It's not theoretical, and I demonstrated it in the other thread in a stepwedge scanline example. What's more, the improved highlights were apparent to me when I started using the 1Ds Mark III, but then I may be more sensitive to the improvement than the next guy. It also matters whether one downsamples for output, or requires full resolution for large format output or when significant crops are inevitable. It's just that the Canons do not allow to reduce the bit depth used in the ADC conversion process, that prevents me to demonstrate it with the same camera. Besides, nobody is paying me to do such a comparison, so I have to prioritize how I spend my time.
Without trying to cause any offense, I have to say I that am having some difficulty in understanding your line of reasoning here. On the one hand, when I demonstrate using the same camera
how much more detail exists in the 13th and 14th stops of DR with camera in 14-bit mode, you write, "....differences between sensors and support electronics can play a large role here".
Since you don't have a camera which provides both a 12-bit mode and a 14-bit option, and since you haven't provided any real-world images, shown at 100%, or 200%, 400% or even 50% enlargement, how can you claim that you have demonstrated in practice
that 14-bit processing improves highlight detail?
What you've provided, in the other thread you refer to, are charts of a scanline from a synthesized stepwedge with Poisson noise added. That seems very theoretical to me.
I recall some time ago there was a discussion about ETTR in which Emil Martinec criticised certain explanations for the benefits of ETTR, which made reference to the fact that half of the available levels, whether from 12 bit or 14 bit processing, were applied to the brightest stop in the dynamic range.
Emil, who surely knows a thing or two, claimed that the additional number of levels made available in the brighter stops, as a result of ETTR, were irrelevant because the sheer number of such levels in highlight areas was far above what the eye is capable of discerning. The real purpose of ETTR, Emil claimed, was to increase Signal-to-Noise so that noise which would have been apparent in an underexposed shot, would no longer be apparent, or would be less
apparent in an ETTR shot.
That seemed reasonable to me. However, Emil went even further with this line of reasoning and claimed that 14-bit processing for images which had less than 14 stops of DR, served no purpose.
I disagreed because I'd seen such differences, and felt they were very real. As I recall, when I posted similar images to those comparisons I've posted at the top of this thread, Emil dismissed both 12-bit and 14-bit images as being so bad that no-one would include such areas in a photograph. He had a point, and I think the discussion ended there.
However, upon reflection, it has occurred to me that such additional detail, however degraded, could prove to be very useful either scientifically or forensically.
Imagine a scenario where I have taken just a single shot of a high-contrast lanscape containing trees and dark undergrowth. As I process the image in Photoshop, raising the deepest shadows experimentally to see if they contain detail sufficiently interesting to include in the final image, I notice what appears to be a very strange, feathered bird in the undergrowth, which I didn't notice at the time I took the shot because I was too far away. So I crop the area, clean up the image and enhance the detail to find out what species of bird this is, just out of curiosity.
Fortunately, I'm using a Nikon camera, so the level of detail should be sufficient to identify the bird. If I'd been using a Canon camera, then forget it! If I'd been using the Nikon in 12-bit mode, the job of identifying the bird would have been more difficult, but perhaps still possible.
As it is, my camera was set to 14-bit processing, so there's sufficient detail for identification purposes. After extensive searching and consultation with Biologists and Naturalists, I find that I have discovered a new species of bird
that is subsequently named after me. I become famous.
Or, if you prefer an alternative scenario, I've rediscovered a species which was long considered to be extinct, the Long-toed, white-faced, yellow-bellied Honeyeater.