I suspect you are right, and the issue here is, that it will enable shadow lifting in post processing with less penalty noise wise, but it will not enable dialing in more exposure, like a new sensor with more highlight headroom would. So you will not actually be able to expose dark parts of a scene longer to get a better starting image, and also to capture more information in dark regions. There is a limit to shadow lifting with the Sony A7rII, and it is not so much noise, but very dark parts tend to lose information and look unappealing. Less noise when lifting the shadows would not do that much good I fear. Longer exposure would.
So I still feel there is room for a next generation BSI sensor that has more highlight capacity, unless the new readout circuitry somehow managed to move the highlight clipping point....
I wonder what the practical use will be of a higher raw DR with the same sensor. The new circuitry will not give the sensor more highlight room I assume. So the additional DR should in practical use mean: more shadow raising capability? That is not the same as a new sensor with larger well depth and more highlight headroom, so I don't have to underexpose the midtones and shadows as much as I now have to do with the A7RII in high contrast scenes. So there the Nikon D850 should still be superior? If the case, then I will wait for the next Sony BSI sensor.
I imagine the gain is from the readout circuitry being improved resulting in lower read noise. This would mean less noise in shadows and hence higher DR.
How they improved the read out I don't know as they haven't really said how.
Dynamic range is the difference between the point at which highlights clip and the read noise floor.
Moving either of them produces additional DR (that is assuming you raise the clipping point or lower the read noise), so in that sense, it doesn't matter which you move, just that the distance between them is increased as a result.
Regarding the concept of dynamic range I assume you mean ratio rather than difference (because difference is wrong).
In terms of electrons (implying photons) it's true that both ends (high and low) determine dynamic range.
Raising the high end requires a higher Full Well Capacity (FWC); per unit area if you care about normalized measures as I do.
Lowering the low end requires lowering the read noise.
But in practice we are constrained by the Digital Numbers (DNs) that come out of the Analog to Digital Converter (ADC).
In that regard we cannot raise the high end; we're already clipping the ADC at it's capacity. (Even if we raise FWC we then have to lower gain.)
Our only option is to lower read noise
And, because of Quantization Error, there's a limit to how small a read noise we can utilize as a function of the bit-depth of the ADC.
Generally an N-bit ADC can measure about N+0.5 stops of dynamic range.
The ILCE-7RM2 actually only uses 13-bits (reported as 14-bits) and pretty fully utilizes them for an Engineering Dynamic Range (EDR) of about 13.3
For a new sensor to have significantly higher EDR than the ILCE-7RM2 it would have to truly use a 14-bit (or more) ADC.
There's a good chance of that since after using less than 14-bits for quite some time both the ILCA-99M2 and the ILCE-9 use 14-bits.
So, if the ILCE-7RM3 produces true 14-bit output then it
might accomplish an EDR of as high as 14.5 which would be about one stop better than the ILCE-7RM2.
However, in recent years, no one has accomplished this large a jump between models; so I'm expecting less.
Regardless of what happens it's an exciting announcement and I look forward to having some hard numbers some day.
