Berto asked: >what type of detectors do you use for measuring in >the J and K system? Silicon becomes transparent around 1100nm and so cannot detect near-IR photons. Germanium has been proposed as a CCD material since its cutoff is around 1600nm and therefore could be used for J & H measures. However, most IR arrays are hybrids, where an infrared detecting material is sandwiched on top of a silicon readout (which looks remarkably like a CCD itself). Photons are absorbed in the detecting material and conducted down to storage sites in the silicon, and then clocked out. The two popular IR detecting materials are Mercury-Cadmium-Telurium (HgCdTe), normally doped so that it is sensitive from 1000-2500nm (JHK), or Indium-Antimonide (InSb), normally doped so that it is sensitive from 1000-5500nm (JHKLM). Platinum Silicide has also been used, but has low quantum efficiency. HgCdTe can be cooled with liquid Nitrogen just like a CCD. HST used the Rockwell 256x256 HgCdTe NICMOS array (what we have in our current IR camera and what made the sn1999clj image on our anon ftp site). Rockwell has also made 1024x1024 arrays and is in the process of making 2048x2048 arrays. InSb requires liquid Helium cooling (the array works best around 30K) since it can detect thermal photons. The largest InSb arrays are the SBRC-NOAO-USNO Aladdin 1024x1024 ones as used in our new camera that will be commissioned in August, with 2048x2048 devices in the research stage. Don't rush out and expect to buy any of these 1k or larger devices unless you have a *really* big pocketbook -- the Aladdin InSb costs US$250K and requires another US$750K or so to build a camera around it! There are, of course, single channel devices made with these materials, but virtually all professional near-IR photometry today is performed with arrays. Once you get beyond 5 microns, you need specialized hybrids, but arrays have been made to 30microns or so. Then you get into bolometer arrays for longer wavelengths, a really specialized regime. Precision IR photometry is still in its infancy (which is why I'm having so much fun with it). Not only are the arrays much poorer in quality than CCDs (nonlinear, bad pixels, high dark current and readout noise, nonuniform QE response, etc.), but the atmospheric windows are not completely transparent. Your transformation coefficients have to be determined nightly, and your extinction coefficients have to be determined hourly. We routinely get 3 percent photometry per observation, and can beat the errors down to 1 percent or so with multiple nights of data. Starting at K (2.2microns), we can detect thermal photons from the night sky. By 3.5 microns (L), the sky is bright night and day (a million photons/sec/pix are typical numbers). So IR professionals are having to work through many of the same problems that amateur CCD systems are plagued with. Arne