FORMING DIFFERENCES AND CORRELATED
|
Most software packages do not permit the between-satellite differencing strategy
to be defined by the user. In general, one of two strategies is "hard-wired"
(Figure in section 7.2.4):
At first glance there would appear
to be no major difference in these approaches,
and the one implemented in a
package has generally been selected on the
basis of personal preference.
However, one remark concerning the influence
of differencing strategy on
the coordinate results should be made. In the
case of the base satellite
option, the choice of "base satellite"
is critical. A poor choice
(for example, a satellite that is low on the
horizon, or one with an
unstable onboard oscillator, or a satellite whose
track relative to the
receiver induces multipath reflections) will "corrupt"
all
double-differences formed. This same influence would not be as
widespread
in the case of the sequential satellite differencing approach,
as it will
"corrupt" only those double-differences that include
that satellite.
To induce a change in the differencing strategy that is employed (or at least the order in which the satellite differences are generated), one option may be to reverse the order of the stations. The first solution would identify station "A" as receiver 1 and station "B" as receiver 2. The order of the satellites recorded at station "A" may be different to those at station "B", and the order of satellites for receiver 1 generally define the differencing pattern. A second solution, in which station "B" is nominated as receiver 1 and station "A" as receiver 2, may lead to a better (or worse) baseline solution than previously. In extreme cases, it may be possible to even successfully resolve ambiguities in one case, but not if the station order is reversed!
Another test of solution "sensitivity" is to attempt double-difference solutions first taking the correlations into account (if it is an option), and then repeating the solution with the correlations "turned off". By replacing the diagonal VCV matrix of the double-difference observations with the correct (non-diagonal) form (eqn (7.2-18)), two effects may be mentioned:
|
|
For baseline solutions, not taking into account the observation correlations does not have much effect. In the case of network solutions for scientific applications (for example, high accuracy over long baselines), observation correlations must be considered as they can be employed to advantage in multi-station ambiguity resolution.
Back to Chapter 8 Contents / Next Topic / Previous Topic
a change in the baseline solution
--> of the order of a few parts per million