Composite Look-Up Table Gaussian Pseudo-Random Number Generator

Abstract

Simulation of digital communication systems for the evaluation of bit error rate (BER) and other performance characteristics can be accelerated if the processing is implemented in programmable gate array (PGA) hardware. These simulations often require one or more Gaussian distributed pseudorandom number sources. Although uniformly distributed pseudorandom number sources can be readily implemented, Gaussian sources are not as easily configured. A typical method is to build a uniform source and transform the distribution to Gaussian. The inversion method accomplishes the transformation by the application of the inverse Gaussian cumulative distribution function (IGCDF). The IGCDF is easily obtained by the use of a look-up table (LUT). However, the memory required for the LUT can become large if it is to accurately represent the IGCDF. In this paper we demonstrate a method that can reduce the size of this LUT while allowing for control of the accuracy.