Hardware Architecture of a Gaussian Noise Generator Based on the Inversion Method

In this brief, we present a hardware-based Gaussian noise generator (GNG) with low hardware cost, high generation rate, and high Gaussian tail accuracy. The proposed generator is based on a piecewise polynomial approximation of the inverse cumulative distribution function (ICDF). We propose to avoid the area-demanding barrel-shifter of the ICDF approximation by means of creating a new uniform random sequence from the uniform random number generator output. The GNG architecture has been implemented in field-programmable gate array devices, and the implementation results are compared with other published designs, achieving a higher deviation with fewer hardware resources. Our GNG generates 242 Msps of random noise and achieves a tail of 13.1 sigma with 442 slices, two multipliers, and two Block-RAM of a Virtex-II device. The generator output successfully passed commonly used statistical tests. © 2012 IEEE.
This work was supported by Fondo Europeo de Desarrollo Regional and the Spanish Ministerio de Ciencia e Innovacion under Grant TEC2008-06787 and Grant TEC2011-27916. This brief was recommended by Associate Editor Y. Ha.