Memo: Baseline dependent clocks
Some geodetic VLBI sessions show substantial baseline dependent clock
offsets. Four sources of baseline dependent clock offsets are possible:
- Unmodeled source structure effects;
- Systematic clock misclosure introduced by correlation and
post-correlation procedures;
- Presence of strong outliers;
- Errors during resolving group delay ambiguities what may result in
appearance of permanent clock misclosures at X or S band to be a
multiple of group delay ambiguity spacings.
It is questionable whether we should always estimate baseline dependent
clocks due to the points 1,2,3; John Gipson got evidence that it may even
slightly degrade performance in some specific cases. These baseline dependent
clocks adjustments do not exceed 100 psec.
At the same time we HAVE TO ESTIMATE baseline dependent clocks when
a mistake has been made in group delay ambiguity resolution and some baselines
have clock misclosures to be a multiple of group delay ambiguity spacings. If
the database has a clock misclosure at the X band one group delay ambiguity
spacing it propagates to the ionosphere free linear combination multiplied by
the factor 1.08. If the database has a clock misclosure at the S band one group
delay ambiguity spacing it propagates to ionosphere free linear combination
multiplied by the factor 0.08 (or divided by 12). Baseline dependent clocks due
to the errors in resolving group delay ambiguities are of order 1-200 nsec and
totally degrades fit when not applied.
There several ways to overcome this problem in SOLVE:
- Resolve group delay ambiguities once more and correct databases.
Unfortunately S band data are lost for some databases and it is not
always possible to do.
- Estimate baseline dependent clocks for all sessions. There is
a keyword BASELINE_CLOCKS in $FLAGS section of control file.
It may have three values: YES or NO or IN. "YES" means to estimate
baseline dependent clocks for the maximal number of baselines which
still doesn't lead to a normal matrix singularity. "IN" means to
estimate baseline dependent clocks for the baselines which are
specified in the database itself. "NO" means not to estimate
baseline dependent clocks.
- Estimate baseline dependent clocks for the specific sessions only.
$FLAGS section of the control file should contain the line
BASELINE_CLOCKS NO and the lines in arc-file for the problematic
sessions should be updated. A new option of the format of arc-file was
added to SOLVE 04-JAN-99: a line can have a token BASDEP_CLO with
value YES NO IN. Valve has the same meaning as in FLAGS section,
but it overrides global flag specified in FLAGS for this session
only. Example:
$96NOV04XA 4 FREE BASDEP_CLO YES !
Some lists.
- Full list of the estimates of
the baseline dependent clocks for about 2900 sessions used in global
runs. (956 Kb). The list contains 5 columns:
- Database names;
- Baseline names;
- Adjustments of the baseline dependent clocks in psec;
- Formal uncertainties of the estimates in psec;
- Ratios of the adjustments to their formal uncertainties;
The list is sorted in increasing ratios of the adjustments to its
formal uncertainties;
- The 258 bottom lines of the previous
list which contains adjustments of the baseline dependent clocks
exceeding 5 sigma level.
- List of 24 sessions which has
baseline dependent clock misclosures exceeding 300 psec, probably
due to errors in group delay ambiguity resolution. I recommend always
estimate baseline dependent clocks for these sessions.
Back to Leonid Petrov's discussion page.
This memo was prepared by
Leonid Petrov.
Last update: 04-JAN-99 14:55:16