Post
by **Hibou** » Tue Jul 24, 2018 4:07 am

I have been trying to follow the calculation of dark current using the SharpCap method. As I understand it, you average the mean count per pixel (ADU) for four different exposure times, plot them and extrapolate to obtain the ADU for zero time (-> read noise) with the slope giving the ADU/second increase with time (-> dark noise). After dividing by a scale factor of 16 (?), you convert that to 0.38 e-/pixel/s at 20C by multiplying by e-/ADU, which depends on gain. Naive questions already:

1) Where do you find that number e-/ADU ? Is it 1.0 if I don't use gain ?

2) Where does the scale factor of 16 come from ?

3) Why is the mean count, extrapolated to zero exposure time, so high at 2500 ?

The AAVSO method you show on Excel sheet 2, is simpler, since you just subtract the mean ADU/pixel for short exposures (read noise) from the mean ADU/pixel for long exposures, and divide by the time (not done). The scale factor of 16 is again used. An even simpler subtraction method seems sufficient for a rough comparison of CCDs to CMOS.

The FLIR IMX249 CMOS gives 70 counts/pixel in 4s (max. exposure) and 64 for the minimum exposure with zero gain. (Yes, I switched time stamping of frames off). Then the dark current is of the order 1.5 e-/pixel/s i.e. (70-64)/4s at 45C.

In contrast my Lodestar ICX825 CCD gives 0.51 counts/pixel in 10s and effectively 0 for the minimum exposure, so dark current is of the order of 0.05 e-/pixel/s at ~25C. These numbers seem credible, given the known higher dark current for CMOS and the 45C operating temperature.

Looking at the noise, it has a large standard deviation, with isolated hot pixels for the longer exposures, and even for short CMOS exposures. The CMOS camera also has obvious amplifier glow on one side. Even if I do a double imageJ "despeckle" I can only get the apparent CMOS dark current down to 1.0 e-/pixel/s, more than an order of magnitude higher than the CCD.

So despecking isolated hot pixels and then binning 2x2 is more important for CMOS than for CCD, except for very short exposures.

But I haven't yet found how to implement FLIR's "Mode_7" capture in FlyCap, which is supposed to reduce noise and allow software binning.