[teqc] probable upcoming modification to multipath qc algorithm

Lou Estey lou at unavco.org
Tue Jan 22 12:36:08 MST 2019


All,

For the next release of there, there is a modification to the multipath algorithm that
is being tested and will probably be included.  The modification is to readjust the
DC bias value of any multipath combination (e.g. mp12) of an SV after any time gap
of a certain length or longer.  This modification (in most cases) corrects the multipath
rms estimate if the rising SV (after already having set earlier in the same dataset)
does not trigger a possible multipath slip detection.

First let's review.  There are a variety of linear combinations of phase and pseudorange
observables which are interpreted as multipath.  Doing such a multipath combination for
all possible sampled epochs (that have the necessary observables) is a time series, with
possible gaps.  The multipath combination values wiggle, up and down, about a DC bias.
When a multipath slip occurs (as from a phase slip in one of the phase values being used
in the combination), the multipath values jump and then wiggle, up and down, about a
different DC bias.  What we're after is an estimate of the RMS of the multipath values
about each DC bias.  The trick, then, is to detect multipath slips and to estimate the
initial DC bias of the time series, and then readjust that estimate after each slip.

Also, the multipath algorithm in teqc is designed to work when there is no ephemeris
information for determining the SV positions.  In other words, it's designed to work
when there is no knowledge of whether the SV is rising or setting.  This is done
by detecting time gaps and taken appropriate action when a gap is detected, which is
a reasonable approach because any type of long gap (due to a loss of tracking) results
in a loss of lock and (usually) a substantial change of DC bias when lock is reestablished.

The original algorithms work well when the value of a multipath combination from a rising
SV is substantially different from when the SV had set -- since this situation would
generally trigger a check for multipath slip.  In such a case, i.e. an SV setting and
then later rising in the same dataset, the check would determine this event to be ending
a data gap in the multipath time series -- therefore not a real slip in continuous
observations -- and the DC bias was reset.  However, there was no DC bias reset if
only a data gap was detected.

Investigators with ESA (Raul Perez, Nacho Romero, Jeremy Tournay) have found data
from certain permanent stations where, for whatever reason, the multipath DC bias when
the SV rises is not too much different from when the SV set.  In other words, it's not
different enough to cause teqc to flag the change as a possible slip (which, for mp12,
has a default threshold a little over 5 meters).  Therefore the original logic leading
to a reset of the multipath DC bias in such a case does not occur.

The modification to the multipath algorithm that is being considered is merely a reset
of the DC bias after _any_ multipath data gap.  The following sample from a problematic
site shows the resulting change in the multipath values in the .m12 file (for mp12) in
the first four epochs of a rising SV, G22, which had set earlier in the data:

new code, with the DC bias reset after a multipath data gap:

  77400.0000 11 G10 G12 G14 G18 G20 G22 G25 G31 G32 G01 G24
   -0.053     0.005     0.075    -0.440     0.113     0.138    -0.162     0.268     0.016    -0.030     0.034
  77430.0000 -1
    0.027     0.108     0.056    -0.618    -0.099     0.187    -0.281     0.080     0.004    -0.062    -0.019
  77460.0000 -1
    0.102     0.101     0.022    -0.335    -0.352     0.233    -0.358    -0.047     0.024     0.198    -0.122
  77490.0000 -1
    0.279     0.068    -0.048    -0.121    -0.010     0.370    -0.153    -0.098     0.007     0.065    -0.076

original code, without the DC bias reset after a multipath data gap:

  77400.0000 11 G10 G12 G14 G18 G20 G22 G25 G31 G32 G01 G24
   -0.053     0.005     0.075    -0.440     0.113     3.997    -0.162     0.268     0.016    -0.030     0.034
  77430.0000 -1
    0.027     0.108     0.056    -0.618    -0.099     3.967    -0.281     0.080     0.004    -0.062    -0.019
  77460.0000 -1
    0.102     0.101     0.022    -0.335    -0.352     3.935    -0.358    -0.047     0.024     0.198    -0.122
  77490.0000 -1
    0.279     0.068    -0.048    -0.121    -0.010     3.993    -0.153    -0.098     0.007     0.065    -0.076

With the original code, the multipath values for the rising G22 are almost 4 meters,
although the values drop down to correct levels as the window of the multipath
moving average becomes repopulated.  With the new code, the values are much closer
to where they should be.

Also, to compare the more general qc results:

[1641] teqc +quiet +qcq sample.18o > new.qc
[1642] teqc +quiet +qcq sample.18o > old.qc
[1643] diff old.qc new.qc
79,80c79,80
< Moving average MP12     : 0.206787 m
< Moving average MP21     : 0.183625 m
---
 > Moving average MP12     : 0.194933 m
 > Moving average MP21     : 0.182473 m
91c91
< other msec mp events    : 0 (: 36)   {expect ~= 1:50}
---
 > other msec mp events    : 0 (: 34)   {expect ~= 1:50}
101c101
< SUM 18 11 11 00:00 18 11 11 23:59 24.00  30  25071  25071 100  0.21  0.18  25071
---
 > SUM 18 11 11 00:00 18 11 11 23:59 24.00  30  25071  25071 100  0.19  0.18  25071

... which is about what one would expect.

In the first paragraph, I said this modification "in most cases" adjusts the DC bias
and corrects the multipath rms estimate when the SV rising after already having set
earlier in the same dataset.  When does it not work?  This would be when the time
between the SV setting and its rising is less than the gap threshold, which, by default,
is 10 minutes.

Also I'll mention that the problem found does not occur when the SV is rising for the
first time in a dataset with no prior observations.  In this case, teqc is still dealing
with a blank multipath search (since there has been no multipath combinations of this
type yet) and teqc is hunting for the initial DC bias for the multipath.

cheers,
--lou

~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~
Louis H. Estey, Ph.D.              office:  [+001] 303-381-7456
UNAVCO, 6350 Nautilus Drive           FAX:  [+001] 303-381-7451
Boulder, CO  80301-5554            e-mail:  lou  unavco.org

"If the universe is the answer, what is the question?"
                                                -- Leon Lederman
~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~



More information about the teqc mailing list