03cfcef·Author: Arthur Eigenbrot·Closed by: Erik Johansson·2022-03-25
Description
This is based on feedback from initial ViSP science data. There were two identified failure modes:
Modstate/beam shifts were way wrong from beam 2. This was caused by the rotation angle of each beam, while computed correctly, being removed in the wrong direction. Thus the beams became even more rotated away from each other which caused the matching to get bad results. Easy fix: add a negative sign!
In one set of data (pid_1_78 at 397 nm) the angle was computed completely wrong. So wrong that it actually caused a crash in the SolarGain task. The cause of this was a mis-identification of the slit hairlines that are used to compute the angle. This mis-identification, in turn, was the result of a too-low contrast between the low-signal hairlines and the lowest-signal regions of the solar spectra. Investigation showed that applying the lamp correction caused/exacerbated this low contrast and so the lamp correction was removed from the Geometric task. Honestly, this was something we had been expecting for a while; good flat field is not really necessary for the Geometric task and, as we have seen, using the lamp correction can introduce undesired effects. As a further mitigation effort a function has been included that makes sure the angle is computed on the brightest modstate (previously it was just the first modstate) because maybe brighter modstates have the highest contrast between hairline and spectra.
A consequence of the solution to point 2 (removal of lamp correction from the Geometric task) is that now the Geometric and Lamp tasks can run in parallel. Cool!
This is based on feedback from initial ViSP science data. There were two identified failure modes:
Modstate/beam shifts were way wrong from beam 2. This was caused by the rotation angle of each beam, while computed correctly, being removed in the wrong direction. Thus the beams became even more rotated away from each other which caused the matching to get bad results. Easy fix: add a negative sign!
In one set of data (pid_1_78 at 397 nm) the angle was computed completely wrong. So wrong that it actually caused a crash in the SolarGain task. The cause of this was a mis-identification of the slit hairlines that are used to compute the angle. This mis-identification, in turn, was the result of a too-low contrast between the low-signal hairlines and the lowest-signal regions of the solar spectra. Investigation showed that applying the lamp correction caused/exacerbated this low contrast and so the lamp correction was removed from the Geometric task. Honestly, this was something we had been expecting for a while; good flat field is not really necessary for the Geometric task and, as we have seen, using the lamp correction can introduce undesired effects. As a further mitigation effort a function has been included that makes sure the angle is computed on the brightest modstate (previously it was just the first modstate) because maybe brighter modstates have the highest contrast between hairline and spectra.
A consequence of the solution to point 2 (removal of lamp correction from the Geometric task) is that now the Geometric and Lamp tasks can run in parallel. Cool!