[evlatests] Freq offsets, phase transfer, etc.

[Vivek Dhawan]

Re the current discussion of frequency shifts & phase residuals, I had
3 facts that I was trying to reconcile. I now claim I can do so, and
they confirm:

(A) There is a frequency shift of ~250 KHz affecting the phases applied
     during correlation. (Ed measured 374KHz, I dont think I can be so
     accurate).

(B) That the delay model, prior to converting to phases, is nevertheless
     correct.

This was based on the long track at C band, 0.5 MHz channels, DNC array.

1. The phases on a single source, over a long track, twist by ~1 turn
    in 4.5hr, on baselines of about 20 Klambda.

2. The phases after self-cal on the above source (scan by scan), when
    transferred to other sources, leave a residual of upto ~1/2 turn
    on sources about 10deg away.

3. The phase offset on a given source is completely _independent_
    of sky frequency. This is true for both items 1 and 2, i.e., the
    phase twist with time is independent of subband; and the phase
    transferred from cal to target varies with time, but not with sky
    frequency. This took me a while to digest, but I think it shows that
    the correlator delay model is correct to within measurement error of
    ~10psec. A wrong frequency is multiplying the delay to get phase.

I have put plots in Michael's mail slot. I'm out on Thursday - gone to
Albuquerque for car-related woes (weak pun on 'correlated')

Vivek.