[alma-config] thermal noise and phase errors

[Mark Holdaway]

> However, before we adress this problem, I think we need to be sure that the
> configuration
> choice can also handle two (absolutely) unavoidable classes of errors
>     - thermal noise
>     - atmospheric phase noise
> 
> It is not obvious to me that different configurations will result in the
> same behaviour
> vis-a-vis these classes of errors (I would even bet the opposite...). This
> may turn out
> to be important to guide our choice, specially since we know that quite a
> number of
> projects are going to be sensivitity limited.
> 
> Evaluating the response under various noise levels is essential.
> 


OK, I'm stepping off of my "Pointing Error Soapbox" to make some
comments.


1) Thermal noise.  A long time ago, I assumed that the actual thermal
noise of images produced by a ring array and a gaussian-uv-coverage array
would actually be different.  The ring array has more long spacings than
Gaussian, so the outer part of the main beam of the PSF drops off faster
than Gaussian (ie, compared to a Gaussian fit to its main beam).
When you convolve the deconvolved image with the clean beam (or
taper in the Fourier plane with the Clean beam), it seemed to me
that this extra tapering would result in the loss of sensitivity
on the long baselines.  However, the results of simulations I
did indicated that this effect is not seen, or at least there is
no difference between a Gaussian-coverage and a Uniform-coverage (ie,
ring) array.

Why not?  Because the same Gaussian taper is applied to both
the Gaussian-coverage and the uniform-coverage array's images?

Answers anyone?


2) For phase noise simulations:  even though the rms phase errors
increase as a function of baseline length, I assume that the phase
calibration schemes remove most of that increase, and that the
residual phase errors are independent of baseline length.  Then
what you need is a statistical model for these error's time
behavior (ie, for fast switching the phase errors may be a
constant error over the sitching cycle due to the distance to the
phase calibrator, plus a random term varying over the cycle; when
the phase calibrator is observed again, the constant error is
re-randomized; thats not exactly true, but that is the sort of
model that we should make to simplify the simulations, rather
than simulate the entire phase-cal process).


	-M