This is a response to this essay on Luminous-Landscape.
I work for a large semiconductor company, one that is well-known in high-end personal computing and imaging. We make processors to handle graphics jobs, both 2D and 3D. What I see in common in our devices and modern cameras tell me that Moores’ law extends there as well.
The main issue I take with the argument made by Mr. Maxwell is that digital cameras could only benefit from Moore’s Law by way of higher pixel densities. That is a very narrow view of the role semiconductors play in digital cameras today.
In semiconductor terms, the pixels of a modern 20+ MP camera are so big they are what we call “visible from space”. You could see them easily with an ordinary optical microscope. The pixel sites are not obvious fruits of Moore’s Law (or more precisely, they got off that train years ago).
In contrast all of the following are direct or indirect fruits of Moore’s Law:
- High transistor densities translate directly into higher computing power onboard the camera
*For JPEG quality, not just for JPEG output but also LCD preview in the field, live-view, electronic viewfinders and movie recording which are all dependent on the JPEG engine. This includes more sophisticated noise processing engines in all these modes.
*For movie compression with more space-efficient formats like H.264 (as compared to MJPEG).
*For higher frame rates in live-view or movie filming.
*For more complex 3D tracking or contrast-detect auto-focus in all modes.
- Smaller transistors mean more processing can be moved to the sensor itself in the gaps between silicon pixels (not to confuse with microlens gaps or lack thereof)
* For lower readout interference noise and lower power.
* For faster scanning that supports the higher frame rates of future image processors.
* For new and unforeseen applications, for example perhaps HDR photometry at the pixel level.
- New processes intended for high transistor densities achieve lower power, lower heat and lower signal noise.
* For longer battery life during any operation.
*For longer live-view or video operation without overheating.
*For even higher-ISO operation that yields good results.
So Moore’s Law and its underlying mechanisms play a huge role in digital cameras. Have been from the start of digital photography.
As a minor note to close with, I also disagree with Mr Maxwell that f/11 is a universal reference for measuring maximum resolution of photographic lenses. The optical bandwidth of an ideal lens increases dramatically with larger f-stop, such as f/8 or f/5.6. While it's true that mass-market lenses do not deliver good performance all over the frame at f/5.6, that doesn't mean that it can't be done or even that there are no lenses on the market today that can do it. It's clear that a high-end lens would be needed, but I see no reason why a 50MP sensor can't be made and successfully utilized in the 35mm format.