D600 High ISO in DX

Started Nov 23, 2012 | Questions thread
Leo360 Senior Member • Posts: 1,141
Re: pixel pitch and SNR

noirdesir wrote:

Leo360 wrote:

There are two different things here. There is photon count per pixel and number of photons collected per unit area. The latter does not depend on the pixel pitch but the former does. And SNR per pixel gets larger with more photons collected by that pixel (photon shot-noise per photon gets weaker). For the same exposure larger pixels capture more photos and, thus, have higher SNR. This is why pixel peeping reveals more noise-per-pixel for smaller photosites. The price to pay is reduced resolution.

With proper down-sampling (bicubic, etc) to the same level of detail one does can hope to recover the SNR back by effectively combining outputs of multiple smaller pixels into an aggregate one but doing so does not entirely compensate for read-noise increase.

The photon shot noise per sensor area will be exactly the same for both the smaller and the larger pixel sensors, assuming the same quantum efficiency.

Yes, per unit area photon count and photon shot noise are the same. But measured per pixel the photon count as well as photon shot-noise variance are proportional to the pixel area (assuming constant fill factor).

And the read noise component of the final noise usually is the same or even lower for smaller pixels because noise is not something that gets simply summed up as it is a standard deviation.

If your noise sources are uncorrelated you sum up noise variances (squares of std.dev.)

In an ideal world, read noise is proportional to pixel size.

Why is that? I think that in ideal world read-noise should be independent of pixel size and should only depend on the temperature (even ideal world has thermal noise).

So, let's take a 2x2 μm pixel with a read noise of 6 e- (ie, a standard deviation of 6 e-). Now compare that to a 2x2 pixel array of 1x1 μm pixels, where each pixel has a read noise of 1.5 e-.

I don't think so. What makes you think that read-noise std.dev. scales quadratic with pixel pitch. Please, provide some rational for this behavior.

Simple statistics tell us that the read noise component for a combination of these four smaller pixels is:

rn(com) = square root of (1.5^2 + 1.5^2 + 1.5^2 + 1.5^2) = 3 e-

The formula above is based on a questionable assumption. Following your line of thoughts I will consider 2x2 um pixel against a group of 100 of 0.2x0.2 um pixels with the read-noise sigma= 0.06. This follows from your own read-noise scaling rule! Now based on your summation formula I will conclude that

rn = 10x0.06 = 0.6 which is 10 times smaller that original 2x2 pixel.

After signal aggregation from all 100 micro-pixels the 2x2 pixel SNR increased 10 times. Seems too good to be true or is it?

There is also dark current which neither of us have considered.

Bottom line, sensor noise performance is messy and contains lots of proprietary stuff in processing chain that we are not privy to.

So far from what I saw on Bill Claff's charts is that D600(DX mode) dynamic range outperforms  D7000 at all ISOs. I have no reason to think that with D5200 it will be any different.


 Leo360's gear list:Leo360's gear list
Nikon D5100 Nikon D750 Nikon AF-S DX Nikkor 18-200mm f/3.5-5.6G ED VR II Nikon AF-S Nikkor 24-120mm F4G ED VR Nikon AF-S Nikkor 50mm f/1.8G
Post (hide subjects) Posted by
Keyboard shortcuts:
FForum PPrevious NNext WNext unread UUpvote SSubscribe RReply QQuote BBookmark MMy threads
Color scheme? Blue / Yellow