Lily Zhao (Yale) and I looked at the different centroiding methods, indexed by two integers: One integer is how many pixels to use near the peak of each line for the peak fitting. The other is what order of polynomial fitting (to the logarithm of the flux; quadratic in log flux is Gaussian in flux). She made many diagnostic plots and statistics. Now: How to decide which setting of the two integers is best? We tentatively settled on the mean absolute difference of pairs of frames that are taken within the same day (or night really). This uses an assumed stability of the spectrograph, but I think it's safe, because it's hard to imagine that our centroiding will correct for true spectrograph drifts! More soon; if this works, we can further improve the EXPRES pipelines.
No comments:
Post a Comment