Strange midnight turn for GPS block IIA
A strange behavior has been observed in a specific case of midnight turn in yaw for GPS block IIA.
The test case was built in a similar way as the original tests cases, using
GenerateBaseSamples from the reference attitude generator utilities folder. It was however set at a much earlier data than the existing tests and was found on GPS week 1218 (May 2003), for satellite G07 (satCode G37, a Block-IIA satellite), using SP3 file
ngs12183.sp3.Z. On May 14th, between 07:58 and 09:28, a midnight turn occurs and the Sun crosses the orbital plane (β increasing from negative to positive values). This case should already be handled by Orekit since version 9.3, and there are already test cases for it, but only in the noon turn case as no midnight turn with simultaneous Sun crossing was found until now.
The current output is that the spacecraft starts its linear turn at about 0.1272°/s and the Sun changes side. Then, the satellite keeps turning for about 70 minutes, completing one turn and an half instead of only one half of a turn. One turn and an half seems unrealistic and the reference eclips, even the patched version, does not do that.