The industrial application of digital signal processing
This thesis describes an investigation into the application of digital signal processing techniques to the solution of industrial signal processing problems. The investigation took the form of three case studies chosen to illustrate the variety of possible applications. The first was the computer simulation of a digital microwave communications link which utilised narrowband FM modulation and partial response techniques. In order to ensure that the behaviour of the simulation reliably matched that of the modelled system it was found necessary to have a sound theoretical background, implementation using good software engineering methodology together with methodical testing and validation. The second case study was a comprehensive investigation of adaptive noise cancelling systems concentrating on issues important for practical implementation of the technique: stability and convergence of the adaptation algorithm; misadjustment noise and effects due to realizability constraints. It was found that theoretical predictions of the systems behaviour were in good agreement with the results of computer simulation except for the level of output misadjustment noise. In order to make the mathematics of the LMS algorithm tractable it was assumed that the input data formed a series of uncorrelated vectors. It was found that this assumption is only appropriate for the prediction of misadjustment noise when the reference input is uncorrelated. The final case study concerned the automatic detection and assessment of pressing faults on gramophone records for quality assurance purposes. A pattern recognition technique for identifying the signals due to gramophone record defects and a numerical method for assessing the perceived severity of the defects were developed empirically. Prototype equipment was designed, built and tested in extended field trials. The equipment was shown to be superior to previous equipment developed using analogue signal processing techniques. These case studies demonstrate that digital signal processing is a powerful and widely applicable technique for the solution of industrial signal processing problems. Solutions may be theoretical or obtained by experiment or simulation. The strengths and weaknesses of each approach are illustrated.