Efficient smoothing of digitized measurements in a large real-time system

In a recently developed system, involving extensive real-time measurement and digital computation subsystems, each measurement is quality checked to establish reasonableness, edited to remove "bad" samples, and filtered to attenuate noise. These functions are achieved under severe restrictions on computer time and storage. The quality of a sample is determined by comparison with a simple prediction, which is used in editing the sample if the original sample is deemed "bad." A five-point recursive filter having optimally selected coefficients for each measurement is employed. Comprehensive tests have proven the acceptability of the smoothing process.