[Ground-station] Need some nuclear science help

[Phil Karn]

On 6/3/18 16:03, Bruce Perens wrote:
> Thanks. I would assume that a satellite for anything other than LEO that
> we launch is expected to be long-lived and that the flight computer must
> be rad-hard or it won't survive in orbit due to actual physical damage
> to the component, as AO-10's RAM did not. This was back when RAM used
> much larger silicon features than today, and was probably less
> vulnerable. I need to be able to tell people who know even less than I
> about this why we can't use consumer parts with shielding or redundancy
> as a strategy for anything but low earth orbit. I still would like to
> have more numbers, rather than simply relying on experts or waving my
> hands :-)

That's right; consumer/commercial grade stuff is usually OK for LEO with
some caveats for the South Atlantic Anomaly. Higher altitude orbits are
a different story. You can still survive surprisingly long up there, but
most people are likely to consider a 3 year lifetime too short given the
major investment such a project requires.

Until the total dose got us, the AO-10 IHU worked surprisingly well. It
had a (12,8) Hamming (or was it Hademard?) hardware FEC on the RAM, with
background scrubbing in software. We could see the error counts
increment on each pass through the inner belt, but they weren't frequent
enough to cause any serious problems (i.e., there was very little chance
of a double bit error in a byte before the scrubber fixed it).

Only towards the end as the total dose began to mount did we see the
error counts go up dramatically on each belt passage, and as you'd
expect the computer began to crash increasingly often. But for some time
it was still possible to reset the computer and upload the software from
the ground. (AO-10 had no mass storage, only the 16KB of RAM and a very
simple hardware boot loader reading from the command receiver.)

If anybody needs some really rad-hard ROM, I can describe how Apollo's
core rope memory worked...

Phil