When amplitude-modulated pulses are transmitted through a channel, time dispersion may occur which can cause an increase in error rate due to intersymbol interference. If additive white noise is present, the situation is worse. An optimum linear receiver for this situation consists of a Matched Filter followed by a Sampled Data Filter. The Sampled Data Filter consists of a sampler followed by a tapped delay line, the outputs of which are
weighted and summed to get the receiver output. The optimum weights or gains of the tap outputs can be calculated if the received pulse shape, signalling rate, and signal to noise level are all known and fixed.
As an improvement to initial calculation, several adaptive strategies have been proposed in the literature to optimise the tap gains on a continuous basis. This is done by measuring the receiver performance, using actual transmitted data, and adjusting the gains as required.
This thesis compares several promising adaptive schemes on the basis of: (a) speed of convergence, and (b) accuracy of convergence. The speed criterion used was: the total number of data samples from the start of adaptation that are required to adjust the tap weights to optimum values, with the tap gains initially set to zero. The accuracy criterion used was: the error rate of the system after sufficient samples had elapsed from
the start of adaptation for the error rate to stabilize at an approximately constant value.
The conclusion is that an iterative strategy developed by Dr. R.W. Lucky, of Bell Telephone laboratories, has superior performance and is also the easiest to implement.