On the Channel Memory-Diversity Tradeoff in Communication Systems

Abstract

In this chapter we investigate the tradeoff between channel estimation and channel diversity for channels with memory. Block memory channels are analytically tractable yet provide practical models for many communication systems. We show that for finite block length transmission techniques there is a fundamental tradeoff between the channel estimator and the channel diversity. When the channel is operating at rates much below capacity, smaller channel memory is better since channel estimation is not crucial. This is because the code contains more than adequate redundancy to compensate for mismatch due to the potentially poor channel estimate as well as channel errors. However, at rates close to capacity, channel state information is crucial for successful decoding and therefore longer channel memory, which allows better estimates, is better. We draw these conclusions based on analysis of some simple discrete-input, discrete-output channels using the channel reliability function and on simulation results for more realistic channels (soft output) using low-density parity check codes with a joint iterative decoder and channel estimator.

This work was supported by the the Army Research Office under grant DAAH0496-1-0377, a National Science Foundation Graduate Research Fellowship and an AFCEA fellowship.