[evlatests] More on 3 and 8 bit sampler comparisons

[Vivek Dhawan]

I looked at a 2.5hr long track on 3C84, taken the same day, same setup
as just described by Rick.  Essentially the same report.

Some additional points are:

Closure phase rms on the 3-bit data are higher by almost a factor 2,
compared to 8-bit; LL-3bit closure phase on triangle 15-22-28 was as good
as the 8-bit side.  Moreover, the mean closure phase was non-zero, and
different betweeen R and L, 3-bit and 8-bit, though generally the 8-bit
were closer to zero. However, the 8-bit is not perfect - there is a
sawtooth variation of closure phases on all triads, on timescale ~minutes,
that is completely absent in the 3-bit data.

In a previous test, (Aug 24, perhaps not optimally setup, but still
revealing) the closure noise on 3-bit data was modulated on timescale of
minutes and varied by a factor ~2. That modulation comes and goes, it was
not visible in the recent data.

I looked for the 'flying birdies' (frequency spikes) and saw only the
external ones from the microwave links.

For some reason, FRING found its delays with no trouble, but BPASS and
Calib had trouble on a few stations/subbands, I initially thought this was
from the RFI, but they failed on different subbands in the 3 and 8-bit data,
on a single carefully vetted scan which looked clean. Not sure its worth
too much time on tracking this down, but I will watch for these errors.

V.




Rick Perley wrote:
>     A further review of the 30 minute test taken Friday afternoon 
> reveals some other interesting features. 
>