Memo Review Reply

Memo: 403 - Fast Switching Phase Correction Revisited for 64 12 m Antennas
      Holdaway, 2001Dec17

Reviewer: Dave Woody

Date Received: 2002Aug08

Reply from: Mark Holdaway

Date Received: 2002Aug23


Reply:

1) The glish scripts I used for the calculations are available from

http://www.tuc.nrao.edu/~mholdawa/fastswitching_scripts/index.html


2) OK, I apologize for leaving out some of the relevant suporting
information.  


3) Fast switching does not depend critically on ANYTHING

My experience with these simulations over the past 8 years has lead
me to an understanding of fast switching as the optimization of
a very mushy multi-dimentional system.  For example, lets say that
our 90 GHz (ie, cal frequency) sensitivity turns out to be a factor
of 2 worse.  So, we spend 4 times as long omtegrating on the same
calibrator to achieve our required sensitivity.  Woop-de-doo, the
calibrator integration time was only a small fraction of the entire
cycle time, so the cycle time goes up by 10%.  The phase structure
function goes less than time to the unity power, so the effect is
small in the increased residual phase error.   OR, we could realize
that this small increase in residual phase error indicates that we
don't need to measure the phase on the calibrator so accurately, so
we only integrate 3.4 times longer (as oppose to 4 times longer).
OR, we realize that there is a slightly brighter calibrator source
slightly further away which results in a minimal efficiency penalty
because we have a 1.5 s frequency switching time anyway, and the
residual phases are dominated by temporal changes, not by the
distance between the calibrator and the target source.

So, we could make the 90 GHz system twice as bad, and get like a 5 or
10 percent increase in the residual phase errors on most observations.

Similarly, fast switching doesn't depend critically on the height of
the turbulent layer:  the height effects the distance between the
cal source and target source's lines of site, but as stated above,
the residual phase errors are dominated by the temporal changes
in the atmosphere.

If ALL the assumptions are wrong in a bad direction, the results
could be significantly over-optimistic.


I think 0.5, 1, and 2 km turbulent heights would be fine to use for
revised calculations.


18msec == OK, that was the time required to achieve the required
phase errors on the calibrator.  In the simulations, I probably should
have done something like

 t_cal := max(t_cal, 0.5s)




I don't think anyone is going to try to justify deleting 183 GHz
WVR from the project, certainly not me.