Equivalent focal length for MFT lenses

Started Apr 12, 2013 | Discussions thread
Detail Man
Detail Man Forum Pro • Posts: 16,816
Re: MOSFET Source Follower 1/F Noise as a function of Gate Area

Detail Man wrote:

Detail Man wrote:

draleks wrote:

Does read noise depend a lot on the pixel size? I've seen some people try to argue that sensor size is the only thing that matters for noise/dynamic range, but perhaps when you stack pixels too tightly (especially when pushed to do so by the marketing..) the pixels start to interfear with each other...

The reason that I keep it to, "the Read Noise will depend upon the particular image-sensor design", is from this kind of indication of how the matter is not as simple as photo-site size:


bobn2 wrote:

A non-commital statement regarding per-pixel Dynamic Range:

OK, follow the above discussion, you see that as I say, the DR of a pixel is more or less independent of the size of a pixel, which means that the DR measures qualities other than size. Compare pixels with different size, the one with better DR is the higher quality pixel.

Statements to the effect that per-area Read Noise tends to favor smaller pixels:

The historical evidence is just confirmation of a theoretical result, which is essentially the pixel DR is size independent under strict scaling - see the discussion above. If that is true, we would expect normalised DR of sensors to increase in general as pixel sizes decrease, an this in general is what happens. The reverse argument, often put around, is that pixel level DR increases as pixel size increases, and moreover that it increases more than enough to counteract the resulting reduced pixel count. If that was the case we would expect to see a trend where normalised DR of sensors has reduced as pixel size falls. No-one has ever shown that. The best they manage is to argue, when shown the reverse trend, is that the effect is due to unspecified 'technology improvements'. For that to be convincing, they would need to specify what the technology improvements are, and also, since the theory predicts that pixel shrink would produce the observed effect, why it isn't in practice.

April 4, 2012: http://www.dpreview.com/forums/post/41123445

... In this case, about the role of the capacitance of the read transistor gate and floating diffusion on the input referred read noise, I thought it should be so, developed the argument, had a fairly lengthy discussion with Emil Martinec and John Sheehy where they failed to knock it down, later took the opportunity to have it confirmed by Eric Fossum, then saw it confirmed in the Aptina white paper.

The question I am answering is 'if you take a pixel design and scale it geometrically, what happens' - and what happens is quite clear, the FWC and read noise scale in proportion to the area (although Eric says that's not quite true at very small geometries because other noise sources come into play) which means DR is invariant.

April 5, 2012: http://www.dpreview.com/forums/post/41130657

This statement by EC appears to indicate complexity in smaller MOSFET source-follower geometries:

Eric Fossum wrote:

... I think most of the read noise today comes from residual 1/f noise in the very first transistor, which typically gets worse as the transistor area is reduced if nothing else is done to improve it.

April 10, 2013: http://www.dpreview.com/forums/post/51263400

Section 5 ("1/f Noise in Source Follower Reset Circuit") of "Analysis of 1/f Noise in Switched MOSFET Circuits", Tian, Gamal, Stanford Univ., 2001 (on Page 6) presents an appropriate non-stationary 1/F Noise model appropriate to MOSFET source-follower reset circuits in photo-site assemblies. It seems notable that they state the following regarding the result of numerical analysis utilizing the model:

Note that this result is virtually independent of the capacitance. This of course is very
diļ¬€erent from the famous kT/C reset noise due to thermal and shot noise sources. The
reason is that 1/f noise power is concentrated on low frequencies, and thus is less sensitive
to circuit bandwidth and hence C.


While better analytical models for characterizing MOSFET transistor 1/F Noise may have been developed since this paper's publication in 1999, estimating 1/F Noise appears to be non-trivial:

1.2.  1/f Noise in MOS Transistors.

1/f noise has been observed in all kind of devices, from homogeneous metal
film resistors and different kind of resistors to semiconductor devices and
even in chemical concentrated cells. Because 1/f noise is well spread over the
components, people think that there a fundamental physical mechanism is
behind it. Till now such a mechanism is not yet found.

There is a lot of experimental evidence that several mechanisms are involved
in generating 1/f noise.

The MOS transistor has the highest 1/f noise of all active semiconductors, due
to their surface conduction mechanism.

The result is: several theory's and physical models competing together to
explain the 1/f noise in a MOSFET. These theory's and models differ in detail
but are all based on the mobility fluctuation model expressed by the Hooge
empirical relation, and the carrier density or number fluctuation model first
introduced by McWhorter.


DM ...

Found a couple of DPReview threads containing some interesting information here and there:

News and Rumors Talk, June 2011: http://www.dpreview.com/forums/post/38669131

Canon EOS-1D/5D/6D Talk, March 2012: http://www.dpreview.com/forums/post/41077405

... my personal interest in referring to them being how photo-site characteristics vary with size.

Post (hide subjects) Posted by
Keyboard shortcuts:
FForum PPrevious NNext WNext unread UUpvote SSubscribe RReply QQuote BBookmark MMy threads
Color scheme? Blue / Yellow