[evlatests] Frequency offsets between cal & target.

[Ken Sowinski]

On Wed, 6 Feb 2008, Vivek Dhawan wrote:

> VLA and EVLA now: Each antenna shows a phase change with freq.
> offset of (L df), where L is the LO path length to the pad, or
> ~30 turns per 1MHz offset at the ends of the B array.
>
> Archival data clearly shows that (as expected from the design),
> the OLD VLA was much less sensitive. e.g. AB1219, observed on
> 13-NOV-2006, C array, 2.1 MHz frequency offset, shows only about
> 20deg. phase offset (If the old VLA behaved the same as the current
> VLA, there would be ~10 turns of phase at the ends ~1.5km away.)

I believe I have found why the old and new system are not the same
for VLA antennas.  The answer lies in knowing what delay is.  In
general the instantaneous phase we use to drive the lobe rotaters
is the fractional part of the product of signed sum of the LOs and
delay.  In the Modcomp era we computed:

phase = f*(Del_geom + Del_atm + Del_cable + Del_pec)

where Del_cable tried to account for most of the path between antenna
and correlator.  In the current system we compute:

phase = f*Del_calc

where Del_calc is the total delay returned by calc.

Since all the terms but Del_calc (or Del_geom) are essentially constant
the instruments work equivalently for a constant frequency observation,
but are quite different for observations which make small changes in
frequency between calibrator and target.


EVLA antennas suffer a similar symptom, but the reason in this case is
that the fine LO tuning is done at the antenna rather than the control
building.