Jump to content
UltravioletPhotography

Noise and sensor size references


DaveO

Recommended Posts

They write in that first article about DX versus FX (or smaller versus larger) sensor sizes:

 

The light intensity on the sensor remains the same, but the amount of light you can capture has dropped.

...the more light you capture, the less noisy your image, this is key to why large sensors generally give better image quality than small ones.

 

If light intensity (amount of light per square mm) remains the same over each sensor, then why does total collected light explain noise?

Wnen I switch my D810 from FX to DX, the DZ image has the same amount of noise as the FX image. (Assuming I have not changed distance from subject during that switch from FX to DX.) Doesn't it?

Link to comment

People are confused by spurious correlations.

 

A smaller sensor is more likely to show edge effects of the photo sites and the spatial independency so important for Bayer arrays weakens. On-chip circuitry might occupy relatively higher fraction of the light receiving area. And so forth.

 

FX to DX isn't a very big change in noise behaviour whilst FX to m43 might be. The quality of the smaller sensors increase all the time, but so do the bigger ones. Thus there are likely differences that never can be completely removed. Same goes for the drawbacks relating to size as such.

 

In the film days, "larger was better". To be qualified: For some aspects of the photography, not for others. Digital hasn't really changed that although the actual dimensions are smaller overall.

Link to comment

A smaller sensor is more likely to show edge effects of the photo sites and the spatial independency so important for Bayer arrays weakens. On-chip circuitry might occupy relatively higher fraction of the light receiving area.

 

OK, those observations make sense.

 

So it really isn't about "total light collected" ?

Link to comment
I never could see how noise could be about the "total amount of light collected" given that light per sq. mm. is the same on small and on large sensor. But I have read this "total amount of light collected" argument so many times over the last year or so that I started trying to convince myself to believe it.
Link to comment

BJØRN WROTE: If everyone believes ghosts are real, then sooner or later a ghost will be seen. Apparently.

 

I apologize for accidently deleting Bjørn's comment just before this one. It was not intentional. Sometimes having Admin privileges can be hazardous.

 

********

 

 

 

Always go to the definitions! Lots and lots of qualifiers are missing from all other discussions. Now it all begins to make sense.

 

Here in this discussion of shot noise we see that for equal exposures, equal quantum efficiency, EQUAL PIXEL COUNT, EQUAL FINAL IMAGE SIZE that SNR is proportional to the square root of sensor size.

 

But we also see that if DEPTH OF FIELD and shutter speed are held constant, then shot noise does not depend on sensor size.

 

********

 

Wikipedia: Shot noise

 

Shot noise SNR is given by

 

http://upload.wikimedia.org/math/6/0/1/601fc8b8c5fc414e60f134869d5ba4e5.png.

 

P = photons per second per area, incident photon flux per pixel area

Q = quatum efficiency

t = exposure time

D = pixel dark current as electrons per second

N = pixel read noise in electrons.

 

Apart from the quantum efficiency it [shot noise] depends on the incident photon flux and the exposure time,

which is equivalent to the exposure and the sensor area;

since the exposure is the integration time multiplied with the image plane illuminance,

and illuminance is the luminous flux per unit area.

 

Thus for equal exposures, the signal to noise ratios of two different size sensors of equal quantum efficiency and pixel count will (for a given final image size) be in proportion to the square root of the sensor area (or the linear scale factor of the sensor).

 

If the exposure is constrained by the need to achieve some required depth of field (with the same shutter speed) then the exposures will be in inverse relation to the sensor area, producing the interesting result that if depth of field is a constraint, image shot noise is not dependent on sensor area.

Link to comment

You will arrive at a similar conclusion simply by observering "final image size equal" for two differently sized sensors means the overall magnification is different by a factor of the square root of the area ratio.

 

The problem, as always with these comparisons, is you can get almost any conclusion by massaging the parameters to examine. Thus more depends on what you ignore or overlook than what you include.

 

I've given up these comparisons a long time ago. It's a lost battle anyway.

Link to comment

Well, it certainly is a lost battle.

But I have to read and learn until I understand things. Then I can let it go.

 

I now understand why folks write what they do. But they leave out so many of the qualifying statements that nothing makes sense at all !!!

 

It really is nothing more than physics.

Link to comment

I don't understand why people get so obsessed by "noise" and "equivalence". This has nothing to do with photography, only serving as drugs for the information junkies.

 

Perhaps this is an excuse for not trying to grasp photography as a means of visual expression.

Link to comment

Well, I don't know about other people, but I enjoy both. I enjoy the artistry of photography, and I also like science and learning about how things work.

 

When I was very young, it used to be thought that girls didn't need to know that factual stuff. :) Thankfully, that has changed.

Link to comment

I didn't see anything about pixel size. If you keep pixel count the same but have sensors of different sizes, obviously, the smaller sensor has smaller pixels. "Increasing pixel size improves the sensor by increasing dynamic range and signal-to-noise ratio." [1]

 

"For any fixed process technology and pixel architecture, decreasing pixel size will decrease pixel performance. Without compensating technologies, smaller pixels have lower dynamic range, lower fill factor, worse low light sensitivity, higher dark signal, and higher non-uniformity."...

"The peak SNR declines with pixel size mainly because well-capacity (maximum number of electrons prior to saturation) decreases with pixel size from about 37,500 (5.2um) to 16,000 (2um) electrons. The reduction in well-capacity alone, without any other noise contributions, would produce a decline in the peak SNR of more than 3dB. The SNR drop is larger, however, because the effect of noise is more significant when added into lower signal levels. Thus, even though the technology advances described in Rhodes [4] result in lower levels of read noise and dark voltage, their impact remains high because these noises are superimposed on relatively low signal levels. The peak SNR of the smallest pixel is approximately 8dB lower than that of the largest pixel." [2]

 

 

[1] "How small should pixel size be?" Chen, et al. http://white.stanfor...Size_SPIE00.pdf

[2] "Resolution and Light Sensitivity Tradeoff with Pixel Size" Farrell, et al. http://invensense.co...thPixelSize.pdf (A good read)

Link to comment

It's not really the sensor size what matters but the pixel density. The reason why you don't see noise level change when you switch from FX to DX (on the same Nikon camera) is because it also drops the pixel count for the image, therefore the pixel density stays the same. Which also means that for each pixels the amount of light remains the same.

The comparison only makes sense if both FX and DX has the same amount of pixels, which will make the DX much higher pixel density, which means each pixel will have less light to capture (as each pixel will have a smaller surface to work with). For example the 24 Mpix cameras Sony A7 vs A6000. Or in an all FX comparison the Sony A7 (24Mpix) vs A7r (36Mpix) vs A7s (12Mpix)... The A7s clearly wins in noise due to the much larger surface for each pixel.

Link to comment

To be precise, it is the pixel spacing that has to be different when you keep sensor size and the pixel count varies.

 

Photo sites can, in principle, be equally sized. They commonly were in the earlier stages of digital sensor technology, but today photo sites with larger microlenses are getting more common.

Link to comment

Please sign in to comment

You will be able to leave a comment after signing in



Sign In Now
×
×
  • Create New...