Configuration Redundancy for Enhanced Reliability in SRAM-based FPGAs

TL;DR

This paper presents a novel scrubbing technique for SRAM-based FPGAs that significantly improves their radiation tolerance, enabling reliable on-detector applications in high-energy physics experiments through effective detection and correction of radiation-induced upsets.

Contribution

The paper introduces a new configuration scrubbing method that enhances the radiation resilience of SRAM-based FPGAs for on-detector use, demonstrated with irradiation tests on a Xilinx Kintex-7 FPGA.

Findings

Successfully detected and corrected all radiation-induced upsets after high fluence exposure.

Maintained correct functionality of FPGA designs post-scrubbing and reset.

Demonstrated effectiveness on benchmark designs including TMR and serial links at 5 Gbps.

Abstract

Digital off-detector electronics in trigger and data acquisition systems of High-Energy Physics experiments is often implemented by means of SRAM-based FPGAs, which make it possible to achieve reconfigurable, real-time processing and multi-gigabit serial data transfers. On-detector usage of such devices is mostly limited by their configuration sensitivity to radiation-induced upsets, which may alter the programmed routing paths and configurable elements. In this work, we show a new technique for enhancing the usage of SRAM-based FPGAs also for on-detector applications. We show a demonstrator of our solution on benchmark designs, including a triple modular redundant design and a serial link (without redundancy) running at 5 Gbps, implemented in a Xilinx Kintex-7 FPGA. We performed irradiation tests at Laboratori Nazionali del Sud (Catania, Italy) with a 62-MeV proton beam. The results show that our scrubbing technique made it possible to detect and correct all the radiation-induced upsets after a total fluence higher than $10^{12}cm^{-2}$. For both the redundant benchmark design and the serial link, the correct functionality was always restored after scrubbing the corrupted configuration bits and resetting the circuit.