Path: utzoo!utgpu!news-server.csri.toronto.edu!cs.utexas.edu!uunet!mcsun!ukc!newcastle.ac.uk!turing!n02ll
From: M.W.Hey@newcastle.ac.uk (William Hey)
Newsgroups: sci.space.shuttle
Subject: Re: Shuttle computers break down during glide tests
Message-ID: <1991Apr26.135546.26577@newcastle.ac.uk>
Date: 26 Apr 91 13:55:46 GMT
References: <1593@philtis.cft.philips.nl>
Sender: news@newcastle.ac.uk
Distribution: sci.space.shuttle
Organization: University of Newcastle upon Tyne, UK, NE1 7RU
Lines: 77

Greetings,

munk@cft.philips.nl (Harm Munk) writes:

>computers: that was deducible from the the conversation between pilot
>and capcom, and from some shots of some instrument panel showing a
>light with "Computer Fail" and a computer display with a big X all
>across the screen.
>My question: what happened with the computers during that first (?)
>glide test.

Last vacation I saw (by chance) an (old) Open University program concerned with
software reliability, and their basic worked example was aspects of
the Shuttle 5-tier GPC (General Purpose Computer) system.

They interviewed one of the chiefs of software integration and simulation, and
and he explained that a GPCs failed immediately upon release from the SCA.
However the other computers noted the failure and shut it down.  If I remember
correctly, there are 4 GPCs, which process the flight data continually, and each
one polls the others, and vote on what actions to take.
The remaining GPC is a back-up, which is used only in the case of complete GPC
failure, and it contains only the software for ascent, and de-orbit.

Any NASA personel/buffs out there care to elaborate, for example how are 'Mexican
stand-offs' avoided?

He went on to say, that even though a GPC did fail, the fail-safe system worked
accordingly, ie the others contained the failure, and continued the mission.

He claimed that the integration simulations are taken as seriously as if they
were real flights, and they've only had 2 failures at this level.

Even Richard Feynman (Rogers Commission) was impressed by the software Quality
Control.
(Re "What do you care what other people think").

>Along the same tracks: during the landing of one of the later tests, the
>shuttle made a rather bumpy landing. The commentator remarked that this
>was caused by the inexperience of the pilot, but I recall reading

It was pilot error, something to do with the pilot thinking the nose-down
rate was too fast, and he over-corrected on pulling the nose up, so the
Enterprise skipped as a result.
(Disclaimer - speaking from vague memory here, I left all my books at home.
 For any British readers, I saw this on "John Cravens News Round", circa 1980)

>somewhere that is was caused by the computers being to busy with other
>things that they could not constantly attend to the bussiness of landing
>the ship, thereby inducing quite some roll. Is this true, or was the
>commentator right ?

"Roll" ? Pitch more likely, maybe with a smaller resultant role in consequence.

EXTREEMLY unlikely, and besides all shuttle landings are manual
not automatic; infact the pilot/commander assumes control minutes away
from landing...the idea being, should the GPC's fail suddenly at the
last moment, the pilot would tend to over-react on the controls.  However,
the brakes are automated, but I'm not sure about nose-wheel stearing.
(RE "Space-Flight News" - issue : March/April 1991)

Even if the landing was automatic, can you imagine a GPC being programmed
to land the Shuttle, but being 'too busy' at the time!
Have any of you read Douglas Adams "Hitch Hikers Guide To the Galaxy", when
Eddie the ship-board computer, is unable to defend the ship from an imminent
missile attack, as he's figuring out how to make a decent cup of Tea?

No flames for inaccuracies please, just the facts.

Cheers,

Bill


-----------------------+-----------------------------+
|    William Hey          M.W.Hey@newcastle.uk.ac    |
|    Astrophysics : Newcastle University, England.   |
-----------------------+-----------------------------+