Originally posted by john1973 A more intelligent sensor would be nice for real-time exposure control like the LX could. Even better would be (I'm probably just day-dreaming here) multiple readouts of the sensor per exposure. Why? Less noise, stacking photo's in astrophotography is a methode used commonplace to reduce noise. This imo would also dramatically improve on dynamic range when needed. I think this would be comparable creating a HDR, you use all read-outs in underexposed areas and use less read-outs in areas which would clip in a single exposure. You could also meter for the darkest areas you want to see detail in without (to a certain degree) having to be concerned about overexposing lighter areas of your compositions.
Yes, this method kinda works on the K-7 (shadow and highlight correction) BUT as far as I know this is all done digitally in the processor after the shot has been taken - it results in some noise in the darker areas because it's almost the same as editing a picture in let's say Lightroom and applying the graduated filter to the darker part to lighten it -> you produce noise in that part.
To solve this the sensor would have to be VERY smart (or at least the exposure control behind it) to figure out which pixels in the current shot will be under exposed and which will be over exposed then correct the exposure of each pixel (this means separate voltage/current control for each pixel and I don't think this is possible just yet). I don't know exactly how this would work but I do know it would take a lot of processing power to process each of the 14,6 million pixels. And even if you manage to evaluate all the pixels the end result could be a strange looking picture - like the ones where you exaggerate with shadows and highlights in post editing. Again, this could be solved with user set parameters and again there's a problem - how much correction you'd want for each shot ... a slider option maybe.
I mostly shoot landscape and this kind of feature would be great when you can't use ND filters to correct exposure but I think the technology just isn't there yet. Software solutions produce too much noise.
Still the best way to go is to use ND where possible and do a lot of post editing where needed.