Help me understand dynamic range

[ryan_tsang]

<p>I am trying to learn dynamic range and I would like to know how to capture a scene that preserves shadow details without blowing out highlights, assuming the brightness range in the scene is reasonable. I have a D700, shooting RAW 12bit, MacBook with native gamma 1.8 and D65 white balance setting, Aperture 2 for RAW conversion.</p>

<p>imagine this scene: aperture constant at f4.5. shutter speed of 1/1.6s required to capture shadow detail, minimum speed of 1/60s is needed to not blow out highlights. Am I right in understanding that there is a approx 5.5 stops of dynamic range? I took the shot at 1/10sec to split the difference and what I see on-screen are blown out highs and clipped blacks. I need to do highlight and shadow recovery in order for it to be visible. Why can't I see the entire range if my D700 and NEF files can theoretically resolve more than 9 stops of range?</p>

<p>Is it my display that is the limiting factor?<br>

There is a black computer bag on the right, and I want to see the details in that as well as the high lit hardwood floors in front of the fridge</p>

<div></div>

[joe_a2]

<p>That scene has too much range to ever be captured with one exposure. Unless you use fill flash to make the shadows areas a little brighter.</p>

<p>In simplest terms, dynamic range is the range from the lightest thing captured with detail to the darkest thing captured with detail. A digital sensor or film cannot see as wide a range as your eye can see. You need to pick which part (within the wider range of brightness in a scene) that you wish to capture properly.</p>

<p>I'm not sure where you're getting "9 stops" from. Who said that?</p>

[mich8261]

<p>I think the only way a RAW file will get you a 9-stop range is if you blend two versions of the same image (creating an HDR image). Take the RAW file above and process it to benefit the shadow detail and then take the same RAW file and process it for the highlights. When you blend both images you will arrive close to that 9-stop range. There are all kinds of tool to automate this process or you can do it in PS manually (the blending part).</p>

[james_hudspeth]

<p>Hi Ryan, A jpg does not have enough detail to examine. Do you have the raw file that we can look at?</p>

 

[Peter_in_PA]

<p>btw,</p>

<p>the image in question is pretty cool and mysterious just the way it is.</p>

[john_demott]

<p>Ryan,<br>

Imagine using the spot meter in your camera to measure the exposure of the brightest part of your scene and of the darkest shadow area of your scene separately. In each case, the meter would tell you what exposure would render that portion of the scene as a middle gray. The two readings might be five stops apart for example, and we could then say that the dynamic range of the scene is five stops. If you set the camera at the metered setting for the highlights, the camera would still be capable of rendering some values even brighter than those present in the scene, say 2 or 2-1/2 stops more, and at that exposure the shadows would be 5 stops below middle gray, not 5 stops below the brightest value the camera could render.</p>

<p>In your example, you say that 1/60 or faster is required in order not to blow out the highlights. That is a different thing from saying that the indicated exposure when you meter the highlights separately is 1/60. If the highlights are just shy of being blown out at 1/60, then they are likely more than 2 stops brighter than the middle tones already. Likewise, if you say that 1/1.6 is required to capture shadow detail, you may be saying that the shadows are several stops below middle gray at an exposure of 1/1.6. So, although there is roughly 5-1/2 stops between the two settings you mention, there may be a much greater dynamic range present if I am understanding your statement accurately.</p>

<p>Bear in mind also that the dynamic range of the sensor is one thing, but how the image is presented is something else. If you have a scene with 8 stops of dynamic range, for example, the camera may be able to capture an image with detail in both highlights and shadows. But if you want the middle tones in the image to appear normal and have all the tones in scene appear normal in relation to one another, then you may have to let the highlights and shadows remain at the extreme range so that in a print they will have barely distinguishable detail.</p>

[ryan_tsang]

<p>Joe A: I looked up the test results for the D700 in dpreview.com and they report that it has a useable range of 7.8EV. By shooting RAW, you can supposedly extract more than 1 stop so that's where I figure the 9.<br>

If you guys think that our cameras cannot capture 9 stops then I must be interpreting these charts incorrectly:<br>

http://www.dpreview.com/reviews/NikonD700/page20.asp<br>

So, how do you determine how much dynamic range there is in a scene? Do you spot meter the dark and light areas that I expect to see detail in? Basically I want to know before I shoot whether I can do with a single exposure, single exposure + RAW processing, or resort to fill flash, HDR, reflectors and such. </p><div></div>

[Peter_in_PA]

<p>I recommend learning about the Zone System. There is a book called "The Negative" by Ansel Adams. I can't recommend it enough. It is written about film but is still totally applicable in digital.</p>

[james_hudspeth]

<p>Ryan, can we see the raw file? I downloaded the jpg and ran it through NX2 and came up with much more detail then in the jpg you are showing.</p>

[ryan_tsang]

<p>James: I don't have a website to host such a large file.</p>

<p>When you said "ran it through" NX2, what do you mean by that?</p>

[JDMvW]

<p>The <a href="../casual-conversations-forum/00Qwkz">Ozone method</a> , as some of us call it, of using Shadow and Light is usable so long as it is done cautiously. The default settings in Photoshop are much too great, in general.<br>

Obviously first of all, you'll have better luck doing this if you shoot RAW, expose for the highlights, even do two exposures and go HDR on us.</p>

<p>Still, even with a jpg you can do a lot to recover some shadow detail in the Ozone system. Here is going much further than Ryan did above. Note the distinctive 'flatness' however, often seen in postings here and elsewhere as a tattle-tale sign of overdoing this sort of thing.</p><div></div>

[joe_c5]

<p>Many raw converters apply a nonlinear tone curve without even telling you about it. Nearly all in-camera jpeg processing does the same thing, with an “S” shaped curve to increase midtone contrast.</p>

<p>If you can either override this tone curve, or use a different raw converter which allows you to override it, you will probably be able to see much more detail in the shadows. 8 stops of dynamic range is similar to an LCD monitor, which is usually around 8 or 9 stops depending on its contrast ratio.</p>

<p>In this case a single exposure is likely to be enough to see a lot of the shadow detail, but it may be noisy. The advantage of combining multiple exposures into a HDR image is lower noise. Dynamic range is essentially how much darker than the brightest white details can get before they get lost in the noise.</p>

<p>Fill flash will lower the dynamic range of the scene, making it easier to capture without changing your post processing.</p>

<p> </p>

[hawkman]

<p>Ryan,<br>

You got some nice comments from John, if the 1/60 is min exposure time required to avoid blow up, it doesn't mean that it is the correct exposure for a visually acceptable/pleasing photo. The scene you are showing has more than 5 stops of DR if you want to show the detail in extreme shadows and highlights at the same time. However with the D700 it is possible to achieve this with a single NEF file, NEF file has more data than what you see on your screen when you first open the file, in order to pull out all of the hidden data you need to change the tone curve that is being applied to the RAW values. You can adjust the RAW tone curve in Adobe LR by using recovery sliders as well as shadow/highlight sliders. You can also use Nikon Capture NX2 to achiever the same effects. Because you are stretching the tonal range it is better to shoot in 14Bit lossless-compressed/uncompressed NEF mode.</p>

[sam_ellis]

<p>The best explanation I've ever heard of for dynamic range is to thing of each pixel as a bucket and light is water. Too much water/light, and the bucket overflows/blows out to pure white. Too little water/light and the bucket has nothing in it, no information.</p>

[ryan_tsang]

<blockquote>

<p>Many raw converters apply a nonlinear tone curve without even telling you about it. Nearly all in-camera jpeg processing does the same thing, with an “S” shaped curve to increase midtone contrast.</p>

</blockquote>

<p>Ah ha!!! I found that out recently because I wanted the camera's LCD to closer match what I see on the computer (before processing), so I can rely on it more when I'm in the field. The only way I could do that was to apply a normal amount of Active D-Lighting. I shoot RAW so the D-lighting does not affect the image. </p>

<blockquote>

<p>However with the D700 it is possible to achieve this with a single NEF file, NEF file has more data than what you see on your screen when you first open the file, in order to pull out all of the hidden data you need to change the tone curve that is being applied to the RAW values.</p>

</blockquote>

<p>That's what I needed to know, the fact that a scene with at least 5 stops of DR needs post processing. I needed to learn how much DR I can reasonably capture before relying on post. I had no idea and wrong expectations. Thanks for all the help so far! Keep it coming if you've got more :)</p>

<p> </p>

<blockquote>

 

</blockquote>

[ryan_tsang]

<blockquote>

 

</blockquote>

<blockquote></blockquote>

[robert_k1]

<p>You may find this article helpful:<br>

<a href="http://www.ronbigelow.com/articles/dynamic-range-1/dynamic-range-1.htm">http://www.ronbigelow.com/articles/dynamic-range-1/dynamic-range-1.htm</a><br>

It explains DR clearly for a lay person, and describes many post processing methods. But what really sets it apart is the way it lists the pros and cons of each method. Most writings would only talk about the pros, and miss the cons. Once you are aware of both, you can choose as you see fit, for different situations and images. </p>

[ivan_j._eberle1]

<p>Dynamic range is simply the range of brightness that can be captured between the limits of solid black and pure white. It doesn't describe the steps or intermediate values. Curves will have no effect on the camera sensor's ability to capture DR--unless clipping is occuring (keep in mind, if you're using Photoshop, the default values clip). Dynamic range is affected by ISO, however.<br>

Useable dynamic range is somewhat more subjective, in that this also takes into account what shadow detail can be boosted to where it will print or display via modifiying curves to and levels (i.e. how you want the intermediate values to appear), without incurring excessive noise-to-signal.<br>

Can't tell much from the small .jpg but the first thing that occured to me is that if you're looking at the highlight here ie the bright spot on the floor and judging whether or not you've lost the highlight by the detail captured or lack thereof, this apparently OOF highlight at f/4.5 renders this image a poor choice. (Flare might also be a factor.) Did you first measure the scene with the spot meter to determine whether it falls within the DR of 9 stops you're looking for? The other obvious point might be that if it was not, to simply modify your light source with a white bed sheet thumb-tacked over the window or some other simple form of diffusion.</p>

<p> </p>

[robert_k1]

<p>Oops, I referenced the wrong article. What I meant is this older one which talks about many PS tools:<br>

<a href="http://www.ronbigelow.com/articles/shadow/shadow-highlight1/shadow-highlight1.htm">http://www.ronbigelow.com/articles/shadow/shadow-highlight1/shadow-highlight1.htm</a><br>

The other article is more verbose, and more about ACR.</p>

[ryan_tsang]

<blockquote>

<p>Did you first measure the scene with the spot meter to determine whether it falls within the DR of 9 stops you're looking for?</p>

</blockquote>

<p>That's what I'm trying to learn. I don't have a dedicated light meter. I'm not looking to light to scene to have 9 stops of range. I want to know how much range I can capture in one exposure without processing or altering light. Once I know what limitations are inherent, and set my expectations accordingly, then I can apply whatever tools I need to improve the final result.<br>

The chosen image is strictly a technical exercise. It has no intended artistic purpose.</p>

<p>Robert K: Thanks for the link.</p>

[syed]

<p>Starting from 18% (gray) reflectivity, one stop above would be 36%, another stop 72% (off white), and yet another stop 144% (more light than white can reflect, say light from the sun reflected off a shiny surface). Pure white stops at a 100%, an RGB value of 255. For all practical purposes and for everyday shooting, five stops (somewhat black to somewhat white) would be about all that is useful and can be represented on screen and on paper. Anything more and you'd hardly have any detail.</p>

[joel_covey]

<p>Confusing isn't it.<br>

First of all you do have a lightmeter--the histogram in your camera. Shoot to the right--the histogram is as close to the right edge. That's the basic rule--don't clip in your jpeg. Or if you must--don't clip much.<br>

In the RAW conversion you can recover "headroom", about a stop of overexposure. Exact amount depends on your individual camera. One of the dirty little digital secrets is that the jpeg conversions settings are designed to handle the voltage outputs of the <strong> least sensitive</strong> sensor the manufacturer will put in a camera. One of the many reasons to shoot RAW.<br>

As for your jpeg dynamic range--what you see in your almost clipped image is what you have. It's limited to 8 stops by the monitor (0-255 levels) and to 6-7 stops by your eyeball. You can't tell the difffernce between a 0,0,0 level and a 4,4,4 level, for instance.<br>

As for the dynamic range of your camera I hate to trash an icon of photography-dpreview-- but when it comes to the technical stuff they don't know crap. Fancy graphs, yes. A true understanding of what is going on, no.<br>

The dynamic range of your camera is limited by the number of bits in your A/D, 12 or 14 stops in most new cameras. Don't let any one tell you different, especially since it is so easy to check it out for yourself.<br>

The dynamic range of my relatively new D60<br>

<img src="http://farm3.static.flickr.com/2498/3801816918_5130ba410b.jpg" alt="" width="500" height="334" /><br>

Shot at iso 100 with an exposure of 1 sec. <br>

<img src="http://farm3.static.flickr.com/2665/3800996983_e9790f83c3.jpg" alt="" width="500" height="334" /> <br>

Shot at iso 100, 1/4000 of a second exposure. <br>

Noisy yes. But this was shot at the equivalent of iso <strong>400,000</strong> (12 stops less than than the correct exposure).AND it is still a recognizable image of shelf of books. And I suspect (know, done this before with other cameras) that I could pretty it up more with Noise Ninja.<br>

Procedures<br>

Set up the camera at its min iso and the lighting for around a 1 sec exposure. Shoot a series of images with the exposures one stop apart until you hit your maximum shutter speed<br>

Top image--ACR, normal image processing<br>

Bottom image--ACR, exposure, fill and noise reduction max, black min<br>

Opened in photoshop with a levels layer to brighten up the barely visible image.<br>

To sum up. Somewhere between the two extremes you will find an image with a noise level you can live with. That will deternines your dynamic range. Me- based on some of the low light images I shoot and the noise reduction techigues I've learned--I'd say my D60 has a dynamic range of 9 or even 10 stops. Better than I expected.</p>

 

[joe_c5]

<blockquote>

<p>As for your jpeg dynamic range--what you see in your almost clipped image is what you have. It's limited to 8 stops by the monitor (0-255 levels) and to 6-7 stops by your eyeball. You can't tell the difffernce between a 0,0,0 level and a 4,4,4 level, for instance.</p>

</blockquote>

<p>I have to disagree with this. Assume the monitor's contrast ratio is somewhere between 200:1 and 500:1. The monitor would limit the dynamic range to 8 or 9 stops, but that has nothing to do with the 8-bit (0–255 level) encoding. This encoding uses a gamma curve, so that levels near black are much closer together than levels near white. For a gamma of 2.2, there are over 15 stops between level 1 and level 255.</p>

<p>Human vision is much more complicated, but it can handle significantly more dynamic range than 6-7 stops.</p>

<p> </p>

[joel_covey]

<p>Joe C<br />I think we are talking about different definitions of dynamic range. Or I didn't express myself clearly.<br />First the vision. Find an image of Kodak 21 zone, 0-3 optical density step gauge on the web. If it is properly displayed on a monitor or with a projector I, like most people, can see a difference up to about step 14. That's a difference 2 optical density units or a range of a hundred to one or about 6.5 stops. That's what I was talking about--how many stops can you distinguish without the eye adapting to different light levels.<br />As for the monitor obviously you can map large dynamic ranges--the extremely low values of the 12 bit RAW, iso 400000 equiv image for instance-- to the much more limited range of a monitor. Not matter what you call it, contrast ratio, gamma curve, tone curve etc. it is still a mapping function. That is arguably the biggest strength of digital photography.<br />But if you can see a feature with a value of say 5,4,6 against a background of 1.0.1 you have far better eyes than i have. That what I meant to say.<br />One other comment. I ran a second set with the camera's noise reduction off to see how much worse they would be. Turns out they were cleaner. When pushed to this level the noise reduction algorithm adds noise stripes. In the second image they are leaning opposite the angle of the books. Definitely something to investigate.<br>

And in the procedure its brightness max not fill</p>

 

[joe_c5]

<p>Those observations about the image viewed on the monitor are correct, but the <em>reasons </em> are not correct.</p>

<p>Level 0 black on my LCD monitor is clearly brighter than the gloss black bezel, which may not be perfectly black either. The monitor limits the dynamic range based on its contrast ratio, and this has absolutely nothing to do with number of bits, encoding, or tone curves. On an ideal monitor using voodoo magic instead of liquid crystal, levels 4, 5 and 6 would be darker than the LCD's level 0, so all of these levels are expected to look similar if not identical on the LCD.</p>

<p>12 bit raw uses linear encoding, so there are actually 12 stops of dynamic range available. Level 1/4095 is 12 stops darker than level 4095/4095. Level 0/4096 is true black.</p>

<p>8-bit jpeg uses nonlinear gamma encoding, usually either at gamma 2.2 or the sRGB tone curve which is similar. For gamma 2.2, level 1/255 is actually 17.59 stops darker than level 255/255. For sRGB, level 1/255 is 12.3 stops darker than level 255/255. With either of these two gamma curves, 0/255 is true black.</p>

<p>12-bit raw does not offer any dynamic range advantages over 8-bit jpeg; the advantage is that the 12-bit raw will have less posterization due to using 4096 different levels over the approximately 12 stop range compared to only 256 levels. Any of the three file types can encode much more dynamic range than a monitor or photograph can display.</p>

<p>Essentially I'm saying that the monitor is doing all of the limiting. I think eyes are good for at least 10 stops of dynamic range.</p>