Understanding Active D-Lighting

Discussion in 'Nikon' started by tonybeach, Feb 24, 2008.

  1. Active D-Lighting (ADL) should not be confused with D-Lighting. ADL is
    essentially an image optimization setting that utilizes the Expeed image
    processor in the newest Nikon DSLRs; whereas D-Lighting is an editing tool in
    Nikon Capture and Nikon NX -- there is a difference between them and ADL appears
    to be superior to D-Lighting (although I will leave that to others to decide and
    analyze, and any contributions in this thread regarding that would be most
    welcome by me).

    ADL can be turned off in NX and is ignored by Capture One (again, I will leave
    it to others to do their own tests to determine how other RAW converters handle
    ADL files, but please pay attention to the rest of my post before contributing
    your conclusions -- specifically as regards metering). Inexplicably to me, ADL
    cannot be turned on in NX; so it is a "use it or lose it" option. Turning off
    ADL in NX is tricky though, you need to go to the Active D-Lighting tab in NX
    (not to be confused with the D-Lighting tab) and select "Off" and leave that tab
    checked, if you uncheck that tab NX continues to use your ADL setting and
    disables contrast and brightness controls in the Picture Modes.

    Something I really don't understand right now is why ADL cannot simply be turned
    on in NX. What confuses me is that ADL can be adjusted in NX and turned up or
    down, it can be turned off, but it cannot be turned on. This might suggest to
    some that ADL does something to the RAW data, but I have looked closely at the
    files and there is no difference between ADL and non-ADL captures.

    Another peculiarity of ADL is the role it plays in metering. On my D300 ADL
    changes the behavior of Matrix metering and as ADL is turned up the camera
    decreases the exposure value (EV); Normal ADL decreased the EV by 1/3 of a stop
    and High ADL decreased the EV by 2/3 of a stop. This explains the widely
    reported increase in noise using ADL, as it is attempting to preserve highlights
    by essentially underexposing and correspondingly raising shadows. Using Center
    weighted metering EVs were constant at all ADL settings (Low, Normal, and High),
    so obviously the camera is using ADL and Matrix metering together to determine
    what Nikon thinks is the ideal EV.

    My recommendation for using ADL then is to use it set to Low in conjunction with
    Center weighted metering (I set my D300 to "average" at b5 in the menu). This
    will allow you to use ADL in NX, to boost it or turn it off, and this will give
    you the most accurate histogram from the embedded JPEG. Of course, especially
    with high dynamic range (DR) scenes, it is best to shoot RAW as that affords
    about one extra stop of exposure latitude (that amounts to an extra stop of DR).

    For better understanding how much exposure latitude the D300 gives you can check
    this out: http://www.photo.net/bboard/q-and-a-fetch-msg?msg_id=00OPW5 The D300 is
    practically immune from overexposing a RAW shot with no exposure compensation
    (EC) using Center weighted metering. For my own settings which give me the best
    information about how my D300 is exposing, you can find my settings here:
  2. Anthony good idea to start a serious discussion about this topic. If it is all done in software after the data were collected why should the in camera software do a better job than software running on a PC? I noticed that I can apply ADL in camera after the fact. A clear sign that data collection is not part of it. So perhaps at best a thing of convenience for in camera jpg or just marketing hype? (This last comment will give you 30+ comments to this thread^^)

    I would not interpret anything into what NX can do or not - I gave up on this a long time ago. Considering the quality of the UI it could just relate to a bad lunch of the programmer^^.
  3. You may want to have a look here
  4. Thanks for the links. Very useful information to understand ADL.

    Juan Parmenides
  5. Anthony, the article Peter links to states "Its main reason (referring to active D-Lighting) for existence is to help combat dark shadows" which has been my understanding as well. While I have not noticed much difference using ADL (I have not really tested more than casually), I have found turning the contrast lower achieves the desired result. Can you explain the differences between the two approaches? Do you have any sample photos you can post?

    And thanks for the settings link/tips!
  6. "Something I really don't understand right now is why ADL cannot simply be turned on in NX."

    Here's my understanding from a Nikon rep last week.....

    Active-D Lighting can't be 'turned on' in NX because only the newer Nikon cameras support it. In other words, you can't 'turn it on' for shots from a D70 or whatever.

    The rep told me (as I have heard before) that NX is the only software that picks up all the data from the newer bodies because of their RAW conversion algorithms. I suspect this relates to ADL. I haven't tested this myself. Maybe it's true, maybe hype.

    His suggestion was to use NX to -- at the very least -- convert RAW files to TIFFs, then use whatever processing software you like. Obviously he uses NX and says you don't need anything else, though he did have a Nik plug-in.

    Again, I haven't tried this yet but it might be interesting to see if there really is a difference in preserving data using NX vs something else like PS.
  7. I have turned off Active-D (and sharpening) on the D300 after discovering they slow down NX processing considerably. I apply them after the image is loaded in NX. I Nikon's programmers ever get NX back to Ver. 1.2 speed I'll probably turn on Active-D on the camera.
  8. Active D-Lighting samples. Shot with a D3 and NEFS processed in Lightroom v1.3.1
  9. I did not see it mentioned above so I will add that there is still another D-Lighting which is
    available in at least the D80 body within the Retouch Menu.

    From the D80 manual, page 110:

    "D-lighting brightens shadows making it ideal for dark or back-lit photographs. Press the
    multi selector up or down to choose the amount of correction performed. The effect can
    be previewed in the edit display. Press OK to copy the photograph and return to the
    retouch menu or full frame play back."

    The D80 makes an edited copy of the original image, of course leaving the original image

    Sounds like something fun to play with on a long train trips or during extended airport
    layovers. ;<)
  10. "I noticed that I can apply ADL in camera after the fact."

    Hi Walter,

    From what I can tell ADL can only be applied to JPEGs coming out of the camera after the fact, even when the original shot was RAW. On the other hand, if you shoot using ADL in RAW, then you can turn it off or adjust it in NX prior to conversion.
  11. "I have found turning the contrast lower achieves the desired result. Can you explain the differences between the two approaches? Do you have any sample photos you can post?" Hi Elliot, Here is a comparison of a reasonably high contrast scene taken with all four settings: http://photos.imageevent.com/tonybeach/mypicturesfolder/sharing//Active%20D%20Lighting%20OLNH.jpg Here's the same shot and a comparison between the original, an edited version (which shows the edit used), and ADL set to Low from the camera: http://photos.imageevent.com/tonybeach/mypicturesfolder/sharing//Curve%20or%20Active%20D%20Lighting.jpg As for your question about the role of contrast, ADL takes over contrast and brightness and appears to attempt to maintain contrast in the scene.
  12. Thank you Peter for the link.

    Hi Ellis,

    Thank you for the contribution. I am wondering if your example shots were taken with Matrix metering. They appear to me to be exposed differently but Lightroom seems to have otherwise ignored ADL.
  13. yes Matrix metering and aperture priority mode: the top two frames exposure were f/10 @ 1/125; frame 3 (normal) f/10 @ 1/160; frame 4 (High) f/10 @ 1/200. Also keep in mind that for publishing here the NEFS were converted to level 8 JPEGS using the sRGB color space. Here are the Adobe Photoshop Lightroom histograms for the NEF converted to DNG photos those 4 sRGB JPEGS were derived from.
  14. Overall the differences are pretty slight but take a close look at the areas closest to the white and black end points. There was a fairly large contrast range in the subject and you can see in the the image shot with Active D-Lighting set to high that Active D-Lighting has boosted the signal down in the deep shadows (left end) and done the reverse in the upper highlights, preserving detail at the extremes of the exposure.
  15. Here is a photo I took while testing my new D300. I converted the NEF file in NX with defaults, and then turning Active D-Lighting off.
  16. And here is the same photo, with Active D-Lighting turned off in Capture NX. I leave Active D-Lighting turned on all the time as I feel it does indeed improve image quality and improves shadow detail. The same effect can be achieved in post processing, but I prefer not to post process hundreds of photos a day.
  17. From Ellis's histograms above and looking at his photos, it appears to me that Active D-Lighting reduces the exposure program when using matrix metering.

    I don't see the 'boost' that Ellis is referring to in the shadows, as the peaks in the shadow area of the histogram for high ADL are all to the left of the corresponding peaks in the lower ADL settings. Does 'boost' in this context mean the same thing as don't decrease exposure in the shadows by the same amount it is decreased in the highlights?

    Is there a way to confirm whether or not ADL changes the raw capture? I don't know of any utility that would let me do that without demosaicing the Bayer pattern.

    And if ADL doesn't change the RAW data, then how is it different from reducing exposure and applying a curve to bring up shadows?
  18. "From Ellis's histograms above and looking at his photos, it appears to me that Active D-Lighting reduces the exposure program when using matrix metering."
    I can tell you for a fact is not what is happening with Active D-lighting. While the change in exposure levels has a lot to do with decreasing the highlight exposure but the pixels that only are eighth and lower quarter tone (middle and deep shadow tones) exposed are more or less having a little more "gain" applied t oopen up the shadows while the highlight tone exposed photo sites are being throttled back to preserve highlight detail a bit. This is being done directly on the CMOS sensor.

    "I don't see the 'boost' that Ellis is referring to in the shadows"

    Then you are not lookign very closely at the differences in the left end of the histogram. histograms. If you can't see these differnces i nthe samples I posted I suspect it is becasue a.) JPEG compression crushes that a bit and b.) po the compression applied by photo.net is doing more damage c.) I didn't post 1:1 crops of the shadow areas, and d.) I have no idea how good anyone else's computer's display is.

    Active D-Lighting definitely does change the captured "raw" data but it is at the sensor level on a photsite by photosite basis before any other processing like Bayer color interpolation and coding is applied. is applied. BAsically it is looking at the signal and saying if signal is S then do Y and if signal is H then do Z.
  19. I suspect ADL enables each photodetector's analog gain response to be subtly changed on
    the sensor die on a pixel-by-pixel basis, as a function of intensity (non-linearly), before
    digitizing. If so, that's not something that can be "enabled" in post.
  20. Looks like I doubled with Ellis' last paragraph...
  21. Brad got right to the heart of the matter in his 10:24 post.
  22. Hi Tom and Ellis,

    I cannot say what is happening to the unadulterated RAW data; the fact is that for most of us we use our preferred RAW converter and it gives us what we have to work with from the file -- so the proof of the pudding is in the eating, or the conversion as it were.

    This reminds me of the 14 bit versus 12 bit debate, we are getting down to the most minute level of details and it is doubtful that any of this will ever show up in any print. Anyway, just in case anyone is wondering, I have been doing all my testing in 14 bits and despite its insignificance I choose to use it unless I am shooting faster fps.

    Following Ellis' approach of using ADL with Aperture Priority and Matrix metering I took a rather boring but predictable and repeatable shot, the primary virtue of it being that it had an extreme DR; then I took another shot using Center weighted metering with ADL turned off. I made certain that both exposures ended up being identical, they were: 1/10, f/5.6 and ISO 200. I shot them using MLU from a tripod, so everything appears as controlled as conditions would allow.

    After opening both files in Capture One I applied all the same adjustments to both files just as I would in my normal workflow. Both files responded the same way to auto adjustments and with some additional tweaking I came up with a result I found acceptable and optimal for both files. Upon close inspection I could see a difference in the shadow detail and I isolated that difference to the red channel.

    Based on this testing, there does appear to be something happening to the RAW data when using ADL -- at least inasmuch as it is handled by Capture One (there is far too much going on in NX to untangle what is actually happening to the RAW data before it becomes an NEF file using that program). The news as I see it is not good for ADL. The aforementioned red channel clipped earlier using ADL than it did without it. In the link here: http://photos.imageevent.com/tonybeach/mypicturesfolder/sharing//C1_AB07462_1.jpg you can see that the red channel is more clipped in the left crop which is the ADL shot than it is in the right crop which is the non-ADL shot.
  23. Anthony,

    It looks like that is a queen of spades taped to a lampshade. is the lightbulb in the
    lampshade a low color temp household tungsten) ? If so and if that was the only light
    source in the photo that could account for the red spike if that was the case -- there is
    going to be a lot more red channel data n the photo than blue or green.

    Choice of raw processors is critical as you point out . I don't use C1 Pro anymore so i don't
    know how it handles raw data or what types of processing it does as default, I know you
    used to have to set sharpening or USM down to -132 to actually turn it off completely,
    among other things. Capture NX as a default uses however your camera was set up as the
    default for each frame. Adobe Camera Raw 4.3.1 , the raw processor in Lightroom v1.3.1 ,
    looks at exposure and color temperature and seems to ignore ignores most other camera
    presets like sharpening, etc. This might be one reason why Lightroom is faster than
    Capture NX -- it simply doesn't have to do as much processing on each file. It can't
    ignore Nikon's Active D-Lighting for the reasons given above. I suspect that somewhere
    in the raw data fro ma D3 and D300 and in other future Nikon's that will use it is some
    code that says "Active D-lighting was used with this setting and here is a map of how it
    was applied to every photosite.

    (This is getting of of the subject but I notice that I have far fewer problems with color
    fringing (Chromatic Aberrations) with the D3 and D300 even with lenses I have seen it
    with in the past. This leads me to believe that this is also being handled on the CMOS
    sensor on a photosite by photosite basis.)

    I also agree that we are really getting down to minute but critical details here. Of
    depending on how you look at things either God or the Devil is in the details.

    For the record, the only processing I applied in Lightroom was my baseline settings of
    "capture sharpening" at Lr's Landscapes preset and Clarity at 10. the watermark was
    applied upon export as a JPEGS (level 80%, sRGB) in photoshop I composited all 4 images
    and then used Photokit Sharpener set for web/email 1024 super fine edge. The composite
    was saved as an sRGB level 8 JPEG.
  24. Tom, if you would look at the actual photos I posted in this thread, not some computer graph, you would see the obvious boost in the shadows that Active D-Lighting provides.
  25. Hi Ellis, I'm going to investigate the tungsten lighting issue -- there was also some fluorescent lighting coming from behind me, so it was really "wicked lighting". Nonetheless, this is lighting that we sometimes have to cope with, but I want to make sure the problem isn't coming from the hertz cycle and that may have affected the ADL exposure more than the non-ADL exposure. If I see different results I will report back to this thread; otherwise I will stand by my initial findings. Regarding Capture One, version 4 has changed from version 3.7, and sharpening can now be simply turned off or set using USM parameters. That shouldn't be an issue though as the sharpening and NR settings were the same for both the ADL and non-ADL files. You and I agree that something appears to be happening to the RAW data when using your settings (Matrix metering and ADL set to High). Our disagreement may come down to its significance whether it is a positive or negative effect. I had an opportunity today to do an afternoon daylight test. The results are even more subtle than last night's "wicked lighting" test. I would say that pixel peeping today's images yielded an extremely slight advantage to your settings over mine -- but the difference was so slight that I would personally be more inclined to stick to my approach because it gives me better information with which to evaluate exposure (see the link in my original post regarding evaluating exposure for specific details). Below are three histograms illustrating the difference in the test shots I took this afternoon. The full images (resized for web viewing) can be seen here: http://photos.imageevent.com/tonybeach/mypicturesfolder/sharing//Untitled-1_2.jpg After doing an adjustment to the luminosity tonal curve the differences become practically indistinguishable, but there are some and which one is better and which one isn't probably becomes a matter of individual preference -- I do not think the pixels suffer using either approach. Thanks again Ellis for your contributions to this thread, I would otherwise never have discovered the Matrix metering connection to ADL.
  26. Thanks Anthony.

    I think the real advantage of Active D_lighting is pretty straightforward: give the photogrpaher more usable data to work with.

    Fluorescent flicker at shutter speeds shorter than 1/60th I've to be a problem , notab it for exposure but more so for Auto color balancing whether in camera or at the procesing ( if shooting raw) or in post processing if shooting JPEGS).

    it is also worth noting that Active D-lighting possibly ( I haven't tested this) is applied differently at diferent ISOs. ISo settigns on a D-SLR are really signal gain / amplitude settings.
  27. ISO settings? Oh no Ellis, now you're killing me. I'll leave it in your capable hands to figure that one out. One thing is certain to me though, base ISO which is ISO 200 is going to have the most DR.

    It has also occurred to me that I may not be pushing the DR enough in these test shots; while that could make a difference, I suspect that the results would be consistent with what I've seen so far and a curve adjustment in post processing should be able to accomplish very nearly identical results.

    Now I have another issue that may or may not answer the question about where ADL is being implemented by the D3 and D300. Recall that you can apply ADL to RAW files in the camera but they can only then be saved as JPEGs; this suggests to me that ADL is being performed by the camera's ASIC and not by its CMOS sensor circuitry. On the other hand, ADL performed on JPEGs could be different in some way from ADL performed to RAW files; otherwise we should be able to assign ADL to the RAW file after the fact. Maybe you can ask Nikon what the hell is going on here.
  28. Sony's five levels of DRO seem to be more useful and obvious in offring a longer tonal scale. Look at the example in this A700 review to see it in action:

  29. >>> Sony's five levels of DRO seem to be more useful and obvious in offring a longer tonal
    scale. Look at the example in this A700 review to see it in action:

    As can be seen, there are severe consequences of changing the linearity of the pixel's
    response. No doubt why Nikon kept the change subtle.
  30. "Recall that you can apply ADL to RAW files in the camera but they can only then be saved
    as JPEGs"

    As that is simply not true, you are starting from a false premise Anthony.

    "Maybe you can ask Nikon what the hell is going on here."

    i have. Twice. Once in Tokyo back in August when the D3 AND D30 were officially
    announced and once again a few weeks ago when I was writing up my review of the D3 for
    "Professional Photographer".

    What Capture One may be doing with your files is a mystery.
  31. Ellis, what I meant to write was that if you apply ADL after the shot has been taken, it is applied to a JPEG but does not alter the already captured RAW file. What is baffling me is why you can presumably apply ADL to a RAW shot that was not set to ADL for producing a JPEG from the camera but you cannot change the RAW file. I'm not sure what is flawed about that premise, unless what is being done to the out of the camera ADL file is different (which I also suggested in my previous post).

    Did Nikon have a response to your inquiry?

    As for what Capture One is doing to ADL files, it seems to recognize that the ADL RAW files are different from the non-ADL files, which I suspect means they are; how that might differ from how Lightroom (ACR) handles ADL files might merit further investigation.
  32. Hi Brad,

    Sony's DRO implementation appears different than Nikon's ADL; it says in the linked review that, "If you're shooting in RAW mode, you can adjust this settings using Image Data Converter, as well." Now I'm not sure how different that is from Nikon's implementation because you can also do that with the D300, but only if ADL has been set prior to taking the shot. Since both companies appear to be using the same sensors in the A700 and D300, this lends weight to Ellis' suggestion that this is actually happening at the sensor level rather then afterwards.

    Addendum for Ellis:

    I'm getting a little clumsy here, in the last sentence of the first paragraph in my previous post I should have specified that I was referring to the out of the camera ADL JPEG file being in some way different from what Nikon is doing to ADL RAW files.
  33. If ADL is not turned on, and you clip some highlights, there's nothing in post that can be
    done to recover the blown highlights.

    If ADL is turned on, and the top end has been compressed to avoid a highlight clip, then
    operations in post can expand what was compressed. There will be non-optimal trades in
    the process.

    Just my guess on the above, I'm speculating on how ADL works...
  34. What if 1/160 is not 1/160? What if 1/160 is 1/150? The EXIF data would still say it was 1/160. Why when I started this thread did I say there was no difference using ADL? The answer was because I was using Manual Exposure mode. So I went back and did two things to this afternoon's test images: 1. I adjusted the ADL file to -.12 EC in NX. 2. I turned off ADL. Just like that, what I got was the histograms you see below, of two images taken 55 seconds apart. My conclusion is that the actual exposure of the ADL shot was not 1/160, it was 1/150 and that did not show up in the EXIF data. So there you have it, no change in the RAW data.
  35. Anthony,

    per your 4;14AM (man what what you doing up then?)post:

    You have now proved that by doing more work in processing you can produce a histogram that is a near match for what Active D-Lighting is.

    Of course a histogram only tells what the global distribution of tones is, not the individual pixel values.


  36. One more note:
    "Beware the paralysis of analysis."
  37. Hi Ellis,

    4:14am is Eastern time, but I'm on the west coast which is three hours earlier. Of course, I'm now up because I've been sick lately and couldn't sleep -- part of the reason I have had time to dedicate to this topic these last two days.

    What I have proven (to myself anyway) is that the difference between using ADL and not using it is EV, in the particular test shot it was a difference of .12 of a stop. In fact, once I realized what was going on, the only post processing I had to do was turn of ADL and adjust EC (I consider that "pre-processing" as it is performed before conversion, essentially analogous to what is done to film before and during development).

    There are consequences to changing EV, even slightly. WB is usually altered, by moving the EV significantly towards underexposure which is one of the things ADL does (especially when it is set to High), the WB is set more towards the shadows than towards middle gray, and that could one of the biggest impacts on apparent changes in image quality.

    I can testify to the equality of the pixels between the test shots I was working with in the daylight shot. This also explains the differences in the lampshade shot, as subtle changes in the shutter speed would be most noticeable under fluorescent and mixed artificial lighting.

    Finally, regarding "paralysis of analysis", it was a significant effort on my part to learn something here that I hope others can benefit from. Every auto setting takes control away from the photographer and gives it to the camera. Some of those auto setting changes are so subtle they are easily missed and cannot even be identified looking at the EXIF data (stops of less than 1/3 of a stop for shutter speeds and apertures, 1/6 of a stop for ISO when using auto-ISO). ADL is an auto-setting and one of its effects is to change EV, even when you fight that auto setting by adjusting EC, you can only get it within 1/6 of a stop and the camera will select its own EV at increments smaller than that.
  38. Anthiny to sorry to hear that you are sick. Get better soon.

    I am not knocking your efforts. I applaud them; but sometimes we all find ourselves missing the forest for the trees.
  39. Thanks Ellis.

    Speaking of trees: http://photos.imageevent.com/tonybeach/mypicturesfolder/february2008/large/C1_AB07500_web.jpg I got that one yesterday just before the sun faded -- I thought it was kinda pretty.

Share This Page