Any guesses as to what engineering issue might be involved in the difference between Canon sensor DR and Nikon sensor DR?
Guesses, sure!
One big obvious technological difference is that:
- Canon still does analog to digital conversion off the sensor, so the analog signal (charge/current) has to be moved from photosite to the edge of the senor and then moved at far higher speed along the edge of the senor and off-board, where about four or eight ADC units operate at high speed to process the pixels.
- recent sensors used by Nikon (and Sony, and in the Panasonic GH models cameras, and maybe in the Olympus E-M5) instead use a column-parallel on-sensor approach to ADC: they have ADC units on the sensor, one at the bottom of each of the thousands of columns of photosites. The analog signal only has to make the one, slower, trip from photosite to sensor's edge, where with thousands of ADC units operate in parallel, so that each can operate at a far lower rate.
These differences seem to allow greatly reducing the read noise and thus improving the deep shadow handling, dynamic range, and high ISO/low light performance. It has been shown with some high-end Canon SLRs (like the 1DMk3?) that the noise floor in the final output at base ISO speed is far higher than the noise floor in the signal produced within the photosites, so the villain seems to be additional noise that enters during the analog signal transportation and/or in the high speed ADC operation.
With so many major rivals now using column-parallel ADC (Samsung also has that technology at least in some video cameras) and getting excellent results with it, my guess is that Canon is developing something similar, and we will see it in the next generation of high end Canon SLR models in a few years' time.