Determine the cause of maximum-load power faults in the RFSoC readout

Determine the exact root cause of the intermittent power-related fault events that occur in the Xilinx ZCU111 RFSoC Gen2 KID readout during maximum-load operation with four chains and 4×1024 active tones, and identify measures to prevent these faults.

Background

The Gen2 KID readout reaches approximately 82 W at maximum load, with four independent chains each carrying 1024 active tones. Under these conditions, the board intermittently enters a protection or error state, loses SSH access, and requires power-cycling to resume normal operation.

The reported RFSoC temperatures are approximately 70 °C, below the stated failure thresholds, so the authors suspect a localized temperature or current-draw issue rather than a global over-temperature or over-power condition. Resolving this issue is important for reliable operation at the intended detector and tone capacity.

References

At the maximum load ($4\times1024$ tones), we have observed intermittent power-related fault events: the board enters a protection or error state in which SSH access is lost, and normal operation is restored only by power-cycling the board. During these events, the PS and PL temperatures reported by the RFSoC is around $70{\circ}$C, well below the failure threshold of either, so we suspect a localized temperature or current-draw issue rather than a global over-temperature or over-power condition. The exact root cause is under investigation, and reducing the overall power draw and resource utilization is an active development priority.

— CCAT: a two-octave 1.024 GHz KID readout featuring an overlap-channel polyphase synthesis filter bank on RFSoC  (2608.13852 - Ruixuan et al., 14 Aug 2026) in Section 3, FPGA Implementation