Random Number Generation using LFSR and Cellular Automata

There are several methods possible to generate random numbers. Most of the true random number generators are based on natural phenomena, which are supposed to be random like variation of instantaneous amplitude (voltage / current) of thermal noise. There are several methods of generation of number sequences which appear to be random, but deep analysis brings out their predictability characteristics. They are referred to as Pseudo random number generators. There are program (or) software based random number generators using a complicated transfer function. There are also hardware based random number generators (or) random vector generators which find extensive applications in VLSI testing. One of the methods that is adopted is based on using LFSR (Linear Feedback Shift Register) with appropriate feedback polynomials. In LFSRs the feedback is generally from output to the intermediate stages. And another method similar to the process that is used in LFSR, in which instead of providing the feedback from output to individual nodes, the feedback is provided from the neighboring nodes. This is called Cellular Automata. It is believed that the randomness of the sequences generated are very close to the natural random phenomena and there is considerable interest in generating random numbers / vectors for several applications in VLSI testing (or) communications etc. I propose to study LFSR as a random number generator in the first instance with an aim to generate test vectors, which are very close to truly random phenomena. Thereafter, I would investigate the characteristics of the sequences generated by different Cellular Automata and compare them between themselves and also with the sequences generated by LFSR. Most of the work will be carried out through simulation.