Synchronization of a Trigger-Less Data Acquisition for the PANDA Experiment
M. Kavatsyuk
TA-CON-2016-048.pdf
(3.6 MB)
The PANDA experiment at the FAIR facility in Darmstadt, Germany, is aiming at precise measurement of hadron properties in the charmonium region and investigation of exotic-matter states. This will be possible because of employment of the annihilation of antiprotons with very low momentum spread of 2e-5 and high luminosity, up to 1e32 1/(cm s). At the highest luminosity the annihilation rate will rise up to 20MHz. Moreover, the decay topology of events of interest is very similar to the one of background events. This excludes any reasonable performance of a classical triggered readout system. Therefore, a new readout has been developed where self-triggered front-end electronics is able to autonomously detect hits and extract all relevant physics information, a dedicated custom network builds events based on hit time-stamps, and an FPGA/GPU/CPU-based computing system, which selects events of interest on-line using the complete reconstruction, in order to achieve a data-reduction factor of about up to 1000. Such a readout system is called “trigger-less”. Since there is no trigger signal which forces simultaneous readout of acquired data from all front-ends, the electronics should use the same clock signal and should be synchronized. Therefore, a precise clock-distribution and -synchronization platform SODANET has been developed. The SODANET is a bi-directional data-transfer protocol for optical links which foresees transfer of detector- and slow-control data and synchronization messages with fixed latency. The synchronization messages have the highest priority and can be transmitted even if transmission of another packet is not yet finished. The synchronous operation is implemented only in the downstream direction and requires defined latency configuration of serializers/deserializers. A complete prototype of the readout system which employs the SODANET protocol has been constructed. During the presentation the SODANET platform and its performance will be discussed.