Detection of coded and distorted QAM signals

In a digital-transmission system operating at 9600 bit/s over a telephone circuit, the transmitted 2400-baud QAM signal can be convolutionally encoded from a 16-level signal into a 32-level signal, in order to obtain a better tolerance to noise. However, the available improvement in performance is only achieved if a maximum-likelihood detector is used at the receiver and if the detector takes full account both of the coding applied at the transmitter and of the signal distortion introduced by the channel. The paper uses the results of extensive computer-simulation tests to compare the performances of a wide range of different detection processes, for both coded and uncoded signals. The techniques considered include both novel and conventional detectors, and their comparison illuminates the factors in the design of a detector that lead to a good performance. The convolutional encoder used in the tests is the most promising of the trellis coders recently proposed for the given application, and the results of the tests are used to assess the improvement in performance likely to be achieved through the use of the given coder.