If you've been following the discussions over the past....ever....the trend has definitely been toward image taking advances coming from software, not hardware. At least not in the same sense as hardware wars in megapixel wars (still alive on the forums), BSI vs FSI (functionally moot with the current generations of both). For a while now, the physics of sensors and lenses has been such that gains there are very slight and mostly incremental, as we see here. Stacked sensors come with some trade offs, also seen here (an often inconsequential drop in DR for a massive speed increase). Global shutter has some significant tradeoffs as well (noise, DR, for sensor speed). This is what allows us to look back at something like a 5D4 and say 'not much has improved'. There have been improvements, just not where you are looking for them. There are real, physical limits to capturing data on a sensor - bigger well depth defines a hardware limit on dynamic range on the top end, and sensor noise on the low end, but the physics of the number of photons coming in during an exposure, and how much data is really captured there are is a limit as well. So while we love to push our shadows 6 stops in post and analyze to death the performance of the sensor, the reality is the gains to be had on the hardware side might not be as much of a difference maker as you think. Perhaps a future breakthrough will change that, but it hasn't happened yet for any of the major manufacturers.
The breakthroughs have been in software solutions. What people broadly (perhaps too broadly) refer to as Computational Photography. If you went back to the 5D4 release and said that 2 cameras later it would be smart enough to know what the ball is, predict where its going, and focus on that for you, or recognise and track a specific face....you might have received some curious looks. The software available for noise reduction in post making ISO 52k and beyond usable was also not there at that time. Remember pushing things at ISO 3200?
I find the 'fear' of 'computational photography' an amusing discourse online. I myself am not a big editor of my photos. I do things for exposure, color, contrast etc in DXO (a lightroom alternative) for sure. But many of those online that decry in camera noise reduction or other computational effects applied are also the ones who replace skies, manipulate facial features, add light sources that didn't exist in the original photo, clone in our out various features, and otherwise 'compute' an image that mostly did not exist in the first place.
There is currently no photography without computation (unless you still buy film, and don't scan any of it). And the software side is where the innovation has been highest in the last couple of generations of cameras.
Brian
