A similar situation was in r1774 experiment one week before. The scatter of post-fit residuals was large at all baselines, but with ISHIOKA. Why the postfit residuals at ISHIOKA/SEGONG were normal (well, normal in VLBI2010 system means only a factor of two worse than it was 10 years ago)? The correlator output did not have cross-spectrum for IFs 12-16 for baselines with ISIHOKA. Therefore, group delays at ISHIOKA/SEGONG were computed using IFs 7-11.
I re-fringed data with PIMA and confirmed large scatter
for baselines with SEJONG. I disabled phase-cal at both ISHIOKA (the phase-cal
signal was not present, and I am not sure whether the unit is installed at all)
and at SEJONG. But I found another peculiarity: a saw-like pattern in group
delays at ISHIOKA:
.
Group delays computed with Fourfit did not show such a pattern.
I made step forward and ran two fringe fitting: with using IFs 7-11 and
with using IFs 12-16. The reference frequency for the first variant is
8.20499 GHz and 8.73299 GHz for the second variant. Using the estimates
of the group delay I transferred fringe phase from 8.73299 GHz down
to 8.20499
( φ2(f1) = φ2(f2) +
2π * (f2 - f1) * τgr ) and
made a plot of differences in fringe phases computed over IFs 12-16
minus fringe phases computed over IFs 7-11. The plot of dependencies of
these phase differences in time is shown below:
.
For comparison a similar plot for the baseline NYALES20/WETTZELL in
the same scale:
.
Similar plot at ISHIOKA/KOKEE reveals the saw-pattern, similar to that
we saw in the group delays:
.
After removal of a trend -1.007 × 10-3 rad/s, the residual
phase differences look normal:
.
This looks very strange. IFs 7-11 and IFs 12-16 are fed with different cables. It looks like phase in the cable that brings IFs 12-16 drifts with respect to the phase in the cable that brings IFs 7-11. But the most puzzling is that why we do not see this phenomena in Fourfits output.
Compare it with the plot at the beginning of the document (they are in the same scale). And the saw-like pattern at ISHIOKA is gone.
And yes, this trick helps to rescue SEJONG data in r1774 experiment as well.
Experiment r1773 experimetns with SEJONG shows changes during last three
hours, specifically after 2016.12.26_14:02. Phase cal shows jumps:
Post-fit residuals also shows jumps and large scattering. Re-fringing
witth pcal disabled for SEJONG does not bring residuals to normal.
Phase differences of IFs 7-11 versus IFs 12-16 at SEJONG/TSUKUB32 without
pcal applied shows linear trend and jumps after 2016.12.26_14:02
Applying phase-cal reduces trend by a factor of 4 but does not eliminate
jumps:
These jumps spoils X-band group delays. Here are X-band post-fit residuals:
but they do not spoil S-band group delay residuals:
Which group of Ifs: 7-11 or 12-16 is to blame? Here are X-band residuals computed
from IFs 7-11
and here are X-band residuals computed from IFs 12-16
They look normal.
We see phase shift between IFs 7-11 and 12-16. While the phase-shift is constant, it is calibrated when complext bandpass iscomputed. A drift of phase shift was mainly compensated with phase-cal. However the phase shift started to jump after 14:02 and these jumps were not compensated by phase-cal. The phase shifts did not affect group delay computed over 7-11 and 12-16 IFs, but it affected group delay computed over IFs 7-16.
Last update: 2017.02.16_21:15:44
Web page was prepared by Leonid Petrov
()