On 25/02/2008, Jaime Perea <[log in to unmask]> wrote: > Thanks a lot David, > > I think that resampling is correct, but just for curiosity which > header variables do we eventually have to edit?. (fits header > wcs variables are a bit arcane some times) You need to set 1 - CRPIX1 to be the pixel coordinate at some reference channel. 2 - CRVAL1 to be the wavelength at the reference channel 3 - CTYPE1 to be "WAVE-LOG" 4 - CUNIT1 to be "Angstrom" 5 - CDELT1 to be equal to the value of the expression "S0 * ln( S1/S0 )", where "ln()" is the natural logarithm function, S0 is the wavelength at the reference channel (i.e. CRVAL1) and S1 is the wavelength at the next higher channel (i.e. at CRPIX + 1 ). Full details (very full!) of the FITS-WCS standard for spectra are available in FITS-WCS paper III at http://www.atnf.csiro.au/people/mcalabre/WCS/. I notice that the SDSS header includes several different forms of WCS information. If you modify the WCS as described above, it would be safer to remove the other forms of WCS information that could confuse DIPSO. Specifically, I would remove WAT0_001, WAT1_001, and CD1_1. Also, there seem to be a few headers describing a non-existent second axis: e.g. CTYPE2, CDELT2. I would remove these too. David