From:   jwf@ipac.caltech.edu
Subject: IRAS HIRES
Date: August 8, 2008 11:28:29 AM PDT
To:   fmasci@ipac.caltech.edu

Frank,

    The attached PDF is for your archive and possible interest in leisure moments. I hope you won't let it interrupt your work, but since you mentioned not being familiar with IRAS, the properties of its data, etc., this paper can fill you in a bit; section 2 is a good concise description with useful figures. This paper is essentially the journal-article version of Yu Cao's Ph. D. thesis; it includes his additional destriping method, which became part of the standard IPAC HIRES product. It has the nice feature that it does no damage when there are no stripes to be taken out. Stripes in IRAS data result from the baseline drift between detectors, which the main data products didn't care about, since the point-source detection templates included baseline in them, so they fit point-source amplitude above baseline along with baseline itself, and only the former were involved in the primary data product (i.e., the IRAS Point Source Catalog). The data preparation program Laundr took out most of the baseline variation, and it generally isn't visible in co-adds, but even a little bit of residual baseline inconsistency between detector tracks shows up in Hi-Res (that is its job after all), and Yu's method helped a lot.

    Tom Prince was Yu Cao's thesis advisor. I don't know if Yu ever published a solo-author thesis summary.

    As you can imagine, when Yu was around IPAC, a lot of confusion resulted from his name. For example Dianne Engler would say "That HIRES that Yu requested is ready", and I would say "What do you mean? I didn't request any HIRES" and many variations on that theme. We took to using his full name, which had its own problems, since he pronounces it "chow" but most people said "cow".

    By the way, there's an error in the paper after equation 4:


The noise was not equal on all footprints, just for those in the same detector track (or "leg"). Typically there are many such tracks that go into an image, so the MCM operation was not equivalent to RL.

Regards,  John