Actually, we are still trying to help you understand what dynamic range is.
Dynamic Range is well defined, PTP and DxO use very specific definitions with images normalised to a certain print size. Their definitions make the measured DR values usable only for comparison within respective websites, that was my point.
Also they don't actually print the images - PTP operates with a circle of confusion input parameter that corresponds (as far as I understand) a certain print size viewed at a certain distance.
In practice, as above, paper prints can't accommodate more than 7 stops of DR, and before printing, there will be digital downsamling which will involve additional antialiasing processing (and hence some additional information loss). This obviously isn't taken into account by PTP and DxO, as far as I can see. So the resulting DR numbers are 'ideal' and not even applicable to physical prints of the given sizes.
I'm sure you would be surprised to learn that it's actually15 stops of DR you are talking about.
To be precise, log2((2¹⁴-1.5)/0.5) stops of DR.
Where does it come from? In the raw files, the signal is linear and discrete, you can't get more than 2^14. If your formula implies the noise level less than 1 bit, I doubt it.
Is it even a thing - downsampling in 14 bits?
Why wouldn't it be a thing? It happens say in camera when you set it to raw but not to the highest resolution. Don't be confused with CPUs who process data in blocks of 8, 16, 32, 64 bits.
The question is: if the DR at 45Mpix or 60Mpix resolution is important to you, how are you going to get it from a 30Mpix sensor?
Not sure if I understood your question. I feel it implies the DR value is somehow derived from resolution but that's not the case, obviously.
If you convert it on a scanner with the resolution enough to resolve single silver grains (which are submicronic in size, so you will need a 100k+ DPI scanner), you will get a binary image. You can convert it into 8, 14, 16, 24, 32, 48, 64, or whatever number of bits per pixel you want, but it will be binary.
The silver crystals, first of all, are all different in shapes, sizes, and they overlap a lot. In fact there's several layers of them within the emulsion layer. They're not laid out nicely in a regular 2D grid. Your scanner will need to keep the information about the shape and 3D position of each crystal, and it's much more than 1 bit, there's no such a bizarre scanner that can capture all that information.
You can scan to a normal 2D matrix with very high resolution, but each point will have more than two levels of brightness due to the crystals overlapping.
Such digital 1-bit-per-pixel sensors are already proposed.
They're not 1 bit per resulting image pixel. They don't even call the receptors 'pixels', they call them 'jots'.
https://www.gigajot.tech/
If this idea works, it'll probably be a very high DR sensor, but the resulting image will have the normal N-bit pixels:
Photon bits fields are binned over temporal and spatial dimensions during the image processing to create image pixels.
Anyway I'm not sure what exactly you're trying to prove with these 1-bit-jots and 1-bit-crystals. That the 14-bit raw image can have more than 14 stops of DR? No it can't.
You can convert it to 16 bit, then carefully downsample, reduce noise and thus get some additional DR, but that won't be the original image, that'll be a
different, smaller image. Halide crystals and jots don't prove anything here, they're simply irrelevant.
