Speaker
Description
The CBM experiment at FAIR will operate with high-rate, slowly extracted beams from SIS100, reaching interaction rates up to 10 MHz in quasi-continuous spills. To support this, a free-streaming data acquisition architecture is employed. Spill fluctuations under high data load can lead to event fragmentation and data loss. A statistical model of the data flow, incorporating data response of multiple detectors, has been developed. Based on this model, potential throttling strategies for the Timing and Fast Control system are simulated and evaluated to optimize data integrity under bandwidth constraints.
Summary (500 words)
The CBM experiment is being built to operate with slowly extracted particle beams produced by the SIS100 accelerator at the future FAIR facility in Darmstadt, Germany. The beam will be delivered in long spills on the order of seconds, during which the interaction rate may reach up to 10 MHz. Data selection in the experiment requires the identification of rare and complex decay topologies with relatively complex trigger signatures. Under conditions of overlapping events and high track densities, the required trigger signatures cannot be effectively detected using a conventional trigger architecture with hardware triggers. For that reason, CBM is being designed with a free-streaming data acquisition (DAQ) architecture. In this approach, the detector signals are continuously read out and time-stamped, so that complete events can later be reconstructed and selected in real-time in software, without relying on a hardware trigger.
However, slow beam extraction introduces significant fluctuations in the spill structure. As the data throughput in the DAQ system of the experiment approaches the system bandwidth, these fluctuations cause uncontrolled loss of event data, leading to collection of fragmented data that cannot be reconstructed into complete events. It has been previously demonstrated that controlled data throttling can increase the number of reconstructable events collected by the DAQ system under high data load. To implement the throttling functionality, the Timing and Fast Control (TFC) system will collect the information about occupancy of the front-end electronics from approximately 200 FPGA boards that aggregate the data from GBT links, and broadcast a throttling command when data loss is inevitable. Nevertheless, identifying an effective throttling strategy requires a solid understanding of the response of the data acquisition system to particle interactions.
This contribution presents the development of a statistical data flow model in the DAQ system of the CBM experiment. Data response of multiple detector systems are taken into account, as well as the spill fluctuations. Based on this model, a simulation of potential throttling strategies are evaluated.
