Larger sensors do not have less low light noise. It is a myth. part 2

[Woland65]

D Cox wrote:

bobn2 wrote:

The total amount of light that falls on the sensor is given by the size of the lens aperture in millimeters. The total light per pixel, and thus photon shot noise, is given by the total light dividd by the number of pixels. The area of a pixel does not matter, only the number of pixels.

Sorry, I had lost the context of the discussion, and the universality of the great truth. Lens aperture in millmetres, yards or parsecs and angle of view. A 10mm aperture on a 10mm lens is going to put a lot more light on the sensor than a 10mm aperture on a 50mm lens (talking same size sensors).

This is why f numbers were invented.

It is the f number that decides how much light falls on a given area of sensor in a given time, not the aperture diameter in mm.

Yes, and vice versa, f-numbers where not invented to make it easier to understand where low light noise comes from. f-numbers make it easy to handle a camera, but at the same time it causes a lot of confusion as to how the camera actually works.

If aperture where commonly measured in mm it would be more obvious to people that aperture is what controls noise, not sensor size, but using mm for aperture would of course make it much harder for photographers to get the right exposure.

 

[Woland65]

bobn2 wrote:

D Cox wrote:

bobn2 wrote:

The total amount of light that falls on the sensor is given by the size of the lens aperture in millimeters. The total light per pixel, and thus photon shot noise, is given by the total light dividd by the number of pixels. The area of a pixel does not matter, only the number of pixels.

Sorry, I had lost the context of the discussion, and the universality of the great truth. Lens aperture in millmetres, yards or parsecs and angle of view. A 10mm aperture on a 10mm lens is going to put a lot more light on the sensor than a 10mm aperture on a 50mm lens (talking same size sensors).

This is why f numbers were invented.

Yes

It is the f number that decides how much light falls on a given area of sensor in a given time, not the aperture diameter in mm.

No. It is actually, physically, the aperture diameter in mm and the extent over which the light is collected. The f-number provides an approximate parameter to represent that.

 

[D Cox]

Roland Karlsson wrote:

Larger sensor cameras have larger lenses.

Only if you set a fixed angle of view.

You can use exactly the same lens on (for instance) an APS-C and a FF sensor. The lens doesn't change size.

You can easily make good 50 mm F1.4 for FF.

To gather the same amount of light and get the same FOV for FourThirds, you need a 25 mm F0.7. I promise you, it is not as easy to make, actually almost impossible, at least if you want high IQ.

Put a 50mm lens on a FF film camera (so we avoid ISO questions) and on a half frame film camera. Set them up side by side. Set both shutters to 1/60.

They will want _the same_ f number for matched exposure.

To give the half frame the same total amount of light as the full frame over its area, you would have to increase the exposure by a stop. But what we want to know is why the larger format gives better quality at the same exposure.

I think the answer is that the noise in the image increases with spatial frequency (in line-pairs per mm). Therefore the more an image is enlarged for viewing, the more noise can be seen.

If there is no enlargement (contact print from 10x8 film), the frequency at which the amplitude of the noise is enough to make it visible is too high to be resolved by the eye. If there is great enlargement (16mm negative blown up to 10x8) the noise is enlarged enough to be visible, and you see the granularity.

The same thing is happening in digital equipment. Noise increases with spatial frequency on the sensor. Is this affected by the division of the sensor into more or fewer pixels? Probably not, but signal (image) strength at those frequencies is affected - more pixels give a stronger high frequency signal, so a better signal to noise ratio.

 

[Detail Man]

Woland65 wrote:

bobn2 wrote:

D Cox wrote:

bobn2 wrote:

The total amount of light that falls on the sensor is given by the size of the lens aperture in millimeters. The total light per pixel, and thus photon shot noise, is given by the total light dividd by the number of pixels. The area of a pixel does not matter, only the number of pixels.

Sorry, I had lost the context of the discussion, and the universality of the great truth. Lens aperture in millmetres, yards or parsecs and angle of view. A 10mm aperture on a 10mm lens is going to put a lot more light on the sensor than a 10mm aperture on a 50mm lens (talking same size sensors).

This is why f numbers were invented.

Yes

It is the f number that decides how much light falls on a given area of sensor in a given time, not the aperture diameter in mm.

No. It is actually, physically, the aperture diameter in mm and the extent over which the light is collected. The f-number provides an approximate parameter to represent that.

I am not sure that anyone really disagrees here, but... given a constant FOV, the f-ratio gives light per unit area of the sensor, whereas aperture size in mm gives the total light falling on the sensor.

And, heaven knows, it is just so hard to multiply light per unit area by image-sensor active-area ...

 

[bobn2]

Woland65 wrote:

bobn2 wrote:

D Cox wrote:

bobn2 wrote:

The total amount of light that falls on the sensor is given by the size of the lens aperture in millimeters. The total light per pixel, and thus photon shot noise, is given by the total light dividd by the number of pixels. The area of a pixel does not matter, only the number of pixels.

Sorry, I had lost the context of the discussion, and the universality of the great truth. Lens aperture in millmetres, yards or parsecs and angle of view. A 10mm aperture on a 10mm lens is going to put a lot more light on the sensor than a 10mm aperture on a 50mm lens (talking same size sensors).

This is why f numbers were invented.

Yes

It is the f number that decides how much light falls on a given area of sensor in a given time, not the aperture diameter in mm.

No. It is actually, physically, the aperture diameter in mm and the extent over which the light is collected. The f-number provides an approximate parameter to represent that.

--
Bob

I am not sure that anyone really disagrees here, but... given a constant FOV, the f-ratio gives light per unit area of the sensor, whereas aperture size in mm gives the total light falling on the sensor.

It's interesting just to consider the problem from the point of view of the light collection only, what happens in front of the entrance pupil, and regard everything behind it as a black box which produces an image from the light collected. Then you rapidly see that it is, as you say, the aperture diameter and the solid angle from which it is collected which matters. In the context of this discussion (different sensor sizes) the f-number confuses rather than clarifies.

--
Bob

 

[Woland65]

D Cox wrote:

Roland Karlsson wrote:

Larger sensor cameras have larger lenses.

Only if you set a fixed angle of view.

You can use exactly the same lens on (for instance) an APS-C and a FF sensor. The lens doesn't change size.

You can easily make good 50 mm F1.4 for FF.

To gather the same amount of light and get the same FOV for FourThirds, you need a 25 mm F0.7. I promise you, it is not as easy to make, actually almost impossible, at least if you want high IQ.

Put a 50mm lens on a FF film camera (so we avoid ISO questions) and on a half frame film camera. Set them up side by side. Set both shutters to 1/60.

They will want _the same_ f number for matched exposure.

To give the half frame the same total amount of light as the full frame over its area, you would have to increase the exposure by a stop. But what we want to know is why the larger format gives better quality at the same exposure.

I think the answer is that the noise in the image increases with spatial frequency (in line-pairs per mm). Therefore the more an image is enlarged for viewing, the more noise can be seen.

If there is no enlargement (contact print from 10x8 film), the frequency at which the amplitude of the noise is enough to make it visible is too high to be resolved by the eye. If there is great enlargement (16mm negative blown up to 10x8) the noise is enlarged enough to be visible, and you see the granularity.

The same thing is happening in digital equipment. Noise increases with spatial frequency on the sensor. Is this affected by the division of the sensor into more or fewer pixels? Probably not, but signal (image) strength at those frequencies is affected - more pixels give a stronger high frequency signal, so a better signal to noise ratio.

What you are discussing here iare the consequences of using a lens that is not made to fit the sensor. If you use a crop sensor with an ff lens then a lot of the light from the lens will miss the sensor, and because less light is hitting the sensor you will get more low light noise.

Likewise, if you use a crop lens with an ff sensor the light will not cover the sensor.

These are not effects of any fundamental physics going on, it is the effect of a mismatch in the design between lens and sensor.

However, if you only use lenses and sensors that are built to match each other, then a crop sensor will not give you more low light noise, as long as the lens aperture has the same size.

"But what we want to know is why the larger format gives better quality at the same exposure."

Isn't that obvious? At equal exposure the larger sensor will be receiving more total light, and therefore low light noise will be smaller. At equal exposure a sensor with twice the size receives twice the amount of light.

 

[Woland65]

Detail Man wrote:

Woland65 wrote:

bobn2 wrote:

D Cox wrote:

bobn2 wrote:

The total amount of light that falls on the sensor is given by the size of the lens aperture in millimeters. The total light per pixel, and thus photon shot noise, is given by the total light dividd by the number of pixels. The area of a pixel does not matter, only the number of pixels.

Sorry, I had lost the context of the discussion, and the universality of the great truth. Lens aperture in millmetres, yards or parsecs and angle of view. A 10mm aperture on a 10mm lens is going to put a lot more light on the sensor than a 10mm aperture on a 50mm lens (talking same size sensors).

This is why f numbers were invented.

Yes

It is the f number that decides how much light falls on a given area of sensor in a given time, not the aperture diameter in mm.

No. It is actually, physically, the aperture diameter in mm and the extent over which the light is collected. The f-number provides an approximate parameter to represent that.

I am not sure that anyone really disagrees here, but... given a constant FOV, the f-ratio gives light per unit area of the sensor, whereas aperture size in mm gives the total light falling on the sensor.

And, heaven knows, it is just so hard to multiply light per unit area by image-sensor active-area ...

There is nothing wrong with the units photographers use today. They were chosen for good reasons. It just happens that they sometimes lead to confusion when understanding how a camera works.

 

[mike703]

D Cox wrote:

But what we want to know is why the larger format gives better quality at the same exposure.

I think the answer is that the noise in the image increases with spatial frequency (in line-pairs per mm). Therefore the more an image is enlarged for viewing, the more noise can be seen.

The noise (except at extremes of low light) is the shot noise which is the uncertainty in counting the number of photons. It depends on the number of photons captured and is nothing to do with the resolution of the sensor, just the area. A sensor 2x larger will capture 2x as many photons (at the same exposure settings) therefore the statistical uncertainty in measuring that will decrease by square root of 2.

You can get the benefit of this in two ways: print the images at the same final size (in which case the larger sensor camera will have lower noise). or you can print the images so that they have the same noise level, in which case the larger sensor camera will allow you to print a larger picture.

Best wishes

 

[RussellInCincinnati]

It is indeed interesting to point out that small sensors don't have to be so bad in low light as we tend to think they are, maybe they could be within shouting distance of being as good as larger sensors, if only people would buckle down and demand and pay for and use the same-diameter-glass lenses, with much lower F-stop-numbers, with those tiny imagers.

As a practical matter I don't think the reason for the lack of large-diameter lenses for tiny sensor cameras, is because nobody's thought of their ability to make small sensors more useful in low light.

Nice discussion topic though.

 

[Woland65]

RussellInCincinnati wrote:

It is indeed interesting to point out that small sensors don't have to be so bad in low light as we tend to think they are, maybe they could be within shouting distance of being as good as larger sensors, if only people would buckle down and demand and pay for and use the same-diameter-glass lenses, with much lower F-stop-numbers, with those tiny imagers.

As a practical matter I don't think the reason for the lack of large-diameter lenses for tiny sensor cameras, is because nobody's thought of their ability to make small sensors more useful in low light.

Nice discussion topic though.

The main reason for the lack of large lenses for small sensors is probably that for short focal lengths they are harder to make. For short focal lengths it is probably considerably cheaper to use a larger sensor than to make large lenses for a small sensor.

The second reason is probably that if you are willing to carry a large lens, then there is very little reason not to use a large sensor. The larger sensor makes it cheaper to make many of the lenses, and as an important bonus a larger sensor usually has a better dynamic range.

So I think cameras that are really good in low light have large sensors for good reasons, just not the reason that most people seem to believe in.

 

[oscarvdvelde]

RussellInCincinnati wrote:

maybe they could be within shouting distance of being as good as larger sensors, if only people would buckle down and demand and pay for and use the same-diameter-glass lenses, with much lower F-stop-numbers, with those tiny imagers.

As a practical matter I don't think the reason for the lack of large-diameter lenses for tiny sensor cameras, is because nobody's thought of their ability to make small sensors more useful in low light.

Lenses need to resolve increasingly small details to make sensors with high pixel density worth using. To do this while increasing their maximum aperture to f/1.0 or better makes them very expensive and large. At the same time people like small formats because such system is small compared to DSLRs. A full frame DSLR user would probably not object to the size, but might as well shoot cheap f/1.4 lenses on his system for the same result.

 

[Christof21]

bobn2 wrote:

Woland65 wrote:

As long as the image circle of a lens is manufactured to fit the sensor, a large sensor does not produce less low light noise. Low light noise is determined by the amount of light hitting the sensor. The amount of light hitting the sensor is determined by the lens aperture in millimeters. Sensor size does not matter.

I still don't understand why you bring the image circle into this. The image circle simply determines the coverage (i.e. vignetting). It does not affect the illuminance within the image circle.

Full frame cameras produce less low light noise because they use lenses with a larger aperture, aperture as measured in millimeters.

Needs specifying. They use larger apertures for the same angle of view and depth of field. Obviously, you can use the same size aperture on larger sensors, should you so choose.

However, larger sensor pixel size is good for dynamic range.

Not correct, dynamic range is at least in theory invariant under pixel scaling. In practice, there are so many design and process variables, one cannot talk about a universal law. You can see a trend in measurements of real sensors, which indicate DR increasing as pixel size reduces.

What is the conclusion of all the discussion about DR on this thread ?? DR seems to increase when sensor size (not pixel size) increases, right ? This is confirmed in dxomarks when comparing sensors.

This is according to me the only real advantage of FF over smaller sensors.

(Speed Booster has perhaps made it clearer to people that it is the lens, not the sensor that determines low light noise.)

Vast simplification. The QE of the sensor certainly is one of the determinants of noise.

--
Bob