30D & 58EX - general theme of underexposure

I've had my 30D for 14 months and my 580EX for 12. During that time I've had what I would describe as generally disappointing results from the two, with a common theme of underexposure for many of the results. There have been three major events where I have shot a lot of photographs that have nearly all needed a boost to exposure in software....

Wedding - first in large tungsten lit registry office with cream cloloured waals and bounced flash, and then in a restaurant environment with low white ceilings and bounced flash;
Christmas - indoors with very large white painted living room
Outdoor party - at night with lots of coloured lights adorning walls and marquee.

I've tried various ISOs to balance ambient light and flash light, different exposure modes - mostly AV and P and bounce and direct flash. Although I am aware of the EC and FEC features available to me I have never noticed the problems in the field, only after downloading to my laptop, so have never been alerted to the need to make manual adjustments.

Results with the builtin pop-up flash are always much better in terms of exposure success, within its limited constraints of power.

I'm beginning to wonder whether there is a problem with the flash gun or the way it interfaces with my camera. I have about two weeks left on the warranty for the flash so I'd like to understand whether there is a problem with it or if there is a reason why I need to manually override the automatic ETTL2 functioning. Maybe there are some standard tests I can try to see whether things are working as they should and that I simply need to be more aware of the conditions and what my histogram is telling me. Any advice would be appreciated. Here are some samples of uncorrected shots which have me baffled as to why they are so dim....

Exif data is included but here are the stats....

1) Bounce flash used on a low white ceiling across the dinner table.
1/60, f/5.6, 400 ISO, Program, Evaluative, 0EC, 0FEC

2) Bounce flash used on a medium height white marquee ceiling
1/60, f/2.8, 800 ISO, Program, Evaluative, 0EC, 0FEC

HOSTED PHOTO
please log in to view hosted photos in full size.

HOSTED PHOTO
please log in to view hosted photos in full size.

Two more here, this time with direct flash (not bounced)....

HOSTED PHOTO
please log in to view hosted photos in full size.

HOSTED PHOTO
please log in to view hosted photos in full size.

And yet, without doing anything any different I can also get results like these, where quite honestly there appears to be no problem with the exposure....

HOSTED PHOTO
please log in to view hosted photos in full size.

HOSTED PHOTO
please log in to view hosted photos in full size.

Use on manual for interrnal flash shots. Set speed 60th, 30th & whatever aperture you require, f8-5.6 etc. I'll wager you'll not suffer the same underxposure.

Welcome to the dark and unpredictable world of Canon's E-TTL metering system.

You might want to start with some tripod testing to compare the 580EX with the built-in flash under controlled conditions. Use direct flash for this comparison, and do several tests. If the results are consistently different then your flash unit needs service.

For indoor shooting with flash as the main light, most people find that Average flash metering (custom function 14-1 on your 30D) does a better and more consistent job. And most users of E-TTL flash use +2/3 or +1 FEC as a "default" setting.

I know. You shouldn't have to do this. But Canon inexplicably designed their flash system this way.

Derbyshire Weddings wrote in post #3844258
Use on manual for interrnal flash shots. Set speed 60th, 30th & whatever aperture you require, f8-5.6 etc. I'll wager you'll not suffer the same underxposure.

I'm not averse to using manual exposure for the ambient light but I don't really see why a Manual setting of 1/60 and f/5.6 should give results any different from 1/60 and f/5.6 in Program mode. I know about NEVEC and the pitfalls of AV mode but I thought Program mode was immune from this quirk.

Why do you feel manual would work better? What am I missing?

Curtis N wrote in post #3844259
Welcome to the dark and unpredictable world of Canon's E-TTL metering system.

You might want to start with some tripod testing to compare the 580EX with the built-in flash under controlled conditions. Use direct flash for this comparison, and do several tests. If the results are consistently different then your flash unit needs service.

For indoor shooting with flash as the main light, most people find that Average flash metering (custom function 14-1 on your 30D) does a better and more consistent job. And most users of E-TTL flash use +2/3 or +1 FEC as a "default" setting.

I know. You shouldn't have to do this. But Canon inexplicably designed their flash system this way.

I hear what you say. I've read through the flash bible a few times but I have to say I'm really unimpressed with the prospect that average metering is better than evaluative and even less impressed that a default of +ve FEC is a requirement. The thing is, if I'd used +1 FEC then my two "good" shots would have been blown. I'm sure people have got better things to do than wait patiently while I fanny about tweaking FEC and retake shots. That's not really going to capture the moment.

I can't believe this is as good as the Canon flash system gets. I really hope my flash gun is faulty and can get sorted under warranty. I'm also happier to understand that it is Canon's crap flash system rather than my own incompetence that is getting me these poor results but that won't really help me get better pictures unless I give in and bow to Canon's stupid system setup.

I will try some comparative testing and post back if I get any meaningful outcome.

Thanks for replying

tdodd wrote in post #3844350
I'm really unimpressed with the prospect that average metering is better than evaluative...

Canon gives you that option for a reason. Evaluative E-TTL works great for outdoor fill but not so great when the background is close.

I don't think so. Even your "good" shots look a little underexposed to me. The white clothing isn't anywhere close to 255/255/255.

Set your FEC before you start. Any metering system requires adjustment for conditions, like a big white tablecloth.

Curtis N wrote in post #3844505
1. Canon gives you that option for a reason. Evaluative E-TTL works great for outdoor fill but not so great when the background is close.

2. I don't think so. Even your "good" shots look a little underexposed to me. The white clothing isn't anywhere close to 255/255/255.

3. Set your FEC before you start. Any metering system requires adjustment for conditions, like a big white tablecloth.

1. But I thought ETTL II gave a preflash and surely with segmented (evaluative) metering it can determine which areas within the frame are picking up the flash illumination and which areas are not. If anything, with the background failing to return the flash light surely the likelyhood is that the foreground gets overexposed, not underexposed. Also, I thought ETTL II was supposed to have some intelligence related to the distance information fed back from the lens on focus distance. Surely between the ambient light measurement and the distance to subject data this top of the range flash system could do a little better - no?

2. When I say "good" results, I mean about the best results I have had from this camera/flash combination.

For picture of the two girls there is a strong 255, 255, 255 specular highlight on the bottle she is holding and also on her pendant. I'm not sure where in the book it states that any white cloth should resolve to (or near to) 255, 255, 255. The blouse and the strap of her bag are at around 90%-93% luminance according to Lightroom. In my view that seems pretty reasonable overall. The faces look plenty bright enough to me. This is a scene shot at night - I don't think faces (or anything else) should look quite as bright as when illuminated by brilliant daylight/sunshine, unless you want that P&S style bleached face appearance. Of course, with "shoot to the right" histogram management I guess the shot could be a bit brighter but for the natural look it needs to be calmed down to about where it is now, I think.

Maybe the guy could stand a little more exposure but again, I prefer the more authentic look of the night scene rather than the "blasted with light" appearance that goes with hitting the 255, 255, 255 target. I've attached this photo run through auto tone in Lightroom, which did boost brightness and add +0.6 stops to the exposure but to me it looks overcooked. I think somewhere between the two would be about right. But then, once again, we appear to have another result that is underexposed (a bit). I've also attached a before/after comparison of the photo, unretouched and with auto tone applied.

3. Fair enough. But these exposure are way off the mark and really not acceptable in my opinion for "default" results. Plus, in the group shot I see nothing remarkable in the photo that should cause such poor results - no predominance of white or bright tones that might quench the flash early. There's a fair bit of dark background though, so if anything I would expect a bit of a bump in the exposure, not the opposite.

HOSTED PHOTO
please log in to view hosted photos in full size.

HOSTED PHOTO
please log in to view hosted photos in full size.

The Canon flash system is a disgrace ETTL 2 is very inconsistent to say the least.

there are many factors which effect it, these include:

1, The lens you are using and weather it transmits distance info to the camera ( the 50mm is the only in your sig that doesn't )

2, how much ambient light is around.
Generally speaking, in good light ( using the flash as just a fill ) no FEC will be needed.
In complete darkness FEC of + 1 1/3 will be needed ( in my experience )
For in between amounts of ambient light vary FEC accordingly

3, whether the flash is pointing straight ahead.
Seems odd but the distance info is only used in the flash calculations when the flash is pointing straight ahead ( assuming your lens is capable ) so exposures will be different if you bounce or don't.

4, If your using a diffuser
The diffuser will have to be adjusted for if your using the flash straight ahead but not if bounced. This is assuming the lens is capable of distance info.

These are only mine ( and my brothers ) experiences - try them for yourself. I know that others here disagree, but most agree with the Canon flash inconsistencies in the ETTL 2 mode.

The best workaround I've found is to use average flash metering ( as Curtis suggests ) when indoors and only use evaluative flash metering when outside in good light, as a fill flash.

I wouldn't send the flash back to Canon as they all do the same ( I've tried 4 580EXs on 3 20D/30D bodies )

Either find workarounds for yourself or switch to Nikon ( their flash system is far better )
Sorry I can't be more helpful but I believe this to be the reality of the Canon flash system.

I have posted an album of test shots here - http://picasaweb.googl​e.co.uk …hTest?authkey=w​kAM9Va4wJ0 and attached a summary display of all the results together, here. The summary results are crops from a much larger photograph, simulating a human in a darkened room. The album in the link above shows the original scene. The test shots were taken as follows....

Popup evaluative (P, Tv, Av, M)
Popup average (P, Tv, Av, M)
580EX evaluative (P, Tv, Av, M)
580EX average (P, Tv, Av, M)
580EX evaluative bounce (P, Tv, Av, M)
580EX average bounce (P, Tv, Av, M)

The conclusions I can draw are....
- the popup flash gives brighter results than the 580EX
- average metering give slightly brighter results than evaluative
- automatic exposure gives fairly consistent background alrhough P mode seems to dim it slightly.?. I thought Av and Tv were the modes that were supposed to suppress the background !?!
- manual exposure gives you control over ambient (background) light - no surprise!
- I'm barely any the wiser about how to combat Canon's bizarre interpretation of how to operate a flash system - the popup flash exposes better than the £300 professional flash gun. This is crazy.

I guess I'll have to wait for dusk and fire off some more test shots with a background too distant for the flash to have much/any effect. The annoying thing is that none of these test results seem anywhere near as bad as my real world efforts. That's not really helping me come up with any answers

HOSTED PHOTO
please log in to view hosted photos in full size.

This link may help in understanding how the flash works in different camera modes:

http://photonotes.org …ash/index2.html​#confusion

The flash shots do look better in average mode.

While I agree that this kind of performance is unacceptable if you can get consistency ( even if it's underexposed ) then at least you can set the FEC and get on with shooting .

Next time you're out ( and indoors ) set flash mode to average and check the histogram, dial in the required FEC and get on with the shoot.

Hopefully you'll get good consistent exposures but remember to keep an eye on the histogram as the normal preview often will misrepresent the shot.

Maybe others will have better suggestions but this is what I do and is the best I can offer.

tdodd wrote in post #3844649
1. But I thought ETTL II gave a preflash and surely with segmented (evaluative) metering it can determine which areas within the frame are picking up the flash illumination and which areas are not.

Remember, all it can do is measure light. Something close and white might give an identical reading as something a bit further away and gray. A tiny area that's extremetly bright might reflect the same amount of light as a larger area that's not so bright. Indoor scenes present very complex sets of segmented light readings that are difficult for any computer to evaluate. Your camera has only pre-programmed algorithms, not intelligence. With all due respect, providing the intelligence is your job.

P mode will always limit the shutter speed to 1/60, which will underexpose the ambient in dark environments unless you're using a fairly high ISO setting and a fast lens.

One of the things that may be influencing some of your "real world" shots is direct reflections of the flash from glass and glossy surfaces. This is particularly problematic with Evaluative E-TTL.

I understand your frustration. I think most people can learn to get reasonably good results using the advice I have already given. If you're not satisfied, you may want to try a flash with a dedicated auto mode like the Metz 58.

Thanks guys. I'll keep plugging away until I get it right. At least it seems my flash is operating to spec.

I have completed further test, this time at dusk and with a distant background and present further results here, this time with histograms to better quantify the exposure differences.

These five images are all with the 580EX at 400 ISO. The first four are with evaluative metering in exposure modes - Program, Tv, Av, Manual. The fifth is with average metering and Program exposure.

It seems very clear to me that in Program mode with evaluative metering the flash exposure is badly underexposed, especially compared to the other results. This seems to be absolutely the worst combination of metering and exposure modes. I really don't understand why Manual exposure fares so much better, especially since in this example the aperture is smaller in Manual mode (f/5.6 vs f/2.8 for Program) so the flash is having to work harder in Manual but delivers better results.

I have many more results, not depicted here, and it does seem that average metering delivers better flash exposures than evaluative - I still fail to see the logic in that one but there you go.

Another odd thing I found is that the flash exposures are better at low ISO (100) than at 400 or 1600, with the 100 ISO results showing less underexposure than the higher ISO shots. This seems a bit backwards to me - surely you ncrease ISO to yield brighter photographs, not dimmer ones. Canon has some seriously perverse logic in play here.

Anyway, I'm off to bed now to ponder on all this. Maybe by tomorrow I'll have figured out what I'm supposed to do to get reliable automated flash exposures. So far the advice to use average flash metering and/or bump up the FEC seems spot on.

HOSTED PHOTO
please log in to view hosted photos in full size.