E-TTL, active focus point, and compensation

[Tom W]

I did a little experimenting today with the flash, and thought I'd share the results. While many here are already familiar with the "quirks" of E-TTL, there are quite a few here that might benefit from this.

Anyway, here it is:

Mr. Tan Hat and Miss White Door have decided to get married, and have invited Tom W as their photographer. So Tom, always eager to the task gathers up his 10D, appropriate lenses, 550EX flash, a few sets of batteries, and assorted other equipment and heads off to take some wedding photos.

In the first photo, the active focus point is centered right over Mr. Tan Hat's tan hat. This provides a satisfactory, balanced exposure with some detail showing in both Mr. Tan Hat's dark jacket and Miss White Door's white dress.

http://images.fotopic.net/?id=4790681

In the second shot, the active focus point lands right over the dark jacket that Mr. Tan Hat is wearing. Due to the way E-TTL biases the flash exposure to the active focus point, this results in good exposure of the dark jacket, but overexopsure of the picture in general. Note the blown-out detail in Miss White Door's garb.

http://images.fotopic.net/?id=4790679

After checking his histogram, Tom the photographer quickly adjusts his flash exposure compensation. Since he usually shoots at +1/3 FEC, he subtracted 1 whole stop from that and took the same shot with a net FEC of -2/3 stops. This brought an overall satisfactory exposure again, with detail in both dark and light areas.

http://images.fotopic.net/?id=4790677

Minutes later, Tom the photographer takes another picture, this time placing his active focus point right over Miss White Door's bright white dress. The camera exposes the dress a little under, but leaves the rest of the composition very underexposed.

http://images.fotopic.net/?id=4790678

Again, after checking the histogram, Tom the Photographer readjusts the FEC from the +1/3 setting that he usually uses to +1 1/3 stops and re-takes the picture. This time, the overall exposure is balanced with detail showing in both dark and light areas.

http://images.fotopic.net/?id=4790680

Because Tom the photographer effectively used his histogram and his flash exposure compensation when dealing with very light or very dark subjects, he was able to take several hundred wonderful pictures of the new Mr. and Mrs. Tan Hat's wedding. They loved the pictures, paid him a handsome sum, and everyone lived happily ever after.

[Cadwell]

As long as you got paid, that's alright ;) :lol:

Interesting example :)

[Tom W]

Why Mr Tan Hat gave me the coat off his back. ;)

[robertwgross]

Good example. You better write a "how-to" book.

---Bob Gross---

[Mike H]

Tom, it looks like from those results that unless we are dumb enough to put the active focus point over a pure white or a pure black, we'll probably land on our feet when using E-TTL2 (especially if shooting RAW).

I wonder what would have happened if you used bounce flash instead? Maybe the happy couple will need some honeymoon photos.

Thanks.

Mike H

[Tom W]

Tom, it looks like from those results that unless we are dumb enough to put the active focus point over a pure white or a pure black, we'll probably land on our feet when using E-TTL2 (especially if shooting RAW).

Well, if II is as good as promised, you should be able to compose the picture as you wish without worrying about extreme dark or light subject material. And sometimes, in order to get the composition you want, the active focus point unfortunately does fall over the bride's dress or veil, or the groom's navy tux (or the black Mercedes in the garage or whatever).

I shot these RAW, but didn't do anything to them in the conversion - they're pretty much straight out of the camera.

I wonder what would have happened if you used bounce flash instead? Maybe the happy couple will need some honeymoon photos.

Thanks.

Mike H

Bounce flash would have brought similar results (without the shadow of Mr. Tan Hat's hat, of course). E-TTL meters on the reflected light of the pre-flash, and will automatically adjust output to compensate for bounce, or diffusers and such, so that the same amount of light will hit the target.

I've shot a few with bounce on this flash, but this was adequate to illustrate the effects of dark and light subjects and the necessary FEC that helps overcome these effects.

If I get a chance after dinner this evening, I may fiddle around a little more with bounce. The 550 is new for me, but it does resemble my 380EX enough that it has been pretty easy for me to adjust to.

[Mike H]

Thanks for the response.

I'm not so sure about bounce giving you the same results, since this new system is working distance into the equation, something that is complicated by the use of bounce flash. I will have to convince myself that you're right with some tests.

You would likely get different results if custom function 14 were set to average exposure over the entire AF elipse (page 151 of the manual), rather than evaluative metering. Several users on another board report better results when setting the flash metering to the averaging pattern. In your sample photos, my guess is that it would work better (since it would average the blacks and whites). It would be critical for wedding shooters to know which pattern works better.

I'm planning to do some tests when my neighbors come back into town in a few days. They're such good sports. :D

Mike H

[Tom W]

Thanks for the response.

I'm not so sure about bounce giving you the same results, since this new system is working distance into the equation, something that is complicated by the use of bounce flash. I will have to convince myself that you're right with some tests.

You would likely get different results if custom function 14 were set to average exposure over the entire AF elipse (page 151 of the manual), rather than evaluative metering. Several users on another board report better results when setting the flash metering to the averaging pattern. In your sample photos, my guess is that it would work better (since it would average the blacks and whites). It would be critical for wedding shooters to know which pattern works better.

I'm planning to do some tests when my neighbors come back into town in a few days. They're such good sports. :D

Mike H

I apologize, Mike - I probably should have noted that I'm shooting with the 10D. All I have is "regular" E-TTL, not the newer version. Which is why bounce flash really won't change the exposure in my case.

Given that you have E-TTL II, I would think that your results will be more consistent, especially if you have an average exposure option for tough situations.

I am looking forward to seeing how the newer E-TTL II does in a similar situation, especially since I am of the mind that Canon will include it in the next consumer grade digital SLR, just as they've included it in the Elan 7N. Unless I suddenly change professions (and start making money with cameras), I'm not likely to splurge on the 1D Mk II any time soon, even if I crave it.

[Mike H]

Tom, no need to apologize. I didn't read your original post carefully enough; you did say that you were using the 10D.

If it makes you feel any better about not getting a Mark II, it's heavy, expensive, and much more complicated to use than the 10D. I'm going to need a lot more time to learn this camera than the 10D. That's one of the downsides of having a ton of features.

Mike H

[Sendide]

thanks for this very instructive example tom. as you said, there are a lot of people who will take great advantage of this post (myself included). but that makes me need to learn more about how to correctly read pictures histogram since that was how you guessed that the picture was over or under exposed. is it about how homogenous the histogram is? moreover, how do you know that the shot is over or under exposed (I mean for more complicated picture composition).
please feel free to giave some documentation references for better understanding histograms
regards

Khalid

[CyberDyneSystems]

Very helpfull post!

Thanks Tom!

[Tom W]

Khalid, sometimes, you don't even need to look at the histogram to see if a shot is going to be way over- or under-exposed. The overexposed highlights will flash on the LCD after the shot, and the underexposed will just look too dark. But the LCD isn't the best tool for looking at the actual image except for composition. Better to use it to view the histogram.

Anyway, I always keep my camera LCD set to show the histogram along with a small version of the image. Unfortunately, I don't know how to get the histogram image into a JPG so that I can post it here, or I could show you just what I was dealing with.

EDIT: With the help of PacAce, I was able to preserve and create copies of the histograms for each of the 5 shots so I will include them with the following text as needed. Pardon the nasty JPEG artifacts on these - they aren't to be viewed as artwork. :)

Anyway, I'll "borrow" this simple chart from the Luminous-Landscape web site:

http://www.luminous-landscape.com/images7/dark-light.jpg

As you can see, the chart is simply a graph showing the quantity of pixels that exist in the image at the various levels of "darkness" or "lightness". The extreme left of the chart is black, while the extreme right of the chart is white and everything in between is varied levels of brightness.

Now, using the chart against my photos above, I'll try to describe what the chart looks like for each image (since I can't post them).

http://images.fotopic.net/?id=4834051&outx=600&noresize=1&nostamp=1

The first image is fairly well exposed. The histogram shows a fairly even distribution of dark and light pixels. More importantly, there are no pixels bunched up at either end of the chart.

EDIT: I probably could have exposed about 1/3 stop higher without overexposing, but in my opinion, there's enough leeway in a RAW image (or a JPEG really) to tweak 1/3 stop without any serious consequences.



http://images.fotopic.net/?id=4834050&outx=600&noresize=1&nostamp=1

Image #2 exposed to show detail in the dark jacket, but overexposed the overall image. On the histogram, this shows up with a large "lump" bunched up against the right end of the chart. This is indicative of "blown out" highlights, and will result in loss of detail in the brighter parts of the image (in this case, the grain of the closet door is lost). Blown out highlights will also flash in the LCD image display. (EDIT: See above)



http://images.fotopic.net/?id=4834052&outx=600&noresize=1&nostamp=1

Image #3, I have corrected the overall exposure by reducing flash output by 1 stop (in my experience, one stop is roughly equal to 1/5 of the histogram's chart or the length from one vertical bar to the next). This took care of the "blown" highlights on the closed door, and still kept some detail in the jacket. Virtually all of the histogram information is within the boundaries of the chart, with none bunched up against either end. Plus, there are no large areas at either end of the chart with little or no information. All the camera's output data is visible on the chart. This is a good exposure. (EDIT: See above)



http://images.fotopic.net/?id=4834172&outx=600&noresize=1&nostamp=1

Image #4, I've taken a shot of "Miss Closet Door" without compensating for the darker parts of the image. This produced good detail of the woodgrain on the door, but underexposed "Mr. Tan Hat's" jacket. The histogram shows a good deal of information in the center of the chart (the closet door), but also a large lump of data near the left (dark) end and almost nothing towards the right end of the chart. This is underexposed. (EDIT: See above)



http://images.fotopic.net/?id=4834049&outx=600&noresize=1&nostamp=1

The last image, I used flash exposure compensation to increase the exposure level by 1 stop. This resulted in a better exposure, with more detail on the jacket while retaining the detail of the white door. The histogram for this shot showed a large amount of data towards the right end of the chart (the door), but not bunched up against the end which would indicate blown out highlights, and the data that was close to the left end is more spread out and moved away from the left edge of the chart. This darker "bunch" of data is the dark coat. By getting it away from the left edge, it gives the data some "space" to show greater detail. (EDIT: See above)

One last note - once data is bunched up at either the dark or the light end, there is no differentiation between levels of lightness or darkness. It cannot be darker than black nor lighter than pure white. So, any differences in brightness in those parts of the photograph are lost.

Anyway, since I "borrowed" their chart, here is the link to an article at Luminous-Landscape that gives some information of the use of the histogram:


http://www.luminous-landscape.com/tutorials/understanding-series/understanding-histograms.shtml

[Sendide]

Top contributer , top explanation :wink:
thanks a lot TOM

[slin100]

It would have been interesting to see the effect of CF 4-1, which remaps AF to the * button, on ETTL. On a 10D (and probably the 300D), ETTL is supposed to operate more like an averaging auto flash in this case.

[SDK^]

Just going back to Custom Function 14 – Auto Reduction of Fill Flash
Can someone please explain the difference this makes and the possible situations you might use either option

Thanks :)

[PacAce]

Just going back to Custom Function 14 – Auto Reduction of Fill Flash
Can someone please explain the difference this makes and the possible situations you might use either option

Thanks :)

As is explained in the 10D maual on page 150, when auto reduction of fill flash is enabled, the flash output is reduced so that it does not light up the subject that you are flash filling as brightly as the nature light source does of the background. This usually results in a more natural look, especially if the subject is in the shade.

If you need to light the subject with full flash power, say when you don't really care about the background but you do want the subject to be properly exposed, then you would disable the auto reduction.

[PacAce]

It would have been interesting to see the effect of CF 4-1, which remaps AF to the * button, on ETTL. On a 10D (and probably the 300D), ETTL is supposed to operate more like an averaging auto flash in this case.

I just did an experiment on this and came up with the following conclusions:

Case 1: When shooting in Autofocus mode with CF4=0, Evaluative metering seems to be used for metering the flash.

Case 2: When shooting in Autofocus mode with CF4=1 ("*" button sets focus), Center-weighted average metering seems to be used.

Case 3: When shotting in Manual focus mode (doesn't matter if CF4=0 or CF4=1), Center-weighted average meterng seems to be used.

This is very interesting because, for the subject I used--a very dark brown stuffed bear filling the entire center circle that makes up partial metering area against a white wall filling the rest of the viewfinder, the flash using center-weighted average metering underexposed by about one stop.

Both EC and FEC were set to 0 in all cases.

[Tapeman]

Great post Tom.

ETTL is a mystery to many of us. The manual for the 550EX is very confusing.

My congratulations to the happy couple! :D

[u02bnpx]

ETTL is a mystery to many of us. The manual for the 550EX is very confusing. :D

As a frustrated 550EX user, I heartily agree with this sentiment. And yet these and other forums are filled with their share of true believers---those who swear by their 550s, those who go on to blame its failures upon well-intentioned camera users and manual readers. "Get to know your equipment," they say, "and all will be well." In other words, do a lot of trial and error, and then you'll understand which settings you'll need to make in any given conditions.

However, as we all know, no two exposures are made with exactly identical conditions, and so the trial and error just seems to go on and on. I feel that I should be wearing a T-shirt with a message that reads: "I'll need to take at least 3 flash shots, so stay right there after the first one."

It amuses the bejeebies out of me to conclude that all this talk about plus or minus 1/3 and other methods of compensation reveals how truly non-automated the 550EX is. There's a heckuva lot more tweaking that needs to be applied by intuition and hunch on Canon's flash than any of us needed to use with our old Vivitars on rangefinder cameras.

OK, 550EX defenders, have at me now!

Floyd Lawrence

[PacAce]

It amuses the bejeebies out of me to conclude that all this talk about plus or minus 1/3 and other methods of compensation reveals how truly non-automated the 550EX is. There's a heckuva lot more tweaking that needs to be applied by intuition and hunch on Canon's flash than any of us needed to use with our old Vivitars on rangefinder cameras.

OK, 550EX defenders, have at me now!

Floyd Lawrence

And it amuses me when I read people ranting and raving about a piece of equipment that's gotten the best of them. As far as I'm concerned the 550EX is a flash just like any other flash I've used with any of my cameras, digital or otherwise. And I've used a heck of a lot of them, manuals and auto, Vivitars, Sunpaks, etc. They all serve one purpose and that's to provide light for the camera. And the same laws of physics apply to all of them, whether there's "intelligence" built into them or not. I'm surprised you're not complaining about the Evaluative metering used by the camera when shooting pictures without the flash. Or are you?

And for the record, I'm not a defender of the 550ex or even ETTL for that matter. I'd be just as happy using my old non-dedicated Sunpak flash if it worked with my 10D.

[Tom W]

ETTL is a mystery to many of us. The manual for the 550EX is very confusing. :D

As a frustrated 550EX user, I heartily agree with this sentiment. And yet these and other forums are filled with their share of true believers---those who swear by their 550s, those who go on to blame its failures upon well-intentioned camera users and manual readers. "Get to know your equipment," they say, "and all will be well." In other words, do a lot of trial and error, and then you'll understand which settings you'll need to make in any given conditions.

I really don't share your frustation. I've encountered but a few rare instances where the E-TTL is apparently confounded and I've tried to illustrate at least some of them here for the benefit of others.

However, as we all know, no two exposures are made with exactly identical conditions, and so the trial and error just seems to go on and on. I feel that I should be wearing a T-shirt with a message that reads: "I'll need to take at least 3 flash shots, so stay right there after the first one."

I haven't encountered that much trouble with the system, even as a relative novice.

It amuses the bejeebies out of me to conclude that all this talk about plus or minus 1/3 and other methods of compensation reveals how truly non-automated the 550EX is. There's a heckuva lot more tweaking that needs to be applied by intuition and hunch on Canon's flash than any of us needed to use with our old Vivitars on rangefinder cameras.

OK, 550EX defenders, have at me now!

Floyd Lawrence

Well, what on earth is exposure compensation for but to compensate for unusual conditions. If a flash exposure is off by 1/3 or even 2/3 stop, it is easily corrected in EVU before it is even converted to JPEG and without any significant loss of fidelity.

Yes, the system requires some input from the user on occasion, but it is a heck of a lot more useful than than having to change the aperture for every shot based on distance to the subject (which is often unpredictable). E-TTL is not without its faults, but it is quite spontaneous, or at least it has been in my use. I can basically pick up the camera, compose, and stand a very good chance of getting a good exposure almost every time.

If you want something that requires no thought at all, then you probably need a point-and-shoot camera.

[robertwgross]

Floyd, you are absolutely right. That 550EX must be a piece of design junk.

You better pack it up and send it off to me right now. I'll take it off your hands, purely as a friend would do.

(By the way, I have one already, and also one 420EX.)

---Bob Gross---

[u02bnpx]

[/quote]And it amuses me when I read people ranting and raving about a piece of equipment that's gotten the best of them. [/quote]

[PacAce]

And it amuses me when I read people ranting and raving about a piece of equipment that's gotten the best of them. [/quote][/quote]

Err, cat got your tongue? ???

[u02bnpx]

The response to my deliberately grouchy post was predictable. First, I'm probably some anti-technology Luddite who dislikes getting stumped by a piece of equipment.

Of course I do. And so do all of us. Or am I the only one whose Windows XP has crashed for no apparent reason? The only one who has grumbled about the lack of compatibility between my cable company and the TV and other equipment that I own? (This has stumped three of the cable company's technicians.) The only one who has ever noticed a discrepancy between a statement in a user's manual and a real-life experience with a piece of equipment?

Another post suggests that my wish to enjoy brainless photography (an unwarranted conclusion) indicates that I'd be happier with a point and shoot. It might be more accurate to say that at times (e.g., when antsy grandkids won't sit around waiting for me to make multiple histogram checks and compensation adjustments) I'd like to put my 10D shooting mode on "P" and shoot flash with at least point and shoot results. The manual for both the camera and the flash implies that I should be able to do that. But on these occasions, I usually reach for my S400.

Third, someone has suggested that they would take my 550EX off my hands. Thanks for the offer. I'll consider giving it to you if you'd be willing to pay me $350 for shipping and handling. But enough of the cynical and sardonic.

On a more serious note: I've read enough posts in this and three other forums to be assured that I'm not alone in my 550EX frustrations. And goodness knows, I've tried every conceivable combination of custom functions, metering patterns and exposure modes, diffusers, and compensations that have been suggested. I've read the links to Canon and others.

Through it all, I've learned one thing for sure. The "fix-it" solutions for proper use of the 550EX run a very broad spectrum indeed. For this reason, I can't conclude that all who complain should either use a point and shoot or give away our 550s. Too many of us are willing to take the time to master an instrument that, for all of its fine points, has not met with the acceptance one might have reasonably expected. Else why would we be reading and writing to forums such as these? I could ebay the darn 550 in a flash, but I'm determined to improve my results with it.

Trust me, my reasons for writing here have nothing to do with vanity. I appreciated Tom's "experiment," which made more sense than much in the 550 manual. But, in the long run, it's of minimal help when I have one chance to take a shot and little time to input more than the most basic adjustments for existing conditions. On this score, I'd love to hear from photojournalists who have had experience with the 550.

As for the speculation that I'm probably dissatisfied with the 10D's evaluative metering, let's save that for another thread. Let's do the same for the focus point discussion that always seeems to come down to "use the center point only with custom function 4.1." My oh my, the fun we could all have kicking those cans around again.

Floyd Lawrence

[Tom W]

The response to my deliberately grouchy post was predictable. First, I'm probably some anti-technology Luddite who dislikes getting stumped by a piece of equipment.

....

Through it all, I've learned one thing for sure. The "fix-it" solutions for proper use of the 550EX run a very broad spectrum indeed. For this reason, I can't conclude that all who complain should either use a point and shoot or give away our 550s. Too many of us are willing to take the time to master an instrument that, for all of its fine points, has not met with the acceptance one might have reasonably expected. Else why would we be reading and writing to forums such as these? I could ebay the darn 550 in a flash, but I'm determined to improve my results with it.

And that is why we do this - to improve our use of the tool. I don't believe that I need to tell you that you should read http://photonotes.org/articles/eos-flash/ a couple of times, as I believe that you've already done so. But I think that sometimes, in doing so, one gets too many ideas as to how to take the perfect picture. This is a very flexible flash system, and a very intelligent one, but it is also complex.

Trust me, my reasons for writing here have nothing to do with vanity. I appreciated Tom's "experiment," which made more sense than much in the 550 manual. But, in the long run, it's of minimal help when I have one chance to take a shot and little time to input more than the most basic adjustments for existing conditions. On this score, I'd love to hear from photojournalists who have had experience with the 550.

And I'm glad that you appreciate it - Its simply a matter of me familiarizing myself with the E-TTL and sharing my experiences. My experiment here simply deals with shooting extreme dark and light subjects. Get your center focus point over the face and this won't be an issue most of the time.

I believe that a couple of photojournalists loiter around here - I'd like to hear from them as well.

As for the speculation that I'm probably dissatisfied with the 10D's evaluative metering, let's save that for another thread. Let's do the same for the focus point discussion that always seeems to come down to "use the center point only with custom function 4.1." My oh my, the fun we could all have kicking those cans around again.

Floyd Lawrence

Well, I haven't touched CF 4 - and I don't particularly like using FEL as I don't want to flash people twice - that kind-of destroys the sponteneity of the situation. But I do use the center focus point most of the time and I will dial in a little compensation on occasion if my subject is exceptionally dark or light. The results of this experiment are representative of how effective that trick is.

Hopefully, we can keep this thread on the learning side of things. While I understand the dissatisfaction of some with the reduced predictability of this system, I also think that it isn't nearly as bad as people make it out to be.

[johnanthony]

Is there a way to meter prior to focusing?

What does FEC and EV stand for?

I too shoot 10D, 550EX for about a year now.

Thanks for your pictoral explanation for flash adjustment.

[rodbunn]

You left off the most important feature when shooting with flash;
FEL !!!!!!!!!!!!!!

Much less work and much more ballanced results.

Rod

[Tom W]

Is there a way to meter prior to focusing?

What does FEC and EV stand for?

I too shoot 10D, 550EX for about a year now.

Thanks for your pictoral explanation for flash adjustment.

FEL - flash exposure lock allows metering before focusing. It also produces a flash when you meter along with the actual "real" flash that you get when you take the picture. I haven't messed with it yet, since it takes a bit away from those "spur of the moment" shots. That'll be next month's experiment.

FEC is "flash exposure compensation". Its similar to exposure compensation except that instead of varying the aperture and/or shutter speed, it varies the flash light output.

EV stands for "Exposure Value". Its just a measure of "brightness" of images, based on f-stops and shutter speed. Here's a link that gives a brief explanation:

http://www.chem.helsinki.fi/~toomas/photo/ev.html

Hope this helps a bit.

[nosquare2003]

Tom, great post.

ETTL is not a good system. It is even "worst" in a D60 (similarly a D30). I've got terrible exposure by using flash for the first few months. After studying and doing trial and errors, even my wife can do a consistent exposure for using it. (My wife usually does not care about any technical issues of cameras -- though she knows something.)

However, I have to repeat -- ETTL is not good. But we can overcome it:
- by reading Tom's post; and
- by practice.