i see it this way.
canon must have a good reason to choose 18 MP for itÂ´s new camera.
if they could have produced the camera with the same specs but 30 MP im sure they would have done that. donÂ´t you think so?
so it seems that CMOS (that includes foveon like sensors) and CCD sensor technology is slowing down.
and itÂ´s not as easy as producing faster CPUÂ´s for a computer.
for faster CPUÂ´s you make the DIE bigger or shrink the lithography process to put more transistors on the DIE. you can also hardcode special functions to speed up calculations.
but even CPUÂ´s running against a wall, for years we are around 4 GHz.
a 35mm FF sensor must have a certain size, you canÂ´t make it bigger.
unlike CPUÂ´s shrinking is not the holy grail.... smaller photosites will not help you increasing DR or reducing noise.
there is not much you can do when you reach a certain point of optimization.
you canÂ´t influence the absolut number of photons that reach the sensor area (of course you can, for example with flash... but i think you get what i mean ) and you can not increase the absolut sensor area.
you can only decrease the size of additional circuits to make more room for the photosites (as canon has done for the new sensor. sony has done it with itÂ´s backlit sensor), so the photosites can catch a greater percentage of photons.
sure you can always try to reduce the SNR but that will be expensive at a certain level.
new materials could help or a new sensor technology.
but i think we should get used to smaller steps in sensor development.Quotehowever it will not outperform the 5D MKII in resolution, after proper sharpening is applied.
i say that depends on the AA filter thatÂ´s build in. but i have not done the math.
how is optimal resolution of the new 18 MP sensor compared to real world resolution of the 5D MK2 (i know that is unfair ).
it may not outperform the 5D MK2.. but it could be a imperceptible difference.
I mostly agree.
For a fast action camera the processing power and data bandwidht limits are the main constraints; larger files require longer times for processing and saving, while in-camera buffers are limited in size.
You also need a large photosite area to allow for a decent SNR ratio and support the framerate.
So the specs of a pro camera aimed to sports and wildlife shooters are more or less capped under 20 Mpxls; Nikon will probably follow the same route for the D4s or whatever they'll call the new pro sports camera.
Up to this day Canon has showed incredible skill in optimizing their sensors for catching the maximum amount of light for a given photosite size; they are working very close to theorical limits of actual electronics.
The incredible performance of EOS 7D, given its sampling rate, is a proof of this ability.
Resolution wise however I don't see a real boost in dropping the antialias filter: of course it would boost sharpness (sensor would behave closer to an order 2 MTF rolloff than actual order 3) but top resolution after proper sharpening would be more or less the same, i.E. very close to the same Nyquist limit.
Moreover I'm scared about high MTF values close to Nyquist limit of a sensor: artifacts become very frequent, specially for Bayer matrix sensors.
It's a bit better with Foveon, as there is no color Moire, but artifacts are present anyway (I've seen it happen frequently with my Sigma DP2x, a camera with a very fine normal lense).
I could break it down with a few simulation scenarios when I get the time.
Safer zone to start dropping antialias filters for Bayer sensors is around 4 micron photosite pitch, i.e. close to 50 Mpxls; MTF should not be much higher than 10% there, even for an excellent full frame lense.
I believe Canon actually owns the CMOS technology to produce a 32 Mpxls full frame sensor with noise/dynamic range performance equal or superior to EOS 5D MK II, providing a good resolution boost with no practical drawback.
It would be targeting a different, probably more profitable, market share than 1D-X