Enhanced TRNG based on the coherent sampling

The paper presents several modifications of the True Random Number Generator proposed by Kohlbrenner and Gaj in 2004. The generator is based on a coherent sampling principle and it is aimed at cryptographic applications. It uses the timing jitter present in two clock signals with close frequencies as entropy source. The proposed enhancements are related to the setting of generator parameters (e. g. the frequency of the clock signals) and configuration (simple or mutual sampling) depending on size and composition of the clock jitters. Several versions of the generator have been implemented in two FPGA families, giving the output bit-rate up to 2 Mbits/s. Three generator configurations deliver a high-quality unbiased raw bit-stream, so that post-processing is not necessary and the generated raw random data pass both the FIPS 140-2 and NIST statistical tests. One of the proposed configurations can be fully automated and does not need manual intervention during placement and routing.