[evlatests] 3-bit rms - repeat.

[Vivek Dhawan]

To check the repeatability of 3-bit performance, the sampler rms
measurement was repeated. The answer is that

(1) Many of the outliers have been improved by repeated resets.

(2) Not all samplers have rock-steady behaviour over months.

(3) The bulk of them are much the same, i.e. within +-5% for about 80%
     of the 224 samplers. (I'm eyeballing these stats, so sue me!)

Measurement conditions:

K band, center at 24GHz. 3-bit samplers at 23-25GHz, all overlapped. 8-bit
samplers are setup to overlap the central 1GHz of the 3-bit sky bands.

2 observation blocks were run so as to get all samplers with this scheme,
(A1C1A2C2 3bit + B0D0 8-bit) and (B1D1B2D2 3bit + A0C0 8-bit).

Calibration consists of SETJY, FRING(delay), BPASS, CALIB on one single
chosen scan, applied to all data including the off-source scans. The rms
of the calibrated off-source data is then extracted.

The plots show the ratio of 3bit-rms to 8bit-rms, for each 3-bit sampler
(4 samplers x 2 pol. per antenna), matched with the 8-bit unit from the
same antenna/pol. Thus antenna gain curve, pointing offset, SEFD, Tsys,
elevation changes, opacity, are all divided away, all that remains are
the sampler properties (and T304 differences, if any, that survive the
bandpass calibration step).

The whole exercise was done twice, Aug 16 and Oct 24.

plot 1 & 2: histograms of rms ratio, old and new. The new data look about
1-2% worse than old, on average. Many power outages and resets intervened.
I have no more wisdom on this - that is what they do.

plot 3: Old and new data, identified by antenna and pol.(New has ea27, 28).

         Old = black A1, red A2. New = blue A1, orange A2. and similarly
               for B,C,D.

         e.g. Black higher than blue (or red higher than orange) means that
         sampler improved between the 2 dates. Some moved the other way.

plot 4: scatter plot of new (y axis) vs. old. The different symbols are
         A1A2 etc. not that it matters. I don't see a lot of insight here,
         beyond the fact that the upper left of the plot is sparse, so
         the general trend is the samplers are improving from lower right.
-------------- next part --------------
A non-text attachment was scrubbed...
Name: rms_rat_hist.pdf
Type: application/pdf
Size: 4058 bytes
Desc: not available
URL: <http://listmgr.nrao.edu/pipermail/evlatests/attachments/20121115/8c3515d2/attachment.pdf>
-------------- next part --------------
A non-text attachment was scrubbed...
Name: zhist-oct24.pdf
Type: application/pdf
Size: 4035 bytes
Desc: not available
URL: <http://listmgr.nrao.edu/pipermail/evlatests/attachments/20121115/8c3515d2/attachment-0001.pdf>
-------------- next part --------------
A non-text attachment was scrubbed...
Name: r1+r2.ps
Type: application/postscript
Size: 75491 bytes
Desc: not available
URL: <http://listmgr.nrao.edu/pipermail/evlatests/attachments/20121115/8c3515d2/attachment.ps>
-------------- next part --------------
A non-text attachment was scrubbed...
Name: r1-cor-r2.ps
Type: application/postscript
Size: 57202 bytes
Desc: not available
URL: <http://listmgr.nrao.edu/pipermail/evlatests/attachments/20121115/8c3515d2/attachment-0001.ps>