External metering & 600 EX - no improvement

"Acceptable results"?
Most people wouldn't call two stops off, compared to a well exposed picture, acceptable. I don't, for one.

apersson850 wrote in post #15222425
"Acceptable results"?
Most people wouldn't call two stops off, compared to a well exposed picture, acceptable. I don't, for one.

That is completely out of context. The "acceptable results" I was referring to were the results of Wilt's test which show that the ambient, TTL and Auto metered examples were within a 1/4 stop of null. That is acceptable.

What you are referring to is apparently personal experience that I've not seen any examples of so can't speak to them at all.

To validate your comment I did a quick test;
first with ETTL. 5DIII, 600EX-RT in hot shoe, head zoomed to 200mm, aimed to camera right to bounce off wall. It has my "standard" modifier which also let light go up to the ceiling with is about 18" above Mannie's head. No flash directly toward Mannie.
Second, only change was to switch 600EX-RT from ETTL to Auto.
Third, was same as second but with +2 FEC dialed in on camera.

In Lightroom, all three images were imported with camera profile, all sliders set to zero. Performed white balance on second gray square from right. Then used slider to get the same reading (69.5 ± %) for the square.

Both the TTL and FEC adjusted Auto are zero with essentially the same exposure. The slider adjusted required less than 2 stops.

and here is the FEC adjusted image, and the TTL image with the second square exposure adjusted to standard which was just about 1/4 stop more exposure.
This may be because there was a picture under glass on the wall about 4 feet behind mannie to camera right that has a reflection of the flash bouncing off the ceiling.

Color me slow, today, dmward...

in the first pair, it seems that External mode flash needed +1.35EV adjustment in post, in order for it to achieve the same density as ETTL flash did. Did you make any effort to shoot ambient-only shot, as a non-flash baseline value for density, to see how ETTL density compared to that baseline?

I'm not following what the second pair of photos illustrates. Can you explain?

Wilt wrote in post #15223074
Color me slow, today, dmward...

in the first pair, it seems that External mode flash needed +1.35EV adjustment in post, in order for it to achieve the same density as ETTL flash did. Did you make any effort to shoot ambient-only shot, as a non-flash baseline value for density, to see how ETTL density compared to that baseline?

I'm not following what the second pair of photos illustrates. Can you explain?

Wilt, This was a quick attempt to replicate the situation the OP describes in his first post. i.e. that the Auto mode flash metering in a 600EX-RT is two stops under exposed compared to ETTL metering for the same scene. No ambient only, as that is more or less irrelevant to the OP's stated situation which is comparing exposure via flash and the two options for metering the flash output from the unit during exposure.
I did my normal bounce flashing. My recollection is that ambient would have been more than 3 stops under exposed. So, all these images are flash only illumination.
First image is ETTL baseline.

Second image is Auto metering with adjustment applied to get it to same reading in second square from right.

Third image is Auto metering with plus 2 FEC dialed into the camera. (It is essentially the same reading in the second square as the ETTL baseline.)

Fourth image is the first image adjusted so the second square reads as it should based on the Color Checker standard readings. i.e. "proper exposure" which implies that the images are about 1/4 stop under exposed

dmward wrote in post #15223764
Wilt, This was a quick attempt to replicate the situation the OP describes in his first post. i.e. that the Auto mode flash metering in a 600EX-RT is two stops under exposed compared to ETTL metering for the same scene. No ambient only, as that is more or less irrelevant to the OP's stated situation which is comparing exposure via flash and the two options for metering the flash output from the unit during exposure.
I did my normal bounce flashing. My recollection is that ambient would have been more than 3 stops under exposed. So, all these images are flash only illumination.
First image is ETTL baseline.

Second image is Auto metering with adjustment applied to get it to same reading in second square from right.

Third image is Auto metering with plus 2 FEC dialed into the camera. (It is essentially the same reading in the second square as the ETTL baseline.)

Fourth image is the first image adjusted so the second square reads as it should based on the Color Checker standard readings. i.e. "proper exposure" which implies that the images are about 1/4 stop under exposed

OK, the sequence of four makes total sense now! Nevertheless what is puzzling is why Auto Image 2 is +1.35 in post, while Auto Image 3 is +2EV in FEC...one wonders why the discrepancy of 0.65EV for esssentially two approaches (post after vs. FEC before) using the same External mode?!

Wilt wrote in post #15223776
OK, the sequence of four makes total sense now! Nevertheless what is puzzling is why Auto Image 2 is +1.35 in post, while Auto Image 3 is +2EV in FEC...one wonders why the discrepancy of 0.65EV for esssentially two approaches (post after vs. FEC before) using the same External mode?!

I cropped the image to just show Mannie. As mentioned there is a glass covered print in the background with a substantial reflection that may have influenced the exposure. If you add back the .25 adjustment to get the reference square to proper exposure its just about 2 stops.

Tomorrow, if time permits, I'll try to do the same test without the background interference, and also include an ambient reference shot.

I started to do a test using ambient, ETTL and Auto. The results were closer than I expected given earlier comments. The one thing I did was to have the constant light source and the flash as near as possible to the same as the lens axis.
One thing I noticed was that the Color checker gray scale was visibly different in each lighting scenario. That brought into question which was the "proper" wedge to use for exposure.

That led me to do a series of virtual copies for the image exposed using the Auto exposure option on the 600EX-RT.

The first composite is the same image with exposure set to the value for one of the gray scale squares. Upper left is Sq 1 0 adjustment, next is Sq 2 -.9 adjustment, next is Sq 3 -1.15 adjustment. Second row left is Sq 4 -.95 adjustment, next is Sq 5 -1.35 adjustment, last (lower right) is Sq 6 -.65 adjustment.

This implies that the valid square for exposure adjustment is the white square. It should read 93.33% in Lightroom with all sliders null. Its Prophoto RGB values are 238/238/233 which is well below a blown out highlight.

Based on this information, I would be inclined to use either the white or black square for exposure determination. Probably the white square is better since it tends to support Expose to the Right which minimizes noise in the lower range of the histogram.

The second composite is the three light source images with the gray squares "corrected" in the tone curve module so they read the appropriate % based on the Color Checker standard readings for ProPhoto expressed as a percentage. Those are; 93.33% for the white square, then 74.12%, 56.47%, 40.39%, 25.49% and finally 14.51% for the black square.
Its not hard to see that applying a tone curve that is correct for the Color Checker leads to some weird color shifts.

And, to bring things full circle, here are the three test images; Left image is ambient, center image is ETTL and right image is Auto. None of the images have any exposure adjustments in Lightroom. They are converted from raw with camera profile, all sliders in null value and tone curve is linear. They were corrected for white balance.

The ambient has the lowest reading on the white square (88.5%) it requires +.60 EV on the exposure slider to get it to 93.5%. The middle image is the ETTL exposure, it reads 92.1% on the white square and requires +.15 EV on the exposure slider to get to 93.5%. The right image is the Auto image, it reads 93.5% on the white square which is proper exposure.

The lighting was an incandescent light just to camera left for the ambient reading, and a 600EX-RT in the hot shoe of the camera for the flash exposures. The flash head was pointed directly toward to subject, as was the incandescent light. The meter was set to evaluative for both ambient and flash readings.

Based on this quick test, the ETTL, Auto exposure options for the 600EX-RT appear to deliver the same results and are close to the ambient metering. Naturally, because the metering is evaluative and reading the entire image with a strong bias toward the center (The metering circle was just about the same as mannie's face.) its impossible to speculate about how a more natural scene would bias the metering.

Purolator messed up my delivery yesterday. Should get it today. I'll test it out when I get it.

I'll show you some pictures from my experiments as soon as I can.

Dave,
For reference about my prior technique, so that others understand my rationale...

• all of my prior tests used a linear tone curve...no adjustment at all, except for same WB and then adjust only Exposure...all other variables identical.
  
• And I used an exceedingly simple 'scene' consisting of ColorChecker and 18% grey card so as to not complicate things with flash metering patterns (ETTL) vs. no pattern (External photosensor).
  For your test I would have positioned the camera so that rear wall was at an angle, so as not to reflect back the sheen of the flash, which might have biased the two flash outputs. Even better, zoom in tightly so as to exclude as much of the scene as possible, leaving only your mannequin and ColorChecker tightly framed.
  
• Since square 3 on the ColorChecker (black, dark grey, grey, light grey, very light grey, white) is same density as 18% grey card, I would have measured density on that sample, then adjusted Exposure until the two flash shots matched the eyedropper percentages of the ambient shot.
  
• Contrast characteristic of ambient-only vs. flash would naturally be different, and since a meter assumes 'average' computes to the mid-tone 18% grey, that is why I use that sample, rather than the white whose values change for the contrast imparted by the light source.
  

Wilt,
I agree with most of your statements. Indeed, having the wall at an angle would eliminate the glare.
Relative to the Color Checker grey scale. There should be, by definition no contrast imposed on the scale. That's the intent of the Color Checker as a reference. The black square in not intended to be black without detail, nor is the white square intended to be specular white. My inclination toward using the white square for exposure is my interest in keeping whites from clipping. I've found, for example, that if I start to see blinkies on camera LCD that I have about a stop and a half of headroom before the whites really clip.

I've found that the Lightroom Strong Tone Curve with the white square adjusted to 93% brings the black square to 14% which is just about where it should be based on the reference values. That permits me to use the tone curve to manage mid-tones. Lightroom V4 and process 2012 offer substantial power for tone control. I've done some testing and now use -50 highlight, +50 shadows, -50 white, and +25 black. Along with +15 clarity which is a mid-tone contrast control.

While technically using any square for exposure reference should work, I've found practically speaking that the white square gets me where I want to be with minimum fuss.

Wilt, et al;
Just to bring this discussion full circle;
Here is another quick test that demonstrates a) that selecting the gray scale wedge to use for exposure adjustment can bias the exposure in appropriately, and b) to confirm that ETTL, Auto and Ambient metering result in essentially the same image, exposure-wise when extraneous elements are removed from the frame.

In the composite the ambient light (a halogen flood) was in essentially the same position as the flash head. I had the flash touching the side of the bulb for the flash exposures. The flash head was aimed directly at the subject and tilted down the 7* built into the unit by Canon. No FEC for the ETTL or Auto flash exposures, no EC for the ambient exposure. All metering was set to Evaluative. The meter option was set to (*).

The The first row is ambient, second row is TTL and bottom row is Auto.
The left column is white square, middle column is grey (3rd from black end), the right column is black square.

Here are the exposure slider adjustments necessary to get the appropriate square reading to the CC reference value Blk 14.51%, Grey 40.39%, White 93.33%
This is using Process 2012 in Lightroom 4.2 with Highlight, Shadow, White, Black and Clarity slides at null. And the tone curve was linear


+.4, -.7, -.2
+.1, -.65, -.3
-.3, -1.2, -.95

dmward wrote in post #15227484
Relative to the Color Checker grey scale. There should be, by definition no contrast imposed on the scale. That's the intent of the Color Checker as a reference. The black square in not intended to be black without detail, nor is the white square intended to be specular white. My inclination toward using the white square for exposure is my interest in keeping whites from clipping. I've found, for example, that if I start to see blinkies on camera LCD that I have about a stop and a half of headroom before the whites really clip.

I've found that the Lightroom Strong Tone Curve with the white square adjusted to 93% brings the black square to 14% which is just about where it should be based on the reference values. That permits me to use the tone curve to manage mid-tones. Lightroom V4 and process 2012 offer substantial power for tone control. I've done some testing and now use -50 highlight, +50 shadows, -50 white, and +25 black. Along with +15 clarity which is a mid-tone contrast control.

While technically using any square for exposure reference should work, I've found practically speaking that the white square gets me where I want to be with minimum fuss.

Dave,
Here are two shots of ColorChecker, on under CFL and the other under daylight. The18% grey patch is equalized on both shots to 51-52%. All other parameters except for WB and Exposure are identical for the two shots.

The black patch is 11-12% on the CFL shot, whereas it is 13-14% on the daylight shot...a slight difference but not a signficant one. However the white patch is only 84% on the CFL shot, and 92% on the daylight shot!