In depth analysis of R1542 VLBI experiments on 2012.07.16
There was a message
about poor quality of analysis results of recent VLBI experiment r1542.
This prompted me to re-analyze the experiment using
the visibilities
in FITS-IDI format.
Experiment r1542
A quick analysis reveals three problems in original results, processed
with difx2mark4+fourfit+Calc/Solve, and submitted to IVS:
- HARTRAO (sub)ambiguities;
- noisy NYALES20 data;
- noisy YEBES40M data;
Listing of the original solution is available
here.
I reanalyzed the data using dif2xfits+pima+vtd/post-solve.
Hartrao
Correlation report
mentioned disconnected cable stretcher during experiment. The phase slope
of IFs changed before and after disconnection. This caused a double-peaked
delay resolution function, and as a result ambiguities with spacings
around 10ns at S-band. The best solution would be re-computed complex
band-pass before and after event, but I merely re-fringed the data with
a narrow search window. This worked. No clock offset introduced in the
original IVS solution is needed. I turned phase-cal off during reprocessing.
Cable calibration looks wrong. Turned off cable calibration.
NYALES20
Phase-cal phases look very very different from other stations. They
abruptly change after each scan. David Gordon suggested to re-fringe the
data without phase-cal. This helps a lot. Here are plots of X-band
group delay residuals (with ionosphere contribution from GPS applied)
with phase-cal and without phase-cal:
YEBES40M
David Gordon mentioned that residuals with YEBES40M show some structure.
Here are residuals of the original IVS
solution.
After using mouse for 10 minutes, I succeeded in resolving ambiguities
with spacings ~0.4ns. Here are residuals
after resolving ambiguities in the IVS databases. After resolving
ambiguities, solution with YEBES40M became much better. Here is the solution
listing of the
reanalyzed IVS database.
However, full reprocessing the data with PIMA and VTD/post-Solve
does not reveal any abnormalities at baselines with YEBES40M at all.
Here are residuals of fully
reprocessed r1542 experiment. No re-fringing was needed, except
a couple of points affected by RFI at S-band. I used measured phase-cal
for YEBES40M during reprocessing. Here is the solution
listing of the fully reprocessed
r1542 experiment.
Therefore, we see that ambiguities with spacings 0.4 ns are artifacts
of difx2mark4 + fourfit analysis chain. Similar problem affected
EFLSBERG in
euro112 experiment. I am not aware whether the problem was followed
up.
Back to Leonid Petrov's discussion page.
Last update: 2012.08.10_12:04:48