What the use of 16-bit ADC then? My understanding is that DR is a continuous measure of analog signal while ADC 'bits' is literally the width of digital signal. So whether DR is 13.7 or 2.89 in my particular scenario, I expect to see all sorts of 16-bit values coming from ADC.
With DR below 14 there is probably very little advantage of 16-bit data over 14-bit. So it's not a big issue, but still curious.

With purely 14 bit data, a given pixel could have a value of, say, 1000 in one image and 1001 in another. The average between the two is 1000.5 which cannot be represented by an integer. By multiplying the values by 4, the first value becomes 4000 and the second 4004 with a average of 4002 which can be represented by an integer. The advantage of converting the data to 16 bit this way is that a more natural representation of intensity differences can be achieved when compared to pure 14 bit data.

OK, before the discussion went sideways, I googled somebody's raw image captured on ASI2600MC with NINA. It is a 16-bit FITS containing all possible values (mod 4), not only zeros. Now, it is different capture software and different camera, though specs seem identical for MC and MM.

In what way did the discussion go sideways?

By bringing the DR into discussion, which I am proving irrelevant with the example of ASI2600MC. Do not take it personal in any way, what you said is valid and reasonable. But to me it does not appear as a hard reason to not expect smooth 16-bit data out of ASI2600MM and that sample from ASI2600MC reinforces my expectation even further.

Have a look at the diagram. The DR is expressed in numbers of bits. The maximum is 14 which is why a 14 bit ADC is used and the maximum bit depth of the images is 14. The fact that the numbers get scaled to 16 bits is just what the driver does. So the discussion didn’t go sideways at all. And yes, I know what DR means and how it is determined. All I am saying is that this is how the characteristics of the sensor are published by ZWO.

The ADC is specified to be 16-bit, not 14-bit. The ADC type and all the DR and sensitivity graphs are the same for MC and MM versions. The MC is able to produce full depth of 16-bit values, while MM is not able to do that for me. See the contradiction?

DR =log2(fullwell/readnoise) but it is irrelevant when we want to talk about ADC bitness. I suspect that there is either some config or bug in driver. In manual there is mention that 12bit mode will be used when HW bin is used. Maybe check that?

Right, I already added it to the top post. My problem was HighSpeedMode. Interestingly, this parameter exists for ASI294MC too, but does not seem to affect it in the same way.

I just looked around a bit. For the 294 it's claimed that high speed mode switches the DAC between a 14bit and a 10bit version. Looks like the 2600 is similar? But it is really hard to find an explanation what this mode actually does.... ISTR (but cannot check ATM) that also the 1600 has a high-speed mode that ZWO recommended not to use...

Hi,

i think (almost?) every ZWO camera has high-speed built in. By reducing the ADC bit depth it can speed up captures (in some cameras capture-rate doubles). This mode is not useful for deep-sky, but mainly for high-speed video capture. But this will have impact on Dynamic Range. Sometimes I use that if I want to capture ISS transits. There I need high-frame rate, and can afford less dynamic range. In planetary 8bit recordings are often used and higher capture speeds can be achieved. (However these are mostly dependant on exposure time).

Paul

That is what I read on the internet about high speed mode too. Funny enough, I do not see the ADC width reduction with my ASI294MC (that is MC, not MM). I kinda see a slight difference in dark noise patterns with high speed enabled or disabled. But in both cases the data numbers seem quantized at 14 bit. I guess, this is why I had this mode enabled, apparently harmless. Better disable it for good.