[evlatests] Direct Comparison of 3 and 8 bit gains

[Ken Sowinski]

On Thu, 15 Dec 2011, Rick Perley wrote:

>   Michael R. asked for a direct one-to-one comparison of the visibilities 
> from the 3-bit and 8-bit tests of last month.  I've now done this.
>   The data were taken in parallel, so that the two 3-bit and one 8-bit paths 
> are at the same frequency and taken at the same time.  I did the comparison 
> by generating the CALIB solutions (antenna-based gains) for the three paths, 
> utilizing the subband covering 15.000 -- 15.128 GHz.  The solutions were made 
> from data without applying any prior calibration, utilizing only the central 
> 10 MHz of data, on 3C84, which was taken to have 30.0 Jy flux density.
>   The results are shown in the attached .pdf files:  green is A1C1, red is 
> A2C2, and purple is B0D0, all on the same scale.    The gains are the same to 
> within 1% or so, with the major difference being simply one of normalization. 
> (AIPS clearly has a different idea of normalization than my eye -- I'll talk 
> with Eric to see why this is so).  Beyond the normalization issue, there are 
> some 'steps' in the gains for some antennas/IFs/paths which clearly are due 
> to temporary antenna-based gain changes of <~ 1%.    All the slow changes 
> (due to temperature, antenna gain, etc.) are identical.

I take issue with the last statement.  There are examples for which
the long term trends are the same for both 3- and 8-bit samplers, but
there are examples where the trends differ by about a percent over the
time range shown.  Both behaviors occur with about the same frequency.

It still remains true that the changes in the visibilties are much
smaller than than the changes in P_dif when 3-bit and 8-bit samplers
are compared.