Originally posted by wombat2go I read that the active area of the photodiode in a consumer camera CMOS sensor is not square and may be 60% or so of the pixel pitch. It is buried under light blocking conductors, and may have a chunk dedicated to the switch transistors.
The microlenses help the utilization factor but this effect may fall off at sensor edges and with wide aperture and retrofocal lenses.
The MTF roll-off is related to the pixel pitch for monochromatic and Foveon sensors but is less for Bayer sensors because of the interpolation between the 3 colors.
And it may be deliberately less by the use of the low pass optical filters used to protect against aliasing.
Notwithstanding all the above, the consumer CMOS sensor development over the past 5 years has been an extraordinary achievement. I read its is increasingly driven by the "small hand held device +video" market.
I haven't given much thought about how pixel pitch affects MTF to be honest, but what you say has merit. Its no secret that wide angle retrofocus designs challenge digital photography the most. Especially as you go down in sensor size. I'm not entirely sure (and please correct me if I am wrong), but I think that most DA lenses with modern formulas should be mostly telecentric by design to combat this. For slightly retooled FF designs like the 35/2.4 and the 35 limited I'm not so sure but they are probably mostly telecentric anyways. It seems like it is below 30mm that challenges begin. I believe that the current best microlenses can deal with angles of incidence no greater than 60 degrees. It is true, if you look at electron scanning microscope images of sensors the photodiodes do indeed sit in a deep well with a layer of circuitry that sits directly above. Back side illuminated (BSI) sensor design eliminates this problem and moves the wells much closer to the surface. This is what the Samsung NX1 currently uses and it clearly gives about a half stop to a stop better in light collecting abilities. Sony has long claimed that the pixel pitch of APS-C is sufficient enough that the gains from BSI would be minimal, but clearly Samsung's chip is outperforming their best at least in terms of sensitivity. There is a technology called quantum film that has the potential to eliminate many of the downsides of electronic imaging, but it looks like it is a ways away before they start producing actual sensors based upon that.