How much DR is enough?

3kramd5

EOS R6
Mar 2, 2012
3,084
405
Kit. said:
Actually, that part of the shot noise belongs to the lighter pixels,
Yes, hence “the more light...”

Kit. said:
while people are more interested in darker pixels when they are talking about sensor's DR.
Likely because people quite discuss the noise which can be addressed by design (that introduced by the system). Shot noise is a property of light.

Kit. said:
3. The scene is repeatable and someone with higher DR gear may capture it better, making your shot irrelevant.

Only the 3rd case is really about "the more the better" in relation to the subject's question
You have me rightly confused now. Firstly, a photo I take is not irrelevant because someone else is capable and equipped to take it better. Were that the basis, literally every photograph I’ve ever taken would be irrelevant because I’m not “the best” photographer nor do I have “the best” gear.
Secondly, the it is the 1st case relates to the subject question as I interpret it. I’ll paraphrase as “how much dynamic range capability is enough that in practicality you won’t encounter scenes which exceed the limit of the photography?” That’s a moving target, but I am not convinced that “more is better” is a flawed position. For me, most scenes I shoot are within gear limits, either because I light them myself or shoot under less extreme conditions. However I often shoot the moon, and have to stack due to the massive tonal range between the illuminated parts and the shadowed parts which I can not affect.

Kit. said:
(and by the way, the binary logarithm of exposure is EV. eV is "electron-volts", a measure of energy)
It was not worth overwriting the autocorrect.

Kit. said:
In the real world, the details in the dark foreground and the details in the light background are parts of different scenes, and you don't need to be able to match them together. In the picture, they are parts of the same composition.
You do if that’s the photograph you’re trying to create.
 

stevelee

FT-QL
CR Pro
Jul 6, 2017
1,742
605
Davidson, NC
Blown out highlights are also a DR concern. Isn't that what "exposing to the right" is all about? (I can't recall, since it is really relevant mostly around base ISO, and I don't shoot there enough to make it part of my routine. When I shot color slides, I did tend a bit toward underexposure in contrasty situations, which I guess is the analog analogue to that.)

In my pictures I am generally impressed with how much highlight recovery is possible in ACR. Many times there is enough detail to make the picture look unnatural, so I back off. I like the deep, dark shadows to look like deep, dark shadows most of the time, so I use the Shadows slider mostly to give a bit of fill-in flash effect on a backlit subject rather than trying to check the quality of house cleaning in the murky corners.
 

Kit.

EOS 5D Mark IV
Apr 25, 2011
2,028
1,379
3kramd5 said:
Kit. said:
while people are more interested in darker pixels when they are talking about sensor's DR.
Likely because people quite discuss the noise which can be addressed by design (that introduced by the system). Shot noise is a property of light.
The shot noise in darker pixels is more prominent. Even though the absolute noise is lower, the relative noise is higher.

Shot noise is not additive. If it were, the DR of a 30k-electrons-well sensor would be less than 8 EV, not more than 10.

3kramd5 said:
Kit. said:
3. The scene is repeatable and someone with higher DR gear may capture it better, making your shot irrelevant.

Only the 3rd case is really about "the more the better" in relation to the subject's question
You have me rightly confused now. Firstly, a photo I take is not irrelevant because someone else is capable and equipped to take it better. Were that the basis, literally every photograph I’ve ever taken would be irrelevant because I’m not “the best” photographer nor do I have “the best” gear.
"Someone" doesn't necessarily mean "someone else". It could be you with a better camera. Same emotional attachment, just higher DR.

3kramd5 said:
Secondly, the it is the 1st case relates to the subject question as I interpret it. I’ll paraphrase as “how much dynamic range capability is enough that in practicality you won’t encounter scenes which exceed the limit of the photography?” That’s a moving target,
That seems to be a tautological "target".

3kramd5 said:
but I am not convinced that “more is better” is a flawed position. For me, most scenes I shoot are within gear limits, either because I light them myself or shoot under less extreme conditions. However I often shoot the moon, and have to stack due to the massive tonal range between the illuminated parts and the shadowed parts which I can not affect.
But if stacking works for you, then the DR of your sensor is enough for the job.

It could as well be that non-stacking is theoretically impossible, as the moon moves quite fast. It takes about 2 minutes for the moon to move for its all diameter, and if you need to expose for the full range of a night scene and get a nonzero photon count in a meaningful number of your dark shadows pixels, your moon may get motion blur. Still, even if the DR of your sensor is theoretically limited to not doing the job as you would want it to do, it is still practically "enough" for doing the job as you are doing it now.

3kramd5 said:
Kit. said:
In the real world, the details in the dark foreground and the details in the light background are parts of different scenes, and you don't need to be able to match them together. In the picture, they are parts of the same composition.
You do if that’s the photograph you’re trying to create.
You do if that's the photograph one wants you to see. But it's not a natural way for you to see such a scene.
 

stevelee

FT-QL
CR Pro
Jul 6, 2017
1,742
605
Davidson, NC
The moon is not that hard to photograph, at least a full moon. It is in bright sunlight, and can be exposed accordingly. It's when you want other stuff in the picture that you need some other exposures. If the moon has moved in the meantime, as is likely, then just mask out the other moon images on the other exposures. They are probably washed out anyway.

This is a 100% crop of a handheld shot I did at 400mm of the full moon, but it is a stack of one taken at "sunny 16" (f/16 for 1/400 sec. at ISO 400) with a layer at 20% opacity taken at f/8. The sunny one looked pretty good as is. I don't know why I chose to do that rather than just using one of the ones I shot at "loony 11." Maybe I just overlooked them. The moon was too high in the sky for including landscape. I used autofocus and manual exposure.

 

Kit.

EOS 5D Mark IV
Apr 25, 2011
2,028
1,379
stevelee said:
This is a 100% crop of a handheld shot I did at 400mm of the full moon, but it is a stack of one taken at "sunny 16" (f/16 for 1/400 sec. at ISO 400) with a layer at 20% opacity taken at f/8. The sunny one looked pretty good as is. I don't know why I chose to do that rather than just using one of the ones I shot at "loony 11."
The moon is "Sunny 16" illuminated, but it's quite dark grey, so you need to overexpose it a little to look closer to white.

Also, when it's low above the horizon, it's a bit darker (and, obviously, yellowish), because of a longer path the light reflected from it takes through the Earth's atmosphere.
 

stevelee

FT-QL
CR Pro
Jul 6, 2017
1,742
605
Davidson, NC
Kit. said:
stevelee said:
This is a 100% crop of a handheld shot I did at 400mm of the full moon, but it is a stack of one taken at "sunny 16" (f/16 for 1/400 sec. at ISO 400) with a layer at 20% opacity taken at f/8. The sunny one looked pretty good as is. I don't know why I chose to do that rather than just using one of the ones I shot at "loony 11."
The moon is "Sunny 16" illuminated, but it's quite dark grey, so you need to overexpose it a little to look closer to white.

Also, when it's low above the horizon, it's a bit darker (and, obviously, yellowish), because of a longer path the light reflected from it takes through the Earth's atmosphere.
Also, as I now look at the picture, I see that the moon was not quite full. I had just bought the 100-400mm lens and was trying it out with the moon that was out that night. The moon or a filtered or eclipsed sun gives about the worst-case scenario for CA, I think, so that is what I was looking for. I didn't do any lens correction on this shot, and I don't think ACR did automatic ones. So the lens comes off as being as good as people here have said it is.
 

zim

EOS 5D Mark IV
CR Pro
Oct 18, 2011
2,035
217
3kramd5 said:
zim said:
3kramd5 said:
Kit. said:
The topic.

DR is theoretically limited by photon noise (Heisenberg's uncertainty).
Agreed. The more light you record, the more shot noise you encounter.
so I want a camera with less DR? :eek:
Not at all. What that meant was: You can overcome the noise introduced by the electronics by increasing the well capacity (point beyond which you clip), but you can not overcome the noise inherent to the light.
TBH that's over my head, but my key take away is 'not at all'
cheers :)
 

Rowk

I'm New Here
Jul 26, 2017
12
1
I guess people crave for high DR because it is easier to ETTL than ETTR.

At least I want sharp results and the best image quality possible, that means low ISO and fast shutter speed.
Usually there is not enough light to use my preferred settings - ISO 100 at 1/2000 or faster without a tripod.
(it depends on the subject of course)

So most of the time I end up pushing the shadows! ;D
 

Orangutan

EOS 5D Mark IV
Sep 25, 2010
2,140
3
Rowk said:
I guess people crave for high DR because it is easier to ETTL than ETTR.
Some landscape scenes have DR that exceeds all modern sensors, and bracketing/compositing is not always an option (e.g. wind, flowing water). Also, for action (e.g. birds), there's often not time to adjust exposure, so the only choice is to set EC to a likely setting and hope. For me, more DR is always "better" because I don't mind doing a little PP.
 

Aglet

EOS 5D Mark IV
Feb 26, 2012
1,728
15
AB
Pretty much any camera with 11 or more stops worth of DR is capable of rendering an acceptable reproduction of a natural scene, even when there's some lifting of shadows or darker midtones to achieve a desired (HDR-like) effect.

Difficulties arise if the camera used exhibits "fixed pattern noise" (FPN) in areas where the original tones have been raised in post and the read noise becomes visible. This can be vertical or horizontal stripe effect or both, in luminance and-or chroma. It becomes a visual distraction and limits how much darker tones can be processed upwards before becoming a problem.

This is where the ranks split some years ago.

Sony started making sensors that still had some shadow (read) noise but it was random so processed out better with NR software or could be left in for an analog-like grain effect which most viewers do not find objectionable.

Canon, and some other sensor mfrs., had some issues with read noise showing up as repetitive patterns which became visible if the raw files were overly manipulated in post.

Altho effective DR of these sensors may have been close to each other the one with random-like read-noise was generally more malleable in post than the files which came from sensors with repetitive read noise patterns.

IF you weren't pushing the raw files in post, all systems worked pretty well.
If you were pushing them, then some systems provided better results or required fewer work-arounds.

Choose the tools that work with you the way you prefer or gripe about those that do not while continuing to use them.

It's good to have choices and it's good to learn and be aware of the limitations of your gear. :)
 

Orangutan

EOS 5D Mark IV
Sep 25, 2010
2,140
3
Aglet said:
Pretty much any camera with 11 or more stops worth of DR is capable of rendering an acceptable reproduction of a natural scene,
That's not my experience: according to DxO, my 70D has 11.6EV of DR, yet I frequently encounter problems in landscape where important highlights are blown out, and important shadows clipped in the same image. It's not just a matter of exposing for highlights and getting a clean shadow lift.
 

privatebydesign

Garfield is back...
CR Pro
Jan 29, 2011
9,205
3,438
120
Orangutan said:
Aglet said:
Pretty much any camera with 11 or more stops worth of DR is capable of rendering an acceptable reproduction of a natural scene,
That's not my experience: according to DxO, my 70D has 11.6EV of DR, yet I frequently encounter problems in landscape where important highlights are blown out, and important shadows clipped in the same image. It's not just a matter of exposing for highlights and getting a clean shadow lift.
And there in lies the issue with DR, what is a reasonable measurement? Personally I find results listed at photons to photos on their "Photographic Dynamic Range" chart to be the closest to the results I expect from processing RAW files, they list the 70D as having 8.93 stops of PDR, that is usable DR between a noise floor of 20dB and the white point. It seems your results align with theirs too.

DxO use a black box method of measuring but are also looking at a different kind of measurement, they are looking at engineering DR which includes several stops of DR below the 'noise floor' (wherever you set that). Those stops below the noise floor are not, for photographic purposes, generally usable so including them has become a moot point.
 

3kramd5

EOS R6
Mar 2, 2012
3,084
405
Kit. said:
But if stacking works for you, then the DR of your sensor is enough for the job.
But it’s a workaround, and one which as you noted doesn’t always work. I could probably build a 12-key piano with a pedal which progressive tightens each string by an octave, but one with 88-keys to work with would be far easier to deal with.


Kit. said:
3kramd5 said:
Kit. said:
In the real world, the details in the dark foreground and the details in the light background are parts of different scenes, and you don't need to be able to match them together. In the picture, they are parts of the same composition.
You do if that’s the photograph you’re trying to create.
You do if that's the photograph one wants you to see. But it's not a natural way for you to see such a scene.
I guess I’m not sure what you’re saying here. Quite often I encounter real scenarios where my natural (well, modified - I’ve had surgery) eyes see detail where my silicon eye records either white or black.

And maybe it isn’t me, maybe it’s a client who wants to see an extensive book collection in shelves flanking a large window on a bright day in a room lit only by that window. The end goal of the photo is what matters to me.

I could put reflectors behind me. I could bring lights. I could ND gel the window. Those are all ways to compress the tonal range in the capture, but also they wouldn’t fall within the normal way of viewing the room since it. The alternate is to extend the range capability of the gear.

I respect your position but am a bit confused by it.
 

Kit.

EOS 5D Mark IV
Apr 25, 2011
2,028
1,379
3kramd5 said:
Kit. said:
But if stacking works for you, then the DR of your sensor is enough for the job.
But it’s a workaround, and one which as you noted doesn’t always work. I could probably build a 12-key piano with a pedal which progressive tightens each string by an octave, but one with 88-keys to work with would be far easier to deal with.
I'm saying that "88-keys" is physically impossible to make work because of motion blur in one part of the picture vs low photon count in another part. While the your sensor is still "enough" for the workaround.

3kramd5 said:
Kit. said:
3kramd5 said:
Kit. said:
In the real world, the details in the dark foreground and the details in the light background are parts of different scenes, and you don't need to be able to match them together. In the picture, they are parts of the same composition.
You do if that’s the photograph you’re trying to create.
You do if that's the photograph one wants you to see. But it's not a natural way for you to see such a scene.
I guess I’m not sure what you’re saying here.
In this particular case, we are talking about shooting now for some hypothetical high dynamic range output media in the future, and how high tonal range photos would look on such media. I'm saying that for a human eye looking at your picture represented on that media, the dark details of the foreground and the light details of the background would not belong to the same composition.

If you compress the range on output, they might.
 

stevelee

FT-QL
CR Pro
Jul 6, 2017
1,742
605
Davidson, NC
Perhaps if the camera were the size and weight of a grand piano, that would help.

The analogy doesn’t really work because a piano is an output device, not a recording machine.

As it stands now and should mostly for a long time to come, camera sensors are capable of gathering much more information than any output medium can display. Capturing more information gives you more choices of how to map to the output, including what to distort and how much to misrepresent in order to retain other information. That holds for color profiles as well as luminosity. Do you throw away out of gamut colors, or do you remap all the other colors to bring it into gamut?
 

stevelee

FT-QL
CR Pro
Jul 6, 2017
1,742
605
Davidson, NC
Pianos by their nature do a horrific amount of remapping in their output. They can sound good in large part because we have had close to 200 years to adjust to them in our culture, and at that their evolution into modern form was gradual enough for our ancestors to adjust. Pablo Casals would admonish his students not to try to play in tune with a piano.

The short version is equal temperament maps all notes to just twelve different notes to the octave. Fifths are slightly flattened to accommodate, and major thirds are so sharp that they make more of a buzz than beats. On top of that, octaves are stretched because of the physics of strings under high tension. Keyboards would need 53 notes to the octave to play in tune in all keys, but 31 is good enough to fool the ear, at least in Baroque music. I started to say “close enough for jazz,” as the saying goes. But the I recalled how jazz pianists crush notes together and use false fingering to simulate playing in the cracks.
 

Keith_Reeder

I really don't mind offending trolls.
Feb 8, 2014
957
469
60
Blyth, NE England
Orangutan said:
That's not my experience: according to DxO, my 70D has 11.6EV of DR, yet I frequently encounter problems in landscape where important highlights are blown out, and important shadows clipped in the same image.
And which Raw converter were you using?
 

Orangutan

EOS 5D Mark IV
Sep 25, 2010
2,140
3
Keith_Reeder said:
Orangutan said:
That's not my experience: according to DxO, my 70D has 11.6EV of DR, yet I frequently encounter problems in landscape where important highlights are blown out, and important shadows clipped in the same image.
And which Raw converter were you using?
Usually Lightroom, but sometimes DPP. Not much difference between them. I'm not sure that matters, though: the histogram tells me I'm at least 3-5 stops from full coverage of the scene DR. (I sometimes block a highlight area of the scene to see where it spikes on the histogram, then watch the spike move as I change exposure)
 

3kramd5

EOS R6
Mar 2, 2012
3,084
405
Orangutan said:
Keith_Reeder said:
Orangutan said:
That's not my experience: according to DxO, my 70D has 11.6EV of DR, yet I frequently encounter problems in landscape where important highlights are blown out, and important shadows clipped in the same image.
And which Raw converter were you using?
Usually Lightroom, but sometimes DPP. Not much difference between them. I'm not sure that matters, though: the histogram tells me I'm at least 3-5 stops from full coverage of the scene DR. (I sometimes block a highlight area of the scene to see where it spikes on the histogram, then watch the spike move as I change exposure)
You mean in-camera, yes? If so, it is JPEG-based, and not representative of the full capability.
 

Orangutan

EOS 5D Mark IV
Sep 25, 2010
2,140
3
3kramd5 said:
Orangutan said:
Keith_Reeder said:
Orangutan said:
That's not my experience: according to DxO, my 70D has 11.6EV of DR, yet I frequently encounter problems in landscape where important highlights are blown out, and important shadows clipped in the same image.
And which Raw converter were you using?
Usually Lightroom, but sometimes DPP. Not much difference between them. I'm not sure that matters, though: the histogram tells me I'm at least 3-5 stops from full coverage of the scene DR. (I sometimes block a highlight area of the scene to see where it spikes on the histogram, then watch the spike move as I change exposure)
You mean in-camera, yes? If so, it is JPEG-based, and not representative of the full capability.
Correct, but not as relevant as it might appear, for two reasons: (1) even though it's JPEG based, the JPEG histogram is (from my experience and what I've read on the web) about 1 stop different from the RAW histogram; i.e., if the JPEG is clipped by about 1 stop, I can probably recover it in post, and I take this into account in my exposure settings. (2) I've also bracketed and examined RAW histograms after the fact. In deep forest, when exposing for the shadows, I'd need to take a second frame 3-5 stops darker to keep the back-lit leaves (or splashing water) from being blown-out in the RAW files. Of course, this depends on the angle of the sun (time of year/day) and other factors, but I've tried it several times with fairly consistent results. YMMV.

I will also say, that I have yet to get the full DR of a male wood duck in full light: I have several wood duck photos that are clipped at both ends of the histogram.