Jeff and Mike have mentioned most of the various trade-offs with regard to flat-field calibration images. The usual procedure at most observatories is to take twilight flats with the telescope either not tracking or commanded to dither some small amount between each exposure, the former being lots easier. As Jeff noted, often the problem is getting enough flats for multiple filters in the time available. Sometimes one ends up doing a few filters each session, and getting "enough" over the course of a few days. As an example, with our 72-inch + regular CCD set-up, one starts at 14 minutes after sunset with the B filter, and there is just enough time to get 10 exposures in B, V, and R, working with something like 16-second exposures for the final batch of R images, which have lots of stars. If one gets distracted and doesn't start the first flats until say _16_ minutes after sunset, it's too late, and you won't be able to get them all. There is no such thing as a star-free patch to use for flats, despite efforts some years ago to identify such regions in dark clouds etc, so you just deal with the stars in another way. Dave Schleicher sometimes does comet imaging using eight rather narrow passband filters to isolate various emission species, and this includes filters down toward the atmospheric cutoff at 3200A or so. So his first flats are 10 minutes _before_ sunset, and he gets only three or so images for each filter. The problem with twilight flats is that the spectrum of the twilight sky does not resemble the (dark) night sky; the former is a very blue continuum like that of the Sun (unsurprisingly), whereas the night sky is rather red and has emission lines, something like the spectrum of a planetary nebula (the 'nebula' being the glowing ions in Earth's atmosphere). The characteristics of the flats change as a function of color, and the emission lines can produce interference fringes in thinned chips (which you folks probably aren't using). All of this is minimized by the filters, but not completely eliminated. It is worth noting that there is a gradient in the twilight sky, such that it is best to point the telescope at a particular place to minimize this. The location, often referred to as the 'Chromey spot', is 15-20 deg off the zenith in the azimuth direction _away_ from the location of the Sun. Thus this time of year the 'sunset' spot is southeast of the zenith. This is something that's amenable to being programmed given a Sun ephemeris and local date/time, so could be a telescope-control command in software (and is in the Lowell telescope- control systems). Jeff mentioned a specific well-depth level for the flats, but the main thing here is simply getting enough photons in each pixel in the combined median-averaged flat rather than the well-depth per se. With ordinary telescopes, getting enough images with enough signal (per pixel) on blank night sky is difficult. This usually comes about for two reasons: the relatively long f/ratios of telescopes means the sky gives a relatively weak signal on ordinary exposures, and it is common to take a lot of pictures of relatively few fields, so the star pattern remains fixed on most of the exposures. By contrast, with the LONEOS Schmidt we're running at f/2 unfiltered, and we commonly cover two dozen separate fields in a block of exposures. So even on our standard 45-second exposures the sky is very bright, and the flat is generated directly from the median of 20 or so exposures of different fields. The "thousands" of exposures that Jeff mentioned are not needed; more than a couple dozen in our case is total overkill. This is lucky for us: the f/2 optics are so fast that we cannot take twilight flats with exposures long enough to avoid shutter-illumination problems, nor short enough to avoid getting zillons of stars, and the night-sky emission produces fringing on some parts of the two CCD chips in the camera, which can be taken out only with night-sky flats. We also need to make different flats when there is significant Moonlight, since the sky goes back to being like daytime (sky-blue with no emission lines). This business about the change in the character of the night-sky spectrum is "interesting" when considered for an urban location. The sky is relatively stable thanks to the saturation of street lights, business lighting, cars, etc. On the other hand you still have the emission line spectrum from still-prevalent mercury, sodium, and metal-halide lights. Then again, even the Full Moon hardly changes the sky brightness...just one more street light. There are some excellent articles on the art of flat-fields in various issues of the late, lamented "CCD Astronomy" magazine, including an article by Fred Chromey about the brightness gradients in the twilight, and about building various light boxes, dome flats, etc. \Brian -- See message header for info on list archives or unsubscribing, and please send personal replies to the author, not the list.