2008.09.12

I have spent far too much time this week trying to understand exactly what is happening in my astrometry model, and I think I have finally converged.

A little background: psastro (IPP astrometry fitting program) can use a reference model which describes the layout of the chips and the optical distortion. I generate this model from observations for which I have an astrometric solution. One part of the solution is to take a sequence of images obtained at different rotator positions, but with the same nominal pointing, and fit for the motion of the claimed boresite coordinate, in the assumption that the reported location is rolling around the true boresite.

In looking at the rotator sequences from 2008.08.27, I got concerned by the observation that the boresite motion was not primarily a rotation but a more linear motion. That meant that the model was invalid, so I adjusted the program to be able to write a model in which I simply assert the coordinate of the boresite (in pixels relative to some chip), allowing me to generate a model based on a single astrometrized image.

I discovered some low-level errors in coordinate transforms between defining the model and applying the model which resulted in an inconsistency: I was unable to generate a model from an image, and then apply the model back to that image and correctly predict the location of the chips. After a lot of digging, I fixed all of the small bugs that caused this problem. I also added the capability to the astrometry analysis program to compare the model-predicted locations of the chip 0.,0 pixels to the measured positions, and use this information to measure the boresite offset, the rotation, and the scaling difference relative to the predictions of the model.

I now am able to generate an astrometry model which can then be applied back to images and correctly predict their chip positions. Using this fixed up code, I get a much more rotational boresite motion for the rotator sequences. Attached are three plots, equivalent to ones I sent out last week, but using the new analysis of the boresite offsets. The + marks the measured offset between the boresite and the model prediction for a given image; the line is a fit to the motion as an elliptical path; the o marks the prediction on the path for each image. There are three plots, one for each of the altitudes 30, 50,

70. For these plots, I have forced the boresite to be located at the

midpoint of chips 33 and 44.

These now look more like the measurement of a flexure in the x-direction plus an offset error in the boresite location: the x-displacement is maximized at POS angle 217, which corresponded in these sequences to ROT = -89.

aloha gene