[Aavso-photometry] Linarity of SBIG ST-8 camera

[arne]

Lionel Catalan wrote:
> I've been doing tests to determine the linear response range of my SBIG ST-8
> camera. The tests involved exposing the camera (without telescope) to a
> relatively uniform light source for varying times. I used a device that I
> had previously built to take flat frames. I found that the deviation from
> linearity is less than 1% if the ADU values are kept below 50,000. If the
> ADU exceeds 50,000, the deviation from linearity increases very quicky and
> reaches 12% at 60,000 ADU. Since this is a NABG amera, I had naively
> expected that the linear range would extend up to 65,535 ADU (the full ADU
> range of the camera), but this is not the case. Then, after reading the CCD
> detector specifications, I found that the full well capacity is about
> 100,000 electrons. Therefore, with a measured gain of 2.17 e/ADU, the full
> well capacity should be reached at 100,000/2.17 = 46,000 ADU. So this seems
> to be consistent with my measured linearity range (I hope this makes sense).
> 
> My problem is that I've been using this cameras for several months now,
> collecting thousands of images, and in some pictures, the maximum ADU values
> of my target or comparison stars sometimes exceed 50,000 (although I always
> kept it below 60,000). It would be a big pain now to find these images and
> to eliminate them from my photometric analyses. Has anyone any idea on the
> possible error of having the maximum ADU in a star profile slightly above
> the camera linearity range? My intuition is that the error may not be very
> large because the large majority of the pixels in the star profiles are well
> within the linear range. Any comments/ideas would be welcome. 
>  
The answer is as always: it depends: how many pixels are saturated,
how the stellar profile is sampled, whether you are binning or not,
etc.  Usually saturation causes the star to be measured fainter than
its true value.  You indicate a 12% drop at 60K ADU, so the simple
answer would be that the effect would be less than 0.12mag at this
peak value and with reasonable sampling, but certainly an error of 0.05mag
would be possible. You *can* go back and recalibrate your images, adjusting
for the nonlinearity near full well, but this is a software issue and
would take a fair amount of time.  However, science is full of these
"sand traps" (one step forward, two steps back), and you often just
have to bite the bullet and go back and reprocess things.  I've had
two reprocessing episodes on different telescopes when I discovered
that there was scattered light in the system, for example.

As you have learned the hard way, everyone should characterize their
CCD camera before they begin any scientific program.
Arne