External metering & 600 EX - no improvement

The external metering function of the 580 EX II have always been infamous for not exposing correctly, but underexposing quite a bit. Typically two stops, compared to what E-TTL II figures out to be the correct exposure.

I tried the same thing with the 600 EX-RT, just to see if Canon's engineers got it better this time. You have to give them some credit, since they are at least consistent. The 600 EX-RT underexposes by two stops, pretty exactly, compared to the E-TTL II exposure. Easy enough to judge from the histograms.

So if you do plan to use the Ext. A or Ext. M modes with your 600 EX-RT, you can count on dialling in about +2 in FEC (Ext. A) or set the aperture info on the flash two stops smaller than it actually is (Ext. M).

Hmmm. Did they maybe do this on purpose? I mean really how can you get this wrong again? I have not tried my 580 EXII with my new theory I just came up with 10 minutes ago. I have been searching for used 580 EXII for a while now. Today I'm thinking I may just go with the 600.

I purchased a Metz Mecablitz 58 AF-2. Wonderful flash in both Auto and ETTL - great exposures. In low/poorly lit events I shoot high ISO. I have shot up to 6400. Coupled with a wide open aperture the Metz gives me overexposure warnings in Auto mode. If I lower the ISO to abut 800 it is OK. In ETTL the flash pre fires so it provides a pre-etermined duration before the exposure. In Auto the flash decides on duration during the actual exposure. It needs to shut down the tubes and drain them before overexposing thus the warning. Too bad as it is a gem in Auto. I can use it in ETTL but if I do to I may as well work with Canon. Less to think about when you are busy.

So did Canon deliberately set it 2 stops under so you can shoot at a higher ISO in Auto. Even if it does not give a warning overexposure it could be a risk. We know it does underexpose 2 stops but is it consistent. More consistent than ETTL?

To avoid starting another thread I do have a question about the 600. When off camera does the RT control it in both ETTL and manual? The manual part is important to me. I use PCB triggers but it would be nice to control exposure via camera. Logic tells me it does but I thought I would check. Read up on it at Canon USA but could not find that answer.

digital paradise wrote in post #15209538
To avoid starting another thread I do have a question about the 600. When off camera does the RT control it in both ETTL and manual? The manual part is important to me. I use PCB triggers but it would be nice to control exposure via camera. Logic tells me it does but I thought I would check. Read up on it at Canon USA but could not find that answer.

Found the info in my 5D3 manual.

Can you give us more infomration on how you tested and the camera used? My 580exII has been spot on, same with my 580ex and 430exii.

ZXDrew wrote in post #15209939
Can you give us more infomration on how you tested and the camera used? My 580exII has been spot on, same with my 580ex and 430exii.

Really. I thought the 580ex II underexposing in Auto was sort of a known thing. Most people say it underexposes when I read about i. I tested my 580EX II in auto against my Metz and the 580 underexposed pretty much every time. I just shot the same scene and switched flashes with no changes to the camera settings. No FEC because the Metz technical dept told me Auto was not designed to work using FEC in Auto, only ETTL. The Metz produced excellent exposures.

ZXDrew wrote in post #15209939
My 580exII has been spot on, same with my 580ex and 430exii.

The last two you mention don't even have the mode we're talking about here, so you can hardly have tried them with external metering. Since it's the flash which do this on its own, you get the same result with any camera.
I don't know what made Canon implement it like this, but one could imagine that they could have made the exposure equivalent to those well-exposed images you get when E-TTL II has one of its better moments.

It had long puzzled me as to why ambient-only light exposure would expose a grey card to a different density than ETTL flash exposure (camera controlled) would, or why Canon External (Auto photosensor controlled) would do even worse.

The ETTL underexposure was far worse 'in the early days' (of the 20D), but ETTL is much better in the 40D...In the days of the 20D you had to routinely crank in about FEC +1 even in shooting ETTL flash with the 20D. But now, with the 40D, FEC is generally not noticably needed.

Here is a series of three shots just taken with my Metz 54MZ ETTL-comptible on my 40D. The density of the 4th greyscale patch of the Colorchecker is adjusted in postprocessing with LR to achieve the same eyedropper density values for all three shots, adjusted to 60% in the flash shots (to match what the Av mode ambient light exposure result does -- without adjustment), using the Exposure adjustment. I did not try to match the location of the peak in the histogram, since the reflectivity of the photographed surfaces using flash might be different, so I simply wanted to match density of grey card.

First, the shot in Av mode with ambient light only, as the baseline...

I simply took photos of a subject, which gave a characteristic representation in the histogram directly in the camera, when using E-TTL II. I then used the external sensor and cranked up FEC until the histogram looked the same again. That took 2 stops.

It was pretty dark, so I didn't bother trying ambient light too. Would have needed a tripod.

I have attended seminars where Canon Tech Reps state that the maximum aperture of the lens used for the TTL metering will impact its accuracy.

I think we have to consider that ETTL metering is a "system" that incorporates several elements, each with its own tolerances. Add to that the constantly varying tonality and reflectance of the scene in front of the camera and it seems the system is doing a creditable job.

As for the Auto exposure feature, that is reading light reflected from an "unknown" portion of the scene in front of the flash. How it does as well as it does is again a tribute to engineering skill.

Wilt, the gray wedge may have the same readings, but the large gray plastic thing and the blue sofa fabric are significantly different in each of the three. I'd be reluctant to use something that far toward the edge of the frame as an exposure reference for TTL, and since one doesn't know much about the angle of view of the Auto exposure receptor that far of center might be iffy for that reading too. Its probably also not the best place for an accurate reading for the ambient meter reading either since all the Canon options tend to bias toward the center of the frame.

dmward wrote in post #15219655
Wilt, the gray wedge may have the same readings, but the large gray plastic thing and the blue sofa fabric are significantly different in each of the three. I'd be reluctant to use something that far toward the edge of the frame as an exposure reference for TTL, and since one doesn't know much about the angle of view of the Auto exposure receptor that far of center might be iffy for that reading too. Its probably also not the best place for an accurate reading for the ambient meter reading either since all the Canon options tend to bias toward the center of the frame.

I reshot, totally excluding anything reflective other than the target itself (at the extreme right you see thru a sliding glass door, which does not reflect back any flash). Again, used the 4th greyscale patch on the ColorChecker to compare density.

Ambient only exposure shot in Av mode

Wilt, the three are within less than half a stop. That's well within an acceptable tolerance from null value. Since one is -.14 and the one farthest in the other direction is +.34 the range is within plus minus 1/3 stop of null which is the closest settable interval in the camera metering system.

It looks like these are all full frame. If I were doing the test, I'd flip the color checker so the gray scale in in the middle of the frame. That would put the 4th square just about in the exact middle of the frame and most likely in the middle of the auto sensor field as well.

It also looks like there white balance is a bit different for the three. At least the ambient is visibly warmer than the others. That can have an impact on the exposure reading as well.

It also looks like the tone curve is not linear. Not sure exactly what the settings are but they may be influencing the 4th square since it doesn't look like its been pinned.

It also appears that the middle image, ETTL metered, has a bit of sheen on the gray plastic to the right. It could be that higher than normal reflectance is biasing the exposure a bit. Which would probably show up as a + exposure adjustment since the ETTL metering algorithm would be telling to flash to quench early.

There is a shadow in the ambient sample suggesting that the ambient light is camera left. The two flash samples appear to have been made with the flash in the hot shoe. I'm not sure how much, but it may be that the shadow, since its in the area of greatest sensitivity for evaluative metering, is biasing the reading.

Wilt, I realized I should point out that I am making these observations in the interest of getting an empirical measuring environment that is a free from extraneous bias as possible. Since the objective is to determine the accuracy of two flash metering methods relative to a third metering method for ambient light, it seems best to ensure an environment for the measurements that is a neutral as is possible.

dmward wrote in post #15219836
Wilt, the three are within less than half a stop. That's well within an acceptable tolerance from null value. Since one is -.14 and the one farthest in the other direction is +.34 the range is within plus minus 1/3 stop of null which is the closest settable interval in the camera metering system.

It looks like these are all full frame. If I were doing the test, I'd flip the color checker so the gray scale in in the middle of the frame. That would put the 4th square just about in the exact middle of the frame and most likely in the middle of the auto sensor field as well.

It also looks like there white balance is a bit different for the three. At least the ambient is visibly warmer than the others. That can have an impact on the exposure reading as well.

It also looks like the tone curve is not linear. Not sure exactly what the settings are but they may be influencing the 4th square since it doesn't look like its been pinned.

It also appears that the middle image, ETTL metered, has a bit of sheen on the gray plastic to the right. It could be that higher than normal reflectance is biasing the exposure a bit. Which would probably show up as a + exposure adjustment since the ETTL metering algorithm would be telling to flash to quench early.

There is a shadow in the ambient sample suggesting that the ambient light is camera left. The two flash samples appear to have been made with the flash in the hot shoe. I'm not sure how much, but it may be that the shadow, since its in the area of greatest sensitivity for evaluative metering, is biasing the reading.

Wilt, I realized I should point out that I am making these observations in the interest of getting an empirical measuring environment that is a free from extraneous bias as possible. Since the objective is to determine the accuracy of two flash metering methods relative to a third metering method for ambient light, it seems best to ensure an environment for the measurements that is a neutral as is possible.

I will not debate the validity of any of the points raised...my point in doing this test was merely to demonstrate the type of testing that could demonstrate how closely to 'correct' the flash modes (ETTL vs. External) could be validated. The variables mentioned might improve the precision, but I doubt the effect makes much practical difference. After I finish this reply, I'll check to see what variance the WB issue has, if any, and report back!

Wilt wrote in post #15219983
I will not debate the validity of any of the points raised...my point in doing this test was merely to demonstrate the type of testing that could demonstrate how closely to 'correct' the flash modes (ETTL vs. External) could be validated. The variables mentioned might improve the precision, but I doubt the effect makes much practical difference. After I finish this reply, I'll check to see what variance the WB issue has, if any, and report back!

I rebalanced the slightly warm ambient-only shot, that lowered the density values from 56-57% at WB 9100K, down to 54-55% at WB 7400K. After doing that, then I rematched the two flash shots, to give 54-55% RGB values...

The ETTL shot changed from needing +0.32 to instead needing +0.23, or less than 0.1EV change in the evaluation.
The Auto photosensor shot changed from needing -0.32EV to instead needing -0.23, again less than 0.1EV change in the evaluation.

In short, the fundamental need for slightly ADDITIONAL exposure from ETTL flash vs. needing slightly LESS exposure with Metz photosensor remains a consistent finding. We have 'flash density within 0.25EV/0.33EV (depending upon which test you want to believe) of ambient density' demonstrates what kind of results are achievable with (Metz) flash.

We're polishing a turd, methinks! I remain resolute that similar testing will disclose the error or accuracy of Canon External mode flash.

I think we're saying essentially the same thing.
Both methods for metering flash exposure are well within the realm of acceptability.
The WB correction narrowed the differential to within .25 of null. I don't know of many instances where a 1/4 stop exposure variance has a meaningful impact on an image. Considering the processing capabilities that are now available from programs like Lightroom V4 with process 2012 even half a stop is essentially a rounding error.

Wilt, I wasn't attempting to make things difficult, just attempting to point out that as external variables are eliminated, the differences diminish. Which in my view is a good thing.
Its nice to know that there are alternative automated metering options available on my 60EX-RTs that will deliver acceptable results.