Begin forwarded message:
From: Knut Olsen <email@example.com>
Subject: Re: [LSST-MilkyWay-etc] A few more astrometric thoughts
Date: September 22, 2017 at 9:57:06 AM PDT
To: Zeljko Ivezic <firstname.lastname@example.org>
Cc: John Gizis <email@example.com>, "John K. Parejko" <firstname.lastname@example.org>, stubbs <email@example.com>, Ian Sullivan <firstname.lastname@example.org>, email@example.com
I understood this to be a problem from reactions to one of the metrics that came out of the first cadence workshop in Phoenix. The fast transient group wanted a metric that used the difference in flux between snaps, and the feedback we got (from K-T, I believe) was that the snaps would not be differenced for this purpose. But I see the data product indeed defined in the DPDD as you note. However, if you read the intro to section 1.2, it says:
"A raw image (baselined as a pair of successive 15-second exposures, called snaps), delivered by the LSST camera, is processed by the Instrument Signature Removal (ISR) pipeline, to produce a single-visit image with, at least conceptually, counts proportional to photon flux entering the telescope pupil (in reality, there are many additional optical, pixel and bandpass effects, including random counting noise and various subtle systematic errors, that are treated during subsequent processing). This single-visit image processed by the ISR is called a “Processed Visit Image” and its main data structures include counts, their variance and various masks, all defined on per pixel basis. After the ISR step is completed, the pixel values and their variance are not modified any more. These single-visit images are used downstream to produce coadded and difference images. The rest of the processing is essentially a modelbased interpretation of imaging observations that includes numerous astrophysical and other assumptions.”
The paragraph sure makes it appear that the snaps are used to create single-visit images, and then these are used in the DIA pipeline. But good to know that this impression is incorrect!
On Sep 22, 2017, at 9:36 AM, Zeljko Ivezic <firstname.lastname@example.org> wrote:
On Sep 22, 2017, at 9:28 AM, Knut Olsen <email@example.com> wrote:
I’m very likely not up to date, but I thought that there was no intention to use the exposure pairs to look for fast transients, as this would impose a large new burden on DM. Has this changed?
DPDD (ls.st/dpdd) says on page 19 that DIASource table will list diffFlux (and its error),
"Calibrated flux for point source model centered on radec but measured on the
difference of snaps comprising this visit”
(see also footnote 46).
Do you know what is the origin of the claim that "there was no intention to use the
exposure pairs to look for fast transients”? I don’t recall hearing it or reading it. If there are
any docs claiming something like that, they should be fixed.