Suppose that SNAP can observe a 14th magnitude star without saturating. Does that mean that we can use SNAP itself to transfer the calibration from a 14th magnitude star to a 25th magnitude supernova?
Let's look at the process. Consider two stars: A, at V=14, and B, at V=25. Star B is the "target": a supernova at z=1, near maximum light. Star A is the "calibrator": a bright star, perhaps a white dwarf, with known fluxes and/or colors across the visible and IR.
0.4*(25-11) ratio of intensity = 10 = 25,000
In the collaboration meeting in Feb, 2002, Bebek discussed observational strategies for both visible and IR detectors. He suggested combining exposures of length 200 seconds in sets of 4 to reach the faint limit. According to my notes, the visible detectors will have
QE = 60% dark current 0.02 e-/sec/pix readnoise 4 e- 0.1 arcsec per pixelI've made a few reasonable assumptions and calculated the signal from star B which we would receive in a combined 800-second exposure: roughly 1000 electrons within an aperture of radius 4 pixels. The signal-to-noise ratio for this exposure is roughly 25, meaning that the random uncertainty in the brightness of star B will be around 0.04 mag.
Can we simply change the exposure time to 1 second, look at bright star A, and compare the two signals to derive the magnitude of star B? Let's see: in a 1-second exposure with FWHM = 2 pixels, star A will produce roughly 32,000 electrons inside the aperture. Roughly 40 percent of this, or 13,000 electrons, will appear in the central pixel. If the full well capacity is 60,000 electrons (as quoted by Mike Lampton, http://costard.lbl.gov/Snap/get/Calibration/113/1.html) then this sounds reasonable.
If not, then we introduce a systematic error to all observations of faint SNe.
But the 200-second exposures are likely to accumulate an average of 4 electrons in each pixel. This is fraction of the signal from the faint target star B (at least in the central pixels of the PSF). Measuring the dark current accurately enough to subtract it is probably not hard. That's good, too.
In the infrared, however, this may be more difficult.
But the 200-second exposures are likely to accumulate many cosmic rays. One of the reasons to take multiple 200-second exposures instead of a single longer exposure is to reject cosmic rays.
If the cosmic-ray rejection is not done properly, then measurements of faint objects in long-exposure frames will be biased in some manner. This is a possible source of systematic error between short and long exposures.
If we have the luxury of orienting the spacecraft so that the single faint star B falls on exactly the same pixels, we avoid flatfielding errors entirely.
However, this isn't going to happen. The SNAP spacecraft will not be able to orient itself specially for each SN. Instead, SNe will fall randomly on chips across the entire focal plane. Unless the large-scale flatfielding across the entire focal plane is accurate to 1 percent, we risk adding systematic errors to the bright/faint comparison. Note that if the location of star A on the focal plane happens to be a maximum or minimum of sensitivity, then ALL SN measurements will be in error, systematically; averaging together SNe which fall in different locations will not help.