Fault injection is an accepted method for emulating the effect of ionizing radiation on digital electronic circuits. It can be oriented either to ASIC designs or to SRAM-FPGA designs. When the target device is an SRAM-FPGA the injection has to be assessed both in the functional plane and in the configuration plane. It has been demonstrated that the classical protections oriented to the functional structure are not enough, so the configuration plane has to be analyzed, in the same way. This paper describes the adaption of the FT-UNSHADES2 platform as a fault injection system that tests faults in the configuration plane. The mechanism that assesses the effect of faults in the configuration is read-modify-write, in cycles of inject and repair, based on partial reconfiguration.
In this paper the authors categorize that there are four types of possible faults in the FPGA that should be considered: unrelated, non-damage, outer-propagated and inner-propagated. Faults in the unrelated and non-damage configuration bits are affordable and can be fixed using scrubbing techniques. The damage and propagated faults propagate from the configuration plane to the current data processed and a complete scrubbing followed by a master reset should be asserted to recover the functional behavior of the device.
Other result found is the relationship between the faults in the functional observability and the configuration bits. A result that only can be found if the injection system can distinguish between the faults over the above mentioned planes.
A Fault Injection technique oriented to SRAM-FPGAs
FPGAs and Parallel Architectures for Aerospace Applications ; Chapter : 4 ; 49-59
2016-01-01
11 pages
Article/Chapter (Book)
Electronic Resource
English
Timing driven placement for fault tolerant circuits implemented on SRAM-based FPGAs
Tema Archive | 2009
|Power Analysis in nMR Systems in SRAM-Based FPGAs
Springer Verlag | 2016
|Reconfigurable Fault Tolerance for FPGAs
Online Contents | 2010