[evlatests] Bandpass STability

[Rick Perley]

    A short report on BP stability.

    The formal requirement is for a stability of 1 part in 10^4.  This 
is an unrealistically tough requirement, which did not take into account 
that the errors in the bandpass are general independent between 
antennas, polarizations, and vary indepednetly over time.  A more 
realistic analysis (driven by the scientific requirement of a detection 
of a weak absorption line from a strong source with opacity 1e-4) leads 
to a requirement of bandpass voltage stability of 1 part in 1000, over a 
frequency span of 0.1% of the rest frequency, and over a time scale of 
~30 minutes. 

    To measure whether we can do this, I have used the recent 
observations of 3C84 and three nearby objects (one of 2.5 Jy, one of 0.5 
Jy, and one noise field), taken for the 3-bit/8-bit tests.  I used every 
fourth observation of 3C84  to calibrate the bandpass -- this meant a 
bandpass calibration once every 30 minutes.  This bandpass function was 
then applied to generate an 'incremental' bandpass solution for every 
scan -- once every 30/4 = 7.5 minutes. 

    I attach two plots to show the result.  One is for 3-bit, the other 
for 8-bit.  The frequency span for each plot is the same.  (14.872 -- 
15.512 MHz).  These are BPLOT plots, showing the variation of the 
bandpasses over time.  The antenna shown is ea26, in RCP, for 5 subbands 
(128 MHz each).  The spectral resolution is 2 MHz, so the relevant 
frequency scale is about 7 channels (14 MHz)-- the requirement applies 
on this, and smaller, frequency scales (meaning the slopes and curvature 
on frequency scales larger than this are to be disregarded).   There is 
a solution every 7.5 minutes -- note that every fourth one is flatter 
than the others -- this is the 'calibration' scan. 

    The vertical scale is from -0.15% to 0.15%.  Bright red is +0.15%, 
dark blue is -0.15%.  The adjusted requirement of 0.1% stability will 
correspond to orange and dark green.  You'll note that for this antenna, 
the 0.1% requirement is easily met.   Also note that 3bit and 8bit are 
essentially identical.  The vertical 'disturbance' in the fifth subband 
on the 3-bit plot is the 'resonance' phenomenon (which is unusually weak 
for this antenna). 

    Not all antennas are as good as this one -- ea10 and ea14 are 
notably worse (by a factor of perhaps 3) on both 3-bit and 8-bit paths.  
The other four antennas in this experiment are about the same as this.  
And some subbands are worse than others.  The reason for these 
discrepancies will have to be uncovered over the course of time. 
-------------- next part --------------
A non-text attachment was scrubbed...
Name: EA26-8B-STAB.eps
Type: application/postscript
Size: 875914 bytes
Desc: not available
URL: <http://listmgr.nrao.edu/pipermail/evlatests/attachments/20111122/b39bb6d2/attachment.eps>
-------------- next part --------------
A non-text attachment was scrubbed...
Name: EA26-3B-STAB.eps
Type: application/postscript
Size: 873019 bytes
Desc: not available
URL: <http://listmgr.nrao.edu/pipermail/evlatests/attachments/20111122/b39bb6d2/attachment-0001.eps>