Data reduction processes using FPGA for MicroBooNE liquid argon time projection chamber

Abstract

MicroBooNE is a liquid Argon time projection chamber to be built at Fermilab for an accelerator-based neutrino physics experiment and as part of the R&D strategy for a large liquid argon detector at DUSEL. The waveforms of the ~9000 sense wires in the chamber are continuously digitized at 2 M samples/s - which results in a large volume of data coming off the TPC. We have developed a lossless data reduction scheme based on Huffman Coding and have tested the scheme on cosmic ray data taken from a small liquid Argon time projection chamber, the BO detector. For sense wire waveforms produced by cosmic ray tracks, the Huffman Coding scheme compresses the data by a factor of approximately 10. The compressed data can be fully recovered back to the original data since the compression is lossless. In addition to accelerator neutrino data, which comes with small duty cycle in sync with the accelerator beam spill, continuous digitized waveforms are to be temporarily stored in the MicroBooNE data-acquisition system for about an hour, long enough for an external alert from possible supernova events. Another scheme, Dynamic Decimation, has been developed to compress further the potential supernova data so that the storage can be implemented within a reasonable budget. In the Dynamic Decimation scheme, data are sampled at the full sampling rate in the regions-of-interest (ROI) containing waveforms of track-hits and are decimated down to lower sampling rate outside the ROI. Note that unlike in typical zero-suppression schemes, in Dynamic Decimation, the data in the pedestal region are not thrown away but kept at a lower sampling rate. An additional factor of 10 compression ratio is achieved using the Dynamic Decimation scheme on the BO detector data, making a total compression rate of approximate 100 when the Dynamic Decimation and the Huffman Coding functional blocks are cascaded. Both of the blocks are compiled in low-cost FPGA and their silicon resource usages are low.