One of the claims for the new A7R IV is 15 stop 'enhanced' dynamic range, whatever that means. The previous models are also said to have more than 14 bits of DR. Can someone please explain how this is possible with only 14 bits of data in the raw file. E.g. are those 14 bits non linear?
Its just the marketing game.
The truth is DR is very misunderstood and there is no standard for testing it .. so a manufacturer can claim anything they want.
The camera manufactures don't even state a given lens or light/weather condition.
Everyone knows lighting effects DR.
Some photographers claim they get 6 - 8 stops DR in the field and I think that might be about right.
The problem lies in the usability of this dynamic range; if too few bits are spread over too great of a tonal range, then this can lead to image posterization. On the other hand, displayable dynamic range depends on the gamma correction or tonal curve implied by the image file, or used by the video card and display device ..
Regardless .. one needs to control their lighting and post process for optimal DR and not rely on the camera and its claim what ever it may be.
The rating is based on a downscaled image. This is how all companies that rate DR do it, as it tests output DR, not pixel-level DR. So the higher the res, the less the noise is visible once downscaled, and the higher the DR score is.
The biggest thing I wish companies would add to the test is highlight recovery, as that directly affects the amount of shadow recovery that is available. I remember when I owned a D750, it had near 1 stop less highlight recovery than my lowly plastic A6000, so I had to dial in -0.7 EV permanently to match the output of both cameras in outdoor use (avoiding constantly clipped skies and clouds). Even further for more challenging scenes. Significantly cutting down the D750's famed high DR.
The read noise, as a statistical variation, can run under 1. So the dynamic range is not limited by the maximum signal level, even when converted to integer values in the raw files. But we agree that if the raw numbers have not enough resolution it there will be more quantization error (but I think that shadows will still be cleaner).
I initially assumed that Sony was using in the A7R4 the IMX551 sensor (14.85 stops of DR at the pixel level) but that is a 8K sensor. It could also be the IMC455 ("only" 14.4):
Both support 16bit ADC (but not at the highest fps).
So it seems that DR has been actually improved over the previous generation (if Sony is using these new sensors). All of this at the pixel level (measured "the DXO way" after normalizing the 61MP image to 8MP, another extra 1.45 EV of dynamic range will be added on top). Curiously, dpreview has stated today that the 15 stop are after downsizing (we hope that they are wrong).
Sony really needs to upgrade the RAW format, specially to add support for lossless compression (Canon lossless files use about 1 byte per pixel). There is a recent Sony patent on the subject, so we hope they'll address this sooner or later, maybe also increasing the bit depth.
Thanks for the replies. I wasn't aware SNR was quoted on the basis of a down-sampled image. That now makes a lot more sense. And @cgarcia, of course you are right that the noise on average can be a fraction of a bit - seems obvious now.
Regarding the 16 bit ADC versus 14 raw words, is that only in the newest sensors, or did the previous generation cameras also have this, and if so, are those additional bits accessible for shadow/highlight recovery?
Nobody really knows at this point. Other than probably some Sony employees and a few people who did field testing I suppose. Is it 15 stops on DxO, requiring a less than half stop improvement over A7r3, or did they really do something here? Is it some HDR mode? Nobody knows, not really a way to tell, at this point.
I could be wrong, but I don't think bits and DR stops are interchangeable. I mean, the number of colors are affected. Jpegs are 8 bit, 2^8=256, so you get 0 to 255 on RGB on each, so I think you get 256^3 colors? I suppose 14 bit is 2^14, and then that to the 3rd colors so that is where the billions of colors claims come from I guess ( I got great colors shooting jpegs.)
But I don't really think its limited as far as 14 bits can only get 14 stops. Actually I don't think you'll get 14 usable stops. But I am guessing a bit here. But more bits are better, for editing, imo.
Even 14 stops is phenomenal. I have yet to shoot a landscape with the A7RIII where lack of latitude/dynamic range at base ISO has been an issue. Yeah, I’d like more, and I’d prefer less noise at base ISO, but I’m really splitting hairs (of which I have few).
Just go to photons to photos website and check the photo DR, a more accurate measure of real world DR. Bill Claff knows his stuff and is the bible for these tests. A7RIII is about 11.3 EVDR @ ISO 100 IIRC and only a tiny bit more than the A7RII result. Maybe the A7RIV is a bit better but let’s wait for Bill’s results. Apparently per pixel DR is better than A7RIII but not sure of the numbers.
