[evlatests] Settling Times for Referenced Pointing

[Bob Broilo]

> A modern ACU would have a dramatic improvement in this type of performance.

Even just replacing the A4 and A5 cards with even a simple CPU.  The
existing system is a PI scheme that will always overshoot the desired
positon at full slew:
http://www.aoc.nrao.edu/~bbroilo/acu_frm/specs.pdf

Adding a little D (Derivative) would improve things, but would be
tricky to do on the analog loop computer (in fact it looks like this
was the intent but was abandoned, probably due to time constraints or
oscillations).  A slow (like 100Hz) programmable loop could fix the
overshoot and reduce the time to source, and then we could implement
better control algorithms as they are developed and tested.

However, it seems silly to do this without replacing the other guts of
the ACU at the same time.  A PLC could do most of the positioning and
error logic, and a small cheap processor could do the rest.  Now that
the panic of ELVA is calming down, perhaps we should continue to
attack this?

Bob.