Logarithmic Signal Quantization in the Presence of Gaussian Noise

摘要

The report examines the properties of four, five, and six bit logarithmic quantization schemes to sample weak signals in the presence of Gaussian noise. The quantization schemes were chosen to permit easy implementation. Curves are presented showing bias, relative variance, and signal-to-noise ratio loss as functions of average noise power. In particular, the six-bit scheme analyzed suffers a loss of only about .05 db throughout its dynamic range of over 70 db. (Author)