How do you deal with camera Offset?

Hi Christian,

For the output of the camera, the offset is essentially a constant which adds to the whole signal (including thermal noise and read noise). How the value your are setting in application/driver is treated is camera specific. For the Altair AA183M Pro that I had, which didn't scale it's 12-bit signal to 16-bit as others do, the offset value that I set was identical to the DN (digital numbers): if I set it to 16, all values were increased by 16 and a dark frame had a median value of 16.
In case of my ASI294MM Pro, the offset value is given in per mille. If you set the offset to 10, you have 1% of maximum signal, which is (due to the scaling to 16-bit) 65535, and yields a median value of a dark frame of approximately 655 (=65535*0,01). 

The important aspect is that you may have values below the offset value. If you set your offset to 0, these negative values can lead to black clipping because all negatives become zero. The even worse case is if the "negative" value is treated as a positive number and -1 will become 65534. 

For an image calibration its therefore absolutely essential to avoid black clipping or negative values. Hence, one needs to analyze the behavior of the camera and observe how the dark frames behave (signal from the sky will always be additive and therefore lifting any signal value). Typically you will see a gaussian distribution around the offset value if you take a look at the histogram. The spread (e.g., standard deviation) of this distribution depends on gain, sensor temperature and exposure time.


The initial settings of the drivers are usually pretty conservative and can be lowered IMHO. Even a value of 80 (8% here), doesn't really diminish your dynamic range by orders of magnitude). If you need some help with some analysis, feel free to get in touch. I've been doing this lately frequently.

EDIT: my recommendation is to find one offset setting that you always use, determined under worst conditions. An offset a bit too high won't hurt. Overall, if you're hunting faint signal, correct image calibration is of the essence.

Cheers,
Björn

EDIT: PS: Here's a thread where similar points came up. Maybe that's also of your interest.
PPS: I've tried to calculate the offset based in the imaging settings you provided in your last thread (600s, gain 300, T=-25C). For that, the offset value would be 4. Would be interesting to verify this with actual dark image data.

H Christian, I am sure someone will soon answer, if they haven't already with a detailed scientific explanation of gain and offset that may or may not help you with your question.  From a practical stand point I can add some basic insight that might make sense to you.  I have used cameras similar to yours as well as many other ZWO caneras and started image capture with SGP, SharpCap and later SharpCap pro.  The sky analysis and Smart Histogram features in SC taught me a lot about CMOS camera gain and offset from a user perspective.  I imaged a year with using Smart Histogram and SharpCap never recommended an offset greater than 10 for any target I tried, while the default camera offsets from the manufacturer were in the 50 to 70 range.  SharpCap I believe also called Offset "Brightness" so that might tell you a little bit about how offset works.  I am an asiAIR Pro user and another interesting aspect regarding offset is that ZWO offers no control at all for offset or brightness in their asiAIR Pro operating system, at least for their OSC cameras.  It is possible ZWO mono camera users have a different experience.  Hope this helps.

scott

used the mono together with AAP for a while and indeed also no offset control.

1.  Your offset must be greater than your noise level

2. The same offset must be used for your reference frames, darks, flats, etc.

Pick a number that is much larger that the noise in your longest dark and stick to it.

CS