The above shows the advantages of sheer aperture and focal length. The 20D image was simply downsized to yield 1.55 arc-seconds per pixel on your monitor (if you click "original" below), matching the 1Ds crop. The 20D image is also showing the full height of the 20D frame. So what is the maximum focal length one can comfortably use for the 6.4 micron pixels in a 20D if we wish to stay within that "reasonable" 1 arc-sec per pixel image scale (2 to 3 arc-sec FWHM seeing conditions)? Sounds like 1350mm is a good, comfortable maximum. We can be pedantic and insist on making an allowance for the loss of resolution in deBayering and allowing for an antialiaising filter. In chart testing, non-modded Canon DSLRs resolve around 85% of their theoretical Nyquist in white light. I suspect that the loss of resolution is perhaps more due to the antialiaising filter than to the deBayering process. Frames from modded Canon DSLRs tend to look sharper, but I have not chart tested one yet. The antialiaising filter is removed when a Canon DSLR is modded, and thus making an allowance for that 0.85x factor may in fact be overkill. Anyway, if you still insist on it, then 1600mm is a useful upper limit. Beyond that and you will be unnecessarily over-sampling on most nights. In rare, excellent seeing one could go up to, say, 2500mm. But frankly, if you are after sheer resolution then there's no upper limit until you approach the focal ratios used in planetary imaging. You are just compounding your difficulties in achieving precise tracking without achieving much benefit. Stay below 1500mm focal length and be happy. Astronerds coming in from astroCCD experience and who have not used DSLRs tend to make weighty, disparaging pronouncements against DSLRs. These pronouncements can often be debunked very easily by simple measurements. Such disparaging statements often relate to resolution, quantum efficiency, thermal noise because of the absence of chilling and similar issues of yesteryear. Actual data on contemporary Canon DSLRs tend to make the jaws drop. The current crop of DSLRs are that good, superior in most aspects to any astroCCD that an amateur can lay his hands on. Very high QE compared to any one-shot-colour astroCCD, extremely low read noise, extremely low thermal noise (to the extent that you will be skyfog-limited at most sites and with most OTAs), etc. However the chilled astro CCDs (monochromatic) remain kings if you wish to do narrowband imaging using half-hour+ exposures per frame at high ambient temperatures. Nevertheless this discussion of resolution should not be confused with high magnification lunar/planetary imaging using fast shutter speeds. For that, one has to look at the focal ratio. f30 would be a good lower match for the 20D; a full two pixels across the FWHM (Full Width at Half Max) of the Airy Disc. f40 to f45 will eke out any resolution that may still be there. Calculations explained at
http://samirkharusi.net/sampling_saturn.html
Oh, yes. I did composite the C14 image onto the center of the wider-field Canon lens image. The composite gives a beautiful, large, glossy print showing an exquisitely detailed M13 in the middle of a vast, much more empty, field of stars. Next: let's look at Omega Centauri.