An efficient fault-detecting methodology, algorithm-based faulttolerance, may be extended to include error correction of the outputdata in a protected linear processing system by coupling a high-ratereal convolutional code with a smoothed Kalman recursive estimationtechnique. A completely protected fault-tolerant linear processingsystem involving error correction is shown where it is guaranteed thatno miscorrected data leave the configuration if at most onebox-surrounded subsystem fails at a time. The real convolutional codedictates the comparable parity streams computed in two ways, forming thesyndrome stream that is passed to the fixed-lag corrector when valuesexceed the threshold settings. The block processing and down samplingfeatures of the convolutional code permit the overhead area to be from20-50% of the main processing area
展开▼