GnssResiduals2AccuracyDefinition

Compute antenna accuracies from observation inputfileResiduals. The inputfileStationInfo is needed to assign the residuals to the equipped antenna at observation times.

The outputfileAccuracyDefinition contains at first step the same accuracy information for all antennas as the input file. Only the azimuth $A$ and elevation $E$ dependent grid points of the patterns where enough residuals are available ($>$ minRedundancy) are replaced by the estimated robust weighted accuracy \begin{equation} \sigma(A,E) = \sqrt{\alpha\frac{\sum_i e_i^2 p_i}{\sum_i r_i \bar{p}}}, \end{equation}where $e_i$ are the azimuth and elevation dependent residuals and $r_i$ the corresponding redundancies (number of observations minus the contribution to the estimated parameters). The weights $p_i=1/\sigma_i^2$ are computed from the estimated accuracies in the least squares adjustment. Since large outliers are downweighted using huber and huberPower, a correction factor $\alpha$ must be applied to obtain an unbiased estimate of $\sigma$.

The inputfileAccuracyDefinition can be modified to the demands before with GnssAntennaDefinitionCreate (e.g. with antenna:resample).

To verify the results the outputfileAntennaMean and the accumulated outputfileAntennaRedundancy of the computed pattern grid points can be written.

Example: Analysis of TerraSAR-X residuals of one month shows that low elevation GPS satellites are not tracked by the onboard receiver. An estimation of accuracies for these directions is not possible from the residuals and the apriori accuracies are left untouched. The other directions show very low phase noise hardly elevation and azimuth dependent for L2W. A nearly zero mean indicates the use of adequate antenna center variations in the processing.

See also GnssResiduals2TransmitterAccuracyDefinition.