[evlatests] Tcals derived from aggregated tipping scans

[Brian Svoboda]

Hi everyone,

I've aggregated a set of ~100 executions of tipping scans that Pedro has 
run that span C to Q (8-bit), although I've analyzed Ku to Q here since 
that's where the tips have the least sensitivity to elevation-dependent 
ground pick-up. You can find in the attached plot for all antennas. I am 
still not completely sure what issues or biases may be lurking in this 
way of deriving the values, but they seem to be reasonably accurate 
where they can be independently checked, and have the notable quality of 
(in theory) not being dependent on the antenna forward efficiency.

These collect all wide-band tipping scan executions for the latest 
receiver on an antenna. Black shows the reference values and blow shows 
the weighted mean of the derived Tcals, and the color shading shows the 
1-sigma interval of all the epochs that go into the weighted mean. The 
lower panels show the ratio. The labels above a given band's frequency 
range show the receiver serial number, date of installation, and the 
number of epochs (executions) that went into that curve.

An interesting cross check is that there appear to be, interestingly 
enough, a few antennas whose Ka Band receivers (i.e., ea07 L, ea13 L) 
that do not have Tcal values, and the Tcals put into the SDMs CALDEVICE 
table are simply extrapolated from the nearest frequency bin. In these 
cases, the reference Tcals are clearly bogus, but the derived Tcals look 
quite sensible compared to the other feed and other antennas. I think 
this is strong evidence that this method is deriving reasonably accurate 
values.

These Tcals are derived as part of a three-stage fit. The first takes 
each (ant, feed, spw) elevation versus system temperature curve using 
the reference Tcal and derives an opacity. The second fits a single 
atmospheric model to all estimated opacity values (i.e., to all 
antennas). The third stage derives the scaling factor necessary to bring 
the measured opacity for a single curve to match the model. A 
consequence of this is that the Tcals can only be fit with this method 
against some array-wide average Tcal value. If all the Tcals are 
systematically high or low across all bands/frequencies, it will bias 
the array-wide atmospheric model fit that the individual measurements 
are rescaled to. A temperature change could plausibly do that, although 
I'm not sure how large the effect should be, and because these are 
aggregated over many different times of year and hours of day, I think 
that should wash out in the weighted average Tcal fit. I've also done a 
first pass to clear out obviously poor data (e.g., clouds, bad antennas).

It's a lot of data, so I'd be happy to provide anyone with an TSV file 
of the values. An interesting angle for this sort of thing is that if 
the Tcals here can be reliably derived from tips and are not dependent 
on the gain, then they could be used to derive the antenna efficiency as 
a function of elevation tracking an astronomical source.

Clear skies,
Brian

-- 
Brian E. Svoboda, PhD
Associate Scientist
National Radio Astronomy Observatory (NRAO)

Office: DSOC 312
Office Phone: +1 (575) 835-7246
-------------- next part --------------
A non-text attachment was scrubbed...
Name: tcal_spectrum_all.pdf
Type: application/pdf
Size: 890210 bytes
Desc: not available
URL: <http://listmgr.nrao.edu/pipermail/evlatests/attachments/20260903/1f8a63fb/attachment-0001.pdf>