RMPS A REALTIME PARALLEL COMPUTING SYSTEM

Abstract

Research and Test Activities have a continuing need to cope with more and more channels of data and at continually wider data bandwidths. There is a consensus in the test community that compressed EU and derived parameter data presented in realtime can significantly reduce total test costs because test engineers can make realtime judgements on the validity of a given test point (mode). Classical telemetry preprocessors usually cannot handle these more demanding realtime processing requirements because, when they were designed, it was assumed that raw data was in a single PCM word and only a simple mx+b EU conversion, or simple data compression was required. Present preprocessors typically use special bit slice technology to speed up the realtime process and they’re only one or two bus systems whose processing capacity is typically less than 300k to 400k parameters per second. Furthermore, many cannot handle word concatenation (except for adjacent PCM words) and none can handle complex derived parameters such as thrust, lift, gross weight, center of gravity, stall speed, harmonic analysis, etc. To address these limitations, a massively parallel computer system has been developed based on up to sixty, general purpose, 1MFLOP floating point computers operating in parallel to support realtime processing of any type, at aggregate throughputs up to 1.5 Mwps. This system can merge realtime data from up to eight different asynchronous sources having word rates up to 2.0 Mwps from any source. Up to 32,768 different parameters can be accepted as inputs with an additional 32,768 ID tags available for concatenated and derived parameter identification. A powerful realtime software package permits the user of the computer system to apply any, or many algorithms) to any or an parameters being processed.