A Fast and Efficient Compression Algorithm for Nab Waveform Data

The high data output of the Neutron "a" and "b" (Nab) experiment requires a fast and effective lossless compression algorithm to be implemented in firmware, reducing the requirements on both bandwidth and storage. The previously deployed Delta-Rice algorithm has two steps: decorrelation using delta encoding, and entropy encoding using the universal Rice-Golomb code. We report an improved compression algorithm featuring better decorrelation using non-integer convolution filters and entropy encoding tailored for Gaussian distributed waveform data. This algorithm is implemented efficiently in an FPGA architecture to run in real-time as the data is being read out. Machine learning, mathematical, and brute-force approaches were employed to search the convolution space to optimize this approach.