[evlatests] Linearity at P-band
Rick Perley
rperley at nrao.edu
Wed Sep 3 16:07:12 EDT 2014
The P-band test done last week permits a close look at our system
linearity over power ranges of ~ 5.
Short observations of Cygnus A, 3C48, 3C380 and 3C123 were taken.
The power settings (both for the T304 attenuators, and the requantizers)
were set on Cygnus A. This source was chosen for the setup since it
increases system power by a factor of about four, and it seems prudent
to do the system gain setups on the strongest source one observes. (See
below for justification).
Two antennas were out of the array at the time: ea08 and ea11.
Two tests were then done utilizing the data from subband #6 -- 312
MHz (BW = 16 MHz).
1) PDif compression.
2) Flux density transfer from 3C48/3C380 to Cygnus A.
Results:
1) PDif ratios were computed in the usual way: PDif(Cygnus
A)/PDif(3C380). The PDif values from two antennas were not useable:
ea13 has zero mean PDif (diode not switching?), and ea18 has erratic
values on the 'L' side, and very odd values on the 'R'.
For the remaining 24 antennas, more than 80% of the PDif
ratios were between 0.97 and 1.03. This range seems to encompass the
typical variations in these values at this frequency band. The values
outside this range were all on the 'low' (compressed on Cyg A) side, but
none was less than 0.93. Results from L-band indicate that there is no
discernible effect on the visibilities when the compression is above
0.90, so I felt safe in applying the switched power to the data.
This distribution of PDif compression is, I believe, better than
any other band, when the power jump is at similar levels (a factor of 4
to 5).
2) I then calibrated the data, utilizing well-tested methods: delay
calibration, bandpass calibration, gain calibration. For the last step,
the flux densities from Scaife and Heald were utilized for 3C48 and
3C380. Note that these data were taken in D configuration, and the
available models do not include the effects of background sources. The
effects of confusing sources are quite evident in the visibilities. I
calibrated Cygnus A utilizing a simple average over the antenna gains
solutions from 3C48 and 3C380. Minimal (= almost no) editing was
attempted.
The results of this are shown in the attached figure, showing the
visibility amplitudes in the 'RR' (horizontal) correlation, over a u,v
range of zero through 1000 wavelengths. The partial resolution of
Cygnus A is evident.
There are two major points made by this plot:
A) Virtually all visibility points on the short spacings are
tightly clustered and heading for a zero intercept (total flux) of 6500
Jy. The Baars et al. value for Cyg A at 312 MHz is 6305 Jy -- about 3%
less than ours. This is of no concern since:
(i) The Baars scale for Cygnus A is not this accurate,
(ii) No accounting for background sources was made in
the calibration
(iii) The Scaife and Heald scale is also uncertain to
this level, at least.
B) There are some visibility values significantly below the
others. All of these come from ea28. What is special about this
antenna? It did not get a requantizer power adjustment, since it was
out of the array when this setting was made. In consequence, its power
level was set for the 'cold' sky sources -- when on Cygnus A the power
was four times higher than the standard. This result clearly indicates
that visibilities are strongly affected when the (digital) power levels
are higher than about 30 counts. In this case, an overpower of a factor
of four reduces visibilities by about 25%. (If two antennas had the
same overpower, the visibilities from that baseline would have been cut
in half).
I then ran a self-calibration, using a simple trial image. The
resulting visibility function is shown in the other attached figure.
The gain corrections are nearly all in the range of 0.96 to 1.04 --
these are typical variations at this band for any source -- whether it's
Cygnus A or a weak point source. This is a very nice result indeed.
*****************************************************
Finally -- a major appeal to get the requantizer settings made part
of the general system setup for this band:
I reviewed the PSum powers as a function of subband and antenna,
prior to the requantizer settings. The range is astonishing: A high of
over 100, and a low of less than one. (!!!!!). These are extrema. A
fairer test is the typical: A high of 40 to 50, and a low of 5.
Invariably the low is in the 16th subband (highest frequency). The high
is always in subbands 4 and 5. (Subbands 2 and 3 are even higher, but
RFI is part of the reason). What we are seeing here is the spectral
range of input power. The high end is weak both because of the receiver
bandpass/gain, and from the steep spectral index of synchrotron
emission. By overpowering at the low frequency end, and underpowering
at the high end, we are seriously affecting the science quality of the
data taken.
I claim it's time we implement an effective regimen for setting the
requantizer gains at P-band.
-------------- next part --------------
A non-text attachment was scrubbed...
Name: CygA-312.eps
Type: application/postscript
Size: 841457 bytes
Desc: not available
URL: <http://listmgr.nrao.edu/pipermail/evlatests/attachments/20140903/e7a4ff20/attachment-0002.eps>
-------------- next part --------------
A non-text attachment was scrubbed...
Name: CygA-312-Self.eps
Type: application/postscript
Size: 932744 bytes
Desc: not available
URL: <http://listmgr.nrao.edu/pipermail/evlatests/attachments/20140903/e7a4ff20/attachment-0003.eps>
More information about the evlatests
mailing list