Offset
The offset is a value that can be set in CMOS cameras. Occasionally, larger noise components occur in a pixel, and it can happen that with a very weak signal, the value for the pixel falls below the absolute sensitivity threshold due to this noise component. Thus, no signal is generated by the ADC, although a signal is present. Pixels that have no value are completely black and there is no information. To avoid this condition, an electron offset (a certain number of electrons) is put into the memory, so that each pixel already gets a start value.
(Source: https://www.youtube.com/watch?v=jRoVNht3csk)
The numerical value of the offset is not the direct numerical value of the ADU value (signal value), but a factor that corresponds to a certain ADU value at a certain amplification (Gain).
A formula used in forums to calculate the offset depending on the gain is as follows:
Source: https://forum.astronomie.de/threads/meine-asi-1600-ist-da-hilfe.273271/
With a Gain of 120, an offset of 20 should be selected. However, this is only a reference value. The value can also be set somewhat higher.
For the selection of the offset value, there are different procedures. On the one hand, the difference between a dark-frame and a bias-frame can be determined, taking care that there is no value of 0 ADU. (More detailed information about these image types is discussed in the section (Bias, Darks, Flats and Darkflats).)
A faster method is to create a bias-frame and then evaluate the histogram (tone value curve). (Source: https://www.youtube.com/watch?v=jRoVNht3csk) A bias-frame is the shortest possible exposure time of the camera and is usually in the µs range. The offset value and the readout noise can then be seen in this capturing. Therefore, the bias-frame is often called offset-frame.
A histogram is the image statistics of a capturing, where the ADU values are plotted over the number of pixels, showing how often which gray tone values occur in the image.
The idea behind this approach is the following assumption: If the offset ensures that all pixels in the bias frame have an ADU value, all individual images will also have values. These will have more noise and signal values due to the longer exposure time and will be detectable in any case.
(Images are from a custom bias frame and evaluated with the program PixInsight.)
This procedure can be used to quickly determine an offset for your own camera settings. As soon as the tone value curve is no longer cut off from the left edge, the offset is acceptable.