DOF questions

[Mark H]

. . .

I won't even bother reading the rest of your pedantic diatribe because it is obvious that you are just trying to bust Bill's chops at every opportunity.

Anyone that bothers to read his replies is aware of his many vendettas that occupy so much of his time. Strange here though, that he'd revive a thread dormant for a week to vent a little more from his spleen. It's sad that he expends far more time and energy on attacking than helping people.

Complete and utter 'BS'.

Bottom line is this 'Billx08'...

There are people here on these forums who are actually interested in technical discussions/issues, avoiding misinformation, and value the importance of actually getting facts straight.

On the other hand - there are people like yourself...

With barely an original thought in your head, and for the most part plagiarising and cut-n-pasting voluminously from other peoples work, and others' web-pages.

You are engrossed in infantile forum tittle-tattle, and preoccupied with siding up to third parties with desperate attempts at character assassinations.

And when the going gets tough - out comes the only response left to you - cartoons, and 'youtube' clips.

People can judge clearly for themselves.

 

[Billx08]

It's more complicated than this, because the tone curve applied to the sensor data by the camera determines the dynamic range , and theoretically, that sensor whose photosites can only capture 4 photons could have a 10 stop dynamic range...

That is a hyper-pedantic approach if I ever saw one ... Bill is talking about DR the way DPReview itself talks about it. The well size more or less determines the input DR (call it "potential") and the tone curve more or less determines the output DR. It's really that simple.

'Billx08' has not at any point referred to DPReview's JPEG DR tests - which in any case would be irrelevant to the OP's question.

This thread is about 'painterdude's question as to why "...smaller sensors struggle with dynamic range issues... [etc, etc]" .

In the context of 'painterdude's question - 'Billx08's reply that the camera's 'tone curve determines the dynamic range is/was complete nonsense (not to mention his ridiculous claim that "...photosites [that] can only capture 4 photons could have a 10 stop dynamic range" ).

You're the one that is misguided and going on an all out attack to deceive by selectively choosing out of context quotes an ignoring earlier statements such as this :

Are they simply not picking up as much light? Could this be changed if they had fewer pixels..or longer exposures chosen ?

It's the related problems of not capturing enough photons, both by the photosites and by the entire sensor. Imagine an extremely small photosite, only large enough to capture 4 photons at most. Not much of a dynamic range there. Compare that with some photosites that can record the accumulation of more than 50,000 photons. It's more complicated than this, because the tone curve applied to the sensor data by the camera determines the dynamic range, and theoretically, that sensor whose photosites can only capture 4 photons could have a 10 stop dynamic range, but it would be a thoroughly useless dynamic range because it would be able to display such a tiny number of different tones. This would be great for posterization lovers, but it would be pretty useless for general photography unless you're into solarization.

You should be quite aware that tone curves are used to change the displayed dynamic range, but as I noted, even though it can sometimes be effective, it doesn't represent the true dynamic range of the sensor and it can lead to other problems. This is yet another example of how you nitpick, sometimes (as here) grossly and apparently intentionally misunderstanding the nits that you choose to slay. Your vendetta, as always, is running on fumes.

 

[Mark H]

It's more complicated than this, because the tone curve applied to the sensor data by the camera determines the dynamic range , and theoretically, that sensor whose photosites can only capture 4 photons could have a 10 stop dynamic range...

That is a hyper-pedantic approach if I ever saw one ... Bill is talking about DR the way DPReview itself talks about it. The well size more or less determines the input DR (call it "potential") and the tone curve more or less determines the output DR. It's really that simple.

'Billx08' has not at any point referred to DPReview's JPEG DR tests - which in any case would be irrelevant to the OP's question.

This thread is about 'painterdude's question as to why "...smaller sensors struggle with dynamic range issues... [etc, etc]" .

In the context of 'painterdude's question - 'Billx08's reply that the camera's 'tone curve determines the dynamic range is/was complete nonsense (not to mention his ridiculous claim that "...photosites [that] can only capture 4 photons could have a 10 stop dynamic range" ).

You're the one that is misguided and going on an all out attack to deceive by selectively choosing out of context quotes an ignoring earlier statements such as this :

It is entirely within context.

It is also inevitably 'selective' - that is the whole purpose - to specifically address the particularly misguided and bogus assertions that you so regularly make.

The fact that you regard proper discussion of facts as some kind of "all out attack" is really quite childish in the extreme, and speaks volumes of your personal immaturity and insecurity.

Are they simply not picking up as much light? Could this be changed if they had fewer pixels..or longer exposures chosen ?

It's the related problems of not capturing enough photons, both by the photosites and by the entire sensor. Imagine an extremely small photosite, only large enough to capture 4 photons at most. Not much of a dynamic range there. Compare that with some photosites that can record the accumulation of more than 50,000 photons. It's more complicated than this, because the tone curve applied to the sensor data by the camera determines the dynamic range, and theoretically, that sensor whose photosites can only capture 4 photons could have a 10 stop dynamic range, but it would be a thoroughly useless dynamic range because it would be able to display such a tiny number of different tones. This would be great for posterization lovers, but it would be pretty useless for general photography unless you're into solarization [posterisation].

You should be quite aware that tone curves are used to change the displayed dynamic range, but as I noted, even though it can sometimes be effective, it doesn't represent the true dynamic range of the sensor and it can lead to other problems.

I know exactly what 'tone curves' are used for, and have known for many, many, years - you however, very clearly, do not.

We also await your theory, of how a '4 photon capacity photo-site' could somehow 'have 10 stops of dynamic range' - it really should be quite fascinating (but I'd advise anyone not to hold their breath whilst waiting).

This is yet another example of how you nitpick, sometimes (as here) grossly and apparently intentionally misunderstanding the nits that you choose to slay. Your vendetta, as always, is running on fumes.

Quit your childish, and futile, accusations of "vendettas" - grow up - and maybe just for once, try to address the facts.

 

[Rachotilko]

You should be quite aware that tone curves are used to change the displayed dynamic range, but as I noted, even though it can sometimes be effective, it doesn't represent the true dynamic range of the sensor and it can lead to other problems.

Now it's getting interesting: Can you be more specific ? I can think of only two such "problems": posterization (due to the limited pixel bit width) and excessive noise in the shadowy areas.

 

[Rachotilko]

I'm not disagreeing. I am saying that SNR is but one piece of the puzzle when trying to maximize the DR in the final image. DR in a practical sense is not about one concept ...

I still don't agree. Just a moment ago I took a really contrasty (but not overexposed) CanonG12 RAW file and processed it in RawTherapee (a great tool BTW).

I heavily modified the tone curve so as to bring up the shadow details while protecting the highlights.

The result: completelly acceptable and natural picture, bar one thing: the shadow areas noise.

I really don't believe that large sensors have some kind of "natural" soft flattening of the tone curve with in the saturation area. Of course, image sensor technologist may prove me wrong.

But even if I am wrong in that respect, the "soft flattening" is perfectly possible to emulate in small sensor realm by means of the tone curve adjustment - especially in RAW. You are limited in this approach only by the bit width and the shadow noise.

As you can see, all this is very practical; there is nothing pedantic about it. DR is determined by the SNR and the bit width of the RAW data.

 

[Billx08]

You should be quite aware that tone curves are used to change the displayed dynamic range, but as I noted, even though it can sometimes be effective, it doesn't represent the true dynamic range of the sensor and it can lead to other problems.

Now it's getting interesting: Can you be more specific ? I can think of only two such "problems": posterization (due to the limited pixel bit width) and excessive noise in the shadowy areas.

Those are the two main drawbacks that I was thinking of, and when users are give choices to change the tone curves in-camera, they often don't have as much control as they'd like and they could do better using a decent photo editor. Those two problems aren't always apparent when viewing small camera LCD images, and if the user adds -EC to the mix to get bluer skies, that further reduces the IQ of the shadow areas.

 

[Rachotilko]

You should be quite aware that tone curves are used to change the displayed dynamic range, but as I noted, even though it can sometimes be effective, it doesn't represent the true dynamic range of the sensor and it can lead to other problems.

Now it's getting interesting: Can you be more specific ? I can think of only two such "problems": posterization (due to the limited pixel bit width) and excessive noise in the shadowy areas.

Those are the two main drawbacks that I was thinking of, and when users are give choices to change the tone curves in-camera, they often don't have as much control as they'd like and they could do better using a decent photo editor. Those two problems aren't always apparent when viewing small camera LCD images, and if the user adds -EC to the mix to get bluer skies, that further reduces the IQ of the shadow areas.

But that (the lack of in-camera tone-adjustment tools) is irrelevant in RAW workflow. So what limits DR in case of RAW workflow ? The answer is obvious: noise and bit width.

 

[Billx08]

Those are the two main drawbacks that I was thinking of, and when users are give choices to change the tone curves in-camera, they often don't have as much control as they'd like and they could do better using a decent photo editor. Those two problems aren't always apparent when viewing small camera LCD images, and if the user adds -EC to the mix to get bluer skies, that further reduces the IQ of the shadow areas.

But that (the lack of in-camera tone-adjustment tools) is irrelevant in RAW workflow. So what limits DR in case of RAW workflow ? The answer is obvious: noise and bit width.

But for the purposes of this thread, that's irrelevant. My replies have been mainly posted to help answer painterdude's questions and while I may be assuming incorrectly, the assumption is that based on replies that I can recall from past threads and what he has written in this one, he's a JPEG shooter and until confirmed otherwise, getting into messy details about all of the different types of noise, bit width, sensor efficiency, and RAW processing would only take this thread in a direction that he may have little interest in. Since you're obviously interested in it, you may wish to continue the discussion with Mark or start another thread. This one was basically about DoF, and many of the posts that took it off topic weren't designed to really help anyone, if you know what I mean.

 

[Kim Letkeman]

... we're not saying anything so different, since we each are discussing noise and tone curves. If you must have it your way, then please do.
--
http://kimletkeman.blogspot.com

 

[Kim Letkeman]

and many of the posts that took it off topic weren't designed to really help anyone, if you know what I mean.

They never are, and I do

--
http://kimletkeman.blogspot.com

 

[Kim Letkeman]

It is wrong - it's thoroughly, and fundamentally, wrong - and you really need to try a lot harder to follow the plot thread 'KL', before misguidedly leaping in to support your forum buddy 'Billx08'.

I can dislike your awful overly complex, pedantic and disingenuous writing all on my own ... your vendetta interactions with Bill are your own issue. You sure seem to love it, that is plainly obvious.

--
http://kimletkeman.blogspot.com

 

[Rachotilko]

OK, I'll relax a bit now.

Anyway, it was Pdude (the OP) who wanted to know the reasons for limited DR of small sensor cameras. And I honestly believe those reasons have nothing to do with tonal curves, or blown highlights - those are actually symptoms, not causes of limited DR. So I wanted to provide him the clarification.

 

[laquila65]

Question #1
Smaller sensored cameras have deeper DOF ..because ??

It's simply matter of the physics, of geometric optics - i.e. the 'lens equation' in combination with the geometry of the lens aperture.

I don't get this either. The optical characteristics of the lens should affect the DOF and nothing else. Let's say I have a camera with a full-frame sensor. I shoot a picture. Then I put a black tape on the sensor (a mask) to make it look like an APS-C size sensor. I shoot the same subject. Will the DOF magically change? No, it will not. It's the same confusion as with the magnification factor that also magically "changes" with different sensor sizes.

 

[Billx08]

Question #1
Smaller sensored cameras have deeper DOF ..because ??

It's simply matter of the physics, of geometric optics - i.e. the 'lens equation' in combination with the geometry of the lens aperture.

I don't get this either. The optical characteristics of the lens should affect the DOF and nothing else. Let's say I have a camera with a full-frame sensor. I shoot a picture. Then I put a black tape on the sensor (a mask) to make it look like an APS-C size sensor. I shoot the same subject. Will the DOF magically change? I don't think it will. Where am I wrong?

You're not wrong if the picture doesn't matter, but the goal is usually to take a particular picture such as a portrait, for example. When you mask the sensor, the picture changes because the heads and feet of the people in the photo are cropped out of the picture, as are people (or parts of them) on the left and right side of the frame. So the photographer has to compensate by moving much farther away from the subjects. Of course that changes the perspective, but it also affects the DoF. Instead of moving away from the subjects, the photographer could shoot from the same position but use a shorter focal length, either by zooming out or by replacing the lens with one that provides a wider angle. Now the perspective doesn't really change, but the different focal length also changes the DoF. So if the only change is to crop the sensor, while you wouldn't see a different DoF, the photo would be drastically altered in a way that probably couldn't be tolerated by the photographer or by the photo's victims subjects.

 

[Kim Letkeman]

OK, I'll relax a bit now.

Anyway, it was Pdude (the OP) who wanted to know the reasons for limited DR of small sensor cameras. And I honestly believe those reasons have nothing to do with tonal curves, or blown highlights - those are actually symptoms, not causes of limited DR. So I wanted to provide him the clarification.

Well, using said logic, SNR is also just a symptom. What is the actual root cause?

1. Small full welll capacity because of tiny photosites, which is something that Bill already mentioned IIRC.
2. Low total light gathered because of small sensor. Another cause of overall image SNR issues.

Those are the main culprits in low dynamic range I would think. Of course, technology with high read noise is another culprit. Low quantum efficiency sensor is related.

--
http://kimletkeman.blogspot.com

 

[Kim Letkeman]

To directly compare images shot in FF and APS-C and m4/3, you must frame the same from the same distance. That way you have the same perspective.

With smaller sensors, the necessary focal lengths change. This changes the depth of field.

So you shoot an image of a person at 6 feet distance with a 75mm lens at f/2.8.

To get an equivalent image on APS-C, you will shoot a 50mm lens at f/1.8. (1.8*cf1.5 == 2.7 ... close enough)

Top get an equivalent image on m4/3, you will shoot a 37mm lens at f/1.4 ...

If you shoot at the same aperture you lose subject isolation by definition ...

--
http://kimletkeman.blogspot.com

 

[Billx08]

OK, I'll relax a bit now.

Anyway, it was Pdude (the OP) who wanted to know the reasons for limited DR of small sensor cameras. And I honestly believe those reasons have nothing to do with tonal curves, or blown highlights - those are actually symptoms, not causes of limited DR. So I wanted to provide him the clarification.

Well, using said logic, SNR is also just a symptom. What is the actual root cause?

1. Small full welll capacity because of tiny photosites, which is something that Bill already mentioned IIRC.
2. Low total light gathered because of small sensor. Another cause of overall image SNR issues.

Precisely, and I actually referred to both of those points in the same sentence in my reply to painterdude :

It's the related problems of not capturing enough photons, both by the photosites and by the entire sensor.

But they're the type of nits that some nitpickers willfully go out of their way to ignore.

 

[Billx08]

To directly compare images shot in FF and APS-C and m4/3, you must frame the same from the same distance. That way you have the same perspective.

With smaller sensors, the necessary focal lengths change. This changes the depth of field.

So you shoot an image of a person at 6 feet distance with a 75mm lens at f/2.8.

To get an equivalent image on APS-C, you will shoot a 50mm lens at f/1.8. (1.8*cf1.5 == 2.7 ... close enough)

Top get an equivalent image on m4/3, you will shoot a 37mm lens at f/1.4 ...

If you shoot at the same aperture you lose subject isolation by definition ...

Right, but I sometimes hesitate to use the term 'equivalence' because of another type of equivalence, that being the relationship (or equivalence) between DoF and the GoF.

 

[Rachotilko]

The confusion stem from using the bloody 35mm eq. terms when referring to the focal length range of the cropped sensors. They are IMHO invented for purely marketing purposes, as they have no optical significance or interpretation.

I'll illustrate it on example of two cameras:

• think of a full frame camera - for example Nikon D800 - with two prime lens: (a) 4.4mm @f2.8 and (b) 108mm @f2.8. Take one shot with each of the lenses. As you might guess, the wide-angle lens will provide almost none subject separation, while tele lens will do super-creamy sexy-bokehed images.

• now think of an ultrazoom, for example Panasonic FZ200. It's real (physical) focal length range is 4.4-108 mm. Now take your shots (@f2.8) on both wide-angle and tele ends with your FZ200.

• compare the wide-angle (4.4mm) shot taken by D800 and wide angle (4.4mm) shot taken by FZ200. They will be equivalent in terms of subject separation (ie. there will be none, the whole scene being wicked sharp). But they will differ (tremendously) by the size of the scene captured - FT200's shot being a very deep crop.

• similarly the wide-angle (108mm) shot taken by D800 teleprime will produce the same subject separation geometry as FZ200's shot taken at the long end (108mm) of its focal range. The images will again differ dramatically in the size of the scene captured.

 

[Kim Letkeman]

Right, but I sometimes hesitate to use the term 'equivalence' because of another type of equivalence, that being the relationship (or equivalence) between DoF and the GoF.

I remember it very well Of course, while the examples we were shown were awful to the point of saying nothing useful, in fact the concept is quite accurate. Thom Hogan describes it without all the pomp and circumstance somewhere ...

--
http://kimletkeman.blogspot.com