Just a Simple Question About Dynamic Range

kcbrown wrote in post #15060203
It sounds like your standard of usability is, essentially, noise-free. If that's the case then yes, I agree, you're not likely to get more than 5 stops of usable dynamic range from the 5D2 or, really, just about any other camera out there.

But that's not what the manufacturers or testers mean when they talk about dynamic range. What they mean by that is the range of light intensities that result in a signal that is distinguishable at all from the background noise. Which is to say, the figures they publish are for the maximum dynamic range that can possibly be used at all. If one's own personal standards are more strict than that then that's fine, of course, but it does not change the fact that the camera is recording a distinguishable signal over a much wider dynamic range than what you appear to be using in your final output.

Nobody says that you have to use the entire dynamic range that the camera is capable of delivering, and that's as it should be -- we all have our own standards of acceptability. Because we all have our own such standards, the only thing the testers can do is give you a figure that is based on objective and non-arbitrary criteria. It's up to you to determine how much of that is usable for your purposes.

That makes a lot of sense. Its in the manufacturers interest to claim high DR and they are able to "prove" it by showing us graphs when people like me challenge them. Their problem is, I refuse to be persuaded by their graphs when I see the reality in front of me every time I shoot!

I willingly accept that I am only saying what my own experience is, based on my own limits as you say.

KenjiS wrote in post #15055894
Heres attempt number 2, Taking your criticisms into account and some inspiration from both of you:

IMAGE LINK: http://www.flickr.com …unetsukiphoto/8​034244373/
Suntouched Sunflower Mk2 by Kenjis9965, on Flickr

Basically i made the background "natural" (its a wood fence) and tried to get the sunflower to scream against it like you suggested.. i do like this better, except that dark ring which i cant seem to get rid of

Was a balancing act in this instance, For the first time i actually DID see the banding, It wont stop me from shooting, but i do see why some people find it a problem...

I did this only with a Curves adjustment.

First limited the curve in the area that has data (increases the contrast).

Second, I 'pinned' the curve where the highlights start and boosted the highlights only.

Light sharpening.

No masks or other local treatments.

Lowner wrote in post #15060231
That makes a lot of sense. Its in the manufacturers interest to claim high DR and they are able to "prove" it by showing us graphs when people like me challenge them. Their problem is, I refuse to be persuaded by their graphs when I see the reality in front of me every time I shoot!

Understood, but do understand that your judgment involves a usability test that is subjective and specific to you. The manufacturers can't really do anything about that. All they can do is tell you the most that the camera is capable of doing at all and let you take it from there.

Additionally, as far as I know it's not really the manufacturers who are making dynamic range claims, it's various testers like DxOMark. And translating the range values they publish to something meaningful to you is something you pretty much have to do yourself. It requires, among other things, an understanding of what their measurement actually means and how that relates to your own purposes. And even if you have all that, it's still only an indicator of what you can expect.

The only way to really know what a camera is capable of with respect to your needs is to use it.

I really wish there were some objective way for them to tell you what the "usable" DR of the camera is, but unfortunately there really isn't, because each person's idea of "usable" differs from the others.

What might be useful is a signal to noise graph across the dynamic range scale at base ISO. If you happen to know what signal to noise ratio corresponds to your lower bound of usability, you'd be able to determine the usable dynamic range from the graph. But that's going to be complicated by the type of noise involved as well, along with artifacts that don't show up in signal to noise measurements, such as pattern noise (e.g., banding, mazing, etc.). Still, it'd be better than nothing.

kcbrown wrote in post #15060203
It sounds like your standard of usability is, essentially, noise-free. If that's the case then yes, I agree, you're not likely to get more than 5 stops of usable dynamic range from the 5D2 or, really, just about any other camera out there.

But that's not what the manufacturers or testers mean when they talk about dynamic range. What they mean by that is the range of light intensities that result in a signal that is distinguishable at all from the background noise. Which is to say, the figures they publish are for the maximum dynamic range that can possibly be used at all. If one's own personal standards are more strict than that then that's fine, of course, but it does not change the fact that the camera is recording a distinguishable signal over a much wider dynamic range than what you appear to be using in your final output.

Nobody says that you have to use the entire dynamic range that the camera is capable of delivering, and that's as it should be -- we all have our own standards of acceptability. Because we all have our own such standards, the only thing the testers can do is give you a figure that is based on objective and non-arbitrary criteria. It's up to you to determine how much of that is usable for your purposes.

Well, your definition of usable seems to be pretty far off from mine. Take this photo for example.

kfreels wrote in post #15062754
Well, your definition of usable seems to be pretty far off from mine.

Not my definition, for the record. Lowner's apparent definition.

Take this photo for example.

This was shot at f4. The darker area below the armpit and down by her rear metered at f1.4. Her left cheek metered at f22. There is texture still in all of those areas. That's 9 stops and I still had some room on the dark end.

Nice!

Good postprocessing technique really brings out the most from these cameras, and I think your shot here is an excellent example of that.

kcbrown wrote in post #15062765
Not my definition, for the record. Lowner's apparent definition.

Nice!

Good postprocessing technique really brings out the most from these cameras, and I think your shot here is an excellent example of that.

Thanks, but the point here isn't about post-processing. I had to do very little to this shot in post. I made no local adjustments. I converted to B&W, made a slight curves adjustment to intentionally push the edge of the gamut - which effectively reduced the DR of my image from what I had captured. Then I applied some sharpening.

What I am referring to is image data captured by the sensor. The lighting of the subject was spread out a full 9 stops or more and the camera captured every bit of it in a single shot. This is what the camera manufacturers are referring to in their dynamic range claims. On the 7D I think I have about a full 11 stops before things slip off to pure black or pure white. That matches well to the claimed 11.7 or whatever the official number is. If there were only 5 stops, then this would have looked a lot more contrasty with most of my dark gray falling off to black and most of the image on the light end fallen off to white. All those mid-tones wouldn't be there.

It could be that you are referring to something completely different as for how dynamic range applies to your particular use, but that's not what the manufacturers are referring to. They aren't lying to you. They are simply using a definition that most of us work with.

In other words, my 7D has almost 3 more stops of dynamic range to work with than kodachrome. It has better DR than Kodak gold and Fuji Velvia. So if the OP has problems by only getting 5 stops on these modern digital cameras I highly suggest staying away from films.

kfreels wrote in post #15063052
Thanks, but the point here isn't about post-processing. I had to do very little to this shot in post. I made no local adjustments. I converted to B&W, made a slight curves adjustment to intentionally push the edge of the gamut - which effectively reduced the DR of my image from what I had captured. Then I applied some sharpening.

I presume that when you adjusted the curves, you wound up pushing the shadows? Because (depending on your postprocessing settings) I didn't think you'd see 9 stops worth of captured light with default conversion settings. Most certainly, that has not been my experience (7D with Lightroom 3).

Good postprocessing technique isn't necessary to make the entire dynamic range of the camera visible, but it is necessary in order to make that dynamic range look good.

Yes, I completely agree, and that's essentially what I've been saying. But just because the camera has captured the light doesn't mean you'll see it in the resulting conversion. That depends entirely on your postprocessing settings.

In my experience, to get the most out of the dynamic range of the camera, you have to pull your highlights by about a stop (this means, generally, that it's best to overexpose by about a stop and then pull the exposure in postprocessing in order to minimize noise) and then push the shadows by the remainder. In my experience, the exposure range you see in the 7D's meter, 6 stops (3 above neutral, 3 below neutral), is just about what you'll get out of a default conversion.


Different cameras have different amounts of headroom in the highlights, too. The 7D's is about a stop. I was under the impression that the 5D2's was actually more, perhaps 1.5 stops or so.

Exactly. They're telling you the light intensity range over which the camera is able to record anything at all above the noise floor. That's useful in that it tells you the most you can possibly get out of the camera. But it tells you nothing about the quality of what you get at any given part of the dynamic range, and how much of the dynamic range you're willing to use before deciding that the quality at a given point is too low is an entirely subjective thing.

Heh. It actually sounds to me like his quality expectations are very stringent. As you push your shadows, you also push the noise. That can't be helped: you're just amplifying whatever's there, be it signal or noise. The deeper the shadows that you push towards visibility, the greater the noise signature will be. If your expectations are that everything you see in the shot must be creamy smooth without any noise reduction being applied, then chances are you won't be happy with pushing the shadows by any amount and, actually, may wind up crushing your blacks in order to further reduce visible noise in the shadows. Crushing the blacks is, of course, a move that will reduce the dynamic range that you're actually using in the final output.


Most people don't really know what dynamic range is and, in fact, few make use of anything close to the dynamic range their camera is capable of. That's because doing so requires that you pull your highlights and push your shadows, while most people will actually crush their blacks and leave the highlights alone for the most part (assuming they're not making any exposure adjustments). That's because what they're after is a dramatic, contrasty look, and that's the easiest way to achieve it. There's certainly nothing wrong with that at all -- the camera's just a tool, after all -- but such people have no business telling others that they "exposed wrong" when they push their shadows to make greater use of the dynamic range of the camera, precisely because they lack the understanding of what's really going on.

kcbrown wrote in post #15062678
Understood, but do understand that your judgment involves a usability test that is subjective and specific to you. The manufacturers can't really do anything about that. All they can do is tell you the most that the camera is capable of doing at all and let you take it from there.

Additionally, as far as I know it's not really the manufacturers who are making dynamic range claims, it's various testers like DxOMark. And translating the range values they publish to something meaningful to you is something you pretty much have to do yourself. It requires, among other things, an understanding of what their measurement actually means and how that relates to your own purposes. And even if you have all that, it's still only an indicator of what you can expect.

The only way to really know what a camera is capable of with respect to your needs is to use it.



I really wish there were some objective way for them to tell you what the "usable" DR of the camera is, but unfortunately there really isn't, because each person's idea of "usable" differs from the others.

What might be useful is a signal to noise graph across the dynamic range scale at base ISO. If you happen to know what signal to noise ratio corresponds to your lower bound of usability, you'd be able to determine the usable dynamic range from the graph. But that's going to be complicated by the type of noise involved as well, along with artifacts that don't show up in signal to noise measurements, such as pattern noise (e.g., banding, mazing, etc.). Still, it'd be better than nothing.

As you say, it would be great to have some reference point from the manufacturer that a buyer could relate to their own methods/uses. So I've come to the conclusion in this discussion that I should at the very least halve any DR claim I see, not very scientific I know, but probably useable. Of course that means I want Canon to start claiming 24 stops!

Lowner wrote in post #15063860
As you say, it would be great to have some reference point from the manufacturer that a buyer could relate to their own methods/uses. So I've come to the conclusion in this discussion that I should at the very least halve any DR claim I see, not very scientific I know, but probably useable. Of course that means I want Canon to start claiming 24 stops!

Yeah, but then you'd have to start quartering any DR claim you see!

kcbrown wrote in post #15063316
I presume that when you adjusted the curves, you wound up pushing the shadows? Because (depending on your postprocessing settings) I didn't think you'd see 9 stops worth of captured light with default conversion settings.

So what? The discussion is about the dynamic range captured by the camera. If good PP technique can bring out 9 stops of DR in an image then the camera must have captured at least 9 stops of DR in the data. Unless it's one of them special cameras they use in CSI.

hollis_f wrote in post #15064063
So what? The discussion is about the dynamic range captured by the camera. If good PP technique can bring out 9 stops of DR in an image then the camera must have captured at least 9 stops of DR in the data. Unless it's one of them special cameras they use in CSI.

Yes, clearly. The point that you are responding to here is that you must use good postprocessing technique to bring out the dynamic range of the camera while making it look good, and that postprocessing actually matters for making use of the camera's DR.

If you read back a bit in the thread, you'll see that I know exactly what dynamic range is, what it means for the images we can produce with it, and what one has to do in order to make visible the dynamic range captured by the camera. Corrections to anything I got wrong are welcome, of course...

kcbrown wrote in post #15063316
I presume that when you adjusted the curves, you wound up pushing the shadows? Because (depending on your postprocessing settings) I didn't think you'd see 9 stops worth of captured light with default conversion settings. Most certainly, that has not been my experience (7D with Lightroom 3).

* * *

Just a quick, relatively tangential note. LR 4/4.1 does a really superb job in handling 7D RAW conversions, and will allow maximizing both highlights and shadows while minimizing noise, which I think significantly extends the "usable" dynamic range. As you note, this is relative to the user's expectations; as for me, I'm not averse to using appropriate noise reduction when needed (LR 4/4.1 does a great job of reducing shadow noise while retaining detail).

stsva wrote in post #15064235
Just a quick, relatively tangential note. LR 4/4.1 does a really superb job in handling 7D RAW conversions, and will allow maximizing both highlights and shadows while minimizing noise, which I think significantly extends the "usable" dynamic range. As you note, this is relative to the user's expectations; as for me, I'm not averse to using appropriate noise reduction when needed (LR 4/4.1 does a great job of reducing shadow noise while retaining detail).

Is it worth to upgrade to LR4 then? I'm still running 3 because i keep forgetting to go get a copy...

Anyways, Sorry for being away, was busy painting my room, Took a little time in between coats to dust off the 580EXII and take pictures of my "supervisor"

IMAGE LINK: http://www.flickr.com …unetsukiphoto/8​046183002/

IMAGE LINK: http://www.flickr.com …unetsukiphoto/8​046177149/

IMAGE LINK: http://www.flickr.com …unetsukiphoto/8​046176979/

Buy the upgrade on the internet, it would take less time than it took you to post these images. At least download the month's trial and you can buy the number later.

As a side=note that might be of interest to some, I just happened across this exchange on the Luminous Landscape forum:
http://www.luminous-landscape.com/forum/in​dex.php?topic=71068.0