Set white point with Macbeth 2nd patch

[George Machen]

There was a thread a month or two ago, either here or over on Dgrin, where someone spelled-out the reason for setting a digicam's white point using the second patch on the Macbeth Color Checker. But I've searched & Googled 'til the cows came home and can't find it. My specific questions are:

- Why use the second patch instead of the first, white (specular) patch?
- What was the deleterious consequence mentioned in that post of using the first patch?
- I'm bringing this question up again because I was urging a friend with a digital video miniDV camcorder (Canon GL2) not to set WB on a purely white surface but rather a slightly darker neutral surface, but he argued that the instruction manual doesn't say that, and presumably this issue carries over to digicamcorders, too.

[George Machen]

OK, I kept searching, and with a Google search criterion of:
site:luminous-landscape.com macbeth "color checker" "second patch"
... the thread I was thinking about came up as the first hit:
http://luminous-landscape.com/forum/lofive...php/t18026.html
(Shows yet again that a Google site-targeted search often is vastly superior to the target site's own search engine; in my case, LL's kept coming up nil.)

So now I'm able to answer my own questions:
The deleterious consequence of using the near-specular white patch is running the risk of clipping channels, leading to an inaccurate adjustment; the second grayer patch gives one some headroom, but still is light enough to take a proper white balance reading.

Nevertheless, yesterday I was able to play around with my friend's GL2 white balance adjustment, and I must say that despite pointing it a number of very white specular surfaces (completely filling the view, of course), the color balance always came out well on-screen, visually at least. (Perhaps I'm coining a phrase: "pleasing white balance"? Heh!)

On the other hand, looking at the transmissive light source of the screen shining light directly into my eyes, maybe I was being fooled by the perceptual phenomenon of chromatic adaptation, neutralizing any slight casts from the aforementioned possible inaccurate adjustments due to channel clipping, whereas such casts may stick out like a sore thumb were the capture to go to print, where I'd be looking at reflective light without the "protection" of cast-neutralizing chromatic adaptation.

So I guess the practical danger of not using the second patch really depends on whether the final viewing will be on screen or on printed material.

[digitaldog]

It all has to do with how much headroom in the highlights you have. If you're real close to actual clipping, then a WB there can cause a color shift (a poor WB if you will). Try this on a rendered image in Photoshop using the gray eyedropper. You can, with images with severe color casts see one or more color channels bump up to the top of the curves (it can't go any farther, it needs to). With linear encoded data, there is just so much more highlight data in that first stop of highlight, it usually works much better than gamma corrected images. But the point is, while you can WB on specular highlights (sometimes), there will be situations where you'll get undesirable results compared to clicking on a tone that is lower (a bit darker) where all three channels can be placed where necessary to produce the desired color appearance.

[Jonathan Wienke]

Another point to consider is that for some obscure reason, the second-lightest patch is the most neutral; all of the others have slight casts that will negatively affect the accuracy of your WB reading.

The clipping issue is moot when setting WB on a video camera with a white card that fills the entire frame. The auto-exposure of the camera will render any neutral object that fills the entire frame to the same tonal value, regardless of the reflectance of the object or the intensity of the lighting, as long as the camera has sufficient exposure adjustment to do so. The specularity issue only comes into play when the white reference does NOT fill the entire frame, as the auto-exposure may decide to clip some small bright highlights to improve exposure of the rest of the frame.

[George Machen]

I searched around some more, and I was mistaken that the link in my post #2 above was the one I was thinking about (as useful as it was!); the info I was trying to remember turned out to be two posts, #16 and #18 (by Jonathan, no less!), in this thread:
http://luminous-landscape.com/forum/index....showtopic=19496
Following are the salient issues from those two posts that I have combined & shortened (note the addition of problems with using the whitest patch even before actually entering clipping):

> Is there a reason that one should click white balance on the
> second-lightest gray patch rather than the whitest one?

There are 2 actually:

1: The second-lightest patch (for some very obscure reason) is the most neutral patch in the Color Checker. The other patches all have small color casts. I don't know why.

2: Click white balance doesn't work properly when the color channels are clipped or very close to clipping. Using the lightest patch as a guide to ensure you have no color channels clipped means you are guaranteed an accurate white balance when using the second-lightest patch.

Digital sensors typically exhibit a small degree of non-linearity just before clipping. So if you use a white reference for white balance, you have two sources of WB inaccuracy: the possibility of a clipped color channel, and the possibility of one of the color channels being in the small non-linear zone just below the clip point. Either of these can skew the accuracy of your white balance reading if you click white balance.

If you use the second-lightest gray patch in a RAW where the lightest Color Checker patch is about level 250 when the Exposure setting is at 0, you get the most accurate click white balance because you're using the most neutral patch in the color checker, and you're safely below the exposure range where clipping or nonlinearity can skew the accuracy of the click WB reading, but high enough on the exposure scale where you can still get a highly accurate reading. It's sort of the sweet spot in the tonal range where click-WB accuracy is greatest.
    - Jonathan Wienke