ISO and signal to noise ratios

[Auf Reisen]

In the recent Sony interview here on DPReview, the author said

While this a7R II certainly did the job perfectly well capturing this high dynamic range scene on my recent trip to Iceland, a lower base ISO that would have allowed me to give the sensor even more exposure before clipping the highlights to the left, yielding an even cleaner image.

I am not sure I understand the logic behind that statement. Bear with me, I am thinking aloud here. If I understand correctly, a higher ISO means that, for any given amount of light that falls on a sensor pixel, this will be translated by the imaging software into a brighter image pixel in the image file. As far as I understand it, the usual linear relationship between ISO and signal/noise ratio is the wrong way to think about it. What we perceive as, say a uniform blue sky, is in reality the result of many different photons with different wavelength hitting the medium (film, sensor, eye), the average of which is perceived as 'blue'. The less light is gathered, the higher random outliers factor in, and the lower the signal to noise ratio.

BUT: signal/noise is a matter of how much light is gathered, not about how it is translated into brightness by the image processing software (well, it is, but this is secondary). So a picture taken at f/2.0 1/800 at ISO 200 has the same signal/noise ratio as taken at f/2.0 1/800 at ISO 400, with the latter just appearing brighter. How bright is bright enough of course depends on the scene.

So, for any given scene at a fixed brightness level, lowering the iso means I can afford to let more light in by stopping up or prolonging the exposure, which in turn means a better signal to noise ratio.

Is that about correct?

 

[AnthonyL]

In the recent Sony interview here on DPReview, the author said

While this a7R II certainly did the job perfectly well capturing this high dynamic range scene on my recent trip to Iceland, a lower base ISO that would have allowed me to give the sensor even more exposure before clipping the highlights to the left, yielding an even cleaner image.

I'm not sure how valid it is to argue against a camera's design base ISO which I assume is determined as being an optimum.

I am not sure I understand the logic behind that statement. Bear with me, I am thinking aloud here. If I understand correctly, a higher ISO means that, for any given amount of light that falls on a sensor pixel, this will be translated by the imaging software into a brighter image pixel in the image file. As far as I understand it, the usual linear relationship between ISO and signal/noise ratio is the wrong way to think about it. What we perceive as, say a uniform blue sky, is in reality the result of many different photons with different wavelength hitting the medium (film, sensor, eye), the average of which is perceived as 'blue'. The less light is gathered, the higher random outliers factor in, and the lower the signal to noise ratio.

BUT: signal/noise is a matter of how much light is gathered, not about how it is translated into brightness by the image processing software (well, it is, but this is secondary). So a picture taken at f/2.0 1/800 at ISO 200 has the same signal/noise ratio as taken at f/2.0 1/800 at ISO 400, with the latter just appearing brighter. How bright is bright enough of course depends on the scene.
So, for any given scene at a fixed brightness level, lowering the iso means I can afford to let more light in by stopping up or prolonging the exposure, which in turn means a better signal to noise ratio.
Is that about correct?

Well you've gone a long way around basically saying that increasing the ISO decreases the exposure (for the same brightness). Any underexposure will have a poor signal to noise ration.

There will be now 150 differing views many of which will ignore the "Beginners" in "Beginners Forum".

 

[FingerPainter]

With regard to shot noise, you've essentially got it right. The effect of ISO setting on camera-added noise on Sony cameras is slight, so you can afford to neglect it.

 

[NikonNature]

That's correct. Exposure is really the amount of light collected by the sensor. More light equals more signal, and therefore a better signal:noise ratio.

ISO is applied after the fact, when the exposure is written to the image file. Thus a low (under) exposure can be 'corrected' so the brightness matches a proper exposure.

More to your point: If you set your ISO to 1600 you will naturally take photos with shorter shutter speeds and/or smaller apertures. That will result in lower exposures and worse signal:noise ratios. Change your ISO to 100 and you will use longer shutter speeds and/or wider apertures, thereby collecting more light and have better signal:noise ratios. Of course in practice we have to make compromises. Sometimes we need a certain shutter speed to freeze action, or a certain aperture to get the depth of field we want. So while base ISO is desirable you can't always use it.

 

[FingerPainter]

<FingerPainter enters nitpicking mode*>

... Exposure is really the amount of light collected by the sensor.

Exposure is really the amount of light falling on the sensor per unit area.

The exposure multiplied by the surface area of the sensor determines the amount of light falling on the sensor.

The amount of light falling on the sensor multiplied by the sensor's quantum efficiency determines the amount of light captured by the sensor.

The amount of light captured by the sensor determines the shot noise in the image.

Since, for a given camera, sensor area and quantum efficiency are constants (ignoring built-in crop modes), the amount of light captured by the sensor is a constant function of exposure, but it is not the same as exposure.

More light equals more signal, and therefore a better signal:noise ratio.

ISO is applied after the fact, when the exposure is written to the image file. Thus a low (under) exposure can be 'corrected' so the brightness matches a proper exposure.

... but doing so does not improve the SNR.

More to your point: If you set your ISO to 1600 you will naturally take photos with shorter shutter speeds and/or smaller apertures. That will result in lower exposures and worse signal:noise ratios. Change your ISO to 100 and you will use longer shutter speeds and/or wider apertures, thereby collecting more light and have better signal:noise ratios. Of course in practice we have to make compromises. Sometimes we need a certain shutter speed to freeze action, or a certain aperture to get the depth of field we want. So while base ISO is desirable you can't always use it.

 

[csiluk]

In the recent Sony interview here on DPReview, the author said

While this a7R II certainly did the job perfectly well capturing this high dynamic range scene on my recent trip to Iceland, a lower base ISO that would have allowed me to give the sensor even more exposure before clipping the highlights to the left, yielding an even cleaner image.

I'm not sure how valid it is to argue against a camera's design base ISO which I assume is determined as being an optimum.

I am not sure I understand the logic behind that statement. Bear with me, I am thinking aloud here. If I understand correctly, a higher ISO means that, for any given amount of light that falls on a sensor pixel, this will be translated by the imaging software into a brighter image pixel in the image file. As far as I understand it, the usual linear relationship between ISO and signal/noise ratio is the wrong way to think about it. What we perceive as, say a uniform blue sky, is in reality the result of many different photons with different wavelength hitting the medium (film, sensor, eye), the average of which is perceived as 'blue'. The less light is gathered, the higher random outliers factor in, and the lower the signal to noise ratio.

BUT: signal/noise is a matter of how much light is gathered, not about how it is translated into brightness by the image processing software (well, it is, but this is secondary).So a picture taken at f/2.0 1/800 at ISO 200 has the same signal/noise ratio as taken at f/2.0 1/800 at ISO 400, with the latter just appearing brighter.

I would say this is incorrect. Signal/noise ratio is determined by the cameras electronics, but you can affect that ratio by the exposure. Your example of 2 different exposures will not have the same signal/noise ratio. If "optimum" exposure is f2 1/800 at ISO400 then f2 1/800 at ISO 200 will have a lower signal/noise ratio at the cost of under exposure and losing detail in the shadows. You can lift the shadows in post, but you also lift the noise along with the detail. The opposite will happen when reversed, if "optimum" exposure is f2 1/800 at ISO200, then f2 1/800 at ISO400 will increase the signal/noise but at the cost of over saturation/exposure, in other words, clipping. The electronics of the camera won't be able to handle this "over saturation" and highlights will be clipped and detail will be lost. Shooting raw will help in the recovery of blown highlights, but you risk the possibility of not being able to recover them.

You need to determine what is best for the scene and the "look" you are trying to achieve.

How bright is bright enough of course depends on the scene.
So, for any given scene at a fixed brightness level, lowering the iso means I can afford to let more light in by stopping up or prolonging the exposure, which in turn means a better signal to noise ratio.
Is that about correct?

Well you've gone a long way around basically saying that increasing the ISO decreases the exposure (for the same brightness). Any underexposure will have a poor signal to noise ration.

Exactly. This is why you want to use your histogram to ensure you get the highest possible signal/noise ratio. ETTR

There will be now 150 differing views many of which will ignore the "Beginners" in "Beginners Forum".

Exactly :-D

 

[Michael Fryd]

There are a number of issues here. Many are complicated by a general industry trend towards inconsistent use of terminology.

There are two general sources of noise in a digital image - the noise inherent on the light itself ("shot noise") and the noise added by the camera and processing.

With reasonable exposures and modern cameras, the shot noise is the biggest noise factor. Shot noise is dependent on the total light collected by the sensor. "Exposure" generally refers to "Light per unit area". For total light, we need to multiply that by sensor area. If we are always talking about the same sensor size, then we only need to know exposure in order to compare shot noise levels.

ISO enters into this in a number of ways. The most obvious is that ISO influences exposure. When you set your camera/meter to a low ISO, it suggests an exposure with a lot of light. If you set a high ISO, it suggests an exposure with a lower amount of light. Thus, if we are influenced by the meter, we will get lower exposures (more shot noise) with high ISO settings.

As it turns out, with some cameras, the ISO setting will affect the noise levels. Some cameras configure themselves differently at different ISO settings. If you set the camera to a high ISO, it may choose a configuration that it better tuned to a low light level.

If you are shooting raw, then ISO's biggest influence is the brightness of the camera generated image preview. When shooting raw, you have a lot of flexibility in post processing the file. If you separate the ISO setting from exposure, you may need to look very closely to find the difference between an image shot at ISO 100, and an image shot at ISO 800 and boosted 3 stops.

Of course, if you are shooting JPEG, the ISO setting is very important, as it drives the mapping between the light intensity captured and brightness in the JPEG. If the ISO doesn't match the exposure, the JPEG will be blown out, or too dark, and it may not be possible to recover the lost information.

Now let's look at the statement:

While this a7R II certainly did the job perfectly well capturing this high dynamic range scene on my recent trip to Iceland, a lower base ISO that would have allowed me to give the sensor even more exposure before clipping the highlights to the left, yielding an even cleaner image.

A "lower base ISO" would have influenced the photographer to pick an exposure with more light. This would have increased the signal to noise ratio of the capture, and produced an image with less "noise".

The mention of the lower ISO allowing more exposure without clipping suggests the photographer was shooting JPEG. If the exposure is increase without an appropriate decrease in ISO, the highlights in the camera-produced JPEG may very well get blown out.

As a general rule, you can minimize noise by shooting a the camera's base ISO, and selecting the maximum exposure that doesn't blow out any highlights you want to maintain. This is frequently called ETTR or Expose To The Right, as the histogram will appear shifted to the right.

On the flip side, modern cameras are so good, that while you might be able to measure a difference, you may not be able to see a difference in noise in a print from an ETTR exposure and a traditional exposure.

 

[FingerPainter]

In the recent Sony interview here on DPReview, the author said

... If I understand correctly, a higher ISO means that, for any given amount of light that falls on a sensor pixel, this will be translated by the imaging software into a brighter image pixel in the image file. As far as I understand it, the usual linear relationship between ISO and signal/noise ratio is the wrong way to think about it. What we perceive as, say a uniform blue sky, is in reality the result of many different photons with different wavelength hitting the medium (film, sensor, eye), the average of which is perceived as 'blue'. The less light is gathered, the higher random outliers factor in, and the lower the signal to noise ratio.

BUT: signal/noise is a matter of how much light is gathered, not about how it is translated into brightness by the image processing software (well, it is, but this is secondary).So a picture taken at f/2.0 1/800 at ISO 200 has the same signal/noise ratio as taken at f/2.0 1/800 at ISO 400, with the latter just appearing brighter.

I would say this is incorrect. Signal/noise ratio is determined by the cameras electronics, but you can affect that ratio by the exposure.

That's not entirely correct.

Noise is the visible manifestation of variation in signal values from the ideal, correct or average signal value. There are two sources of variation: variation added by imperfections in the camera's electronics, and variation already present in the light itself even before that light is captured. This variation in light is a natural consequence of the quantum nature of the generation of light by the process of photon emission from atoms reacting to an increase in energy state.

This latter form of variation is called shot noise. In most parts of a well exposed image taken by a modern digital camera, the shot noise will be far more significant than the variation added by the camera. So while it is true that the SNR of the image is affected by the camera's electronics, it is far from true that it is determined by those electronics. For most well-exposed images, the impact on image SNR of camera-added noise is negligible compared to the impact of a one stop change in exposure.

Your example of 2 different exposures will not have the same signal/noise ratio.

Probably not exactly the same, but presuming the f/2 1/800 ISO 400 settings resulted in an image that was not dominated by totally dark areas, the one stop change from ISO 200 to ISO 400 will result in a change in image SNR much less than what would have occurred with a 1/3 stop increase in aperture diameter or exposure time. The difference in SNR would likely be undetectable by visual observation, and being less than 1/6 stop, would be negligible.

The SNR WRT shot noise would be exactly the same between the two images, and the camera-added noise, which would be a very small part of the total noise, would change by a very small amount. For most well exposed images, when there is no change in shot nose, there is no effective change in total noise.

If "optimum" exposure is f2 1/800 at ISO400 then f2 1/800 at ISO 200 will have a lower signal/noise ratio

Yes, but negligibly so, On Sony cameras, a one stop change in ISO does not have the same order of magnitude of effect on image SNR as a 1 stop change in exposure.

at the cost of under exposure

Exposure hasn't changed, so there will be no more underexposure. ISO changes brightness, not exposure.

and losing detail in the shadows.

There is a small chance one might be able to detect a change in noise in the darkest shadows, because in those areas the signal is so low than the camera-added noise becomes locally significant.

You can lift the shadows in post, but you also lift the noise along with the detail.

Yes, and that also happens when you raise ISO (using gain), to all the noise except that small bit of noise which is added after the gain stage. In Sony cameras, the portion of noise added after the gain stage is trivially small. When the camera impets an IS chag e digitally, the signal and noise are both increased by the saem factor, so there is no change in SNR. Digital implementation of ISO increases can result in greater quantization error than increases implemented by gain.

The opposite will happen when reversed, if "optimum" exposure is f2 1/800 at ISO200, then f2 1/800 at ISO400 will increase the signal/noise

Yes, but again by a very small amount, not commensurate with the change in SNR caused by the same degree of change in exposure.

but at the cost of over saturation/exposure,

Again, exposure hasn't changed, but any pixels that were within a stop of maximum value at ISO 200 will be blown at ISO 400.

in other words, clipping. The electronics of the camera won't be able to handle this "over saturation" and highlights will be clipped and detail will be lost. Shooting raw will help in the recovery of blown highlights,

Not truly blown highlights, but rather only those highlights whose values have been increased beyond the maximum possible value in JPEG but not to the maximum possible value in RAW.

but you risk the possibility of not being able to recover them.

You need to determine what is best for the scene and the "look" you are trying to achieve.

 

[NikonNature]

In the recent Sony interview here on DPReview, the author said

While this a7R II certainly did the job perfectly well capturing this high dynamic range scene on my recent trip to Iceland, a lower base ISO that would have allowed me to give the sensor even more exposure before clipping the highlights to the left, yielding an even cleaner image.

I'm not sure how valid it is to argue against a camera's design base ISO which I assume is determined as being an optimum.

I am not sure I understand the logic behind that statement. Bear with me, I am thinking aloud here. If I understand correctly, a higher ISO means that, for any given amount of light that falls on a sensor pixel, this will be translated by the imaging software into a brighter image pixel in the image file. As far as I understand it, the usual linear relationship between ISO and signal/noise ratio is the wrong way to think about it. What we perceive as, say a uniform blue sky, is in reality the result of many different photons with different wavelength hitting the medium (film, sensor, eye), the average of which is perceived as 'blue'. The less light is gathered, the higher random outliers factor in, and the lower the signal to noise ratio.

BUT: signal/noise is a matter of how much light is gathered, not about how it is translated into brightness by the image processing software (well, it is, but this is secondary). So a picture taken at f/2.0 1/800 at ISO 200 has the same signal/noise ratio as taken at f/2.0 1/800 at ISO 400, with the latter just appearing brighter. How bright is bright enough of course depends on the scene.
So, for any given scene at a fixed brightness level, lowering the iso means I can afford to let more light in by stopping up or prolonging the exposure, which in turn means a better signal to noise ratio.
Is that about correct?

Well you've gone a long way around basically saying that increasing the ISO decreases the exposure (for the same brightness). Any underexposure will have a poor signal to noise ration.

There will be now 150 differing views many of which will ignore the "Beginners" in "Beginners Forum".

Looks like you were right.

The science and technology is interesting, but it is also a very deep rabbit hole. I appreciate and respect those who understand it much better than I do, but for a beginners forum a simplified, practical answer would be more appropriate.

 

[csiluk]

While I respect your answers, any amount of under exposure from "optimum" is going to result in a lower S/N ratio. The more it's under exposed, the lower that S/N ratio is going to be. Even on a Sony.

ISO is still part of the exposure triangle, and aperture, shutter speed or ISO can be used to change brightness.

 

[Porky89]

In the recent Sony interview here on DPReview, the author said

While this a7R II certainly did the job perfectly well capturing this high dynamic range scene on my recent trip to Iceland, a lower base ISO that would have allowed me to give the sensor even more exposure before clipping the highlights to the left, yielding an even cleaner image.

I am not sure I understand the logic behind that statement. Bear with me, I am thinking aloud here. If I understand correctly, a higher ISO means that, for any given amount of light that falls on a sensor pixel, this will be translated by the imaging software into a brighter image pixel in the image file. As far as I understand it, the usual linear relationship between ISO and signal/noise ratio is the wrong way to think about it. What we perceive as, say a uniform blue sky, is in reality the result of many different photons with different wavelength hitting the medium (film, sensor, eye), the average of which is perceived as 'blue'. The less light is gathered, the higher random outliers factor in, and the lower the signal to noise ratio.

BUT: signal/noise is a matter of how much light is gathered, not about how it is translated into brightness by the image processing software (well, it is, but this is secondary). So a picture taken at f/2.0 1/800 at ISO 200 has the same signal/noise ratio as taken at f/2.0 1/800 at ISO 400, with the latter just appearing brighter. How bright is bright enough of course depends on the scene.
So, for any given scene at a fixed brightness level, lowering the iso means I can afford to let more light in by stopping up or prolonging the exposure, which in turn means a better signal to noise ratio.
Is that about correct?

Or to put it the old fashioned way: high sensitivity (ISO) settings produce worse signal to noise ratios and underexposing also produces a worse signal to noise ratio.

Now wait for the plethora of undergraduate theses saying much the same thing in several thousand words complete with theoretical explanations about Total Light and other such nonsense intended to make you think it is a fiendishly complicated subject and the author is an expert.

 

[Michael Fryd]

Any amount of under exposure from "optimum" is going to result in a lower S/N ratio. The more it's under exposed, the lower that S/N ratio is going to be. Even on a Sony.

ISO is still part of the exposure triangle, and aperture, shutter speed or ISO can be used to change brightness.

I suspect you will see some pushback here.

Exposure and print brightness are not the same thing. While ISO can affect brightness, it does not directly affect exposure.

Loosely speaking, "Exposure" is the light that falls on the sensor. "Brightness" is an attribute of the print/image we are looking at. A simple example of the difference is the brightness control on your monitor. It can make an image look too bright or too dark, but has no affect on "Exposure".

In the film days, "Exposure" had to match the ISO. If they didn't match, you lost highlights or shadow. ISO was a property of the film, and therefore you needed the exposure to match the film that was loaded in the camera. If the exposure was too low for the film, we said the image was "underexposed". With typical processing, this resulted in a dark print, but one could lighten it up when printing.

In the digital world, Exposure and ISO may not need to match at all. Many cameras fall into a category called "ISO invariant". This means the camera actually ignores the ISO setting when recording the image data. If you keep the f/stop and shutter speed the same, you get the same data whether the camera was set to ISO 100 or ISO 1600. ISO doesn't enter it until you process the RAW file to create the JPEG. If you are shooting raw, this might happen the day after the shots were taken.

In other words, with many modern cameras, ISO doesn't directly affect the image capture at all. It is purely a part of post processing the captured raw data.

If you have one of these cameras, it doesn't make much sense to include "ISO" in an "exposure triangle", as ISO has no affect on Exposure nor on the recorded data.

Calling it an "Exposure Triangle" can confuse beginners because the name incorrectly suggests that ISO is a direct factor in exposure.

If you want an exposure "triangle", a better choice would be aperture, shutter speed and light level. Altering any of these three will actually affect exposure.

Disclaimer - A lot of cameras are not strictly ISO Invariant, although they are close. Setting the ISO might change some of the internal optimizations.

 

[Porky89]

That's correct. Exposure is really the amount of light collected by the sensor. More light equals more signal, and therefore a better signal:noise ratio.

ISO is applied after the fact, when the exposure is written to the image file. Thus a low (under) exposure can be 'corrected' so the brightness matches a proper exposure.

More to your point: If you set your ISO to 1600 you will naturally take photos with shorter shutter speeds and/or smaller apertures. That will result in lower exposures and worse signal:noise ratios. Change your ISO to 100 and you will use longer shutter speeds and/or wider apertures, thereby collecting more light and have better signal:noise ratios. Of course in practice we have to make compromises. Sometimes we need a certain shutter speed to freeze action, or a certain aperture to get the depth of field we want. So while base ISO is desirable you can't always use it.

 

[Porky89]

Oh come on! Just what is your agenda?

 

[Michael Fryd]

Or to put it the old fashioned way: high sensitivity (ISO) settings produce worse signal to noise ratios and underexposing also produces a worse signal to noise ratio.

Except that's not how it works.

Generally the ISO setting has very little affect on noise.

At the same exposure. An ISO 100 capture boosted 3 stops will have about the same noise as an ISO 800 image.

The reason high ISO images are typically so noisy is that photographers generally use high ISO with low exposures.
Put your camera in manual and take a properly exposed shot at ISO 100. Keep the shutter speed and aperture the same, and reshoot at ISO 800. Open the ISO 800 shot and bring the levels down 3 stops. It will have about the same noise as the ISO 100 image.

The noise is from the low exposure, not the high ISO.

 

[Michael Fryd]

Oh come on! Just what is your agenda?

Helping beginners understand that noise is from the low exposure, not the high ISO.

When beginners think the noise is from high ISO, they are tempted to improve the image by turning down the ISO and underexposing. If the noise was actually from the high ISO, this would improve noise.

 

[Porky89]

This is going to be a good one.

 

[Porky89]

Oh come on! Just what is your agenda?

Helping beginners understand that noise is from the low exposure, not the high ISO.

When beginners think the noise is from high ISO, they are tempted to improve the image by turning down the ISO and underexposing. If the noise was actually from the high ISO, this would improve noise.

Turning down the sensitivity, or ISO as you call it, does invariably improve the image. And shows just what total nonsense you are spouting in order to make yourself look smart. Sensitivity (ISO) is part of the exposure triangle and if you adjust the other two parameters to suit then the exposure does not change. Reducing sensitivity and failing to adjust the other two parameters would indeed result in underexposure and reduced image quality. But what idiot is going to do that?

By the way, referring to the sensitivity setting as "the ISO" is like referring to the aperture setting as "the F" or referring to the shutter speed as "the seconds".

 

[Porky89]

Or to put it the old fashioned way: high sensitivity (ISO) settings produce worse signal to noise ratios and underexposing also produces a worse signal to noise ratio.

Except that's not how it works.

Generally the ISO setting has very little affect on noise.

At the same exposure. An ISO 100 capture boosted 3 stops will have about the same noise as an ISO 800 image.

The reason high ISO images are typically so noisy is that photographers generally use high ISO with low exposures.

Put your camera in manual and take a properly exposed shot at ISO 100. Keep the shutter speed and aperture the same, and reshoot at ISO 800. Open the ISO 800 shot and bring the levels down 3 stops. It will have about the same noise as the ISO 100 image.

The noise is from the low exposure, not the high ISO.

Just as with the nonsensical Total Light theory, you are confusing cause and effect. And you are certainly not helping the OP with this rubbish.

Now I really must get out of here while I still have my sanity. But before I go, I will leave you with this false syllogism:

All men are mortal.
Plato is a man.
Plato is mortal.
Therefore all men are Plato.

 

[csiluk]

Oh come on! Just what is your agenda?

Helping beginners understand that noise is from the low exposure, not the high ISO.

When beginners think the noise is from high ISO, they are tempted to improve the image by turning down the ISO and underexposing. If the noise was actually from the high ISO, this would improve noise.

Either one, under exposing or high ISO will increase noise. Do both together, well don't, unless you have no choice.