Lily Zhao (Yale) showed me issues late last week with the outputs of the EXPRES spectrograph line centroiding for its calibration (laser frequency comb and ThAr lamp) exposures. So early this morning I hacked up a parabolic fitting code and then a cubic fitting code. The former are analogous to what we used in SDSS imaging for stellar centroids and Vakili and I analyzed a few years ago. The cubic methods might be less stable, but they are able to capture line asymmetries. I passed these on to Zhao for testing against their pipeline, which fits Gaussians. The key idea, from my perspective, is that the central few pixels contain almost all of the information about the line center; as you go further out on the line, you get much more sensitive to details of your line-spread function without adding much to your centroid quality. It's engineering, of the kind I love.
No comments:
Post a Comment