Speaker
Description
The Phase-II upgrade of the China Spallation Neutron Source (CSNS-II) introduces a superconducting linear section, which imposes stringent machine protection requirements by necessitating a total system response time of less than 10 μs. This paper presents a high-throughput, ultra-low-latency data acquisition and processing platform specifically engineered for real-time beam status monitoring and the Fast Protection System. By bypassing standard bus protocol overheads, the architecture employs high-speed LVDS signaling (250 MHz on the backplane) and multi-gigabit serial links (up to 10 Gbps) to achieve deterministic, point-to-point data transmission between front-end modules and the master logic board. During the critical data acquisition process, an RFSoC-based digitizer utilizes 14-bit, 5 GSPS direct RF sampling to capture signals from the low-noise front-end electronics for Fast Current Transformer. To achieve the lowest possible latency, ardware-level DSP algorithms were implemented, including optimized IQ demodulation, pipelined CORDIC, and parallelized fault detection logic. This optimization constrains the processing latency to under 3 μs, enabling an end-to-end system response time of precisely 4.5 μs for superconducting cavity protection. This robust platform provides not only a reliable solution for bunch-by-bunch diagnostics but also a high-performance framework for future AI-driven predictive interlock and edge-computing applications in high-power accelerators.
| I have read and accept the Privacy Policy Statement | Yes |
|---|