##### LTAFILE : node51:/data2/gpuuser/gabtest_1350lo_16aug2013_lsb.lta
##### Settings:
   Observe date 16 AUG 2013
   OBJECT  3C147 and 1330+251
   LO1 1350 MHz
   RF 1280 MHz
   Band = flipped (GUI), meaning LO1 > RF (or LSB)
   S04 is taken for maintenance
   No. of Channels 2048
   Channel-width 97.66 kHz

##### Note from Dharam:
 Re. scan-0, we feel that pointing-offsets were not loaded (missed running "run pmanto"), hence correlation-coeffiscients are low in
this (scan-0) scan.  Also, pl. ignore scan-4 (we missed "gts" command thr'u online). The band-shapes (attached files
gab_gpu_1350lo_lsb_phs-chan.jpg and gab_gpu_1350lo_lsb_amp-chan.jpg) are made using scan-2.

 Re. correlation-coefficients (ccfs), they are typically a factor of 20% low than the expected ccfs (ignore scan-0 and scan-4) for all
scans.  E06 needs attention (Ajit, pl. look into this!), it's self-bandshape is poor and baselines involving it give very low ccfs as
compared to other baselines.

##### Note from Yash:
Good to see these results.  Following a quick look through, I have the following queries and comments :

1.  I request that self amplitude plots be included as a standard part of any such reporting : we will continue to learn a great amount
from looking at and evaluating self data -- please don't ignore them.  Could I ask for the self plots for these observations to be
generated even now and circulated.

2.  Also request that all the plots be made with "lines" option (rather than "points") as finer details of fluctuations are easier to
make out with line plots.

3.  Different plots in this set appear to have been made with different reference antennas (e.g. C10, W04).  Would be better to make
all the plots with the same reference antenna for easier comparison -- it is better to use a central square antenna for this.

4.  I would also suggest a slightly larger separation between the chosen LO and the centre of the band (e.g. 1350 - 1280 = 70) -- 1350
is somewhat marginally above the top end of the band (1280 + 60 for the 3 dB point of the sub-band filter); one could easily use 1370
or even 1380 for the GAB LO for 1280 band -- this would result in the 120 MHz passband being more in the centre of the 200 MHz band of
the GPU corr (instead of being close to the lower edge).  This is what I meant in my earlier email about working out some standard
values of GAB LO for each RF setting : upper and lower LOs for each of 4 subbands for LBand; upper and lower LOs for the existing 610
system; upper and lower LOs for the 250-500 MHz system.

5.  I agree that E6 needs to be looked into.  Its cross bandshape shows clear evidence for a problem (would be useful to look at the
self bandshape also).

6.  In general, the cross-corr ampls and phases with time (and freq) look to be behaving fairly decently -- this is a good sign about
the basic performance and stability of the entire chain from RF to GPU corr.

7.  Some work needs to be done to better undestand the correlation coefficients. Would be useful for any such report to contain the
flux values of the calibrators used, and the expected value of the corr coefficient.

8.  The above issue is best addressed by having GSB data in parallel with the GPU data and pulling out the frequency channel(s)
corresponding to the same RF frequency for detailed comparison of amplitude and phase.  Hope this can start happening in the tests
planned from next week onwards.

I think this is good progress and we should start thinking of "regularising" such tests so that the systems are exercised as much as
possible.

##### More Note from Yash:
Thanks for sending this quick first update on the data of 16th Aug. Couple of points / requests are as follows :

1.  Can the self plots be generated from tax (instead of from the bandmon display) by using the same settings (scan no, timestamp etc)
as used for generating the cross-amp vs allchan plots -- then it will become much easier to compare between self and cross results :
e.g. the small ripples seen in cross of S04 and S02 : what fraction of these are seen in the self plots; the RFI seen near ch 250 looks
very intriguing where it appears to change shape from two features for arm antennas to one feature for CSQ antennas -- what does the
self show for these antennas at the same timestamp.

2.  One important plot that got missed out is the allant_phase_vs_time plot with C4 as the reference.  Would be useful if this could be
added to your present collection (alongwith the selfs generated as described above).

##### And more Note from Yash:
Thanks for sending this missing plot of phase vs time.  If you compare this with the earlier version (which had E6 as the reference
ant) you can see the marked difference (the effect of using CSQ antenna as reference, vs using an arm antenna as reference).  The new
plot with C4 as reference clearly shows that the phase is quite stable with time on short baselines (e.g. C6, C10) -- in fact, you can
see nice things like very flat / constant phase with rms reducing as you go to the stronger calibrator etc.

On the other hand, as you go to longer baselines, one can see increasing phase variations with time, with some of the worst being for
C4 with W4 and E6.  In fact, when it is on the weaker calibrator, the phase of C4 with W4 and E6 shows a ramp with time -- the ramp is
faster for E6 than for W4 (and much shallower ramp is visible with W1) -- looks like some evidence for some residual, uncorrected
fringe?  Sanjay and Harsha : is there some possibility for a small error in correction of fringe (maybe due to time offset or something
else?).