For the CHAMP orbit campaign, the available solutions are compared by computing the difference vectors in 2000 sample points per day. This corresponds to a sample interval just under 45 seconds, during which the satellite travels nearly 350 km along its orbit. In total, the 11-day campaign period from day 140 to 150 of 2001 provides 22000 equidistant sample points, which must be considered as a reliable basis for statistical analysis.

By carefully assessing all available comparisons, the following information is obtained from the comparisons alone:

• An absolute upper limit to the orbit error RMS(A) is given by the lowest of the (N-1) comparison values RMS(A,B). This corresponds to the worst case assumption that RMS(A,B) would be entirely due to RMS(A), namely if orbit B is the true satellite orbit and therefore RMS(B) = 0.
• Some obvious dependencies between two solutions can be identified, namely if one of the (N-1) comparison values RMS(A,B) is distinctly smaller than the other (N-2) values for orbit A. In particular, dynamic solutions that use the same models (e.g. the same gravity field) can sometimes show a common error signal, or two solutions that use the same POD software.
• Possible reference frame deficiencies are identified and corrected, as explained above

Table 1 offers links to statistical summaries of each available solution pair (top triangle) and to graphical representations of the orbit differences in radial, along track and cross-track direction (bottom triangle).

The 22000 comparison points for each pair provide the error distribution curves of Figure 1.

The RMS over these 22000 points is given in the bottom triangle of Table 2. The top triangle shows the order of the curves in Figure 1, from 1 (lowest RMS, i.e. top left curve in the Figure) to 78 (highest RMS of orbit differences, bottom right curve in the Figure).

Table 3 summarises the direct results that can be derived from the comparison analysis, as indicated above.

Any pair of orbits that are close to the true CHAMP orbit will inevitably be close to each other as well, which implies that the norms of the difference vectors for the associated comparison become small. In the distribution plot of Figure 1 this means that the orbits involved in the highest curves are most likely to be near the actual CHAMP orbit. However, this is a necessary but not a sufficient condition for having precise CHAMP orbits. Two solutions that are not sufficiently independent may also show small differences, as explained above.