To test these various encodings for fault tolerance, a circuit was devised which could both generate faults and examine the effect of these faults. The circuit was designed to run on an application board containing a Xilinx Spartan II FPGA (field programmable gate array). Hamming-3 (H3) encoding gave by far the best fault tolerance, recording 0 errors in fault injection tests. However, it required the most resources and was the slowest of the encoding methods. Hamming-2 (H2) encoding had fewer errors than binary encoding, but one-hot encoding had the most errors, due to its large number of target flip-flops. It was also the slowest, and showed poor use of resources.

The results from the tests performed showed that for fault-tolerant designs, H2 was the best compromise in terms of size, speed, and fault-tolerance and is preferred over both binary and one-hot state machine encoding. The results also showed H3 encoding to be fault-tolerant to single faults and, therefore, preferred when ultimate reliability is required in a critical application.

This work was done by Stephanie Taft Naegle, Gary Burke, and Michael Newell of Caltech for NASA’s Jet Propulsion Laboratory.

This Brief includes a Technical Support Package (TSP).

Fault-Tolerant Coding for State Machines (reference NPO-41050) is currently available for download from the TSP library.

Login first to download.

Username

Password

Remember me


Lost Password? Register