Concensus on the S100's Lowered Pixel Density?

Started Sep 16, 2011 | Discussions thread
Shop cameras & lenses ▾
Ron Parr
Ron Parr Forum Pro • Posts: 13,774
Re: Canon said pixel density dropped due to CMOS design

regis3 wrote:

So why camera-sensor-producers give each pixel (or cluster of pixels) micro lenses ,and why do they bother to make circuits smaller?

They add microlenses to focus some of the light that might have hit the circuitry blocking part of the top of the pixel onto the non-blocked part of the pixel so that it can be detected.

They try to reduce the size of the circuitry (or use BSI) because microlenses are not 100% effective.

In both cases, the effect is to increase the amount of light that is actually detected by the sensor, thereby increasing the S part of SNR. Neither of these things decrease noise. Increasing the signal will also increase one kind of noise called photon shot noise, but it increases more slowly than the signal does, so it's still a win. (See more details below.)

Theoretically ,lets say that from 100% of the light coming to the sensor from the lens ,80% of it is a "good light" (meaning it represents the actual picture) and 20% of it is "bad light" (meaning some kind of light-noise).

There is no "light-noise". There is something called photon shot noise, which scales with the square root of the number of photons detected. This is actually due to the fact that photon emission is a random process (Poisson distribution) so that under constant illumination, if average number of photons hitting a surface is n, the standard deviation will be sqrt(n).

If your pixel receives less light ,chances are it gets more "bad light" than "good light" ,but if you attach micro-lenses and increase the amount of light it receives (by reducing circuity-size that "disturbs" the light coming-in) ,it'll get more light.

There's no bad light. If you previously received n photons, you'd have photon shot noise of sqrt(n) plus some other sources of noise (readout, etc.) that we'll call c. Your SNR would be n/(sqrt(n) + c).

Let's say you now capture 25% more photons. Your new SNR is (1.25 n) / (sqrt(1.25 n) + c)). Since the numerator is growing faster than the denominator, this is a win.

-- hide signature --
Reply   Reply with quote   Complain
Keyboard shortcuts:
FForum PPrevious NNext WNext unread UUpvote SSubscribe RReply QQuote BBookmark MMy threads
Color scheme? Blue / Yellow