Don Haines said:
pedro said:
Don Haines said:
bdunbar79 said:
I want to assume the pixels won't get larger. That's really the only way I know to increase the FWC in the current 1Dx sensor. If that won't happen, can the DR improve at high ISO by a 1/2 or a full stop? If so, how will they do it?
One way would be to move to a finer fabrication process. This would result in less "wasted space" and would improve the quantum efficiency... they MIGHT get a third of a stop from that, but it would be pushing it....
Why are Canon increasing MP count in a low light pro beast if it has a worsening effect on high ISO IQ, as guess to understand based on this explanation? Aren't Canon shooting themselves in their footby doing so?
The finer fabrication process is only one way to improve things, and I am sure that there are a lot more. Look at the 7D replacement 7D2.... The megapixel count went up, yet the ISO performance went up by about 1 1/2 stops, only about a half stop could be explained by a fabrication change. I would expect that going from a 1DX to a 1DX2 would be at least a similar gain.... unless Canon has a few more surprises for us...
Aye, more than one way to skin a sensor.
Backside Illumination. Multi-layered Photodiodes. CCD backing stores (i.e. multi-bucket pixels). Deep photodiodes (the photodiode is deeper, and the charge separation layer reaches deeper, increasing it's surface area, which increases charge capacity).
There are also more out of the box concepts, like continuous readout. This is where during exposure, the pixels are basically photon counters and bit flippers, and constantly read out. Rather than accumulating charge, you simply ratchet up an infinite count. To be effective (i.e. to not accumulate massive amounts of read noise at the same time), other technological improvements are necessary to keep read noise extremely low, and Q.E. apparently needs to be very high (basically 100% for them to actually act as photon counters, and I don't know that 100% is actually totally necessary), but it is technically a means of achieving infinite dynamic range. (And that might even be possible if counts were done in/converted to a 32-bit float rather than a 16-bit int.)