FERNANDES, K. K. L.; http://lattes.cnpq.br/3566183785994947; FERNANDES, Kaio Kikelisson de Lima.
Resumo:
Random distributions have become fundamental elements to the science development
in many areas. However, the random number generation in digital systems was and
will continue to be big challenge considering that it’s not possible to convert a deterministic
system into a machine capable of generating stochastic events. To overcome these limitations,
many methods were developed to implement pseudo-random number generators. This
project aims to implement, in hardware, a Gaussian random number generator based in the
inversion’s method and adopting two main architectures: LUT and MCM. To accomplish
that, mathematical models were developed in Python to validate the generation method
and to functionally verify the RTL implementations. Consequently, the two architectures
were build in Verilog and verified by digital simulations, achieving mean and standard
deviation values as expected. Besides that, the proposals were synthesized using a 28nm
and 45nm technology that ended presenting considerable LUT’s implementation superiority
regarding timing, area and power. Nonetheless the results obtained, it was possible to
recognize MCM’s structures application potential in system needing fixed multiplications
with less complexity than the ones addressed in this project.