Abstract:Neural Networks (NNs) are increasingly used in the last decade in several demanding applications, such as object detection and classification, autonomous driving, etc. Among different computing platforms for implementing NNs, FPGAs have multiple advantages due to design flexibility and high performance-to-watt ratio. Moreover, approximation techniques, such as quantization, have been introduced, which reduce the computational and storage requirements, thus enabling the integration of larger NNs into FPGA devices. On the other hand, FPGAs are sensitive to radiation-induced Single Event Upsets (SEUs). In this work, we perform an in-depth reliability analysis in an FPGA-based Binarized Fully Connected Neural Network (BNN) accelerator running a statistical fault injection campaign. The BNN benchmark has been produced by FINN, an open-source framework that provides an end-to-end flow from abstract level to design, making it easy to design customized FPGA NN accelerators, while it also supports various approximation techniques. The campaign includes the injection of faults in the configuration memory of a state-of-the-art Xilinx Ultrascale+ FPGA running the BNN, as well an exhaustive fault injection in the user flip flops. We have analyzed the fault injection results characterizing the SEU vulnerability of the circuit per network layer, per clock cycle, and register. In general, the results show that the BNNs are inherently resilient to soft errors, since a low portion of SEUs in the configuration memory and the flip flops, cause system crashes or misclassification errors.

SRAM-Based FPGA SEU Fault Propagation Model

Fault modeling and characteristics of SRAM-based FPGAs (abstract only).

Modeling the Effect of SEUs on the Configuration Memory of SRAM-FPGA Based CNN Accelerators

An Accurate Fault Location Method Based on Configuration Bitstream Analysis

Design and Verification of Multiple SEU Mitigated Circuits on SRAM-based FPGA System

New Insights into the Impact of SEUs in FPGA CRAMs

CRC Circuit Design for SRAM-Based FPGA Configuration Bit Correction

Comparison of Single Event Effect and Space Electrostatic Discharge Effect on FPGA Signal Transmission

Study of an Automated Precise SEU Fault Injection Technique

A New Approach To Detect-Mitigate-Correct Radiation-Induced Faults For Sram-Based Fpgas In Aerospace Application

Characterization of Single Event Upsets of Nanoscale FDSOI Circuits Based on the Simulation and Irradiation Results

Electron-Induced Single-Event Effect in 28 nm SRAM-Based FPGA

A Novel High-Density Single-Event Upset Hardened Configurable SRAM Applied to FPGA

Radiation-tolerant routing algorithm design for FPGAs

SEU sensitivity and large spacing TMR efficiency of Kintex-7 and Virtex-7 FPGAs

Simulation of SEU Response of Advanced 20 Nm FDSOI SRAMs

Single event upset mitigation for FDP2008

The Research of FPGA Reliability Based on Redundancy Methods

Improving Fault Tolerance for FPGA SoCs Through Post Radiation Design Analysis

Radioresistance Placement Algorithm for FPGA Based on SRAM Structure

Analyzing the Single Event Upset Vulnerability of Binarized Neural Networks on SRAM FPGAs