[mmaimcal] Correlator reconfiguration time

[Mark Holdaway]

> Trying to beat down the quasi- 1/f noise in HEMTS, or sky noise, in
> total power mapping.  And, if there is a switching secondary, something
> related to chopping.
> 
> We should probably have a contest.  I can't claim familiarity with all
> the possible modes that might demand fast dump times, but they always
> turn out to be less than what you planned for.  For the GBT, the new
> standard continuum back end was originally designed to have a ~ 1 s 
> cycle time.  Try using that to switch out 1/f noise.

I agree with Harvey: my simulations indicated that the atmosphere was
actually pretty easy to take out compared to 1/f. My strategy for beating
1/f noise was to observe as fast as possible with OTF ==> very small
integrations per beam made the thermal noise higher and the 1/f noise
lower.  We WANT to be thermal noise limited; to get better SNR required
doing many scans across the source.  What limits you here is if you go TOO
fast, you spend all your time turning around and you have epsilon
integration on source... so there IS an optimal slew rate and dump time
(ie, a point beyond which faster may be worse). Even if 1/f noise
ultimately limits the sensitivity, it should still average down, as the
residual (ie, high frequency part of the) 1/f noise after subtracting a
temporal baseline should still average down.  (Argue with me if you think
this is wrong.)

Anyway, this didn't require any new correlator setups between OTF scans,
just God-awful correlator dump times which I don't remember.  Harvey's
advice is good: faster, faster, faster: while there may be an optimal
speed/dump rate, we might have the calcs off by a factor of 1.4 or 2.
Bryan Butler is the man now.

	-M