How to obtain (predictable) flash photography with Canon's ETTL-II?

[DriesA]

Hi,

I’m a little confused by Canon’s new flash metering system, called ETTL-II. Some weeks ago I bought myself a Canon 20D after having worked with a Canon 10D for 2,5 years. In this time period I managed to get used to the flash metering system of the 10D. In my opinion, using flash photography with the 10D you can’t do without shooting in autofocus mode (I use the external flash 550EX by the way). Especially when you have to react quickly. The reason for this is that flash metering with the 10D is linked strongly to a pretty narrow metering zone around the active autofocus point, i.e. mostly also the zone in which your main subject resides. So, the “only” thing you were supposed to do was making your composition, choosing your autofocus point to focus your subject, making some flash exposure compensation if necessary based on the surrounding zone (never re-compose, unless you used FEL (flash exposure lock)), and you would end up with a nicely exposured picture. On the other hand, when not using autofocus, the 10D only uses the small central part (in and slightly around the partial metering circle) of your composition for metering. Hence it appeared to be difficult to me using manual focus mode because flash metering and focusing is done in different zones withing your composition. But, as said, using autofocus mode, you could make pretty predictable exposures.

So, I managed to master the flash technique of the 10D for my purposes. But then Canon came in with a whole new metering system on the 20D and the professional digital camera series: the so called ETTL-II. As far as I have experienced uptill now, with this system flash exposure metering and focus points aren’t linked to each other any more. Something I regret of course. What’s more, the calculation of the “correct” exposure seems to happen fully automatical, and hence is in no way predictable! Seemingly, Canon has opted with ETTL-II for an automatic flash metering system that tries as much as possible to prevent blowed out highlights. In my experience, this results in often (severely) underexposed pictures. Of course, I know that when you are shooting in a dim place with brightly colored walls, you normally have to apply some positive exposure compensation, but how do you know how much, if the metering system seemingly don’t give you any idea how and where calculating this, since the zone around your autofocus point doesn’t play any role any longer?? I can’t hardly believe anyone can live with this thought. It means you can’t play creatively with flash light anymore, unless on a trial and error basis! At least I can’t live with that … After all, you don’t always have the time to check your histogram after every shot.
Soit, maybe I am fully wrong and don’t understand anything about the new system. At least, my last wedding photography session gave me some experience, but not enough to find out the logic behind ETTL-II. As far as I experienced, it’s hard (if ever) to get an overexposed picture. Mostly they are (seriously) underexposed. Something that is going against the rule of “expose to the right” when using raw.

Hence my question to you is whether there is someone knowing the logic you have to handle to obtain a decently exposed picture with ETTL-II. How do you know in which direction you have to apply flash exposure compensation if you don’t know how the calculation behind it is done? And is there someone who knows the difference between “Average” and “Evaluative” options under custom camera function ETTL-II and the consequences for the exposure? Which one is best used for optimal manual control and how then? Uptill now, all this things aren’t clear to me at all … And yes, I have devoured the article on eos flash photography on photonotes.org without turning out any wiser, alas. Moreover, the author is complaining about Canon not giving any decent explanation to their customers on how to handle with ETTL-II or any of their pre-existing flash metering systems.

So I’m hungrily hoping for a enlightening answer. Maybe worth mentioning still: I’m working with an Omnibounce diffusor attached to my 550EX at a 30° level most of the time, so the distance metering ETTL-II incorporates (one of the new features) isn’t of any use here since this measuring is relevant only in case of direct flash. Further, the program mode used is mostly “M” which enables me to control both shutter speed and diafragm, or “Av” when ambient light is sufficient to hold by hand and only minor fill-in flash is needed. I always use one single autofocus point to focus on my main subject after making my composition (so no automatic focus point selection, nor recomposition afterwards). Maybe these notes turn out to be important …

Greetings.

Dries

[PacAce]

Here are a couple of test shots I took with my 1DmkII (ETTL II) and 10D (ETTL). Review the shots and tell me wish you think is more predictable in terms of the flash results.

The shots below were taken with the 1D2 (top) and 10D (bottom) with flash set to Manual. The subject to camera flash distance was measured and used to determine the aperture needed for proper exposure (as indicated by the distance scale on the 550EX). As you can see, all the images are pretty consistent, not just among the pictures taken with the same camera but also between the 1D2 and the 10D. This should not be a surprise since the flash was set to manual and all the pictures were taken with teh same aperture (shutter on 1D2 was set to 1/250 and 10D to 1/200).
.

[PacAce]

The next set of shots were taken with the 1D2. The top row are the same shots you saw above, taken with flash on manual.

The 2nd row are shots taken with the flash in ETTL mode and the camera set to ETTL Evaluative mode.

The 3rd row are shots taken with the flash in ETTL mode and the camera set to ETTL Average mode.

Although the exposures are varied in the ETTL mode shots, depending on the reflectivity of the subject, they don't vary that much from the top row shot in manual flash mode.
.

[PacAce]

And lastly, these pictures were shot with the 10D.

The top row shows the pictures taken with the flash in manual mode.

The bottom row shows the pictures taken with the flash in ETTL mode. These pictures show that not only do the pictures differ substantially from those taken with the flash on manual but also among themselves depending on the reflectivity of the subject.

I should point out these shots were taken with the focusing set to the * button which were not pressed when I took the picture so AF point biasing was not really used here. That fact just occurred to me after I completed processing the test shots. I'll set up another test with the 10D with the focusing point active to see how much those shots will differ from these.
.

[DriesA]

Thanks, Leo. Nice comparison between the two systems!

This should not be a surprise since the flash was set to manual and all the pictures were taken with teh same aperture (shutter on 1D2 was set to 1/250 and 10D to 1/200).

That's true indeed. With bursting the same amount of light on the scene, exposure is supposed to be consistent, regardless of subjects reflectivity.

Although the exposures are varied in the ETTL mode shots, depending on the reflectivity of the subject, they don't vary that much from the top row shot in manual flash mode.

There seems to be only very minor differences between "average" and "evaluative" ETTL-II indeed. I don't understand Canon offering this option without decently instructing their users the difference between both possibilities.

I should point out these shots were taken with the focusing set to the * button which were not pressed when I took the picture so AF point biasing was not really used here. That fact just occurred to me after I completed processing the test shots. I'll set up another test with the 10D with the focusing point active to see how much those shots will differ from these.

That would be great. Anyway, you can see clearly the difference in metering system between ETTL and ETTL-II. ETTL on the 10D is in manual focus mode only focusing on the central partial metering circle, which explains the big differences in exposure between the shots, based on subjects reflectivity. But at least here you know on which part of the scene measuring is done. This is not the case with the ETTL-II. And then it is very hard to estimate the amount of exposure compensation you have to apply to obtain a decent overall exposure. You'll see that with the use of an single autofocuspoint on the 10D, flash metering is based on the very near surroundings of the respective point. With ETTL-II however, you won't notice barely any difference in exposure when switching between different autofocus points. And if there is a difference, then it's hard to tell why. Hence, in my opinion, the ETTL system with narrow metering around the focus point is more convenient in estimating the amount of flash compensation than in case of ETTL-II, with which you don't know how and based on which parts of the scene metering is done. You simply don't have any insight in the underlying calculation method, which I severly regret. Uptill now I don't see how you "master" ETTL-II unless on a trial and error basis, hoping the predefined (and regrettably undisclosed) algorithms are able to manage your scene. Although ETTL-II produces in most cases a decent exposure, which is more consistent than that based on ETTL (which assumes manual intervention by dialing in the corresponding flash compensation), gives you the freedom of constantly recomposing and enables you to react quickly, the result remains highly unpredictible in the sense that you don't know based on what you have to apply any compensation eventually.

So, if anyone knows how to deal with the uncertainty of ETTL-II in terms of managing differently illuminated scenes, I would like to hear how ... I think that your experience is worth more than any manual about flash photography.

Thanks.

Dries

[PacAce]

Dries, these samples are just one of many varied conditions that you could find yourself shooting in. I think these shots could represent your typical indoor flash shots. Later tonight, I'll shot a couple of test shots that may not be typical but are encountered often and you'll see how ETTL and ETTL-II each handle it and which gives you the upper hand.

As for the difference between ETTL-II Evaluative and Average, are you familar with the auto-thyristor flashes which automatically set the flash for you without interacting with the camera? Well, you can just think of the ETTL-II Average mode to work in a similar fashion. The exposure set by the camera is based on the reading of the flash averaged over the entire frame (well, actually ,over the entire area of the frame where there are light sensors, of course). This is basically what happens in ETTL, too, when there is no active focusing point in use (as in my 10D flash shots above).

With ETTL-II Evaluative, the camera will try to determine where the subject is and then bias the exposure for the subject. So, although you are not using the AF point to determine how much EC you should apply as you would with ETTL, you should be able to look at the overal subject to determine the EC, no? I guess the analogy of this is comparing an exposure determined by using a spot meter and that determined by using a partial spot meter. As long as the proper EC is specified for each reading, you'll be able to get a nicely exposed picture. Of course, the amount of EC required for each will be different. And my take is that you may need more drastic EC when shooting with a spot meter if the spot you're reading from does not truly represent the reflectivity of the overall subject. I think with the partial meter, you'll get a lot closer to the overal subject reflectivity.

Of course, you can argue that you can always position the spot meter over an area that is mid-tone gray and that is a valid argument, too.

[Curtis N]

Thanks for posting these shots, Leo. I know how time-consuming such a task can be.
They tell quite a story when you take the time to study them.
Maybe this thread should be added to a "sticky" somewhere.

[PacAce]

The do tell a story, don't they, Curtis? I found that photo sessoins like these are the best way, at least for me, to actually see what the flash (and even the camera, for that matter) is doing under different shooting conditions. It sort of adds dimension to the theoretical stuff that we read about here in the forum, in other websites and in books.

One other thing I'm going to try to test is how different levels of ambient lighting controls the actual light put out by the flash. From what I understand, ETTL-II is supposed to make the lighting look as close to the natural lighting as possible. So if you are going to be shooting in a dimly lit room, the shot is going to come out a little darker (under exposed) than if the same shot were taken with more ambient lighting turned on in the room.

[DavidEB]

PaceAce --

that's a very useful demo, thanks for the effort.

the middle shots (black jacket) in ETTL but not manual modes seem to have a different white balance. this is puzzling to me -- any thoughts here?

[PacAce]

As promised, here are the test shots, shown below, comparing ETTL with active AF point (top) shot with the 10D, and ETTL-II in Evaluative mode (bottom) shot with the 1D2.

In all shot, the 550EX flash in ETTL mode was used with both cameras set to Manual @ ISO 200, f/8, 1/200. The focusing point used in all the shots was the center AF point which is on the gray card placed under neat the puppet's chin.

I included a reference image, shot with the flash on manual. It's the leftmost image with the black/gray/white card over the puppet's head.

My thought in using the gray card was to provide a neutral spot for the 10D to focus on and lock in the ETTL exposure. With ETTL exposure bias leaning more towards the AF point, the expectation is that the exposures should not vary widely from one shot to the other, no matter what the surround reflectivity is. This behaviour supposedly makes ETTL flash photography more predictable--just determine the reflectivity of the area under the AF point and adjust EC accordingly.

Well, as the top row show, that's not exactly how ETTL behaved on my 10D. It is very evident that the exposure varied substantially. In the top row below, the 3rd shot from the left with the black jacket was overexposed while the last one with the white shirt was underexposed, both more than the ETTL-II counterpart shown underneath them.

As far as exposure and color accuracy is concerned, first prize goes to the 2nd picture from the left at the bottom, taken with the 1D2 of the brown coat. The one above it, shot with the 10D, is a close second.

Another obversation I made is that, although the exposure of images shot with the flash on manual seem to be dead on, the color is not true to life, as they look too cold.
.

[PacAce]

Originally Posted by DavidEB

PaceAce --

that's a very useful demo, thanks for the effort.

the middle shots (black jacket) in ETTL but not manual modes seem to have a different white balance. this is puzzling to me -- any thoughts here?

David,

All the shots were taken with WB set to Auto so I really can't explain it other than to say that it may look that way because it was over-exposed. If you look at the latest set of tests I did, the color of the 10D black jacket shot isn't as bad because the AF point biasing prevented the shot from being too over-exposed. The more accurage colors are those shown in the shots with the brownish colored jacked. Even the reference shots where the flash was set to Manual, the color is off, way off actually.

[PacAce]

Earlier I mentioned something about the ambient EV level affecting the ETTL-II flash results. Below are the results of the test I did to see if there was any truth to that. The first image below was shot with the 1D2 in ETTL-II Evaluative mode. The leftmost shot was taken with all the lights turned on so that the room is relatively bright. The second show was taken with all the lights turned off. And the third shot is a reference shot with the flash turned off and all the lights turned on to show that ambient lighting would not have had any effect on the actual exposure of the shot.

The camera was set to Manual, ISO 200, f/8, 1/200

The bottom image shows the same test done with the 10D. The first one was shot with all the lights on and the 2nd was with all the lights off. There seems to be a slight difference in exposure but it's not as noticeable as that of the 1D2.
.

[PacAce]

Here's my last test for today.

I've had many experiences with auto-thyristor flashes where any type of reflection of the flash in the picture almost guarantees a ruined shot. Well, I have read many writeups that tout ETTL-II's ability to deal with such specular reflections, so I did a test to see how good a job it really does with the reflection of its own flash. I also tested to see how ETTL fairs in a similar situation.

The first image below shows the results of my first test. The one on the left is ETTL-II (1D2) and the one on the right is ETTL (10D). Hmm, maybe they were right after all about what ETTL-II can do with specular reflections. What I was really suprised about was how well ETTL did against the reflection. I was expecting it to be more underexposed than it was because that was my experience when I encountered a similar situation in "real life" before.

So, to investigate the reason for this discrepancy , I took a couple of more test shots. What I did with each one was to change the angle of the camera so that the reflections occured at different spots. The second image is a compilation of the different shots. In each case, the focus point was on the face of the "model" with the reflection occuring nearby.

The top row are those of the 10D and the bottom row are those of the 1D2. As you can plainly see, the performance of the 10D varies widely depending on the location of the reflection relative to the AF point. The 2nd image from the left looks like what I experienced before.

What really surprised me about the last test results shown in the 2nd image was the fact that ETTL-II really didn't handle the specular reflection as well or as consistently as I thought it would, the first image not withstanding. Although the underexposed experienced wasn't as severe as that of the 10D, it nevertheless did underexpose the picture when the reflection was very close to the subject.

So what did we learn here? ETTL-II and ETTL will handle specular reflections very well, as long as you keep them away from the AF point (for ETTL) and the main subject (ETTL-II).
.

[DriesA]

Thanks for posting these shots, Leo. I know how time-consuming such a task can be.
They tell quite a story when you take the time to study them.

It sort of adds dimension to the theoretical stuff that we read about here in the forum, in other websites and in books.

Well, I can but agree on that, Leo. Great job. Very instructive!

My thought in using the gray card was to provide a neutral spot for the 10D to focus on and lock in the ETTL exposure. With ETTL exposure bias leaning more towards the AF point, the expectation is that the exposures should not vary widely from one shot to the other, no matter what the surround reflectivity is. This behaviour supposedly makes ETTL flash photography more predictable--just determine the reflectivity of the area under the AF point and adjust EC accordingly.

Well, as the top row show, that's not exactly how ETTL behaved on my 10D. It is very evident that the exposure varied substantially.

Well, I have to say that it's not merely the zone exactly under the autofocuspoint that is used for flash metering by the camera. It's a bit broader than that, resulting in some underexposure/overexposure in case of the white/black coat respectively. The metering zone is taking in acount more than only the grey card ... How big the metering zone exactly is, Canon isn't telling us nowhere. Once more, alas.

Another obversation I made is that, although the exposure of images shot with the flash on manual seem to be dead on, the color is not true to life, as they look too cold.

That's strange, isn't it. Seems like the white balance if fooled one way or another when setting the flash output manually.

As for the difference between ETTL-II Evaluative and Average, are you familar with the auto-thyristor flashes which automatically set the flash for you without interacting with the camera? Well, you can just think of the ETTL-II Average mode to work in a similar fashion.

OK, that may work to understand the reaction of the camera ... Thanks.

With ETTL-II Evaluative, the camera will try to determine where the subject is and then bias the exposure for the subject. So, although you are not using the AF point to determine how much EC you should apply as you would with ETTL, you should be able to look at the overal subject to determine the EC, no?

Well, yes, I should be able. But I think this will be a matter of experience in the long run. Every new Canon camera series is responding (although sometimes only slightly) different. This doesn't make life easier, isn't it? Fact is that in judging the appropriate FEC, you can be mislead by the camera using totally different calculation algorithms for only slightly different real world scenes. That's what bothers me most ... Unpredictability.

I guess the analogy of this is comparing an exposure determined by using a spot meter and that determined by using a partial spot meter. As long as the proper EC is specified for each reading, you'll be able to get a nicely exposed picture. Of course, the amount of EC required for each will be different. And my take is that you may need more drastic EC when shooting with a spot meter if the spot you're reading from does not truly represent the reflectivity of the overall subject. I think with the partial meter, you'll get a lot closer to the overal subject reflectivity.

I agree on that. That's why I consider the FEL (flash exposure lock) often very useful. It uses only the partial metering circle for flash metering, so you can pretty good estimate the EC required. But in my opinion you can't use it when photographing people, since it implies a nasty preflash which make people believe the shot is already done while the actual exposure still has to come ...

The ambient light and reflection test shots are very instructive too. Thanks for that. But giving ambient light a role in your shot, requires actively exposing for ambient light and use the flash only for fill-in of the subject. This is best done in Av or Tv mode, or in M mode of course. The drawback is that you will need a tripod, although some movement (because of the slow shutter speed) combined with the freezing of the scene by the flash (using second curtain) can give a very pleasing result too. Often I find myself shooting with 1/10s or lower when trying to catch the atmosphere or movement.

Once more, thanks for the effort, Leo. I've never been helped more with such a well-thought answer in any other online forum.
And as I may conclude from all your answers, flash photography will always include a certain uncertainty. Or am I wrong (not mentioning external flash meters as a help)? I will have to reconcile with that, I suppose. I'll just keep building my experience ...

Thanks again and keep posting!

Dries

[DriesA]

By the way, this site of canon is also pretty instructive. Of course far from as deep and technical as this thread, but I consider the pictures, text and audio quite good. It gives you a good preliminary idea of the eos speedlite capabilities. You have to check the tips and tricks section on eos speedlites.

Dries