What affects the "Dynamic range"?

[Alireza S]

Hi,

I'm interested to know:

1. What is the Dynamic range? and...

2. What really affects the Dynamic range? (ISO or any other things...?)

Thx,

Alireza

 

[John Sheehy]

Black is the lowest sensor readout where minimal necessary resolution is still maintained. It is not a fixed value.

Black is where there is no photon signal at all, and nothing to resolve. Perhaps you are thinking of the lowest usable signal level that shapes a particular measure of DR.

That brings me back to the OP's question as to what affects the DR, and one part of the answer is that your definition of DR affects it, if you're looking for a number. If you're looking at the empirical relationship between signal and noise without drawing a line in the sand for lowest usable signal, only the hardware and environment (heat, RF interference, etc) affect it.

You meant to say photon shot noise. Electronic noise also include shot noise component.

What would it be modulating? Are you referring to the photosite thermal noise, as having discrete electron charges as if they were from imaginary photons?

 

[John Sheehy]

To be sure we are talking in same terms, when I say that clipping occurs I mean that the brightest part of a scene that I want to retain detail is clipped by the camera.

On the bright side, the noise in the clipped highlights is zero.

Except when they first start to clip - only the positive outliers of noise first clip, and as you raise the exposure, eventually all the negative outliers clip as well. This is a very narrow band for low ISOs and/or big photosites dominated by photon noise, but for very small photosites and/or high ISOs, this can conceivably extend highlight headroom, much in the way that it is extended with negative films.

 

[Xasan]

Black is the lowest sensor readout where minimal necessary resolution is still maintained. It is not a fixed value.

Black is where there is no photon signal at all

He was talking about something he calls photographic dynamic range.

You meant to say photon shot noise. Electronic noise also include shot noise component.

What would it be modulating?

Dark current contains shot noise.

 

[John Sheehy]

Black is the lowest sensor readout where minimal necessary resolution is still maintained. It is not a fixed value.

Black is where there is no photon signal at all

He was talking about something he calls photographic dynamic range.

You defined black, and defined it wrong. Black is not the bottom end of typical quantified DR. The bottom end of DR is the standard deviation of blackframe noise (engineering definition), or the signal level at which SNR is some arbitrary value, like 1, or 3, or 20. Black is always below that level.

You meant to say photon shot noise. Electronic noise also include shot noise component.

What would it be modulating?

Dark current contains shot noise.

You clipped away the part where I suggest that those fake photons are what you may have meant. In any event, very few people think of that when they think of shot noise. It is an additive noise completely independent of signal.

 

[Xasan]

Black is the lowest sensor readout where minimal necessary resolution is still maintained. It is not a fixed value.

Black is where there is no photon signal at all

He was talking about something he calls photographic dynamic range.

You defined black, and defined it wrong.

Gosh. You are not reading, and taking out of context. Whatever.

 

[The Ghost of Caravaggio]

You meant to say photon shot noise. Electronic noise also include shot noise component.

What would it be modulating? Are you referring to the photosite thermal noise, as having discrete electron charges as if they were from imaginary photons

In incoherent, quantum mechanically open systems, such as digital imaging, quantum noise associated with electrons in a camera's detection and data stream circuitry is alway smaller than quantum noise associated with light amplitudes (usually called photon counts in this Forum).

With strong light levels the "modulation" models would be identical (Gaussian) because both involve large numbers of events. In very low light levels a Possion distribution could apply to photon quantum noise.

Quantum noise contributions from the electrons in the photo-diode array, DC signal routing and amplification stages and the ADC are important. As Johnson (thermal noise) and other uncorrelated electronic noise sources decrease, data stream quantum noise becomes more important. An ideal data stream would only have electron quantum noise.

Quantum noise contributions from the light are fundamentally different and often dominate the uncertainties for imaging raw data numbers.

While an ideal data stream's uncorrelated noise would be determined by electron quantum noise, the quantum noise from the light will be greater.

Details follow.

A wave function's symmetry with regard to its spatial coordinates is related to its quantum noise levels. Wave functions for photons are spatially symmetric. The wave functions for electrons are spatially asymmetric. Useful models for the former are Bose statistics and Fermi statistics for the latter. It turns out particles with symmetrical wave functions bunch together while particles with asymmetric wave functions do not. The bunching increases the quantum noise compared to un-bunched particles.

 

[beatboxa]

Black is the lowest sensor readout where minimal necessary resolution is still maintained. It is not a fixed value.

Black is where there is no photon signal at all

He was talking about something he calls photographic dynamic range.

You defined black, and defined it wrong.

Gosh. You are not reading, and taking out of context. Whatever.

Correct. I was not talking about "black" in the 'rgb = 0' context. I was referring to black as the noise floor in the context I defined.

 

[John Sheehy]

Correct. I was not talking about "black" in the 'rgb = 0' context.

Neither was I. Black in not a solid number, unless it is overly quantized, and it is actually a small tragedy if the mean is zero, with half the read noise swing clipped away.

I was referring to black as the noise floor in the context I defined.

The noise floor is not equal to black.

I don't know why anyone would want to confuse the noise floor and black.

Why are you people fighting so hard to defend a false and confusing statement?

 

[alanr0]

You meant to say photon shot noise. Electronic noise also include shot noise component.

What would it be modulating? Are you referring to the photosite thermal noise, as having discrete electron charges as if they were from imaginary photons

In incoherent, quantum mechanically open systems, such as digital imaging, quantum noise associated with electrons in a camera's detection and data stream circuitry is alway smaller than quantum noise associated with light amplitudes (usually called photon counts in this Forum).

With strong light levels the "modulation" models would be identical (Gaussian) because both involve large numbers of events. In very low light levels a Possion distribution could apply to photon quantum noise.

Quantum noise contributions from the electrons in the photo-diode array, DC signal routing and amplification stages and the ADC are important. As Johnson (thermal noise) and other uncorrelated electronic noise sources decrease, data stream quantum noise becomes more important. An ideal data stream would only have electron quantum noise.

Quantum noise contributions from the light are fundamentally different and often dominate the uncertainties for imaging raw data numbers.

While an ideal data stream's uncorrelated noise would be determined by electron quantum noise, the quantum noise from the light will be greater.

An important point which many (including myself) often overlook.

The conventional shot noise formula applies when the current is generated via random events at a potential barrier such as a PN junction, or by thermionic emission in a vacuum tube. For diffusive flow in a metallic wire, coulomb interactions reduce the fluctuations (noise power) by a factor 3, as described here: Wikipedia - shot noise.

A more clear-cut example is an "electron pump" using a single electron tunnelling (SET) device, in which electrons are "clocked out" one by one, and the current depends almost entirely on the frequency of the clock waveform driving the device. Fluctuations are orders of magnitude smaller than the classical Poisson noise formula.

Here is a paper from NIST where an SET is used to calibrate a capacitor. Current is rather low at 10^8 electrons per second or roughly 1.6 pA, but uncertainty per cycle is 1 part in a billion.

Here is another application, where currents of 100 pA were generated as part of an international project to re-define the Ampere in terms of fundamental physical constants.

 

[alanr0]

Black is the lowest sensor readout where minimal necessary resolution is still maintained. It is not a fixed value.

Black is where there is no photon signal at all

<snip>

You meant to say photon shot noise. Electronic noise also include shot noise component.

What would it be modulating?

Dark current contains shot noise.

You clipped away the part where I suggest that those fake photons are what you may have meant. In any event, very few people think of that when they think of shot noise. It is an additive noise completely independent of signal.

Not only that, but there is a good chance it won't follow Poisson statistics. Leakage currents often exhibit substantial 1/f or flicker noise.

 

[J A C S]

Black is the lowest sensor readout where minimal necessary resolution is still maintained. It is not a fixed value.

Black is where there is no photon signal at all

He was talking about something he calls photographic dynamic range.

You defined black, and defined it wrong.

Gosh. You are not reading, and taking out of context. Whatever.

Correct. I was not talking about "black" in the 'rgb = 0' context. I was referring to black as the noise floor in the context I defined.

Black, in the context of DR, should be black incoming signal, not whatever the sensor outputs. At the bottom, the sensors usually have noticeable non-linearity, channel dependent, and strong non-uniformity so judging what black might be by that output is very questionable by those reasons, and the ones mentioned by others.

 

[beatboxa]

Correct. I was not talking about "black" in the 'rgb = 0' context.

Neither was I. Black in not a solid number, unless it is overly quantized, and it is actually a small tragedy if the mean is zero, with half the read noise swing clipped away.

I was referring to black as the noise floor in the context I defined.

The noise floor is not equal to black.

I don't know why anyone would want to confuse the noise floor and black.

Why are you people fighting so hard to defend a false and confusing statement?

 

[beatboxa]

Black is the lowest sensor readout where minimal necessary resolution is still maintained. It is not a fixed value.

Black is where there is no photon signal at all

He was talking about something he calls photographic dynamic range.

You defined black, and defined it wrong.

Gosh. You are not reading, and taking out of context. Whatever.

Correct. I was not talking about "black" in the 'rgb = 0' context. I was referring to black as the noise floor in the context I defined.

Black, in the context of DR, should be black incoming signal, not whatever the sensor outputs. At the bottom, the sensors usually have noticeable non-linearity, channel dependent, and strong non-uniformity so judging what black might be by that output is very questionable by those reasons, and the ones mentioned by others.

In which context? Are you saying that 'black' does not exist in any rendered photo and is irrelevant to a sensor's dynamic range?

Because a sensor's dynamic range is highly dependent on what the sensor outputs. That's the context of the conversation here.

Ignoring the sensor's output is ignoring the OP's question and the context of the answers.

 

[J A C S]

Black is the lowest sensor readout where minimal necessary resolution is still maintained. It is not a fixed value.

Black is where there is no photon signal at all

He was talking about something he calls photographic dynamic range.

You defined black, and defined it wrong.

Gosh. You are not reading, and taking out of context. Whatever.

Correct. I was not talking about "black" in the 'rgb = 0' context. I was referring to black as the noise floor in the context I defined.

Black, in the context of DR, should be black incoming signal, not whatever the sensor outputs. At the bottom, the sensors usually have noticeable non-linearity, channel dependent, and strong non-uniformity so judging what black might be by that output is very questionable by those reasons, and the ones mentioned by others.

In which context? Are you saying that 'black' does not exist in any rendered photo and is irrelevant to a sensor's dynamic range?

The RAW file is not a photo, it is data for the scene. Black in the RAW file makes sense only related to black input, and the latter makes good sense, no photons.

Because a sensor's dynamic range is highly dependent on what the sensor outputs. That's the context of the conversation here.

Yes, but I would not call it black. People call in noise floor, etc. Also, 1:1 SNR is somewhat arbitrary but convenient.

Ignoring the sensor's output is ignoring the OP's question and the context of the answers.

That is why I am not doing it.

 

[Great Bustard]

Hi,

What affects the "Dynamic range"? I'm interested to know:

1. What is the Dynamic range? and...

DR (Dynamic Range) is the range of light levels where detail can be recorded by a photo. The low end is where detail is drowned out by noise and the high end where detail is blown out by oversaturation.

2. What really affects the Dynamic range? (ISO or any other things...?)

As above, noise on the low end and how much light the sensor can absorb before oversaturating on the high end.

The ISO setting matters only inasmuch as higher ISO settings result in less electronic noise than lower ISO settings (how much less depends *tremendously* upon the particular sensor and the particular ISO settings). Thus, for example, a photo of a given scene at f/2.8 1/200 ISO 1600 will have less noise than a photo of the same scene at f/2.8 1/200 ISO 100, and thus greater DR.

However, when using ISO 1600, you typically use 1/16 the exposure time that you would at ISO 100 for a given relative aperture, and that lesser amount of light increases the noise *far more* than the lesser electronic noise lessens the noise.

For example, f/2.8 1/3200 ISO 1600 is far more noisy than f/2.8 1/200 ISO 100 (and, consequently, less DR). This is not because of the ISO setting, however -- it's because of the exposure time: 1/3200 puts 1/16 as much light on the sensor as 1/200, all else equal, and it is the lesser amount of light that makes the photo more noisy, and thus less DR.

Cameras with larger sensors and/or lower "real" base ISO settings, however, can absorb more light before oversaturating, and this, in combination with a correspondingly longer exposure time, will increase the DR. However, this DR is not always as much larger as one might expect because the noise advantage is eaten into by increase electronic noise (however, the greater amount of light recorded matters more than the increase electronic noise).

Thx,

Alireza

I hope this helps!

 

[beatboxa]

Black, in the context of DR, should be black incoming signal, not whatever the sensor outputs. At the bottom, the sensors usually have noticeable non-linearity, channel dependent, and strong non-uniformity so judging what black might be by that output is very questionable by those reasons, and the ones mentioned by others.

In which context? Are you saying that 'black' does not exist in any rendered photo and is irrelevant to a sensor's dynamic range?

The RAW file is not a photo, it is data for the scene. Black in the RAW file makes sense only related to black input, and the latter makes good sense, no photons.

First, you're changing the scope here--nobody said anything about raw files vs anything else. The context of the 'black' definition was actually specifically for "photographic dynamic range," so if you don't consider raw to be a photo, then it's out of scope for 'photographic dynamic range." Simple as that.

But beyond that, you're not correct about raw files. The raw file is just as much a photo as a JPEG is--the difference is only in quantity, format, and fields of data. And DR is essentially the difference between 'black' and 'white'--in practical terms of a photo, this is the difference between the noise floor & full saturation.

A raw file is not synonymous with raw data.

Because a sensor's dynamic range is highly dependent on what the sensor outputs. That's the context of the conversation here.

Yes, but I would not call it black. People call in noise floor, etc. Also, 1:1 SNR is somewhat arbitrary but convenient.

What do you call black in a photo? Because that's the context that this black definition was found in: "Photographic Dynamic Range." Scroll back up to the post in case you missed it.

Ignoring the sensor's output is ignoring the OP's question and the context of the answers.

That is why I am not doing it.

Yes, you are ignoring the sensor's output. Highlighted & underlined above.

 

[J A C S]

Black, in the context of DR, should be black incoming signal, not whatever the sensor outputs. At the bottom, the sensors usually have noticeable non-linearity, channel dependent, and strong non-uniformity so judging what black might be by that output is very questionable by those reasons, and the ones mentioned by others.

In which context? Are you saying that 'black' does not exist in any rendered photo and is irrelevant to a sensor's dynamic range?

The RAW file is not a photo, it is data for the scene. Black in the RAW file makes sense only related to black input, and the latter makes good sense, no photons.

First, you're changing the scope here--nobody said anything about raw files vs anything else. The context of the 'black' definition was actually specifically for "photographic dynamic range," so if you don't consider raw to be a photo, then it's out of scope for 'photographic dynamic range." Simple as that.

Not that simple because photographic DR is much a more arbitrary concept than black.

But beyond that, you're not correct about raw files. The raw file is just as much a photo as a JPEG is--the difference is only in quantity, format, and fields of data. And DR is essentially the difference between 'black' and 'white'--in practical terms of a photo, this is the difference between the noise floor & full saturation.

No, it is not. DR is a range in the first place, not a number. In digital photography, it is usually defined as ... you know what... but "black" does not appear in the definition. For film however, this definition does not work but range still makes sense.

Now, if you want to define black as the noise floor, OK, but you have to realize that you are in the tiny minority of 1 person here.

A raw file is not synonymous with raw data.

Because a sensor's dynamic range is highly dependent on what the sensor outputs. That's the context of the conversation here.

Yes, but I would not call it black. People call in noise floor, etc. Also, 1:1 SNR is somewhat arbitrary but convenient.

What do you call black in a photo? Because that's the context that this black definition was found in: "Photographic Dynamic Range." Scroll back up to the post in case you missed it.

The definition I have seen takes a higher SNR as acceptable as the noise floor level. I have not heard of anybody calling that black but now, I have one person.

Not everything is black and white but black pretty much is.

 

[Joofa]

Correct. I was not talking about "black" in the 'rgb = 0' context.

Neither was I. Black in not a solid number, unless it is overly quantized, and it is actually a small tragedy if the mean is zero, with half the read noise swing clipped away.

I was referring to black as the noise floor in the context I defined.

The noise floor is not equal to black.

I don't know why anyone would want to confuse the noise floor and black.

Why are you people fighting so hard to defend a false and confusing statement?

 

[Joofa]

Hi,

What affects the "Dynamic range"? I'm interested to know:

1. What is the Dynamic range? and...

DR (Dynamic Range) is the range of light levels where detail can be recorded by a photo.

Not a range, per se. Rather a number. The number of levels a system can support.

 

[beatboxa]

Correct. I was not talking about "black" in the 'rgb = 0' context.

Neither was I. Black in not a solid number, unless it is overly quantized, and it is actually a small tragedy if the mean is zero, with half the read noise swing clipped away.

I was referring to black as the noise floor in the context I defined.

The noise floor is not equal to black.

I don't know why anyone would want to confuse the noise floor and black.

Why are you people fighting so hard to defend a false and confusing statement?