Help Understanding Astronomy Lab Results

[Stub Mandrel]

Hello, can anyone help me understand these results from AP lab?

[wimvb]

My guess: As it says on the bottom; the dark noise (15.2 adu) can't be lower than the read noise (15.37 adu = 4.3 e / 0.28 e/adu). Dark noise = dark signal noise + read noise.

Redo the calibration frames?

Btw, I don't have the program. I'm just interpreting the screen values.

[Stub Mandrel]

I interpret that as total dark frame noise =  read noise + dark noise.

My camera is cooled to well below 0C so I would expect the dark signal to be very low. Here's a portion of a dark stretched by AP lab, aside from hot pixels I don't think noise is a problem for me.What I'm struggling with is understanding how to get from those readings to  working out my optimum settings. I've plotted various graphs and tried image simulation (except I don't know what to put in the magnitude boxes) and get believable results but they are not consistent. It's recommending iso 1800 or 1600 or 100 and subs from 1s to 400s depending on what I put in or which graph I look at!

[wimvb]

Dslrs are tricky. New cmos cameras have variable gain, as opposed to ccd which use a fixed gain, which pins the full well at the maximum adu (64k for a 16 bit adc). Cmos gain is translated to iso number. But cmos also have an offset. This offset should be chosen such that no pixel values are 0 or smaller. Think of the histogram curve where the left tail shouldn't extend below 0. But in a dslr this offset can't be adjusted by the user. I think that dslr manufacturers use this offset dynamically to minimise noise. And as this clips data values, depending on gain setting and possibly averages, a statistical analysis doesn't make sense. Other tools that can analyse ccd characteristics generally show similar results. (PixInsight has such tools, that don't work properly either with dslr data).

In film photography, you had to expose to lift the dark areas beyond the film sensitivity. In ccd/cmos, you should expose so that you stay in the linear range, ie not saturate anything but the very brightest stars.

I hope this makes sense.

[Stub Mandrel]

Makes sense, Wim, the problem is that APLab give you lots of figures but doesn't explain what they mean or how they are calculated.

I know it has many 'typical' figures built in for different cameras (including mine) which it uses to calculate the graphs in addition to the calculated data I pasted above.

I wonder if the 'read noise' figure is an off-the shelf value for an uncooled Canon rather than one calculated from my darks?

[wimvb]

Read noise is measured as pixel value variation of bias frames.  Dark signal and associated noise from dark frames. Gain and full well can be determined if at least two frames of each are used.