[evlatests] Gain Transfer comparison -- 3bit vs 8bit

Tue Nov 22 13:25:20 EST 2011

    As part of preparation for the 'path-to-completion' review, I've 
taken a close look at our ability to transfer gain accurately over a 
wide elevation/temperature range.  The data used were from the long 
3bit/8bit run taken last week -- four sources of widely varying 
strength, over a large elevation range, and (simultaneously) a large 
temperature swing.  (Elevations went from 20 through 78 degrees while 
temperatures declined from +10 to -7 C). 

    The calibration methodology needs to be summarized -- it is the same 
that I have used for the 'flux density' runs: 

    1) All system corrections are applied.  This means that I used 
'TYAPL' to apply the switched power calibration to the data (both 3-bit 
and 8-bit). 
    2) An elevation gain curve was determined, utilizing all four 
sources (3C147, 3C84, J0303+4716, and J0349+4609), for each antenna and 
IF, using the AIPS task 'ELINT'.  This fits a 2nd order (parabolic) 
function. 
    3) This gain curve was applied, and the resulting antenna gains, for 
all sources, averaged over the full time range.  (In other words, no 
temporal gain variation was applied in calibration).  The source flux 
densities were determined from the averaged gain differences, using 
3C147 as the reference. 
    4) The residual gains were determined (to be shown below) after the 
best-fit flux densities were determined. 

     We are interested both in the temporal variation of antenna gains, 
after the application of the best-fit parabolic antenna gain.  Because 
we know there are unexplained variations in the 'PDif' values between 
3-bit and 8-bit paths, comparison of these residuals is expected to be 
'interesting'.  (And it is!).  We should also expect a difference in 
what the derived flux densities of the calibrators are between the two 
paths. 

    Major Results:

    1) For both 3-bit and 8-bit data, there is a large variation in 
visibility amplitudes as the source rose (and the temperatures fell).  
This is not due to a pointing problem -- a large change in total system 
power (PSum) was also seen.  (PSum is hugely dominated by receiver power 
and atmospheric emission, not source power).  The PSum curves are very 
revealing -- the total power first declined (clearly as a result of the 
source rising, and atmospheric contributions declining), reached a 
minimum, then started rising again, presumably as a result of the drop 
in temperature).   The system temperatures throughout this period follow 
a very reasonable curve -- dropping as the source rises and atmospheric 
emission declines, and reaching a steady minimum at the expected values 
(typically 28K) at high elevation.  This clearly indicates that the rise 
in system power at high elevations is due to a gain change, which in 
fact is correctly removed by application of the PDif gain monitoring.  
This is true for both 3-bit and 8-bit data paths.  But not all is rosy 
here ...

    2) If the gain monitoring system (PDif) were working correctly, we 
would expect to see the same parabolic gain fits for each antenna and IF 
between the 3-bit and 8-bit paths.  We don't.  To show this, I have 
attached 6 plots showing the ELINT fits for ea06, ea18, and ea26 
(carefully chosen to display the extrema in behavior) -- labelled with 
'*ELGAIN.eps'.   The 8-bit path fits for ea06 and ea26 show the expected 
behavior, with the maximum antenna gain at 50 degrees elevation, or 
higher.  ea18 is shown here partly because it clearly has an optics 
problem, resulting in a very sharp decline in gain at low elevation.  
(It's Tsys and PSum values look like the other antennas). 
    Compare these to the three plots from the 8-bit path -- ea18 shows 
the same basic function (the optics issue is dominating), but the other 
two antennas have completely inverted gain functions, and poorer fits. 

    3) The temporal stability of the individual antenna gains is shown 
in the remaining six plots (sorry about the size -- AIPS makes big 
postscript files, and converting these to another format loses essential 
resolution).  Plotted are the antenna voltage gains, as a function of 
time (hence, elevation) for the three sources, following the elevation 
gain function removal, and adjustment by the source flux densities (done 
independently for 3-bit and 8-bit).  Yellow is 3C84 (J0319+4130) = 30 
Jy, Green is J0303+4716 = 2.1 Jy, and Purple is J0349+4609 = 0.55 Jy).  
Note that for the 8-bit path, all three antennas easily meet the 0.5% 
stability requirement.  (The remaining 4 antennas meet this also).  
However, for the 3-bit path, the story is more complicated:  for ea26, a 
cubic residual is found (weakly seen in 8-bit also), for ea06, the RCP 
side shows a large offset in gain between the sources, while for ea18, 
the offsets are seen with the opposite sign.  (ea06R has 3C84 with lower 
flux, ea18R has it the other way around).   The antennas chosen for 
display are not by random:  ea06R has a PDif *expansion* (PDif is larger 
on 3C84 than on nearby cold sky), while ea18R is a large PDif 
*compression* (by 8% between 3C84 and cold sky).  ea26 shows very little 
PDif compression.  The effect of this variation is to make an error in 
the apparent flux densities. 

    4) Finally -- is there a difference in the derived flux densities of 
the three sources (w.r.t. 3C147)?  Unsurprisingly, there is.  Relative 
to 3C147, the calibration procedure claims the following:

    Source            3-bit flux                8-bit flux         Ratio 
(3b/8b)
---------------------------------------------------------------------------------------
    3C84                   30.47                    29.99              1.016
    J0303                    2.113                    2.152             
0.982
    J0349                    0.546                    0.553             
0.987
    3C147 (ref)           2.735                    2.735             1.000
---------------------------------------------------------------------------------------

    The ~4% deviation from strong to weak sources in the 3-bit path is 
clearly due to our 'PDif compression/expansion' problem. 


-------------- next part --------------
A non-text attachment was scrubbed...
Name: EA06-8B-ELGAIN.eps
Type: application/postscript
Size: 72139 bytes
Desc: not available
URL: <http://listmgr.nrao.edu/pipermail/evlatests/attachments/20111122/e34a9fd5/attachment.eps>
-------------- next part --------------
A non-text attachment was scrubbed...
Name: EA18-8B-ELGAIN.eps
Type: application/postscript
Size: 71231 bytes
Desc: not available
URL: <http://listmgr.nrao.edu/pipermail/evlatests/attachments/20111122/e34a9fd5/attachment-0001.eps>
-------------- next part --------------
A non-text attachment was scrubbed...
Name: EA26-8B-ELGAIN.eps
Type: application/postscript
Size: 72026 bytes
Desc: not available
URL: <http://listmgr.nrao.edu/pipermail/evlatests/attachments/20111122/e34a9fd5/attachment-0002.eps>
-------------- next part --------------
A non-text attachment was scrubbed...
Name: EA06-3B-ELGAIN.eps
Type: application/postscript
Size: 72022 bytes
Desc: not available
URL: <http://listmgr.nrao.edu/pipermail/evlatests/attachments/20111122/e34a9fd5/attachment-0003.eps>
-------------- next part --------------
A non-text attachment was scrubbed...
Name: EA18-3B-ELGAIN.eps
Type: application/postscript
Size: 71569 bytes
Desc: not available
URL: <http://listmgr.nrao.edu/pipermail/evlatests/attachments/20111122/e34a9fd5/attachment-0004.eps>
-------------- next part --------------
A non-text attachment was scrubbed...
Name: EA26-3B-ELGAIN.eps
Type: application/postscript
Size: 71792 bytes
Desc: not available
URL: <http://listmgr.nrao.edu/pipermail/evlatests/attachments/20111122/e34a9fd5/attachment-0005.eps>
-------------- next part --------------
A non-text attachment was scrubbed...
Name: EA26-8B-GAINSTAB.eps
Type: application/postscript
Size: 3937950 bytes
Desc: not available
URL: <http://listmgr.nrao.edu/pipermail/evlatests/attachments/20111122/e34a9fd5/attachment-0006.eps>
-------------- next part --------------
A non-text attachment was scrubbed...
Name: EA26-3B-GAINSTAB.eps
Type: application/postscript
Size: 3937988 bytes
Desc: not available
URL: <http://listmgr.nrao.edu/pipermail/evlatests/attachments/20111122/e34a9fd5/attachment-0007.eps>
-------------- next part --------------
A non-text attachment was scrubbed...
Name: EA18-3B-GAINSTAB.eps
Type: application/postscript
Size: 3937988 bytes
Desc: not available
URL: <http://listmgr.nrao.edu/pipermail/evlatests/attachments/20111122/e34a9fd5/attachment-0008.eps>
-------------- next part --------------
A non-text attachment was scrubbed...
Name: EA18-8B-GAINSTAB.eps
Type: application/postscript
Size: 3937988 bytes
Desc: not available
URL: <http://listmgr.nrao.edu/pipermail/evlatests/attachments/20111122/e34a9fd5/attachment-0009.eps>
-------------- next part --------------
A non-text attachment was scrubbed...
Name: EA06-8B-GAINSTAB.eps
Type: application/postscript
Size: 2706527 bytes
Desc: not available
URL: <http://listmgr.nrao.edu/pipermail/evlatests/attachments/20111122/e34a9fd5/attachment-0010.eps>
-------------- next part --------------
A non-text attachment was scrubbed...
Name: EA06-3B-GAINSTAB.eps
Type: application/postscript
Size: 2706527 bytes
Desc: not available
URL: <http://listmgr.nrao.edu/pipermail/evlatests/attachments/20111122/e34a9fd5/attachment-0011.eps>