[mmaimcal] Re: WVR question.

[Al Wootten]

Darrel Emerson writes:
 > Hi Al,
 >    In a lunchtime discussion, this question came up:
 > 
 >    If we take the best existing WVR system, whether it's 22, 220, 183 GHz,
 > and look at the residual errors from that system in terms microns of path
 > length, then, using the measured statistics of the atmospheric
 > fluctuations at Chajnantor, for what fraction of time will this WVR system
 > at Chajnantor make a significant improvement in atmospheric phase noise?
 > 
 >    Has anyone calculated this number?
Interesting question.  I don't think it is calculated in exactly the way you
pose it but let me see if I can make an estimate.

We have defined at one point 'phase stable' as better than a radian, and
at another, in the goals listed in the project book, 0.15 radian at 230 GHz.
The current 183 GHz systems come close to achieving this, while the current
22 GHz systems have not yet.  So I will assume the project book goal can be
reached at 183 GHz.

For the project book goal, this is 99% of theoretical sensitivity.  Without
any corrections, we achieve this 30% of the time on 150m baselines, and 1%
of the time on 1.5km baselines, according to MMA Memo No. 169.  If we only
want 50% sensitivity, this is achieved 85% of the time on the shorter
baseline and 43% on the longer baseline.  For 90% sensitivity, the percentages
are 60% and 12%.  If we assume that we are thinking
of 230 GHz, I would say that the WVR system will NOT make a significant 
improvement only 1% of the time at kilometer baselines, and will be used for
all observations on those scales.  In the compact array, I would say that
the WVR system would NOT make a significant improvement perhaps 40% of the
time, and would be used just over half of the time.  I would conclude that
for the average observation at 230 GHz or above, in whatever array, the
WVR system will be in use.  Below 230 GHz in the compact arrays, it might not
make a significant improvement in most cases.

I think that, as in current IRAM software, the system should probably
compare observations of the caibrator with and without the WVR corrections
applied, and when it finds a significant improvement, apply the corrections
to the data.  In practice, the system would then work all the time.  For some
resolutions of frequencies, fast switching may prove superior; I can't
estimate now how much of the time one would choose to use that system
and not use th WVR system.

Al