Subject: for Scoville search file Date: Sun, 27 Apr 2003 17:41:49 -0700 (PDT) From: Greg Aldering To: ALSpadafora@lbl.gov CC: aldering@panisse.lbl.gov Hi Tony, Could you please start a fold for our ruminations concerning the Scoville search? Here is the first entry. >From aldering@panisse.lbl.gov Sun Apr 27 15:34:42 2003 Date: Sun, 27 Apr 2003 15:34:42 -0700 (PDT) From: Greg Aldering To: saul@lbl.gov Cc: aldering@panisse.lbl.gov Subject: Scoville search: RPS2 results on exposure times & CR penalties Hi Saul, I have run some simulated observations through RPS2 for the CDF-S. It appears possible to almost perfectly split an orbit with 3 CR-SPLITs in one band and 1 CR-SPLIT (i.e. just one exposure) in the other. There is only a modest penalty for doing 3 CR-SPLITs in one filter and 2 CR-SPLITs in the other. For 3 CR-SPLITs, the minimum total exposure is 1030 seconds. This leaves 1020 seconds of exposure for an image in a second filter. If you want 2 CR-SPLITs for the second exposure, then it must be reduced to a total of 900 seconds. As for CR's, using the canonical number that 1.5% to 3% of pixels are hit in a 1000s exposure, I calculate that 0.51% to 1.03% of pixels will be hit in a given search image of 1030/3=343 second. So, in the search, as many as 1800 pixels are hit twice, and 18 pixels are hit three times, in each 4k x 4k ACS field. Assuming that you need to hit at least two pixels to trigger an event, a scanner would have to inspect 9 (or less) CR-induced "objects" in each image. There will also be CR hits over as many as 50,000 pixels in the CR-SPLIT=NO reference image, comprised of roughly 11,000 CR hits, in each ACS field. These could be masked after being detected in a "reverse search." Next, imagine that you want to check for the detection of an I-band candidate at a known location on the V-band image. You probably want to have a clean region of 3x3 pixels in order to make a meaningful check. So, about 27% of the time the region of interest will have been hit by a CR if the V-band search-epoch image is a single 1020 second exposure. If the V-band image is made of two exposures totaling 900 seconds, the region of interest will be hit only 0.7% of the time. So, if you believe in the value of the V-band limits, and you want to avoid throwing away 1/4 of the good SNe, you should take two exposures in V-band at the time of the search. Now, on to the exposure time calculations. - Greg