White balance reference frames - exposure

[Doug Kerr]

It is widely considered that, when taking a "white balance reference frame" (for subsequent examination by the camera to determine the chromaticity of the incident light in the setting for upcoming photography), it is advantageous for the area to be measured to have a photometric exposure on the sensor that is fairly high.

The rationale is that a more precise measurement of the ratio of the outputs of the various color channels can be made in such a situation.(The ADC can only read the channel outputs to a precision of one unit. A one unit precision for a value of about 3800 is more precise than a one unit precision for a value of about 1900.)

I'm not sure we know this for certain, but it seems a reasonable assumption. (Michael Tapes in fact cites a similar rationale for recommending a fairly high reflectance for a gray card to be shot "in the scene" as a basis for white balance correction during raw development.)

Now assume that we are using a white balance diffuser for white balance measurement. If we use metered exposure for our reference frame, the metering system of most of our cameras will cause a photometric exposure for the uniform luminance of the "image" on our sensor (let's assume for the moment that we have a diffuser that makes a uniform luminance image) of perhaps 13-22% of the saturation photometric exposure.

We can thus certainly afford to "bump" this by one or even two stops of exposure compensation to move the photometric exposure into the "serious" part of the scale. And I suggest doing so as a general practice when taking a white balance reference frame. (I'm a little cautious, and I generally use a +1 EC.)

Some diffusers may not make a uniform-luminance image on the sensor, at least not for all lens and lens setting situations. (I don't want to mention these by name, but they seem to always want a cracker.)

In the face of that, it might be prudent to use "partial" or "spot" (or at least "center-weighted average") metering when taking the reference frame, especially if we are going to "bump" the metering, as I suggest here.

Incidentally, when using a gray card for white balance reference frames on a Canon EOS camera, we of course must make the card image at least fill the circle in the center of the viewfinder. That being the case, and especially if we are using "partial" metering or the equivalent, we have the same photometric exposure situation described above. (A higher reflectance for the gray card doesn't do anything for us in this usage, owing to the operation of the metering system.)

Thus here, as well as with a diffuser, an EC "bump" would seem prudent.

 

[Asher Kelman]

In addition to your helpful comments on assumptions of brightness needed for white balance reference, I'd like to add a few cautionary notes in the use of any references like the ExpoDisk™, WhiBal™ Color Parrot™ or any other trick you use!

1. Use one of these devices as the light is golden and the sun is dying and you will make a perfect Tuscany scene ordinary!
   
2. In the studio, use this before adding a jel or color to your lights!
   
3. To be fastidious make your camera have none of it's own fingerprints in the first place by profiling your camera!

For quick work, these discs/cards do the job within O.K. limits.


For creative work, it may be not what you want! Do you want to maintain some of the natural light variations or go for what the scene might look like after a lighting crew worked for hourse making the light flux illumitating everything "normal" at whatever temp and distribution you choose.

You have to know what you are doing where light is complex and it might mean several translations from RAW and selecting different parts of the image to have varying contributions where the light is complex.

Do you really want to remove all the green cast from a tree or the red from a girl's dress on a child's white satin dress or the white table cloth? The changes will depend on where you placed the central cross hair assuming you have set it up as I do ffor one spot!

Asher

 

[Kathy Rappaport]

Uh oh - I have been here for a while - I actually understand this stuff now. I used to bypass these threads as over my head. Thank you for adding to the useful information bank!

 

[Bart_van_der_Wolf]

It is widely considered that, when taking a "white balance reference frame" (for subsequent examination by the camera to determine the chromaticity of the incident light in the setting for upcoming photography), it is advantageous for the area to be measured to have a photometric exposure on the sensor that is fairly high.

The rationale is that a more precise measurement of the ratio of the outputs of the various color channels can be made in such a situation.(The ADC can only read the channel outputs to a precision of one unit. A one unit precision for a value of about 3800 is more precise than a one unit precision for a value of about 1900.)

That's exactly what I do, and that rationale is also why I advocate using a 'Babelcolor White Target' reflection reference for the most critical applications. Afterall, we're after a 'White' balance, not a gray balance. Maybe that's more semantics than physics, but I consider the precision argument made above as being valid.

I'm not sure we know this for certain, but it seems a reasonable assumption. (Michael Tapes in fact cites a similar rationale for recommending a fairly high reflectance for a gray card to be shot "in the scene" as a basis for white balance correction during raw development.)

In practice, I've found the differences from "Click White" balancing in a Raw processor between a Whibal and the Babelcolor White Target to be very minimal.

Now assume that we are using a white balance diffuser for white balance measurement. If we use metered exposure for our reference frame, the metering system of most of our cameras will cause a photometric exposure for the uniform luminance of the "image" on our sensor (let's assume for the moment that we have a diffuser that makes a uniform luminance image) of perhaps 13-22% of the saturation photometric exposure.

We can thus certainly afford to "bump" this by one or even two stops of exposure compensation to move the photometric exposure into the "serious" part of the scale. And I suggest doing so as a general practice when taking a white balance reference frame. (I'm a little cautious, and I generally use a +1 EC.)

I agree if we shoot a reference frame that's intended to be used for Click White Balancing in a Raw processor, for the reasons stated above. I do caution against pushing the exposure too far, because the last 1/3rd stop before clipping can be inaccurate due to differences in the clipping point of the R/G/B sensels. So a +1 EC or +2 EC should suffice. However, the click-white method is pretty reliable, and will also produce a good baseline color temperature estimate from medium gray, as long as it's really spectrally neutral. There may be differences between Raw converters, so it won't hurt to opt for a higher exposure.

However, I do not agree if we are taking a Custom White Balance (CWB) reference shot that's to be evaluated by the camera to produce a CWB color temperature setting. I don't know any specifics for other brands, but Canon cautions against it. In the manual for the EOS-1Ds Mark III (page 65) they state the following:

"Photograph a solid-white object.

• The plain, white object should fill the center spot metering circle.
• Set the lens focus mode switch to <MF>, then focus manually. (p.86)
• Shoot the white object so that a standard exposure (gray) is obtained. If it is underexposed or overexposed, a correct white balance setting might not be obtained."

While we do not know the inner workings of of the CWB algorithm, I could assume that it is based on some sort of 'gray world' white balancing heuristic. Depending on the color model used internally, a higher (or lower) luminosity than 'average' can result in a more limited range of saturation and thus a lower precision. Besides that, from a practical point, using the camera metering as-is will on average produce a more solid basis than some less experienced users having to fiddle with EC adjustments (which will probably be forgotten when taking the actual exposures).

Some diffusers may not make a uniform-luminance image on the sensor, at least not for all lens and lens setting situations. (I don't want to mention these by name, but they seem to always want a cracker.)

In the face of that, it might be prudent to use "partial" or "spot" (or at least "center-weighted average") metering when taking the reference frame, especially if we are going to "bump" the metering, as I suggest here.

I agree. When the center spot metering circle area is expected to be of a certain exposure level, it is also required to adapt the metering for the reference shot to that specific region of interest. When filling the full frame with a uniformly lit object it probably won't make a difference which type of measurement is used, but for a non-uniform brightness pattern, one needs to adjust the metering for the CWB reference shot (which can be a disadvantage because one can forget to reset it for regular shooting).

Bart

 

[Doug Kerr]

Hi, Bart,

That's exactly what I do, and that rationale is also why I advocate using a 'Babelcolor White Target' reflection reference for the most critical applications. Afterall, we're after a 'White' balance, not a gray balance. Maybe that's more semantics than physics . . .

Indeed. Note that there are "gray" reflective objects, but no "gray" light!

In practice, I've found the differences from "Click White" balancing in a Raw processor between a Whibal and the Babelcolor White Target to be very minimal.

That's good to hear.

I agree if we shoot a reference frame that's intended to be used for Click White Balancing in a Raw processor, for the reasons stated above. I do caution against pushing the exposure too far, because the last 1/3rd stop before clipping can be inaccurate due to differences in the clipping point of the R/G/B sensels. So a +1 EC or +2 EC should suffice.

However, I do not agree if we are taking a Custom White Balance (CWB) reference shot that's to be evaluated by the camera to produce a CWB color temperature setting. I don't know any specifics for other brands, but Canon cautions against it. In the manual for the EOS-1Ds Mark III (page 65) they state the following:

"Photograph a solid-white object.

• The plain, white object should fill the center spot metering circle.
• Set the lens focus mode switch to <MF>, then focus manually. (p.86)
• Shoot the white object so that a standard exposure (gray) is obtained. If it is underexposed or overexposed, a correct white balance setting might not be obtained."

Indeed. But in contrast, we note that the Nikon cameras, when taking a white balance reference frame, automatically apply +1 EC. But of course the algorithm used there might be quite different.

When filling the full frame with a uniformly lit object it probably won't make a difference which type of measurement is used, but for a non-uniform brightness pattern, one needs to adjust the metering for the CWB reference shot (which can be a disadvantage because one can forget to reset it for regular shooting).

Indeed. I had even at first considered putting a caution about "don't forget" about EC in my note!

Thanks for your insights.

Best regards,

Doug

 

[Doug Kerr]

In addition to your helpful comments on assumptions of brightness needed for white balance reference, I'd like to add a few cautionary notes in the use of any references like the ExpoDisk™, WhiBal™ Color Parrot™ or any other trick you use!

1. Use one of these devices as the light is golden and the sun is dying and you will make a perfect Tuscany scene ordinary!

1. 
   
   All this is well said.
   
   It is important to note that, as I discuss the conceptual premises of the "theoretically ideal" color balance process, and discuss how various technical properties of tools can help us attain it, I don't in any way mean to suggest that the attainment of this sterile ideal should be our objective in all, or even any, photographic situations.
   
   And in any case, cute redheads with slightly blue skin are still cute.
   
   Best regards,
   
   Doug