Numerical alogrithms for PWM modulators.

Abstract

The development of a simple efficient Pulse Width Modulation (PWM) modulator has been a goal for many research workers. In general three techniques have been used, namely; the analogue triangular wave technique; the use of look-up tables, and the use of Analogue to Digital converters together with analogue circuitry. The modulator described in this thesis is based on an iterative numerical algorithm, and is thus fundamentally different from all previous techniques. The algorithm is limited only by the speed and precision of the associated digital circuitry and can achieve higher modulating frequencies with greater accuracy than can be realised using any of the methods that have previously been investigated. The use of high switching frequencies simplifies the design of filters to reduce both unwanted harmonics and acoustic noise. In this thesis, an equation of a multiphase digital oscillator is derived which is simple to implement and will operate over a wide range of frequencies. The conditions for stable oscillation are derived, and two classes of oscillator are developed. It is shown how the frequency and amplitude of oscillations can be independently and continuously varied. The errors in computing the amplitude and frequency are analysed, and are shown to be cyclic. Upper bounds for the amplitude errors are derived. Single and three phase PWM modulators are described and the implementation procedures for their practical realisation are developed. Two specific implementations of the algorithm are investigated and experimental results confirm theoretical analyses. The algorithm can be incorporated in the Space Vector Modulation (SVM) method of PWM, to improve the resolution at low speeds and to enable the SVM technique to be applied at high gear ratios. A 3-phase 16-bit PWM modulator was built and operated satisfactorily with a pulse switching frequency of 20 kHz and an output frequency range of 1000:1.