[evlatests] Strange Delay behavior

[Ken Sowinski]

On Fri, 21 Dec 2012, Rick Perley wrote:

>    In a test taken last night, at Ku-band, using 1 GHz-wide IFs, I find
> that the delays are *not* the same across each of the IFs!  The delays
> change in blocks of 4 subbands ... so that subbands 1 -- 4 have a
> different delay than subbands 5 -- 8, while 9 -- 12 is different than 13
> -- 16.
>
>    I didn't think the correlator could do this!  (And, for this
> experiment, where the 8 subbands per IF are all contiguous, I don't
> think it should!).

This was caused by too long a memory of subband delays that were used
to phase several subbands simultaneously.  The memory was found to be
in the executor, not the correlator.  I will write a script which can
be used to erase this memory until the executor can be made more
forgetful.

If subband phasing is in use when a script ends, those subband delays
will continue to be sent to the correlator; if the script ends with
phasing turned off, all is well.