Improving linear feedback shift registers using similarity transformations

Linear feedback shift registers (LFSRs) are an important component of numerous applications as a source of low cost pseudo-random numbers. The quality of pseudo-random numbers generated by LFSRs, however, is seldom sufficient to meet the requirements of the applications, necessitating the use of additional hardware to improve LFSR performance. We propose a method based on similarity transformations to improve the quality of pseudo-random numbers generated by LFSRs without increasing the hardware cost of the applications.