Lossless Compression of ATLAS Tile Calorimeter Raw Data

Recording and storing the Tile Calorimeter data at 100 KHz frequency is an important task in ATLAS experiment processing. At this moment Amplitude, Time and Quality Factor (QF) parameters are calculated using Optimal Filtering Reconstruction method. If QF is considered good enough, these three parameters are only stored, other wise the data quality is considered bad and it is proposed to store raw data for further offline analysis. Without any compression, bandwidth limitation allows to send up to 9 channels of additional raw data. Simple considerations show that when QF is bad due to the shape differences between standard pulse shape and current signal (e.g. when several signals overlap), all channels are likely to report bad QF while the contained data may still be valuable. So, the possibility to save just 9 samples is insufficient and we have to compress the data. Experiments show that standard compression tools such as RAR, ZIP, etc. cannot successfully deal with this problem because they cannot take benefit of smooth curved shape of the raw data and correlations between the channels. In the present paper a lossless data compressing algorithm is proposed which is likely to better meet existing challenges. This method has been checked on SPLASH events (run 87851, contains 26 SPLASH events) and proved to be sufficient to save ALL channels data using the existing bandwidth. Unlike the common purpose compressing tools the proposed method exploits heavily the geometry-dependent correlations between different channels. It is important to note that the method relies on the only assumption that the registered signal shape is smooth enough and it does not require exact information about the standard pulse shape function to compress the data. Thus this method can be applied for recording pilled-up or unexpected signals as well.