[evlatests] Preliminary Wobble Report

[Ken Sowinski]

On Wed, 24 Nov 2010, Brent Carlson wrote:

> Hi Rick,
>>
>>    For the third test (Brent's algorithm)
>>
>> No obvious wobbles are visible.  The phase wobble on 9 x 27, if
>> present, is less 1 degree pk-pk.  The amplitude wobble on 2x10 might
>> be present -- if so it is of much lower amplitude.
>>    A Fourier analysis is needed for a most discerning judgment.
> If there is no post-correlation residual sub-sample delay correction
> being done, depending on baseline length and integration time, I can
> imagine there being a percent or so amplitude oscillation, and a
> residual phase slope vs freq changing with each integration time, as the
> integration time beats with delay tracking, which could explain what you
> see. (The baseline-based coherence loss at the edge of the band due to
> 22.5 deg phase excursion is ~2.5%, which is ~1% averaged over the sub-band.)

Would you not see this then on all base lines which included
at least one antenna with a slow wrt integration interval
delay rate?  There should be many such in C array.  While
Rick did not report this detail, he usually looks at only
a few channels near the center of the subband; if that was
the case the delay clunking affect on amplitude would be
reduced.

> There's also the question as to whether the correlator chip integration
> time, for this test, was set to some odd ball number, or the standard
> 200 or 250 usec.

250 usec was used.  Bob believes that that would be less
important than the choice of LO offset and wanted to test
that hypothesis.  We should have also taken some data
with both the improved Lo offset algorithm and an odd
CC dump interval, but neglected to do so.