Bandpass arguments: B --> V @ z = 0.25 V --> R @ z = 0.16 V --> V @ z = 0.10 V search is better than R search because decay time is similar to rise time, meaning that fewer declining SNe will be found. Also, BTC50 has no R-band (reference) images. R search is better than V or B serch because moonlight and extinction are less of a problem. Just to get a grip on the scope of this survey, to obtain 15 SNe within 15 days of maximum light roughly 5e14 Lsun of stars needs to be observed (using 1/2 of Pain et al rate). This amounts to roughly 50000 galaxies, about the total number with available redshifts. A single telescope with a minimum exposure + readout + setup time of 1 minute would require 100 nights to observe all these galaxies. This dictates the need for a wide-area survey in which many many galaxies can be observed in one field. A wide-area survey also gives a sample which is relatively unbiased. Here are some telescope and detector combinations which have been considered: Scale m0(R) FOV "/pix m 1e-/s sq deg ELVIS + ESO 2.2-m 0.24 25.40 0.292 (zp estimated from Web page) Mosaic + 0.9-m 0.43 23.01 0.957 (assumed BTC QE) INT 0.33 25.40 0.27 (assumed same as ESO ELVIS) EROS ?.?? ??.?? ?.??? (1 sq deg in 12 min to m(R) ~ 22) BTC + Blanco 4-m 0.43 26.25 0.239 Mosaic + Mayall 4-m 0.27 26.25 0.325 (assumed BTC QE) Curtis Schmidt 2.20 22.00 1.566 Burrell Schmidt' 1.44 22.00 2.695 (this is hypothetical) Epps + duPont 100" 0.73 25.28 0.143 (assumed BTC QE) QUEST 1.0? 2.3xL (2.3 degree high drift scanner) The factors governing the survey efficiency and total time include the field of view (FOV) and zeropoint (m0(R)) given above, as well as the effective seeing (with depends on intrinsic seeing and the scale). The time to reach a given signal to noise in a 2 FWHM diameter aperture is: t = ([fobj + (sky+gal)*pi*(FWHM)**2] / [pi * eff * fobj**2 * R**2]) * SNR**2 An additional factor of 2 in time is needed when subtracting images to find SNe. Also, make sure to account for the fact that: fobj = fobj_total*(1 - exp[-rmax**2/2*sigma**2]) = fobj_total*(1 - exp[-2.77*(rmax/FWHM)**2]) When comparing telescopes one can use the following expression: [fobj + sky*pi*(FWHM1)**2] [eff2 * R2**2] texp1/texp2 = ---------------------------- * ------------- [fobj + sky*pi*(FWHM2)**2] [eff1 * R1**2] There is a trade-off in strategy in that bright SNe are easier to follow, but a huge area of sky must be survey to find enough of them. Here is an approximate breakdown of the survey area needed to find 15 Type Ia SNe within 15 days before maximum light, along with the time needed using the Curtis Schmidt or the BTC on the Blanco 4-m to reach SNR ~ 5 on two equal quality subtractions; the noise from the host has been ignored: <---------- BTC ----------> <----------- CS ----------> Zmax Rlim Area Exp OH Pointings Time Exp OH Pointings Time (sec) (sec) (hrs) (sec) (sec) (hrs) 0.25 22.0 50 77 45 209 7.1 18085 140 32 162 0.16 21.0 200 13 45 837 13.4 2898 140 128 108 0.10 20.0 800 2 45 3347 44 466 140 511 86 0.06 19.0 3200 1 45 13389 167 77 140 2043 123 (OH is overhead for readout and repositioning. For undersampled imagers two readouts are required to discriminate agains CR hits; note I have since found out that readout time on the BTC will continue to be 120 seconds). This shows straight off that only a z ~ 0.25 search makes sense with the BTC, while the most sensible search on the CS would be one to z ~ 0.10. A rolling CS search (the only kind practical given the slow discovery rate) would yield about 1 Type Ia per night. Comparing these "best" searches in terms of supernova per collecting area, the bright CS search is more efficient (by a factor of 3) than the BTC search. This factor, and the larger follow-up effort required, could make this search very hard to justify. Obviously the data of Maza bare directly on this issue. Does this agree with his discovery rate? If the discovery rate is 2x lower, then early on follow-up becomes difficult to schedule due to the small number of active SNe. The followup time needed for a CS search would be almost 6 times less than that for the BTC search. This means all the spectroscopy could be done at CTIO, ESO, and the duPont. All the photometry could be done at the CTIO 1.5-m, and 0.9-m, the Swope 1.0-m, and maybe another telescope at ESO. Here is an example using science-grade CCD's in Mosaic on the KPNO 0.9-m: <---- Mosaic on 0.9-m ----> <----------- CS ----------> Zmax Rmax Area Exp RO Pointings Time Exp RO Pointings Time (sec) (sec) (hrs) (sec) (sec) (hrs) 0.25 22.0 50 2165 120 52 33 18085 140 32 162 0.16 21.0 200 345 120 209 27 2898 140 128 108 0.10 20.0 800 61 120 836 42 466 140 511 86 0.06 19.0 3200 9 120 3344 120 77 140 2043 123 So, this could possibly be twice as fast as the CS. It would generate 16x as much data(!), but would allow a deeper search (z ~ 0.16) than the CS. Of course weather is another issue altogether. Follow-up strategy would shift significantly to Lick 3-m, KPNO 4-m, Lick 1-m, KPNO 0.9-m, WHT, INT, Palomar 60", McDonald (?), with some CTIO and LCO participation a possibility. DLT tapes could be Fed-X'ed to LBL. How do these predictions compare with EROS rate of 1 SNe at V < 22 in 2 hrs with 10 minute integrations with a 1-m telescope equipped with a 1 sq degree camera? This translates to about 0.1 SNe per sq degree (of different types and lightcurve phases). The closest analog is the z ~ 0.16 search covering 200 sq degrees. This implies 20 SNe, of which about 12 should be Type Ia, and about 6 should be on the rise. So, the estimates range from 6 to 15 rising SNe Ia in 2 nights on the BTC, 8 nights on the KPNO 0.9-m, or 18 nights on the CS. I did some testing with CS images and found that if they have to be smeared to get good subtraction, the FWHM goes from < 4" up to about 7.5"! The sensitivity then just plummets! Direct tests of CS images with our current software shows that the subtractions are terrible. Lou Strolger's tests with our revised software showed much better results. -------------------------------------------------------------------------------- After discussions esp. with people at CTIO, we've decided to go with at search to z = 0.1. The calculations below assume a rate of 0.02 SNe Ia / sq degree to z = 0.1, and the zeropoints and FOV's given above. Also, we are deciding on exact exposure scenarios, so these are assumed: Zmax = 0.1; Rlim = 20.0; Time for 15 SNe assuming 0.02 SNe Ia / sq degree: Telescope Detector Area Exp Bin Tread Tmove OH Pointings Time (sec) (sec) (sec) (hrs) KPNO 0.9-m Mosaic 750 2x180 1x1 120 + 0 240 784 131 INT EEV Cam 750 2x 30* 2x2 30 + 0 60 2778 93 EROS 750 2x ? 1x1 ? ? ? 750 ? CTIO 4-m + BTC 750 1x 2** 1x1 120 + 0 120 3138 106 xESO 2.2-m ELVIS 750 2x 15* 2x2 15 +15 60 2568 64 * this is much longer than necessary --> grey time can be used ** this choice could not discriminate against CR's.