Throughput of the APO 3.5-m telescope + DSC Michael Richmond Aug 24, 1995 This note supplants earlier calculations of the overall throughput which appeared in the document, "Results of first run of PHOTO on DSC data from May 25, 1995". Readers will be happy to note that the efficiency of the instrument is slightly higher than originally claimed. I have analyzed images taken with the Drift-Scan Camera in the SDSS r'-band on May 25, 1995, in a region near SA 57. After processing them with the standard PHOTO package -- which does NOT attempt to deal with the known 'spurious charge' problem -- I measured the flux around a set of stars on known brightness on several images. I make the following assumptions: - Bessell R passband is equivalent to SDSS r' (my tabulation of Bessell R yields central wavelength of 6544, equiv width 1521 Angstroms; the documents "The SDSS Photometric System", by Fukugita et al., states that SDSS r' has central wavelength 6250, width 1400 Angstroms) - the night of May 25 was clear, and observations at airmass = 1.05 suffered 0.13 mag of extinction in the r'-band - the gain of the DSC is 2.0 electrons/DN With these assumptions, I compared the fluxes above sky inside apertures of various sizes around stars which have been observed by Connolly et al. (private communication) and by Majewski et al. (PASP 106, 1258 [1994]). I used several very bright stars (V < 10) to look at the curve-of-growth of the light from a star in increasingly large circular apertures. I found that >= 86% of the "total" flux from a star ("total" means as measured inside a radius of 96-arcsecs) falls within a radius of 5.5 arcsecs. I used apertures of radius 5.5 arcsec (20 pixels) for most of my measurements, checking the results with a smaller, 5-pixel radius aperture. The main difference between the present result and the earlier one is that I now correct for the light which falls outside of a 5.5-arcsecond aperture. I measured fluxes from a set of stars in 4 different frames, and compared them with the fluxes predicted from the magnitudes listed in Connolly's catalog as "aperture mag". If we define "overall throughput" to mean "the fraction of photons entering the telescope which register as counts on a DSC image" - using 5 stars with 15.95 < R < 17.0, I find overall throughput <= 15% +/- 3% [where the "less than" occurs because I believe my assumption that only 86% of the light falls inside a 5.5-arcsec aperture is a very conservative estimate] - using 11 stars with 15.95 < R < 18.0, I find overall throughput <= 15% +/- 2% I then used a set of stars from two frames which fall within Majewski's field of standards. - using 5 stars with 17.69 < R < 19.75, I find overall throughput <= 18% +/- 4% The seeing in these frames, when modelled with a sum of two gaussians, had an inner gaussian FWHM of 1.3 arcsec. It may be of some use to note that, under these conditions, I measure directly from the images (that is, I don't just assume a true gaussian profile) - 13% of the total flux of a star (after correcting for extinction) falls within a radius of 5.5 arcseconds - 10% of the total flux of a star (after correcting for extinction) falls within a radius of 1.4 arcseconds Using these values for the telescope + instrument's sensitivity, I calculated the quantity "signal-to-noise" (S/N) using circular apertures around a set of the same stars. Here, I used the standard deviation from the mean pixel count in "empty" areas to determine the "readout + sky noise", and added to it the variance due to counts from each star. Using a circular aperture of radius 1.4 arcseconds, I find - S/N ~ 10 for stars with R ~ 20.0 Nonetheless, when I look at the positions of stars known to be fainter (from a deep image kindly provided by Connolly), I have trouble finding stars fainter than 21.0 with my eye on the DSC images, even though one would expect S/N ~ 5. PHOTO also fails to find faint stars: when its "ffo_threshold" parameter is set to "3" [which ought to mean something like a 3-sigma detection], it finds 50% of the stars at about R = 20.5. Why are the faint stars so hard to find? I have not been able to find a good answer, but my current theory is that the pixels in these DSC images may not exhibit true random, gaussian fluctuations, but instead may feature "clumps". I measured the mean of pixel values inside 10x10 pixels boxes placed in "empty" areas in a number of frames, and compared the distribution of mean values with those expected from a truly gaussian distribution of pixels. The set of measured means had a significantly larger width than expected. I currently interpret this to mean that there are "features" in the images which confuse the eye (and PHOTO) when it attempts to find faint stars. As a further test, I smoothed one image with a gaussian that had the same FWHM as the typical star. In the resulting smoothed image, I was able to pick out several stars with R ~ 21.3, but fainter stars were either invisible or had "peaks" which were not very different from spurious peaks in empty areas of the image. Tim McKay has told me that he has fixed some of the DSC electronics, and that the "spurious charge" problem should not occur in future DSC runs. I have a run at APO upcoming, and plan to use the DSC with only a single amplifier in slow-scan mode. If I manage to collect useful data, I will subject it to some of the same tests I have made here. APO APO APO APO APO Apache Point Observatory 3.5m APO APO APO APO APO This is message 2 in the apo35-dsc archive. You can find APO the archive in /u/strauss/apo/mailer/apo35-dsc on astro.princeton.edu APO To join/leave the list, send mail to apo35-request@astro.princeton.edu APO To post a message, mail it to apo35-dsc@astro.princeton.edu APO APO APO APO APO APO APO APO APO APO APO APO APO APO APO APO APO