Regarding the sensor...very disappointing. Sounds like a re-purposed 70D sensor with a DPAF improvement. I was REALLY, REALLY hoping Canon would really show something impressive on the sensor front with the 7D II. If the camera really does hit the streets with a 20mp sensor, I fully expect it to have the same DR limitations as all of Canon's previous sensors. Extremely disappointing. Guess we'll have to wait for the 5D IV to see if Canon can actually step up their sensor IQ game or not...which is just...so far down the road...Bleh.
Also worried about the "fine detail"...I really don't want them to start removing AA filters. That is just a dumb trend that photographers like simply because they do not understand the value of an AA filter, or the ease by which AA softening can be sharpened.
it sorta almost leads one to believe that Japanese Canon Fangirls post here where they were claiming that Canon feels they have Canon users trapped enough that it won't matter if the bodies they push out can't keep up as per sensors and even other features at times (still not a hint that they are actually moving any DSLR sensors to new fabs and the panny gets 4k and yet the super new 7D2 which was promised to have revolutionary video and this and that is still 1080p)
and yeah the AA filter-less stuff I am not a big fan of, maybe when we get to 180MP FF or 60MP APS-C or something.
Yeah, we really need sensors to significantly oversample the lens before we can legitimately start dropping AA filters. Otherwise we just end up WITH aliasing, and that's never good.
I was not really interested in the 7D II being a big video DSLR anyway...I don't really know that anyone truly was, you just don't get that cinematic look with a smaller sensor...not without having very wide apertures anyway (like a lot of expensive cinema lenses do).
The thing that I think Canon really needed to nail, and which increasingly appears as they will not, is producing a truly new sensor with a fundamentally new design on a smaller fabrication process size. It just isn't happening. If this thing is still a 500nm transistor part...I mean...WOW. That technology is about fifteen years old!! What is Canon doing? It's one thing to be conservative, but now it's just getting ludicrous...
I have always wondered about this, and you may be the guy to answer. Intel's next series of chips is what, 14nm process? I understand that Intel is purely in the microprocessor business, and Canon has to do a lot more than just optimize processes for sensors, but is there any practical reason why sensor transistors are / should be / need to be on such a different scale? Or is it just a matter of business and not wanting to make the necessary investment to keep shrinking? The fact that intel shrinks every other year has just made me wonder... because clearly there's an advantage to a smaller process.
no one creates large sensors using the latest technology - the A7R / D810E sensor for instance is on 180nm. which is speculated to the be the same as the 70D sensor. D700, D4, etc were even on larger than that (350nm to 250nm)
the toshiba sensor uses 65nm and sony was looking at and just starting to use 90nm for it's APS-C sensors, but unless you're talking the smart phone / compact sensors - there's just no benefit to the smaller geometries over the cost of production with the pixel granularity where it is.
canon's current line of lithography systems can produce chips under 90nm - far exceeding even really what is required by sensors - so it's not as if canon can't if they feel they have to. also to add to that, canon now has the equipment to product down to under 10nm geometries.
To be honest, people are humping on this as the core reason - not really. and most of them don't have a freaking clue, but all of a sudden turn into electronic and chip designers (not to mention camera designers too). canon certainly has a problem "downlevel" from the pixel - but their QE from their current 70D isn't that much off than the D5100's QE and even cutting the pixels in half they improved the QE by 10% over the 7D sensor level spec.
Your partly right, but your largely missing the point. A 500nm transistor is actually HUGE by todays standards. Think about it, that is half a micron. For a pixel that is surrounded by half-micron transistors, that is a FULL micron off all side of the pixel. A 4µm pixel is then only capable of, at most, a 3µm photodiode. The full size of the pixel itself is 1.78x larger in area than the photodiode. However, if you move to a 180nm process, your losing less than half a micron in total. That means the photodiode can be 3.6µm in size. The pixel is only 1.2x larger in area than the photodiode. Your photodiode area has increased by a factor of 1.5x, which reduces noise by 1.2x. That is significant. It's a stop gained in noise performance.
However, the REAL point about moving to a smaller transistor size is the ability to put more logic on the sensor die. At 500nm, Canon would have to make the sensor die itself quite a lot larger in order to move all the ADC logic onto the sensor itself, and make it column-parallel. At 180nm or better 90nm, they could move the ADCs on-die and need less than half, maybe less than one quarter, of the die space that would be necessary to do that with a 500nm process. Yield would remain high, so the cost of moving ADC onto the sensor would be much lower in the long run. THAT is the real point of moving to a smaller process. To allow more logic to be placed on the sensor die itself. The biggest gain there would be allowing full blown, high performance CP-ADC (and, maybe, also employ some of the other patents Canon has, such as dual scale ADC, power decoupling, etc.)
There is also a significantly greater per-pixel transistor requirement for stacked pixel designs. Canon recently released a patent for a five-layer sensor design. I honestly don't know how they would pull that off with a 500nm process. Not without a very low fill factor which would push noise levels sky high. However, with a 90nm fabrication process, creating a five layer sensor would be much, much easier, without running into serious problems with noise.
To add about 500nm vs 14 nm, I have read elsewhere (an article on Intel processors), that anything beyond 65nm is extremely difficult to produce and the yield is about 70% (which means 30% scrap in production, which in turn means very costly production, imagine if Canon had to scrap 30% of their sensors). The main obstacles are due to van der Waals forces - the force necessary to break the particle is larger than the force to clean it from a contamination for example (i.e. it is easier to break it than to clean it) and also the lytography is made with ultra-violet light and not with normal light, because the wave-length of the white light is comparable with the geometry of the particle at 14nm. The conclusion was that the efforts and costs to go down from 150nm to 90nm to 65nm to 25nm etc. is not a linear but geometrical progression and only 2-3 companies in the world can produce technology at 14nm. For example Intel is the only producer of processors at 14nm, their competitors are far behind.