I've been noticing for some time that the white balance on my D800 tends to be significantly different from my D4. OK, I lack a grey/white card to tell for sure which is right, but my eye tells me the D800 is the one that is wrong.
The attached three shots were D4 then D800 then D4. They are all "as shot" in Lightroom. The settings lightroom interpreted as as-shot were:
The reason for the D4 a second time was to make sure the light did not change substantially. These are through a 200/F2 @ F16 on a tripod, and taken sequentially as quickly as I could switch bodies. They are at ISO 100.
White balance is A1 on both. I have checked a dozen times to make sure I didn't have one on A2 (warm). I went further into the adjustments for "A" and it is centered on both.
Both cameras were on Matrix metering (but that doesn't seem to matter), and on the center point.
The D4 exposed slightly differently also, at 1/30th vs. 1/25th for the D800, and I did not force it either pre- or post- to be the same, so the center is slightly brighter in addition to being a different color.
This is pretty consistent. I see it worse at night, e.g. shooting two bodies for soccer, but the first thing I do in soccer is go wipe out the "as shot" anyway. In fact most sports I end up tweaking - bright uniforms, water, signs. So this hasn't been a big issue.
But I finally decided to dive in and see if I had a setting different.
And I can't find one.
What might I be missing? Or might these two cameras be just that far apart?
>You not only have possible differences between cameras, you >have possible differences in how LR reads the two cameras. I >don't think Auto WB is always that perfect no matter what the >camera. > >In any event, it's very easy to change WB in LR, if you shoot >RAW.
I do shoot raw, and that's why this isn't a big deal, I am always adjusting white balance anyway. But it's a LOT of difference.
It's not lightroom, here's the same sample with ViewNX:
Hmmm... no. They are packed up and I'm running out to a soccer match, but I will. I ignore that mostly due to shooting raw, but maybe it's possible. I know a too high setting there can affect metering. But because of that I thought I had it off or low on both. but will check.
While the LR settings won't be the same, they should look visually similar. Have you compared the EXIF data in View NX2 for the images to see if something is different in the WB settings? It's pretty good at providing details regarding settings.
>While the LR settings won't be the same, they should look >visually similar. Have you compared the EXIF data in View NX2 >for the images to see if something is different in the WB >settings? It's pretty good at providing details regarding >settings.
Yes, I did that early on. I can't see in ViewNX how to tell the recorded value, but the recorded settings both say A1. I've flashed back and forth between images looking at what changes in ViewNX, and the only thing I noticed is HighISO Noise Reduction was on for the D4 (which I ignore since I don't shoot jpg, and this was low ISO).
I may do some more tomorrow in a more controlled environment (i.e. not outside but inside under unchanging light).
Incidentally if I force the shots to the same temperature in lightroom they look almost the same (the 25th vs. 30th exposure is a slight difference).
>I can't see in ViewNX how to tell the recorded value.......
As far as I can tell the View NX and Capture NX interfaces only show the general white balance mode used by the camera (e.g. Auto1, Preset D1, etc) - I can't see anything that gives the WB in actual degrees Kelvin. That doesn't mean it's not hidden somewhere in the exif though.
Your d800 and d4 are clearly generating different white balances with your raw nef images of the same subject matter under the same illumination using the same lens and the same importing software. The variation in mired degrees kelvin of your d800 clearly shows a shift toward blue on the standard blue to amber mired-axis. I agree with others that the pix you have posted, as rendered, show the d800 to be farther off-axis toward the blue than the more amber-shifted d4. There are possibly also a number of other variables (that you have not specified) regarding your camera metering and camera focus settings of your d4 and d800, that could be introducing this blue shift with your d800, and not with your d4? I would suggest setting a white balance manual preset for both your d800 and d4, with the white card placed under the same general illumination. Easiest is to place the white card, if outdoors, within the shade, so no direct sunlight illuminates the card. You will be creating two different manual white balance presets from the same white card under the same kelvin illumination. Once you have saved these presets to your d800 and d4, then at your leisure, take photos with both cam bodies using the same lens of any subject matter (preferably a scene under the same kind of skylight shadowed illumination you have created your manual presets for - but it doesn't matter - we are just trying to test your cam bodies' white balance algorithms to see if the same image under the same degree kelvin light produces a similar white balance for the different bodies, all other things being equal) Assuming you have used the same camera metering and focusing settings for both d800 and d4 manual white balance presets, the images you create on the d800 and d4 at those identical white balance presets should produce raw nef images - such that when those images are rendered through lightroom or nikon software, the images from the different bodies should have a very close white-balance resemblance? I would also urge you to record duplicate jpeg optimal fine images along with the nef images. And to transfer those same optimal fine jpeg duplicates through your lightroom and/or nikon software to your computer along with the raw originals. The jpeg fine images will have all of your nikon proprietary manual preset white balance settings 'baked-in' and your image importing software should not touch these jpeg white balance settings on import. Your imported jpegs should have the approximate same white-balance as well as your raw nef rendered imports. If your white balance is still greatly different between your d4and d800 images with your software imported jpegs - just go back to your d4 and d800 and reset all your cam settings to factory default, and try this same manual white balance preset setting test once again. But remember, there will always be slight white balance differences even among the same nikon cam body models, as well as between entirely different cam body models. I rent different bodies and different lenses all the time, shooting with manual white balance preset using strobes and a mix of sunlight and skylight and reflectors. But even using on-location manual white balance presets based on either white or grey card references always produce slight, but still noticeable white-balance differences between flesh tones from the same model cam bodies. good luck
I think you are right, you need to get a white-balance card - partly to simplify the target scene.
I don't have those camera bodies, so I am not sure what "on matrix metering ... and on the center point" means, but metering right on the edge of a white car hood and a green lawn may be emphasizing any WB algorithm differences in the camera models.
To me "matrix metering" means averaging over the frame and so I agree with you that it shouldn't make any difference exactly where the focus point is aimed . . . a white-balance card will give more clues.
Thu 17-Oct-13 10:53 AM | edited Thu 17-Oct-13 10:56 AM by Gromit44
Linwood - have you got Photoshop?
If so, the Adobe Camera Raw dialog box shows the white balance of a NEF file in degrees Kelvin. It's the colour temperature as set by the camera and it seems to work for all of the D800's white balance modes (auto & preset manual), although I haven't tried the 'K - choose color temp' option.
There's a slider below it marked 'Tint' - I'm not sure whether this has come from the camera's exif data or it's an adjustment suggested by ACR itself.
Page 349 of Thom Hogan's D800 guide: Thom advises that the Kelvin values reported by RAW converters are not to be trusted. He says Adobe converters in particular report values that he does not believe to be correct.
Years ago, Nikon implemented a practice of encrypting their white balance settings to make it difficult for others to work with their raw files and force people to use their software. Some hacked their encryption, whereas Adobe and Nikon eventually came to an agreement that Nikon would provide the means for Adobe to get white balance info via a Nikon-provided app. This happened because Nikon got a significant black eye in the process. How white balance info is displayed in different programs varies because they work with a variety of camera bodies, and you'll find that different brands and different models are all displayed differently in different raw converters. It doesn't necessarily mean the visual results are different. It just means the temperature and tint settings that are displayed differ.
What's happening with Linwood's white balance doesn't have anything to do with this interesting artifact of raw processors. It's something else yet to be determined.
I can understand Nikon not wanting to give away info to their rivals but it doesn't help those of us who bought Capture NX2 and Nikon cameras. We should be able to see the exact Kelvin value in the NX2 interface, rather than some vague 'group' setting.
If Linwood could read the exact values he'd be able to narrow down the possible reasons for the dramatic difference in colour between shots 1&3 (the D4) and shot 2 (the D800). As it is, NX2 is no good for his purpose and the values given by ACR are probably wildly inaccurate.
OK, experiment #2. I'm going to forego looking at K values as it sounds like that may be confusing between reporting software.
Here's what I tried -- in a darkened room, I lit a subject by incandescent light only, so it had a steady temperature through the experiment.
First I took a white(ish) door and did a pre-set from it for each, labeled "D-1".
Then I shot 5 images of the same scene. I used an 200/f2 which IMO renders colors more accurately than any other I have. I shot in order:
Auto-1 Auto-2 Incandescent Forced Kelvin = 3750 (chosen at random to be similar to the light) Preset D-1
I did it with the D800 (top) and D4 (bottom).
The short version is that by eye they are all the same but the preset. The preset on the D4 is MUCH warmer (too warm). Since I don't use preset generally I am not too worried, but it is a bit strange.
But all the auto and the two forced presets (Incandescent and K) match.
Which is very different from what I saw outside. It's cloudless outside, let me go do something then. Though in doing this I noticed the fine tune is way off on my D4 and 200 -- off on a tangent for a bit to check it at more reasonable distances.
I'd agree, Preset D-1 on the D4 looks very different to Preset D-1 on the D800 - but I can't see how you'll ever get to the bottom of the mystery without knowing the actual Kelvin values used by each camera (assuming all other camera settings are the same).
I set up a target on a driveway. Is it a good target -- don't know, but it's a pretty plain scene, the entire view (and I was on matrix metering if it matters -- I still don't know for sure). It was bright sun in a cloudless blue sky, so the light was very constant.
I took four shots each, just to confuse you (but really because I didn't swap cameras first) the D4 is on top, the D800 on the bottom:
Auto-1 Auto-2 Direct Sun Force temp = 5000 K
The D800 looks warmer, and if you look at the histograms there is a pretty consistent red (and only red) shift:
At first I thought that one was more sensitive generally, and it was an exposure difference, but what I'm seeing is the D800 having more red, and less yellow from what is otherwise the same image (on auto, so it is perhaps shifting). These are from ViewNX by the way.
Adobe shows those two at (4400,-7 for D4) and (4750, +3 for D800).
Below is the histogram for the fixed "K" setting, where in theory the color rendering on conversion should be identical. They are closer, but they are not the same -- the red is shifted a bit higher on the D800. That's consistent throughout the shots.
It seems as though the sensor is actually more sensitive to red on the D800? Is that how to interpret this?
But perhaps NOT on the AF sensor, which sets the "auto" (and maybe a bit the reverse), so the result is auto is warmer?
And in warmer (lower temperature, oddly) light like incandescent it doesn't matter as much in some fashion?
I went back and looked at the indoor shots, though, and the red peaks are shifted right on the D800 there also, including on the fixed presents like incandescent.
I guess if there's a conclusion (unless someone has better ideas), there appears to be a real difference in the two cameras, it isn't a settings difference. I've been through things so many times I think I can say that for sure.
I could try to adjust for it, but it appears to be worse at higher temperatures, so decent incandescent, significant at sunlight, and probably worse under florescent though I haven't tried (it is worse under arc lights but I'm not quite sure why as they average a lower temperature).
So I'll probably leave it alone, as most of what I do is sports and I manually adjust anyway. I usually just blow in a specific temperature that I've figured out for each venue, then adjust as needed for different areas and cycling.
But there's definitely a difference here, subtle at times, more extreme at times.
On a related note -- anyone know what area is analyzed for color temperature? And I'm leaping to the conclusion it's done at focus time, maybe it's calculated from the raw image as it is processed?
>Auto 1 and Auto 2 are now very different in the driveway >shots - whereas they weren't in the bookcase shots. If it were >me I'd ask Nikon why.
You mean just A1 vs A2 for same camera? Both cameras seem to show relatively little difference outside.
I think that's what is expected; A2 is listed as "preserve warm tones", which to me means for incandescent it leaves it a bit warmer, but I would assume for daylight there should be little difference. Which is correct, right?
>Yes, I agree. It appears the difference in the cameras comes out as the color temperature goes up (cooler).
Yes, that's what it looks like.
The problem is it might not be a difference in the way the two cameras are handling white balance (i.e. the D4 is applying a different K value to the D800) - it could be something else we haven't even thought of.
If we knew the Kelvin values were exactly the same for each pair of shots (D4 vs D800) we could rule out white balance.
No, I'm assuming the exterior light source stayed the same while both cameras did the driveway shots - and the interior light source stayed the same while both cameras did the bookcase shots. So (for instance) if the Kelvin value of D4 driveway shot A1 proved to be exactly the same as the Kelvin value of D800 driveway shot A1 - then we could rule out white balance errors for that pair.
But we can't test for WB differentials because we have no accurate method of viewing the Kelvin values on any of the Auto, Incandescent, Direct Sun or Preset D-1 shots. The only WB mode we can read is Forced K in Capture NX2.
>Now is the sensor (whatever is doing the WB calculation) >looking at exactly the same part of the image, don't know. >But that shouldn't matter when I set the WB Kelvin in the >camera, right?
We don't know if WB is done using the main imaging sensor or the 91,000 pixel RGB metering sensor. I don't think it matters provided both cameras are seeing exactly the same image under exactly the same light (and they're both using the same lens, aperture & shutter speed).
>We don't know if WB is done using the main imaging sensor or >the 91,000 pixel RGB metering sensor. I don't think it matters >provided both cameras are seeing exactly the same image under >exactly the same light (and they're both using the same lens, >aperture & shutter speed).
Well, not exactly. If (hypothetically) the main sensor on one camera is recording a bit warm, or a bit cold, and if (hypothetically) the wb is done by the AF sensor, and if (hypothetically) it is the same and not customized to the camera, then sensor recording difference will show up on the result.
If the readings are from the imaging sensor, and use the same algorithm in the two cameras (again, hypothetically), you would expect any difference in recording by the sensor to be wiped out by the calculations in "auto" but not in the present temps.
Another thing you might try is making a best guesstimate of the cold temperature and set the cameras to the same temperature and compare images. Even if the colors are not right, they should be nearly identical and it would evidence any differences in the cameras and not their settings?
Scott Chapin Powder Springs, GA, USA Nikonians Team Member
I'm not certain but I don't think the AF sensor has anything to do with WB.
As far as I can tell the D4 & D800 have three sensors:
1. The main image sensor 2. The 91K-pixel RGB sensor 3. The AF sensor module (Multi-cam 3500FX)
Nikon's website seems to indicate that WB is done by a combination of sensors 1 and 2. It says: "With the 91K-pixel RGB sensor and the image sensor working together, the camera renders white as white with supreme accuracy." (Source page - http://imaging.nikon.com/lineup/dslr/d800/features01.htm)
Even if sensors 1 and/or 2 vary slightly between samples, surely any differences would be ironed out during WB processing. In other words all D4/D800 camera bodies must have the same set of fixed reference points built in, otherwise the K values in the menus (and p145 of the manual) are meaningless. In theory, 4K on my D800E should look identical to 4K on your D4 if both cameras take the same shot under the same lighting. Ditto for all other K values.
Indeed, but... Nikon comes close to being alien technology left for the natives to puzzle over and experiment with. I've used Nikon since the early 70's, and never seen them step forward and discuss questions and issues with their customers, have you?
No, I've never seen them step forward on a forum like this but I suspect one of their experts would be happy to explain things over the phone (assuming we knew where they were located - the serious experts are probably all in Japan!).
Your latest tests with your d4 and d800 with the bookcase incandescent and the driveway mixed sunlight and skylight show that your d4 and d800 are reasonably close in reproducing the actual scene-illumination color temp. As I mentioned earlier in this thread, these slight differences should be expected among different cam bodies of the same nikon model, as well as between cam bodies of different nikon models. However, the differences your manual presets are showing between the d4 and d800 with the bookcase under incandescent lighting are clearly greater than should be expected. Most likely due to the minor difficulties involved in properly doing a white-card or grey-card manual preset. Remember that the white or gray card need to be placed within the lighting that is incident on the scene. Just zoom into the reference card, keeping camera position the same, and make the manual presets with both cam bodies using the same lens. I always do my white/grey card reference shots in manual exposure mode, and zero-out the exposure on the cam body meter before pressing the shutter and setting the white balance. The easy thing to remember, whenever using multiple cam bodies, on location or in the studio, is that since the white balance is always slightly different between cam bodies, just set your cam bodies manually to the same white balance option in both cam bodies WB menu. Of course you will need to fine tune your white balance later between different cam body images in post-processing. But do not use any auto-white balance settings, as they will exaggerate the differences among and between the cam bodies.. Select an actual color temperature for your on-location scene and set the same temp for all your cam bodies. On-location, I always use a lot of strobe mixed with skylight, sand reflection, water reflection and whatever else I throw in, so I always set the different rental nikon bodies to the cam body manual white balance setting for flash, which is 5400 degrees kelvin, which is an option available on the d4 and d800 and d3x and every other nikon body I have ever worked with. Of course there are always slight differences between cam bodies regarding color temp, but they can all easily be smoothed out later with whatever image processing software you may be using. And if you are not using flash as your main illumination source, just set all cam bodies to the same white/grey card manual preset and you can correct the expected differences later in post-processing. Thanks for all your white-balance testing and thanks for taking the time to post your efforts, all your work can only help make all of us better nikon shooters. good luck.
It took me a couple of days to do a comparison between my D800e and D4 because of weather. Shooting in sunny conditions (which seemed to be where the performance was most different), I got the same auto white balance results from those two cameras - I can't see a difference. As I mentioned above, the exact temperature and tint shown in ACR/LR is slightly different, but that's normal between different camera models or brands.
>It took me a couple of days to do a comparison between my >D800e and D4 because of weather. Shooting in sunny conditions >(which seemed to be where the performance was most different), >I got the same auto white balance results from those two >cameras - I can't see a difference.
Thank you. I suspect I have some sample variation in the camera. One of them. One day I'll get either a color checker of some sort and do a bit of testing.
Or... I wonder how meaningful it would be to shoot a monitor with a specific color pattern on it, then load both images and see if they come out at the same color pattern. If I get far enough away that I don't see the rgb pixels, that might be interesting. Maybe this isn't about the AWB calculation so much as different color or intensity responses of the sensor?
This is more curiosity than a real problem I need to solve. It's not far enough off to really be sure which camera is wrong anyway. Though I did some soccer shots the other day for daylight, mostly D4, and they really seemed too cold. So I'm starting to think the D4 may be the culprit (if there is a culprit -- I could simply have two cameras from opposite ends of the bell curve).
Some of you older darkroom folks who used to print color, and older photogs who shot medium format color negative or transparencies surely must remember the macbeth color chart? This remains a very useful tool on location or in the studio or wherever. For those of you who feel more comfortable just shooting with auto white balance DSLR nikon cam settings, the color chart helps immensely in post-processing when trying to match the colors accurately from different nikon camera bodies. Remember the days when shooting color neg, we simply placed the macbeth chart in the frame of the shot, made one exposure under the lighting we were using, and then pulled the card for the remainder of the shoot. Once the negs were developed, it was easy to use the macbeth chart in the image frame to set ballpark color-filter settings on the enlarger. The 'old' macbeth chart (8.5x11 inches - roughly the same 35mm aspect ratio for still and cine) remains available: http://www.filmtools.com/maccol.html?gclid=CIbippm0o7oCFQZyQgodlnsAnA You can also get the chart from x-rite through adorama. If you are going to use the macbeth chart, you also should buy grey and white reference cards to standardize your manual preset white balance settings. But mount all these cards on larger 4-ply museum quality mattboards to keep the cards from bending and warping.
Well, I shot Golf today, which unusually meant I was shooting the D4 and D800 together, about equally.
It's the D4 that's off. Quite far actually.
Bright, sunny daylight and it pretty consistently shot at 4500-4700 degrees (as read after the fact in Lightroom). The D800 was about 5000-5500, about where I expect and more to the point it looked right. Setting it with the dropper in lightroom kept it very similar to the D800.
About half way through today I switched the D4 from A1 to A2, without much if any effect (probably correct as it's aimed at warmer tones at lower temperatures).
The difference frame to frame (between the cameras) in similar scenes was huge, so I'm having to adjust all the D4 shots.
One of the few times I've really relied on auto white balance, as usually I'm on a particular field or place, not roaming a fairway. Kind of a pain. Going to have to see if I can use the fine tune to adjust (I've heard it is pretty finicky as you need larger adjustments at once end than the other, so not sure).
>So is the D4 only off (i.e. too cool) when using Auto WB >under sunny daylight conditions? > >If so, fine tuning Auto WB (by tweaking the A-B axis to make >it warmer) might not work since it'll affect all lighting >conditions, not just sunny daylight.
Yeah, maybe, though I shoot mostly sports, and most non-daylight sports I'm going to be using a preset white balance, so that's less important (if shooting family or friends or something inside I'd be using the D800).
I made a by-eye estimate today on the golf course, using the LCD. I'm just starting to look at it now in lightroom, and they are much closer - warmer and more saturated. I'm not sure it is right - they almost seem a bit over-saturated (I moved the select point directly right toward yellow/orange, but what resulted was strongly more green saturation as well).
I need to do this more carefully in a more controlled environment, just been spending four days in a row shooting all day tournaments, so haven't been able to actually pursue carefully.
But I think I might be able to fine tune it enough, and because I don't need it much at night, I might not care. Also, a lot of scenes at night that I might use it (say street scenes) often have weird colors to begin with, so "right" is a bit less well defined.
I agree that if you are not going to use a white-card manual preset for both cameras, shooting in the same lighting situations, just use the SAME in-camera white-balance factory setting default menu-option for the d4 and d800 that most closely match your actual scene illumination color temp. As you have clearly shown by your careful tests, the color temp differences between your cameras are in the "ballpark", and you can fine-tune/match the images color temps later in your post-processing.