[teqc] qc of "highrate" data
lou at unavco.org
Tue Dec 12 08:33:55 MST 2006
There have been some recent questions about qc-ing of "highrate",
1-sec sampling or less (i.e. 1-Hz or higher). Some of the default
qc parameters in teqc will probably not work well for highrate,
since the default qc parameters were "tuned" long ago for 30-sec
The first parameter that probably needs adjustment is the multipath
moving average window length, which by default is set to 50 epochs.
(So for 30-sec data, this is nominally a 25-minute window.) The
default of 50 epochs works pretty well even for 10- to 15-sec data,
and maybe even for some 5-sec data. But when you are sampling at
around 1 sec intervals, you probably need to increase this with the
-mp_win option. The rule-of-thumb for starting is to try a number
of epochs which corresponds to 10 - 25 minutes. So for 1-sec data,
you'd want to see how teqc does, say, with "-mp_win 600" to "-mp_win 1500".
The next two parameters have to be considered together. These are
the "jump" conditions for the ionospheric combination and its time
rate of change (time "derivative") for detecting ionospheric slips.
By default (again, for say 30-sec data), teqc is set to use
"-ion_jump 3.403e38 -iod_jump 400" -- which is basically ignoring any
epoch-to-epoch ionospheric change (ion_jump) but detects an ionospheric
slip if the time derivative is 400 cm/minute or greater. This works
reasonably well for all known receivers at around 30-sec sampling,
mainly because of internal phase smoothing in the receiver.
As best we can tell, this internal phase smoothing is reduced or
skipped all together when the sampling interval decreases to a certain
value. We don't know any of the details, and the smoothing probably
depends on the receiver manufacturer, the receiver model, maybe the
receiver firmware. The current rule-of-thumb is that phase smoothing
done by the receiver is gone or much reduced when the sampling is down
to 1 sec (plus or minus). At that point, you are looking at much
more "raw" phase values, and the default jump conditions in teqc
don't work well. The best approach at this point seems to be to
set the time derivative jump condition to a very large value, and
set the ionospheric change to about 8 cm (plus or minus). (Note:
400 cm/min = 6.667 cm/sec.)
Some preliminary experiments with data from a NetRS suggest the
following should be used with 5-Hz data (0.2 sec sampling) -- at
least from a NetRS:
-mp_win 4500 -ion_jump 8 -iod_jump 3e38
The multipath window therefore is nominally 15 minutes long for the
averaging for this 5-Hz data.
The suggestions here should be applied for other receivers and other
highrate sampling intervals as a starting point, but you'll probably
need to experiment a bit.
Hope this helps.
More information about the teqc