[evlatests] Tcal accuries at P-band

[Rick Perley]

     Vivek's point is (of course!) a good one.   So I looked at the 
ratio of the PSum (digital total power) between Cygnus A and the 
calibrator 3C48.   Providing the digital power is itself linear, this 
ratio gives us a number proportional to the antenna efficiency.

     The result is that there is indeed considerable variation between 
the antennas -- and that the antennas which are clearly low in 
efficiency are often the ones requiring a higher-than-normal gain 
correction, and vice versa.   Hence, some (but not all) of the 
variations in the PDif-corrected data are due to an incorrect efficiency.

     At 312 MHz (the subband I utilized), the great majority of the 
antennas have a ratio of Cygnus A - to - 3C48 power between 3.8 and 
4.2.  Those significantly low (poor efficiency) are:

     ea06  The ratio is 3.6 and 3.4 in R and L, respectively. However, 
the gain corrections for this antenna are small, so if the efficiency 
really is low, then the effect is neatly offset by the TCals being too 
high.

     ea10 and ea12.  Both of these antennas have power ratios of about 
3.3.  The required gain correction are exactly as expected for low 
efficiency, so I conclude that for both of these, the Tcals are about 
right.

     ea14.  The ratios are 3.4 and 2.1 (!) for 'R' and 'L'.  However, 
the gain corrections (0.96 and 0.71) are incompatible with low 
efficiency.  Something else is wrong here, particularly with the 'L' side.

     ea18.  Very low on the 'L' side only (3.10).  No gain correction 
possible, as the PDif values were nonsensical.

     ea19.  Ratios of 3.1 and 3.4.  But the gain correction of 0.65 on 
the 'L' side is incompatible with low efficiency.  On the 'R' side, the 
correction of 1.07 is in the right direction for low efficiency, but is 
not large enough.  The listed Tcal of 21K is probably too high.

     ea23.  Ratios of 3.7 on both polarizations -- nicely matched with 
the gain corrections of 1.12.  Conclusion:  Tcals are o.k., but 
efficiency is too low.

     ea25.  Ratios of 3.81 and 3.19 for R and L.  The gain corrections 
of 1.25 and 1.5 are larger than needed for the apparent efficiencies.  
The listed Tcal values of 10 and 11 K are the lowest for any antenna, 
and are probably lower than actual.

     ea27.  Ratios of 3.65 for both sides, but not matched by the unity 
gain corrections.  Presumably offset by similarly incorrect Tcal values.

An extra piece of information is illuminating.  ea28 did not get its 
requantizers set in this experiment, since it was out of the array at 
the beginning of the run.  The power levels for this subband were 
(remarkably!) a factor of four too high -- PSum of 60 rather than 15.  
Despite this, the efficiency ratio (3C405/3C48 power ratio) was 4.15 for 
both polarizations -- right where it should be.   This is an indication 
(but not proof) of linearity in the power measurement.  However --- the 
visibility data from this antenna for Cygnus A are *definitely* 
compressed.   (More on this, later).

Recommendation:

We might consider the following, as an experiment:  Utilize the power 
ratio data to adjust the efficiencies utilized in the correction 
algorithm (TYAPL).  (Currently, these are all the same value).  Then 
repeat the operation which provides the gain corrections, and utilize 
these values to adjust the Tcal values.  If this works (as it should) at 
P-band, consider doing this at other bands.


On 09/02/2014 10:43 PM, Vivek Dhawan wrote:
>
> Hi Rick,
>
> I'm sure you appreciate this, but its worth reiterating, since we
> assume in TYAPL that all antennas have the same gain, (Jy/K), the
> CALIB gains are measures of (G/T) not T alone.
>
> I would not be surprised if the gains spread by many%, so I would not
> change the Tcal values without an independent measure of the G's via
> say an on/off measurement on a strong known source. Even if the absolute
> flux is uncertain, the relative gains can be measured.
>
> And if we are doing the Tcal fudge at P band, why not at all the other
> bands as I have lobbied for before?
>
>
> On Tue, September 2, 2014 18:06, Rick Perley wrote:
>
> |      Here I report on the accuracy of the Tcal values.
> |
>
> |      The not-so-quite good news is that there is a large scatter. Below
> | are the CALIB gains needed to bring the data into alignment with the
> | correct flux densities.  Presuming the utilized efficiencies are
> | correct, and are the same for every antenna, the error is in the values
> | of the Tcals.  The needed correction to the Tcal values is the square of
> | the gain factor listed.  Hence, if the gain factor is 1.1, then the Tcal
> | value should be increased by 1.21.
> |