home *** CD-ROM | disk | FTP | other *** search
- Newsgroups: sci.space
- Path: sparky!uunet!charon.amdahl.com!pacbell.com!ames!saimiri.primate.wisc.edu!zaphod.mps.ohio-state.edu!cis.ohio-state.edu!news.sei.cmu.edu!fs7.ece.cmu.edu!crabapple.srv.cs.cmu.edu!roberts@cmr.ncsl.nist.gov
- From: roberts@cmr.ncsl.nist.gov (John Roberts)
- Subject: Re: Shuttle computers
- Message-ID: <Bxw8ru.Kus.1@cs.cmu.edu>
- X-Added: Forwarded by Space Digest
- Sender: news+@cs.cmu.edu
- Organization: National Institute of Standards and Technology formerly National Bureau of Standards
- Original-Sender: isu@VACATION.VENARI.CS.CMU.EDU
- Distribution: sci
- Date: Wed, 18 Nov 1992 04:05:28 GMT
- Approved: bboard-news_gateway
- Lines: 47
-
-
- -From: prb@access.digex.com (Pat)
- -Subject: Re: Shuttle computers
- -Date: 17 Nov 92 04:58:22 GMT
- -Organization: UDSI
-
- -In article <BxsMA3.D2o.1@cs.cmu.edu> roberts@cmr.ncsl.nist.gov (John Roberts) writes:
- ->Do you recall how tight the lockstep is? I don't believe I've seen anything
- ->written on that since before STS-1.
-
- -I think it was 1/2 a clock. i believe every instruction is run at the same
- -time on each processor and when a computation is ready they go to the master
- -for comparison and validation.
-
- That *is* a tight lockstep - obviously a pure hardware implementation (at
- least for detection). Do the processors have measurement "hooks" into
- their internal workings for the lockstep comparison?
-
- -one of the first shuttle flights was screwed up because IBM tested the
- -software ona ground test bed which had different cable lengths then the
- -shuttle and the propagation delay was putting the clock edges enough out that
- -they kept locking out the slave in each pair. i think it was something
- -like this. there are two control channels, each has a primary and backup
- -computer. the backups kept going off-line and the mission got scrubbed.
- -i am sure henry knows the true details but thats how a guyt from IBM
- -explained it to me.
-
- If the new GPCs have a performance of 1.2 Mips, and the old GPCs have
- "a third" of that, then it is very unlikely that the main processor clock
- ran at more than a few megahertz. Since signals travel through cable at
- roughly 1/2 - 1/3 of the speed of light, that sounds like either an
- extraordinarily long cable or an unreasonably tight timing tolerance
- relative to the clock speed.
-
- The first launch of the first Shuttle (STS-1) was delayed a few days
- because when the system was powered up on the launch pad, an unacceptable
- phase relationship was established between two of the system clocks. The
- relationship was such that it was extremely unlikely to happen at any given
- powerup, but no provision had been made to prevent it (the self-checks
- *were* designed to detect it).
-
- The solution at the time was to look at the phase relationship after powerup,
- and if detected, the processors would be turned off and powered up again. :-)
-
- John Roberts
- roberts@cmr.ncsl.nist.gov
-
-