Quantization Noise Characteristics Resulting from Gaussian, Negative-Exponential, and Sinusoidal Random Input Signals.

摘要

The purpose of this study was to investigate the trade-off between the number of quantization levels and the resulting noise characteristics for three classes of commonly occuring input signals, namely, those signals possessing Gaussian, negative-exponential and random sinusoidal distributions. This study derived expressions for the mean-squared error, output spectrum and error spectrum by expanding the nonlinear quantization function into a summation of orthogonal polynomials matched to the corresponding input signal distribution. Once accomplished, orthogonality properties were applied to provide usable expressions. A set of three Fortran 77 programs were developed-each of which applied to one of the studied input signal classes. The appropriate program produced upon demand either a mean-squared error value and a signal-to-quantization noise ratio or quantizer output spectrum data and quantization error spectrum data. Typical input power spectral densities were applied in order to produce the spectra data. The study resulted in a set of tables which provided mean-squared error and signal-to-quantization noise ratio data based on various numbers of bits used for the quantization process. Also,a number of plots displaying the power spectral densities under consideration were produced as based on similar numbers of bits. Keywords: Theses, Quantization, Noise, Power spectra, Probability density functions, Digital communications. (jhd)