CCD pixel profile
-
- Posts: 1944
- Joined: September 23 2004
- Affiliation: University of Sussex
- Contact:
CCD pixel profile
Presumably you need to know the pixel profile for accurate astrometry from CCD images. Does anyone know what the sensitivity profile for a typical CCD pixel is? i.e. is it fairly uniform so that you are just measuring an integral over a square pixel, or is it something more complicated? (e.g. more circular, concentrated near the centre).
Thanks.
Thanks.
-
- Posts: 183
- Joined: September 24 2004
- Affiliation: Brookhaven National Laboratory
- Contact:
CCD pixel profile
Please, do correct me if I speak bollocks, but in general the pixel size is chosen to match typical seeing conditions on ground (ie. 1'' or so) and difraction limit in space, so it shouldn't matter what exactly the pixel profile is. I would suspect you would be limited by optics accuracy much before hitting ununiform sensitivity of a pixel anyway.
-
- Posts: 1944
- Joined: September 23 2004
- Affiliation: University of Sussex
- Contact:
Re: CCD pixel profile
Certainly the seeing is likely to be larger than a pixel, so I'd agree it doesn't make much difference for big sources.
But if you are trying to accurately measure the seeing, and your stars are only a few pixels across, the pixel profile may not be totally negligible. Programs like SExtractor give barycentre and fwhm values of sources to tiny fractions of a pixel, which at the sub-pixel level I'd have thought might depend at some point on the pixel profile? (if you fit say a Gaussian profile to the seeing, you can pinpoint the centre - i.e. best guess at the actual position of the star - to within around 0.05 of a pixel in many cases).
But if you are trying to accurately measure the seeing, and your stars are only a few pixels across, the pixel profile may not be totally negligible. Programs like SExtractor give barycentre and fwhm values of sources to tiny fractions of a pixel, which at the sub-pixel level I'd have thought might depend at some point on the pixel profile? (if you fit say a Gaussian profile to the seeing, you can pinpoint the centre - i.e. best guess at the actual position of the star - to within around 0.05 of a pixel in many cases).
-
- Posts: 183
- Joined: September 24 2004
- Affiliation: Brookhaven National Laboratory
- Contact:
CCD pixel profile
Obviously none of us is an observer... :) ; I agree that PSF might be varyiing across the CCD at the sub-pixel level but this is probably a combined effect of optics, CCD aligment and pixel profile and I am not sure which is the main effect and also I would be surprised if one could distinguish between the various sources of this systematics...
-
- Posts: 7
- Joined: February 24 2005
- Affiliation: Princeton
- Contact:
CCD pixel profile
I don't build optical arrays, buut... The PSF of a given pixel is the convolution of the telescope beam (Airy disk) with the pixel area (weighted by the response over the pixel). The response of a given pixel is then the convolution of the pixel PSF with a point source. If it's a true point source, and seeing isn't swamping things, then you can in principle measure the PSF of each pixel by rastering over a source, and therefore get some information about any sensitivity variation within a pixel by deconvolving the (calculable) telescope beam. My guess is that there is very little variation from pixel to pixel (and perhaps swamped by seeing anyhow), so that image centroids are effectively unbiased. Other non-idealities in the CCD array likely contribute much more to the error in the measurement (variations in sensitivity, striping, etc.).
cheers,
Bill
cheers,
Bill