[evlatests] VLA image -- and an interesting clue ...

[George Moellenbrock]

Folks,

Rick, Michael, and Barry encouraged me to calibrate and image in
CASA the blank field (near 0217) in the L-band data
(256 chans/subband) taken Jul 10.  The intent here is to isolate
the details of Rick's reductions in AIPS from intrinsic problems
with data itself, as regards recovery of the weak (10s mJy) background
sources in the "blank" field, and to better characterize frequency-
and bandwidth-dependent effects in the imaging.

SUMMARY:  I believe I have pretty much reproduced Rick's results.
Namely, the blank field image is poorer than what Rick achieved
with VLA data;  I (barely) recover a background source 692" West
and 92" North of the blank field phase center (I've NOT made any
correction to the global phase sign).  I have not verified the location
of this source with Rick's reference VLA image. For narrow (1 MHz)
channels near the band center, the strength of this source is ~0.45%
of 0217 itself.   For wider bandwidths (8 and 64 MHz), the strength of
this source decreases, though not as dramatically as reported
by Rick.  Also, 64 x 1 MHz and 8 x 8 MHz cubes (middle 50% of
the 1436 MHz subband) show that the apparent strength of this source is
a significant function of frequency, tending to decrease away from
the band center.  There is no convincing evidence of systematic
motion in the source position as a function of channel, but the peak
posiiton fluctuatates a bit in the 1 MHz resolution image due to
the general crappiness of the image.

DETAILS:

1. Entire reduction done in CASA, starting with the MS
generated by the archive (from Michael).  I split out the middle 2 subbands,
which are the cleanest and most sensitive.  I've calibrated both subbands,
but imaged only the first one centered at 1436 MHz). For calibration, I
assume a 1 Jy point for 0217, and make no attempt to scale this to proper Jy.

2. I quacked the first 15 seconds in each scan, along with several dead minutes
at the beginning of the first scan.  Also flagged channels 29 and 72 (counting
from zero) on all baselines in the 1436 MHz subband; these channels contain
birdies (the one at ch 72 is the 1408.25 MHz birdie reported by Rick
on Monday).
Otherwise the data are very clean.

3. As there is no formal delay calibration in CASA (yet), I relied on bandpass
calibration from scan 9 (fifth 0217 on-source scan) for all freq-dep
calibration.
The solution interval was the whole scan (the time-dep phase variation was
relatively tame) and the solution was normalized in amp and phase post-solve.

4. Gain calibration was determined (using the bandpass obtained above) from
the on-source scans both per-integration and per-scan.  The per-integration
solutions reveal considerable time-variation in the phase (up to ~rad) on
timescales shorter than the scans; the long-timescale phase variation is much
smaller.  Closely spaced antennas show very similar phase fluctuations, as
expected.

5. Both the bandpass and gain calibration were applied to all of the data
(the gain calibration differently for the two fields as noted below).  I had
no trouble calibrating the bulk delay (cf Ed), but I note that the constant
delay calibration performed here leaves behind a residual of a few 10ths
of a nsec that fluctuates on timescales of a few seconds to minutes.  This
residual is nominally too small to explain the freq-dep effects discerned
from the images and described below.

6. 0217 on-source imaging:  Calibrated with the per-scan gain solution, the
dynamic range is ~500 per 1 MHz channel and ~constant  over the 128 channel
cube.  This image is limited by poor phase calibration.  Calibrated with the
per-integration gain solution, which tracks the detailed phase fluctuations
much better,  the dynamic range improves dramatically at the band center,
and deterioriates toward the edges.  E.g., 32 MHz from the band edges, the
(1 MHz) dynamic range is 1100-1200, and it increases to 1600 at the center.
Closer to the edges (including where the bandpass drops precipitously), the
dynamic range drops to 460 at one edge and 700 at the other.  Even at the
band center, I see no background sources in the 0217 field.

7. Blank field imaging:  The only reasonable option for calibrating
the blank field
is to use the long-timescale solutions.  I used my per-scan solution (Rick used
a constant solution, smoothed from a short-timescale solve, in fact).
There is no
doubt that there are more rapid phase fluctuations that remain uncorrected in
this data.  In the resulting images (1 x 64 MHz, 8 x 8 MHz, and 64 x 1 MHz),
I consistently find a background source 692" W and 92" N of the phase center
(which is probably backwards due to the phase sign convention error).  This
source is probably strongest in the narrowest central channels at
0.45% of 0217.
At 8 MHz resolution in may be a little weaker (0.4%); at 64 MHz, even
more (0.36%).
(Rick saw more substantial decreases with bandwidth, I think, which may
have to do with the fact that he averaged the data in frequency before
gridding.)
More obvious is the decrease in power away from the band center, which is
most significantly detected in the 8 MHz/ch image.  This variation is
presumably
the origin of the deterioration with increasing bandwidth.   I haven't
looked carefully
enough to find any additional sources; I gather Rick expects we should
find many based on the VLA imaging.  They are not at all obvious....

That's all for now.

-George







On Fri, Jul 10, 2009 at 5:20 PM, Rick Perley<rperley at nrao.edu> wrote:
>    I calibrated the VLA data, taken simultaneously with the WIDAR test
> data.
>
>    The images are very nice, with loads of the expected background
> sources.
>
>    I then returned to the WIDAR images, specifically the set of 'noise'
> images (blank field).  There are three, made with 1, 10, and 100 MHz BW
> each.
>
>    The peak background source in the 1 MHz wide field is 11.2 mJy
>    The same object in the 10 MHz field is now only 7.1 mJy
>    The same object in the 100 MHz field is now essentially invisible,
> at 1.2 mJy!
>
>    The VLA dataset shows this background source at 17  mJy.
>
>    For the VLA correlator dataset and the WIDAR 100 MHz dataset, the
> expected thermal noise limits should be about the same  -- and they are,
> at 0.15 mJy.
>
>    From this, I'm guessing that the act of coherently summing over the
> bandpass is actually destroying the signal (but not the noise).
> Something is backwards?
>
>
> _______________________________________________
> evlatests mailing list
> evlatests at listmgr.cv.nrao.edu
> http://listmgr.cv.nrao.edu/mailman/listinfo/evlatests
>
>