COLORRIGHT works better with flash than competition

[Doug_Kerr]

Hi, Drew,

Originally Posted by drew

I put it in there for you, Doug.

Because I knew you'd want a reference. This is one you like, no?

And I was just pulling your leg, given that you had accused me of being Michael Tapes' dupe.

But, why bother going down that rabbit trail again.

I think perhaps "rabbit hole" would be a more apt literary metaphor.

Best regards,

Doug

[drew]

However, actually, in that setting, we should expect to get a fairly good WB measurement with just the naked lens, aimed so that only background fell within the "guide circle".

Sure that might work. But that's not the point at all here.

Yes, thanks for acknowledging that the narrow directivity pattern helps when using a flash.

However, the biggest differences here appear to be related to exposure level with the other product.

In regular "Program" mode with a 550EX or a 580EX attached the exposure shutter speed and aperture are identical across both reference shots. In this mode Canon does not seem to want to vary the exposure setting. Both reference shots were 1/60 and f/4.

Here is another series with a 550EX flash attached and all else equal.

Canon 40D, P Mode, Spot metering, bare bounced flash off ceiling, etc. All the info is in the exif data of the raw files.

COLORRIGHT

Full Size RAW File.



Expo

Full Size RAW File.



Expo +2 EV

Full Size RAW File.



It's kind of hard to tell because it is still off a good bit, but this third image where we overexposed the expo reference shot (+2 EV) the color temp came closer into line. This also happened with the other series you saw. This third reference image and resulting image were exposed by the camera at 1/60 and f/2.8. Since the lens wouldn't go below 2.8, I guess it really wasn't a full +2 EV.

Go ahead and download all the raw reference images and simply balance the reference images yourself in LR, or whatever. You will see how far off the second and third one are from each other no matter where in the image you click.

In a nutshell, what I think is happening is that since Canon does not really change exposure in P mode with a flash attached, the expodisc just cuts down the light too much to get an accurate reading.

Open to hear any alternative hypotheses. Of course, as you know, doesn't mean we'll agree on it.

Try it out yourself with your own external flash on your 20D. Vary the backgrounds, etc.

[Doug_Kerr]

Hi, Drew,

Originally Posted by drew

Yes, thanks for acknowledging that the narrow directivity pattern helps when using a flash.

I said no such thing. I said that the narrow pattern of the ColorRight should be advantageous if:

a. We were measuring from the camera position, and

b. The "set" had a large neutral background.

This had nothing to do with the use of flash.

In regular "Program" mode with a 550EX or a 580EX attached the exposure shutter speed and aperture are identical across both reference shots. In this mode Canon does not seem to want to vary the exposure setting. Both reference shots were 1/60 and f/4.

Well, in program mode with a flash attached, the Canon EOS cameras will not set a shutter speed slower than 1/60 sec nor an aperture smaller than f/4.

So in Program mode, I guess you are right - the "balance" between flash and ambient light for a metered exposure would vary with the transmission of the diffuser. Good point. I guess I was somehow assuming that the exposure was "fully metered" (not in the "limited" way in Program mode with flash). Good of you to catch that.

In a nutshell, what I think is happening is that since Canon does not really change exposure in P mode with a flash attached, the expodisc just cuts down the light too much to get an accurate reading.

That seems a very reasonable interpretation.

And so, in an ambient plus flash situation, with the ambient fairly low (as in typical indoor flash shooting), and the user (for some reason) using P mode, the higher transmission probably would be an advantage in getting a meaningful WB measurement.

You recent test frames seem to illustrate that principle.

My apologies for not recognizing that aspect of the deal.

Best regards,

Doug

[MikeA]

Originally Posted by drew

Depending on where I click on the whibal I can get a pretty good variety of readings.

Drew, that had been my experience with the Whibal, but I have found it much better (usable even) if I increase the color picker's sample area. On this particular image, doing so reduces the WB variation down to 100 or so. The default small picker gives variations all over the place.

[Doug_Kerr]

Hi, Drew,

Now the telling test would be the very same thing with a blue or red background of that same configuration.

(It's not hard to expect "from the camera position" to work when the set is one giant gray card!)

Best regards,

Doug

[Doug_Kerr]

Hi, Drew,

Just as a matter of interest, I tried a shot with Carla against a non-neutral background, with custom WB set from a reference frame taken from the camera position with the Color Parrot 1.2. The camera was a Canon EOS 20D, with a Canon EF 24-105 mm f/4L IS lens. Lighting was wholly by flash, using a Canon 580EX II in ETTL mode. Ambient exposure was essentially suppressed by the use of a fast shutter speed and moderate aperture (set in M mode).

Here is the in-camera corrected image (JPEG output); it has been downsampled but is otherwise unmodified:



Carla's pendant is a noncommercial laboratory white reference target. It has not been laboratory certified, but its neutrality is believed to be within 0.0007 u'v' unit.

The chromaticity of its image (read from the corrected JPEG output) differs from theoretically ideal correction by du'=+0.0036, dv'=+0.0111, du'v' (total discrepancy)=0.0112. (Note that on my familiar chromaticity charts, the limit of the scale is 0.01 unit, so this correction error would have been off the chart.)

This is, as we would expect, an error in correction in the yellow direction (the inevitable result of making a WB measurement with a mostly blue target - that is, the scene). (This is of course why the "from the camera position" technique isn't reliable - it uses any old thing as the target.)

On an L*a*b basis, the reference target came in at 90, 0, +9.

The white areas of Carla's dress came in at an average of about du'v'=0.0128 (again in the yellow direction). Of course, we don't know the reflective chromaticity of the fabric itself.

The image of course doesn't "look that bad", since the extensive blue background provides for some visual chromatic accommodation (somewhat as if we were viewing the subject directly).

We of course made no corresponding test with any other diffuser, since we have none intended for use with the "from the camera position" technique.

Note that since the flash output would have been different for the WB reference frame and the actual shot (owing to metering system compensating for the attenuation of the diffuser), the chromaticity of its output could have been slightly different between the two situations. I forget which way it shifts with output, so I can't tell whether that might have aided the discrepancy or otherwise. (I have that data, but can't put my hand on it just now.)

Best regards,

Doug

[Doug_Kerr]

Hi, Mike,

Originally Posted by MikeA

The default small picker gives variations all over the place.

This is probably mostly the result of noise (fixed-pattern and temporal) in the image.

Best regards,

Doug

[Pro Photo Store]

Visit the B&H Pro Photo Store for the best Gear Prices and regular specials!

Brilliant Color for any Digital SLR
White Balance Cap
The Pros Agree on ColorRight
What People Are Saying