Optimizing Speed of a True Random Number Generator in FPGA by Spectral Analysis

Security and speed are two important properties of today's communication systems. In order to generate initialization vectors and keys for such communication fast enough, a true random number generator (TRNG) with a high bit rate is needed. In this paper an FPGA implementation of a TRNG based on several equal length oscillator rings that achieves a high bit rate is analyzed by using spectral analysis. The design is examined by defining a MatLab model of the TRNG and by investigating the frequency spectrum at different locations in order to find the speed increasing potential of the TRNG. Experiments performed on an Altera Cyclone II FPGA have shown that a TRNG, whose parameters were optimized by means of such a model, achieves a bit rate of 300 Mbit/s. Experiments with repeated restarts from a known state have shown that the output of the TRNG contains true randomness and not only pseudo randomness.