Abstract
In an earlier paper2, the adaptive quantization was suggested to improve the performance of digital systems. The theory of this scheme was developed and by using deterministic inputs, the effectiveness of the adaptive quantization was illustrated. In this paper the characteristics of the adaptive quantization are developed for a stochastic input. The mean square error in the output, the probabilities of the maximum percentage error, and the cumulative probability distribution of the percentage error for a stochastic input have been calculated for (a) an adaptive quantizer and (b) a uniform step quantizer.
The above quantities are calculated for a Gaussian input with zero mean. It is observed that the mean square error can be reduced from 0.04715 to 0.00317. The probability curves also show an appreciable improvement in the system performance.