Canon ETTL-II users only - Flash question

[pgriz]

I have a T1i and a 580EX-II flash. The following question is ONLY about the mechanism Canon uses to do the ETTL-II flash metering.
Scenario: When the camera and the flash are set up for ETTL-II metering, I usually have the flash head angled for bounce, and may or may not have a flash diffuser on it. I also usually shoot in M(anual) program mode because I don’t want the camera trying to expose for ambient (which it will in Av or Tv modes). Many of the resulting shots are underexposed. If I increase the flash compensation (FEC set at +2), the result is still underexposed. If I shoot the exact same scene in full manual, I get a good exposure, or even have to dial in a smaller aperture to prevent over-exposure. It is obvious that the flash has more than adequate power (as can be seen in manual mode), but using the camera’s metering usually gives under-exposure.

I’ve looked this issue up on the various documents posted by Canon and also on other on-line sources, and from what I’ve been able to figure out, the T1i uses evaluative metering over the entire scene, based on the pre-flash. There is no option to restrict the through-the-camera flash metering to the center area or even to the spot-meter area. So… if I want to use the ETTL metering (because it would be very convenient, especially in a moving and changing shooting environment), what should I be doing to get better exposures?

 

[Big Mike]

On my cameras, (20D), I can set a custom function to change the flash metering from Evaluative to Average (I think). I didn't see such a function in the manual for the T1i, so it looks like you're stuck with whatever metering type you have.

Flash metering works similar to the camera's ambient meter, in that it tries to turn everything to a mid tone. So if it's 'seeing' something that is rather bright, it doesn't know 'bright', it just sees a lot of light and gives you less exposure. Thus you must increase the exposure with FEC.

So if you are having to increase the FEC by +2, I would think that you were trying to shoot a scene that is highly reflective. If that's not the case, then maybe there is something weird going on.

Can you show us examples?

 

[Derrel]

One area that would be worth investigating is the presence or absence of the flash diffuser. I just pulled my 580 EX-II out of its case to confirm this: it has NO sensor to key it into the presence of a snap-on type of diffuser like a Sto-Fen cap...this is different from say the Nikon SB-800, that has a button-type sensor that is depressed when its snap-on diffuser cap is fitted. I am not sure if you have a snap-on type of diffuser or something else, but the problem you are experiencing is one Nikon has figured out a solution to: namely, the flash unit needs to be "told" that an extrernal diffuser is on the flash, so that the pre-flash reading can be interpreted correctly. So, are you using a snap-on, external diffuser that is altering the way the light is dispersed? And, is your flash (designer? flash maker's company?) smart enough to sense the presence of such a device?

During the development of through the lens metering over the decades, several manufacturers have run into difficulties with this bounce flash type of problem because their simple, old-school flash metering analysis was geared toward DIRECT, straight-ahead flash metering.Your Manual versus Automatic results seem to indicate that an old problem has not been solved adequately with the T1i and the 580 EX-II flash combo. Maybe you need to buy a higher-level Canon that has better flash metering? I don't really know.

So...there are two areas explaining some of the "mechanisms" that are probably resulting in your T1i's flash exposure problems.

 

[pgriz]

Thanks, Mike and Derrel. Mike, don’t have any to share as I delete on the spot anything that doesn’t look decent. I’ll have to run some simulations and post those. Derrel, thanks for looking at this. You’re pointing me to the possibility that the preflash is interpreted incorrectly in a bounce situation. So I will figure out some tests/experiments to tease out what’s going on. As for snap-on diffusers, when I do use a diffuser, it is a Fong clear inverted dome, and it seems to work better than a full, non-diffused bounce. That may be due to some of the preflash being directed forward…

I will spend some time on the weekend to set up the following test: In all tests, the camera will be set to Manual, and the shutter speed will be set to 1/100, and the initial aperture setting will be at whatever the flash-meter recommends.

Scenario 1: Test of manual vs. ETTL auto flash in direct mode.
A room in my house, with white ceiling and walls, in which I will take a well-exposed direct manual flash. Then I will let the camera do the ETTL exposure, with no exposure compensation. Then, if the two are not the same, I will dial in enough compensation so that the two ARE the same.
Scenario 2: Test of manual vs ETTL auto flash in bounce mode off ceiling.
I will bounce the flash (no diffuser) off the ceiling and shoot manual, with the aperture chosen using the flash meter. Once I have the correct exposure, I will record the settings. Next, I will switch the flash to ETTL, and again shoot. If the exposure is different, I will dial in enough Flash Exposure Compensation so that the two ARE the same.
Scenario 3: Same as with scenario 2, but this time with a Gary Fong flash diffuser.
Scenario 4: Same as scenario 2, but now the flash will be pointed backwards, bouncing off the back wall.
In each case, I will compare the fully manual exposure against the ETTL exposures (with and without compensation) and see if there is a useful pattern.
I also have a 430EX II, so maybe scenarios 5 through 8 will be the same as 1-4, but with the smaller flash.

Stay tuned.

 

[Derrel]

pgriz,
I did a little bit of digging on Google, and came up with these four links. The first two are pretty good "technical source" type links, the second are more user-based, anecdotal experience type links. One thing the anecdotal user reports seem to share: many users automatically dial in a + 2/3 EV flash compensation!

Chuck Westfall on Canon E-TTL II

http://photonotes.org/articles/eos-flash/

and also


20D underexposure using flash in green mode - Photo.net Canon EOS Forum

Canon 580EX Flash problems (underexposure) - Photo.net Canon EOS Forum

 

[pgriz]

Thanks, Derrel. The first link was new to me, the second I already had in my library. The two threads on the Photo.net Canon forums are interesting. I am going to test the flash as I outlined on the weekend, and will share whatever results and conclusions that I come up with.

 

[davebmck]

Using ETTL, your camera should meter correctly even if a diffuser or flash card is used, and/or if the flash is set to bounce. This is because the flash fires an extremely short pre-flash and the camera uses that to meter the exposure. So no matter how the flash is set up, the camera sees the light available and adjusts accordingly. Some lighting situations may confuse the camera, but for the most part it works properly.

I'm not sure what you mean by "ambient", but when using flash, you should set your exposure for the background (ambient?) and the the flash will set its power level to expose what the camera assumes to be the subject.

 

[Josh66]

I just pulled my 580 EX-II out of its case to confirm this: it has NO sensor to key it into the presence of a snap-on type of diffuser like a Sto-Fen cap...this is different from say the Nikon SB-800, that has a button-type sensor that is depressed when its snap-on diffuser cap is fitted. [...] the flash unit needs to be "told" that an extrernal diffuser is on the flash, so that the pre-flash reading can be interpreted correctly.

If you're using the little flip out diffusion grid thing (supposed to give you 14mm coverage), it does know when that is out and automatically sets the zoom to 14mm.

Like davebmck said, if you're using any kind of slip on diffuser, the flash will still 'know' because the pre-flash fires through it too. It will see whatever light is coming through it, and adjust accordingly.

I use a diffuser maybe half the time, and use TTL maybe half the time too. I use the flash on a 350D and a 1N RS (though, on the 1N RS the flash can only be used in TTL, not E-TTL) and have honestly not had problems like what you're having. From my personal experience, TTL has been very accurate... Bounce or direct, doesn't seem to make much difference.


Though... Now I can't remember if it was in E-TTL or E-TTL II on the 350D, and that camera is pretty much dead now. I could probably charge up some batteries and at least see which mode it wants to use.

Does it make any difference if you switch it from E-TTL II to E-TTL, or do you have the same problem in both?

 

[pgriz]

@ davebmck: I know the theory, but the flash performance in practice was different. Ambient means whatever light is present other than the flash. To keep the results consistent for the test below, I used a dark room, so that ambient light would not affect the results.
@ josh: yes, when you pull out the diffuser screen, the flash does assume it is going for 14mm coverage. However, in the test below, it was set by the camera to 24mm which was the actual focal length used. Regarding setting E-TTL or E-TTLII, according to my manual(s) it is ETTL II when paired with the T1i.

So below is the test and the results. The numbers are the frame numbers (which I will upload sometime later…)
Camera settings: ISO 100, Flash WB, 1/125 shutter. 24-105mm lens set at 24mm. Camera is T1i, flash is 580EX-II, mounted on camera hotshoe. Camera on tripod. Batteries of flash freshly charged.
Scenario 2: bare flash, pointed at ceiling (ceiling bounce)
6417: Flash pointed to ceiling. Flash: Manual, full power. Aperture of f/4.0 according to flashmeter.
6418: Flash set to ETTL mode. No flash compensation. Approx. 1 stop underexposure.
6419: Flash set to ETTL mode + 1 stop Flash compensation. Exposure similar to 6417.

Scenario 3: Gary Fong clear diffuser with opaque dome, flash pointed at ceiling.
6420: Flash: Manual, full power. Aperture of f/4.0 according to flashmeter. No difference in overall exposure, slight difference in shadows compared to 6417.
6421: Flash set to ETTL mode. No flash compensation. Approx. 1 stop underexposure.
6422: Flash set to ETTL mode, +1 stop flash compensation. Exposure similar to 6420.

Scenario 4: Bare Flash head aimed at wall behind camera.
6423: Flash: Manual, full power, aperture of f/4.0, according to flashmeter. Overall exposure similar to 6417 and 6420.
6424: Flash set to ETTL mode, no flash compensation. Approx. 4 stops underexposure.
6425: Flash set to ETTL mode, +2 stops flash compensation. Approx. 2 stop underexposure.
6426: Flash set to ETTL mode, +3 stops flash compensation. Approx. 1 stop underexposure.

Scenario 1: Bare Flash head aimed directly forward.
6427: Flash set to Manual, full power. f/stop recommended by flashmeter was f/11. Tried 9.5 Too bright.
6428: Flash set to Manual, full power. f/stop recommended by flashmeter was f/11. Tried 11 OK, but too much glare.
6429: Flash set to ETTL, f/11, no compensation. Same exposure as 6428.

Conclusions:
1) ETTL and manual flash when the flash was aimed forward gave the same result. Direct flash is not attractive, but ETTL and manual gave same result.
2) ETTL gave about 1 stop underexposure when the flash was bounced up, compared to the manual result.
3) Using a diffuser with the flash aimed up gave no change in exposure (compared to bare flash), but the overall look was softer.
4) Aiming the flash to the back gave a 4-stop difference between manual and ETTL reading.

This means that, at least according to my tests, Canon's ETTL II implementation does not work when the flash is bounced from the back wall.

 

[Josh66]

according to my manual(s) it is ETTL II when paired with the T1i.

Have you tried manually changing that, and if so, did it make any sort of difference?

 

[table1349]

Flash Photography with Canon EOS Cameras - Part I.

 

[pgriz]

@ Josh: Custom function 5, option 1 allow the choice of TTL for the 580 EX II flash. So I will have to experiment with that setting to see if there is a difference.

@ gryphonslair99: That's the same link that I already had, and Derrel posted as well. The document says that the bounce flash, as in Scenario 4, should work, but my actual example shows it does not. So either something about my setup and test isn't correct, or ETTL-II implementation does not work well with the setting(s) I used.

The album with the test shots is here: http://s1224.photobucket.com/albums/ee377/pgriz01/Flash%20Test/

The two shots that show failure of ETTL in bounce mode are here:
Manual flash, bounced off back wall:

Exact same setup as above, but flash switched to ETTL mode, 0 FEC.

 

[pgriz]

Sigh. I don't get it. I repeated the tests from last night, at ISO 100 and 400, with the 580EX II flash bare and with the Fong diffuser, at all angles of flash from 0 (direct) to 90 (ceiling bounce), to 180 (back wall bounce), and with the result that ETTL-II exposure was either exactly the same as the manual exposure (as measured with a flash meter), or at most one stop less. 34 exposures, and I cannot get the flash to seriously miscalculate the exposure, as it did yesterday in the 180-degree bounce flash. Yet, under non-test conditions, the backward bounce flash had serious under-exposure with the ETTL-II setting. Did Canon somehow update my flash firmware overnight while my flash was sleeping in its bag? :scratch:

I'm gonna try something else...

 

[dylanstraub]

Check in the menu of this camera to find out if your flash is firing on first curtain or second. I was fooling around with my camera today and got lousy results with it set on second curtain. I have the Light Dome on mine and I've yet to see any issues with the few bounce shots I've taken.

 

[Derrel]

Hmmm...just a wild shot...I wonder if the flash foot is absolutely 100 percent tight and fitted perfectly. A year or so ago, I read of a pro wedding shooter whose camera's flash was not 100 percent solidly seating in the shoe on "talls"...the weight of the flash was enough to keep it from making perfect contact.

Problems like this that occur intermittently can be very tricky to isolate.