karn@eagle.UUCP (06/02/83)
A message from the UoSAT folks at the University of Surrey: The bulletin/telemetry/digitalker program which showed memory error correction on UoSAT during the weekend has been re-run today and displays no further problems, at a preliminary glance. It seems that a byte became corrupted during the program execution and this was continually accessed, toggling the error counter, until successfully re-written at the re-load. Further telemetry on Thursday morning (GMT) will confirm this diagnosis. G8NEF, G8NOB, University of Surrey. Note from KA9Q: This message refers to an apparent primary computer memory problem that appeared last week. The 16K byte primary computer memory, which consists of twelve 4116 dynamic rams, uses a Hamming error-detection coding scheme. Whenever a read cycle is done, the hardware checks the redundant bits and produces a correct data byte even in the event of a single bit error. The computer continues to work normally, except that a three bit hardware counter (displayed in the telemetry) is incremented. There should have been a software routine periodically scanning memory, rewriting any bad locations, but it had apparently not been loaded. The cause of the "continuous errors" was probably due to either the memory being commanded off and not being reinitialized, or to a single soft error perhaps due to a radiation particle. The Phase 3-B onboard computer uses an identical memory system design and UoSAT is serving as the first flight application of this technique. Phil Karn, KA9Q