[Guest]

From an engineer standpoint, adding more pixel actually adds more faults than benefits:

 

Closer pixel increase leakage current, charge in one pixel can leak to others -> bad.

 

High pixel density lower the amount of light fall on each element -> lower sensitivity.

 

16 MP aps-c already pushed many lens to the limit. High pixel count without extraodinary lens to resolve it would have results similar to an interpolated high MP photo from a lower resolution one, what for?

 

And, ugh, we are dealing with optical wave here, so the diffraction limit would be half wavelength. Taking around 700nm, the minimum optically resolvable distance would be 350nm, which would be the limit between two pixel to achieve meaningful results. Pixel density would be 1/0.35^2 ~ 8.16 pixel per micron square, or 8.16 megapixel per mm2. The maximum meaningful MP on an APS-C sensor would be 370 * 8.16~ 3000 MP or 3 Gigapixels It would be far until we hit that point indeed. (The optical resolution would be the first physical wall we would hit, since the smallest size human can put on silicon is less than 18nm, much smaller than the optical wall thingie)

 

FF are always expensive. A FF is 35.8 x 23.9 ~= 855 mm2 of silicon. A Core i7 (high bin, no defect) at 294mm2 (3 times smaller) costs 583$ by comparison. So the sensor alone costs at least arount 583*855/294 ~ 1700USD. (of course the manufacturing process is different, defect in cpus and in sensors are also treated differently, hence different and difficult to compare yield, but you get the point).

 

A 6x7 sensor at 60x70=4200mm2 would cost at least 8400$ in silicon alone, not to mention excruciatingly low yield, tiny market, and without economy of scale.

 

Don't forget about diffraction of the aperture. In the tests here you will see that with higher density sensors, performance of top lenses in the center already starts to fall at f4 or f5.6.       

What would be the optimal pixel density on a sensor when the lens has f2.8? or f2.0?