Just thought a heads up might be useful to someone. I have done some tests using a 20D modified by Hutech for astro use. The mod involved replacing the Canon UV/IR blocker with clear glass and anti-reflection coatings. This 20D now records everything from about 400nm to 950nm wavelengths (presumably what Michael refers to as Visible+IR). Visible light is between 400nm and 650nm; the longer wavelengths are Near IR. Michael's article was for a camera that is meant to record this Near IR. His has the IR-pass filetr installed right over the sensor. I use my camera for IR landscapes by putting a Hoya R72 or Wratten 89B somewhere in the optical train. You see nothing in the viewfinder! but the camera is even more sensitive to the IR than the stock version is to visible light. Since the only light reaching the innards of the camera is now IR, the autofocus works excellently and delivers dead-on focus. No need to use the IR mark to correct focus. There is however a very big qualifier. Many lenses are quite lousy in IR. They were not designed for it and the lens elements made from different glasses no longer cooperate to bring together good focus. This shows up more in ultrawide lenses (eg a 14mm) than in teles. But if the lens is capable of excellent images in IR, eg the Canon 100/2.8macro USM, focus is nailed dead-on. OK I did not analyse this for the nit-pickers but you can judge the results yourself on the webpage below. I have much more info for astrophotographers on the website, but the autofocus data may be pertinent to those on Luminous Landscape. My 20D can of course also be used for normal, visible light photography, with autofocus and autoexposure, but no AWB. Details here:http://www.pbase.com/samirkharusi/hutech_mods
Closing down the aperture to f8 seems to take care of image quality in IR in all my shooting so far (I have not used all my lenses in IR yet...).