CCD pixel-to-pixel variations

[Paul.Jorden at eev.com]

Posted to CCD-world:
-+-+-+-
Steve, and ccd-world,
 >
 > Has anyone any knowledge on how pixel-pixel QE varitions
 > changes as a function of
 >
 > 1) Wavelength
 > 2) Time
 >
 > I'm looking for quantitative real results, but theory guesses
 > are welcome !
 >
Don't forget you need to state whether the chip is frontside or backside; I 
assume you are mainly interested in backside chips.

I have measured intra-pixel QE as a function of wavelength before, as I 
guess you know.
I can merely 'guess' about (2)

 > These variations are traditionally removed (or at least attempts
 > are bravely made) by high-quality flat-fielding at the wavelength one is
 > observing at. But has anyone done any quantitaive tests of how dependent 
the
 > RATIO's of pix-pix variations are on wavelength ?
 >
I guess the intra-pixel variation can give you an upper limit on how bad it 
can be, especially if the projected slit width is quite small.
Frontside chips, could be quite dependent on wavelength, since interference 
effects in the electrodes are very strong.  For backside cillumination, 
these effects come into play around the same wavelengths as fringing 
(>800nm).

However at the very shortest wavelengths (<400nm) you will be very sensitive 
to surface effects.  So pixel modulation will increase at these short 
wavelengths, and decrease as you move longer.  Hence PRNU is worst at short 
wavelength.

So I would guess, on thin chips results could be quite good for wavelengths 
of 400-800nm especially.  I think you ought to be able to take 'superflats' 
at a few wavelengths, and interpolate simply between them.
 >
 > Also, the pix-pix variations would need to be constant a
 > function of time. Has anyone any experience of this also ?
 >
I would expect them to be stable.  Dominant effects are the intrinsic 
spatial structure of the device.  I could only imagine slow changes with 
time on a low spatial frequency, if you had contamination build up on the 
chip for example.
Basically most things are stable with time (temperature, chip operating 
conditions, hopefully illumination conditions,...)
You might have problems if the cryostat or spectrograph setup changed (eg 
dismounting/remounting cryostat.
I suspect the dominant factor will be wavelength stability.

 > We are intending to do these experiments at ING, but any
 > advice or results that has already been done would be
 > very much appreciated. We will try this out with a 1k Tek
 > and an EEV42-80.
 >
I'll be very interested in the results.

Cheers,
Paul
_______________________________________________________________
Dr Paul Jorden,  CCD Applications, EEV
Waterhouse Lane, Chelmsford, Essex, CM1 2QU, UK
Tel:      44 (0) 1245 -453458 (direct), -493493 (switchboard)
Fax:      44 (0) 1245 453224
see also:-     http://www.ccd.eev.com/
 

-+-+-+-
For information about CCD-world, send email to owner-CCD-world at cfht.hawaii.edu.