CCD pixel-to-pixel variations

[ATWOOD at mps.ohio-state.edu]

Posted to CCD-world:
-+-+-+-
										
You must separately correct for the additive and multiplicative errors.  Subtract
off the BIAS frame, which may or may not have structure, then you must divide by the
true response for each pixel to the wavelength that will be used for an observation.
Great care must be taken to be sure that this FLAT FIELD correction is done
from data that is taken with the same wavelength that is used for your observations.
For example, the night sky is very hot compared to filament lamps and it is easy
to see the difference between "sky flats" and "lamp flats".  (Obviously you must subtract
a bias frame from your flat fields.)

The general question of the stability of the QE(x,y,wavelength) is complicated and,
I think, not well understood.  We have a large data set from the YALO telescope,
a years worth of every night imaging, that should be analyzed to see if the night
by night calibration strategy is best or if the system (not just the CCD) is  stable
enough that flats and darks should be averaged over a longer time scale.  Even so, results 
one detector should not be generalized to any CCD.  The CCD on the YALO (a LLICK3 
2k square) shows very little, or no, QE hysteresis some other devices have very large changes
in QE with exposure history.  The problem of "fringing" (high spatial frequency sensitivity 
variations when a CCD is illuminated with narrowband red light...what is narrow and red varies
with the device) is more complicated and requires even greater stability of the wavelength vs
x and y.

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