A saw-like pattern in residuals and phase-cal phases

Analyzing IVS databases at end of 2015 I noticed excessive noise in group delay residuals at baselines with YEBES40M. I requested the visibility data from two experiments, eur138 and r1720, observed in December 2015 and reprocessed them.

Group delay residuals in both eur138 and in r1720 experiments show a saw-like pattern. This pattern increases the rms of residuals by roughly a factor of 2.

Analysis of phase calibration phases shows interesting things. Here is the plot of YEBES40M phase calibration at S-band and its zoom.

Here is similar plot of YEBES40M phase calibration at X-band and its zoom.

We see a superposition of two components: saw-like pattern that is not scaled with frequency and a slowly variating components without teeth that is scaled with frequency. The spacings of teeth in phase calibration is approximately correspond to spacings of group delay residuals. Similar situation in r1720 experiment.

What if to solve for admittance of phase-calibration phase, i.e. to use phase-cal phases as partial derivatives? I wrote a simple user partial program in Solve and re-ran a solution. The saw-like pattern was significantly (a factor of 5) reduced, although did not disappear entirely. Analysis of listings of reference and advance solutions shows a remarkable improvement in rms of postfit residuals. Roughly speaking the saw-like pattern contributed 50 ps noise in quadrature. Similar situation with experiment r1720: reference and advance solutions. Compare group delay residuals of the reference and advanced solutions:

eur138 reference advance
r1720 reference advance

Phase-delay solution in r1720 at RAEGYEB/YEBES40M baseline

Experiment r1720 involved 193m long baseline RAEGYEB/YEBES40M. Without estimation of phase-cal admittance I could resolve S-band phase ambiguities, but not X-band. Using estimation the admittance factor, I was able to resolve ambiguities at both bands. Here are phase-delay residuals+clock function at S-band and X-band. Here are listings of phase delay solutions at S-band and X-band.

The estimates of the admittance factors agree between three different solutions in r1720 experiment, but a factor of 2.3 greater in eur138 solution:
r1720 X/S Group delay -20.2 ± 0.7 ps/rad
r1720 S band phase delay -19.7 ± 0.5 ps/rad
r1720 X band phase delay -22.6 ± 0.4 ps/rad
eur138 X/S Group delay -46.3 ± 1.8 ps/rad

What is the reason?

Since I have never been at YEBES I am cautious to conjecture. I have disabled phase-cal at YEBES40 in a trial solution. The saw-like pattern in residuals remained. That means that saw-like path delay occurs not in the path from the receiver to the data acquisition terminal (DAT), but somewhere before.

I suggest the station staff to check every air-condition unit and investigate which on injects unwanted path delay.


Back to Leonid Petrov's discussion page.

Last update: 2016.02.14_20:17:31
Web page was prepared by Leonid Petrov ()