Auto White Balance - is it really all that bad?

[ATJ]

I don't use Auto White Balance and generally wouldn't because I can't see how the camera can possibly get it right, especially if the scene does not average to grey.

I was in a discussion on a UW photography forum about using in underwater and I would think underwater scenes would cause even more issues. Anyway, I just did some experimental shots with my D300 (above water) but was surprised how right the camera got it - in fact, I can't even work out how it did it.

I initially did it with Sunlight, but the varying cloud cover today meant that I couldn't trust the results, so I redid it using the SB-800 as a light source. I took shots of my gray card, and then shots of various coloured objects such that the object dominated or filled the scene. I then loaded the images into Lightroom and look at the White Balance settings that the camera chose. In all cases, regardless of the colour of the scene, the white balance settings were either identical or extremely close - so close that you'd not be able to pick the difference with your eyes.

How does the camera do it?

[Murray Foote]

I don't know but I understand accuracy with auto white balance essentially started with D3/ D300/ D700.

Regards,
Murray

[gstark]

I understand accuracy with auto white balance essentially started with D3/ D300/ D700.

That's essentially true, and the 5D2 is also bloody good in this regard.

That said, there will still be likely to be variations across different images shot under a given set of conditions, and to me, that means potentially some PP issues that I need to be aware of at some point further down the track, and that's where it loses my favour.

[ATJ]

What would those "given set of conditions" be? As I said, I was getting extremely consistent results across a wide range of simulated subjects and those results are what I would have considered as accurate as the white balance of the gray card was correct.

[mickey]

reflections of different colour light on skin tones is what I usually have to change AWB for otherwise it's not too bad. Such occasions are on fields of grass with the subject close to the grass there's alot of green tinge. Sometimes the camera will choose red tint to compensate for all the green when it really was that green.

Also, bear in mind that LR's raw converter isn't fantastic at getting the colour right, although the file itself might be right given a grey callibrated image. But because of LR you sometimes have to change the colour a bit anyway to compensate.

At the end of the day it's why you shoot in raw

[ATJ]

reflections of different colour light on skin tones is what I usually have to change AWB for otherwise it's not too bad. Such occasions are on fields of grass with the subject close to the grass there's alot of green tinge. Sometimes the camera will choose red tint to compensate for all the green when it really was that green.

That is what one might expect but I am not seeing that at all. Please reread my comments above. I shot a variety of subjects and the white balance was nearly identical if not identical for all of them. I'm not talking about Lightroom getting colours right. I'm talking about it merely reporting the white balance values stored in the raw file.

For example, see below. The colours are for close-up shots of the colour squares on my mini ColorChecker chart. The square took up around 50% of the image.

Grey card: 5900K -3
Cyan: 5900K -3
Orange yellow: 5900K -3
Bluish green: 5900K -3
Moderate red: 5900K -4
Red: 5900K -3
Yellow: 5700K -2
Green: 5900K -3
Orange: 5700K -7
Light skin: 5650K -8
A green bag: 5650K -2
A yellow envelope: 5650K -2

[gstark]

What would those "given set of conditions" be? As I said, I was getting extremely consistent results across a wide range of simulated subjects and those results are what I would have considered as accurate as the white balance of the gray card was correct.

Maybe.

Let's consider, though, an example where you're doing a product shoot. Smallish items, and you're shooting with a constant light source, directed from the outside of a light tent. Maybe the sun?

Thus, your lighting is totally consistent from go to whoa. Same temperature, same intensity, same diffusion characteristics. Basically, I would set my wb and exposure all manual, then forget about that aspect of my shoot and get on the the job of getting the images made.

Let's now consider the subjects. Small items, perhaps toys or jewellery. Each of these individual items may be of very different colours from one item to the next. Toys - perhaps several, each of a similar size, and one each of a bright yellow, bright red, bright blue, bright green, and a bright orange.

One series of images with a blown white background, and a second set with a rich black background. With the constant setup that I've described, I can just go ahead and shoot this, and not worry about the results; I know they're going to be consistent and my workflow in post should be fairly simple. And, yes, as always, very minimalist.

Let's now alter the parameters ... just one thing at a time ... so let's switch back to AWB. While there's no way for us to accurately predict what the outcomes will be (and yours will most likely be different from mine because of shooting variations that we each, as individuals, may bring to the picture) My expectation is that the outcomes from this part of the exercise will demonstrate a variation in the wb actually used within each image. Each image may still look ok, but I would be expecting to see different colour temperatures being applied across the range of images made, due to the effect of the inherent characteristics of the items being photographed.

Let's again alter the parameters, this time switching to Aperture exposure mode, and again we should repeat each of the shots. I would expect that we would see variations in the EV applied to each of the images, and especially when we compare those with the white background to those with the black.

As an adjunct to this last one, let's expand this part of the test to use different metering modes. Let's shoot one set using CW, and a second using matrix or evaluative, for Canon shooters.

I think it's a worthwhile exercise to discuss what the expected outcomes might be here, and I think that it's a further worthwhile exercise to grab the implements of image-making and actually do this shoot, so that y'all can actually see the results, and compare those results with what the expected outcomes were.

As an observation on your tests, I would generally accept and expect that the awb settings should come somewhat close, and perhaps our eyes might not be able to pick the differences.

Or perhaps they might: shoot a person wearing a loose white dress or shirt in the open shade under a tree in a park, using AWB; I'd like to see that one.

[chrisk]

What would those "given set of conditions" be? As I said, I was getting extremely consistent results across a wide range of simulated subjects and those results are what I would have considered as accurate as the white balance of the gray card was correct.

my d300 was spot on aswell. i have to say that the d700 has been fooled a bit though. its not often but often enuf for me to take notice and i've never done that b4.

[ATJ]

Let's now alter the parameters ... just one thing at a time ... so let's switch back to AWB. While there's no way for us to accurately predict what the outcomes will be (and yours will most likely be different from mine because of shooting variations that we each, as individuals, may bring to the picture) My expectation is that the outcomes from this part of the exercise will demonstrate a variation in the wb actually used within each image. Each image may still look ok, but I would be expecting to see different colour temperatures being applied across the range of images made, due to the effect of the inherent characteristics of the items being photographed.

Let me first make one thing perfectly clear. I was not evaluating the images by whether they "look ok". I know not to do that. I have been looking at the actual white balance numbers set by the camera. What I am finding is these numbers are either not changing at all or changing ever so slightly - when the lighting conditions aren't changing.

I very much agree that if the lighting conditions are constant then setting and using the same white balance settings for the whole shoot is the right way to go - that's exactly how I shoot now. (Note that if you shoot raw, you don't actually have to set the white balance in the camera as long as you have some way to set it in post processing so that it is both accurate and consistent. This is achieving the same thing in a different way.)

The point I'm making is that, at least with my D300, that under the testing I have done so far, under consistent lighting I am getting consistent white balance numbers from the camera set to AWB even though the colours of the subjects are changing.... i.e. the same (or very similar) colour temperatures and tints. Like you I was "expecting to see different colour temperatures being applied across the range of images made, due to the effect of the inherent characteristics of the items being photographed" but that was not the results I was getting. I don't understand how the camera is doing this because it is non-intuitive but it is doing a fantastic job.

I have been testing this under consistent light because this is the only place you can test it. As I said above, I probably wouldn't use it under consistent lighting. However, given how well it is working, it may well be of value when working in inconsistent lighting conditions. i.e. where the colour of the light is changing from shot to shot and it is either not possible or impracticable to use a single white balance setting across the range of images.

Last night I did some test shots under fluorescent lighting (which is often quite difficult) and I saw much the same thing - different subjects, different colours but almost identical white balance settings between images.

I plan to try some different lighting conditions (e.g. incandescent lighting) just to see how the camera does.

One further point. I always shoot raw. I often forget to set the white balance on the camera. I am in the process of setting up bank of Lightroom presets for known lighting conditions. e.g. I have one for my D300 when shooting with twin DS125 (underwater strobes) so when I come back from a dive I can apply the preset to all the photos and not have to worry about what WB was on the camera. I am now thinking about leaving the camera on AWB knowing that when I shoot with consistent lighting I can easily apply a set WB across all the images, but also know that I may actually have closer WB for other conditions (compared to if I have the WB on the camera set incorrectly).

[aim54x]

I always knew that this would be an interesting thread...just to add my two cents...I use AUTO WB most of the time *ducks slap from Gary* and have no real problems with it in either my D300, S5 Pro (does it even better) nor my D60. If it becomes an issue I switch to a preset...or change it in the RAW.