NRAO/ALMA Imaging Group Meeting, March 1 2005
present: Mark Holdaway,
Sanjay Bhatnagar, Kumar Golap,
Ed Fomalont, Al Wootten
Reconstituted Minutes (ie, not in the order the meeting happened):
Sanjay is working on a pointing self-cal algorithm. Currently
a primitive pointing error solver works, and Sanjay is working
on the code that corrects the data for the pointing errors in Imager.
(This is a huge step forward.)
Kumar is currently working on mosaicing, combining total power and
Mark's vision is to be able to simulate all relevant imaging errors
with AIPS++: thermal noise, realistic atmospheric phase errors,
atmospheric amplitude fluctuations, pointing errors, voltage pattern
errors, illumination offsets, and probably some other errors we haven't
written down yet (eventually we will get to polarization). The
big results we will get from this vision are:
1) we should be able to say what effects will be most limiting for
imaging as a function of frequency and atmospheric conditions,
2) we will be able to exercise our soon-to-be-written self-cal
algorithms and determine how to use them iteratively (ie, the
errors are not orthogonal, so the iterative solution for different
effects -- ie going back and forth between deconvolution and
the different sorts of self-cal -- may be problematic).
(Ed thinks that sooner or later we will have a Single Grand Solver.
He's been hanging out with geodetics folks too long?)
At some point, we realized the obvious: in order to get a good model
for whatever sort of self-cal you are doing, we will need to get
the (error free) total power + interferometric imaging under
We will have at least two different regimes to worry about:
a) weak source -- self-cal is not possible
b) strong source -- self-cal is possible
Thinking now, it seems there may be several regimes:
c) source too weak to use self-cal, but with images limited
by the various errors nonetheless
d) enough SNR to accomplish one sort of self-cal, but not
another kind (HEY! We need to develop some rules about
what sort of pointing corrections you can solve for as
a function of SNR and source structure.)
A zeroeth order approach to these errors: flag antennas with the
worst pointing (or other error) if we cannoy correct for the errors.
For single pointing imaging: doing a small mosaic may actually help
with the pointing self-cal. We'll find out.
Need to consider snapshot simulations in addition to long track
Kumar was interested in knowing what sort of dynamic range is possible.
Mark said you can have a 10 Jy unresolved quasar and easily have
a thermal noise level of 10 uJy -- ie, 10^6 potential DR
Ed volunteered himself & Steve to do simulations at a future date.
(( Ed: how about this: I simulate the data, and you reduce it?
you'd become the world's expert on running pointing self-cal
beofre it is really needed. ))
Ed still thinks non-random residual phase errors will dominate the
imaging process. We will look into this:
==> Ed will investigate phase errors as a function of cal-target
separation at the VLA (Ed: I know that I investigated the RMS
of the residual phase errrors -- it is possible that I didn't
look at the average phase error!) The 65 micron antenna position
spec leads to 1deg phase errors ( for a typical cal-target setup,
at 950 GHz ??? ); calibrator position errors can lead to systematic
phase errors. What else? Thermal effects in the fiber optics?
Fiber optics torsional effects?
Mark didn't mention this, but he's got atmospheric simulation code
in Glish (uses the full AIPS++ data structure for the MS, gain tables,
etc, should be easily integratable for the whole burrito simulations.
Our next meeting will be March 15, 4pm EST, 2pm MST (when do you
guys go onto Daylight Savings Time?)
- 07 Mar 2005