[evlatests] P-band Correlation problem -- nonexistent!

[Rick Perley]

I reported last week evidence for a correlator-based problem, based on 
P-band data taken over the weekend.

The report was completely erroneous.  There was nothing wrong with the 
correlation products.

The erroneous report was based on incorrect calibration method, 
involving the calculation of the delays.   In the hope that a short 
description might help others avoid falling down the rabbit hole, I 
offer the following explanation.

At the Cassegrain bands, we often utilize 128 MHz-SPW windows (64 at 
L-band).  This, combined with the high SNR, permits single-SPW delay 
solutions to be made.

However, at P-band, with much narrower SPW widths, and much poorer SNR, 
use of single-SPW delay solutions is very risky.  And indeed, this is 
what led to my premature conclusion that the correlator was somehow 
destroying coherence.  Examination of the delay solutions showed wild 
variations from solution to solution for many antennas.  Application of 
these incorrect solutions caused the decorrelation.

The correct method is to use multi-band delay solutions.  Examination of 
these shows very steady delays and phases for all antennas, for all 
time.  Application of these resulted in excellent calibrated data.

Thanks to Paul for looking into the data, and insisting that all was 
normal.  Apologies to anybody else who followed me down the rabbit-hole.

Rick