120 dB of dynamic range -- meaning one pixel is ~20 bits hotter than the one next to it -- sounds pretty amazing. I had no idea modern image sensors were capable of anything close to that. Is that the correct interpretation of "dynamic range" in this context, or is there a more specialized definition that leave some room for handwaving?
The last time I messed around with digitizing video, you got 8 bits of each R/G/B component and you were happy about it.
Your interpretation is correct. You need multiple exposures or so called "partial resets" for high dynamic imagers. Having an ADC with more than 8 bit should be accompanied with proper gain and offset compensation, minimum for the columns. If you search for those keywords, you can find articles like https://ieeexplore.ieee.org/document/4621212
The trick is that you get some log like sensitivity curve and not a linear one. This is ok for almost all applications, since you don't need the super fine greyvalue resolution in bright areas.
The last time I messed around with digitizing video, you got 8 bits of each R/G/B component and you were happy about it.