[Ground-station] Need some nuclear science help

[Phil Karn]

Radiation was one of the first projects I did for AMSAT when I joined in
1980. The problem then was to estimate the lifetime of the 4116 16k x 1
bit dynamic RAMs to be flown on AO-10.

There were a *lot* of uncertainties. My final estimate was that the RAM
would last anywhere between 1 and 6 years. It actually failed after 3.

There are two basic kinds of ionizing particle radiation at issue here:
solar wind particles trapped in the earth's magnetic field and cosmic
ray particles.

The trapped solar wind particles are better known as the Van Allen
belts. There are two: an inner, nastier one composed mostly of protons
and an outer one that's mostly electrons that are more easily stopped.
Both are densest along the geomagnetic equator.

The inner belt is densest at 1-2 earth radii; that's why you don't find
very many satellites at those altitudes. Because the earth's magnetic
field axis is at an angle to the rotational axis and offset from it,
both latitude and longitude matter. Because of the offset, the inner
belt essentially touches the atmosphere in the region between south
America and Africa; this is the so-called South Atlantic Anomaly. The
ISS (and other LEO sats) picks up most of its exposure in this region.

The trapped particle energy distribution has a very long tail. As you
add shielding the dose drops rapidly but you reach a point of
diminishing returns. It's impractical to stop it all.

Then you have cosmic rays, mostly hydrogen and helium nuclei but also
heavier ions. These are much less numerous than trapped solar particles
but they have MUCH higher energies. They are essentially impossible to
stop. These are the major radiation source on the moon and in
interplanetary space. The flux does vary about 2:1 with the solar cycle;
a more active sun tends to shield the inner solar system from cosmic
radiation.

The bottom line is that while you can do a lot with a moderate amount of
shielding, there's no way you can stop it all. Some of what gets through
can have very high energies, so random errors and latchup are always a
concern.

Latchup can be handled with fast-acting, self-resetting circuit breakers
in the power supply lines.

ECC (paying attention to geometry) can fix the transient error problem
with RAM but you can always have a transient error in some other part
(like a CPU) where it's not so easy to correct.

If one computer isn't reliable enough, some sort of system level
redundancy seems essential. I seem to recall three computers arranged in
a circular rock-scissors-paper arrangement, i.e, where computer 1 can
monitor and reset computer 2, which can monitor and reset computer 3,
which can monitor and reset computer 1. (Or maybe I just made that up.)

That leaves total dose, which is what got AO-10. There's basically
nothing you can do about it beyond using whatever rad-hard parts you can
get, adding as much shielding as you can afford, and avoiding hotter orbits.

Phil