel Estimation in Coded Modulation Systems

Abstract

With the outstanding performance of coded modulation techniques in fading channels, much research efforts have been carried out on the design of communication systems able to operate at low signal-to-noise ratios (SNRs). From this perspective, the so-called iterative decoding principle has been applied to many signal processing tasks at the receiver: demodulation, detection, decoding, synchronization and channel estimation. Nevertheless, at low SNRs, conventional channel estimators do not perform satisfactorily. This thesis is mainly concerned with channel estimation issues in coded modulation systems where different diversity techniques are exploited to combat fading in single or multiple antenna systems.

First, for single antenna systems in fast time-varying fading channels, the thesis focuses on designing a training sequence by exploiting signal space diversity (SSD). Motivated by the power/bandwidth efficiency of the SSD technique, the proposed training sequence inserts pilot bits into the coded bits prior to constellation mapping and signal rotation. This scheme spreads the training sequence during a transmitted codeword and helps the estimator to track fast variation of the channel gains. A comprehensive comparison between the proposed training scheme and the conventional training scheme is then carried out, which reveals several interesting conclusions with respect to both error performance of the system and mean square error of the channel estimator.

For multiple antenna systems, different schemes are examined in this thesis for the estimation of block-fading channels. For typical coded modulation systems with multiple antennas, the first scheme makes a distinction between the iteration in the channel estimation and the iteration in the decoding. Then, the estimator begins iteration when the soft output of the decoder at the decoding iteration meets some specified reliability conditions. This scheme guarantees the convergence of the iterative receiver with iterative channel estimator. To accelerate the convergence process and decrease the complexity of successive iterations, in the second scheme, the channel estimator estimates channel state information (CSI) at each iteration with a combination of the training sequence and soft information. For coded modulation systems with precoding technique, in which a precoder is used after the modulator, the training sequence and data symbols are combined using a linear precoder to decrease the required training overhead. The power allocation and the placement of the training sequence to be precoded are obtained based on a lower bound on the mean square error of the channel estimation. It is demonstrated that considerable performance improvement is possible when the training symbols are embedded within data symbols with an equi-spaced pattern. In the last scheme, a joint precoder and training sequence is developed to maximize the achievable coding gain and diversity order under imperfect CSI. In particular, both the asymptotic performance behavior of the system with the precoded training scheme under imperfect CSI and the mean square error of the channel estimation are derived to obtain achievable diversity order and coding gain. Simulation results demonstrate that the joint optimized scheme outperforms the existing training schemes for systems with given precoders in terms of error rate and the amount of training overhead.