7D - 580EX II - Bounced Wirelessly - ETTL failure?

Hi AndrewD2 - I'll have a go. I hope I understand your suggestion; Enable a slave first (ETTL, FEC=0) only then enable 'wireless' on 7D & test. Doesn't sound too promising to me, but I'll try later today. I had been testing Canon Manual Wireless before reverting to an ETTL/Autoflash network - without resetting the 7D. So I might try turning off my 7D, re-setting a slave, enabling wireless, then straight to testing ETTL/Autoflash network.
Have you tried the latest firmware?
Cheers

There's a pdf trying to describe how I do when this problem completely ruins my images here:
http://sites.google.com/site/apersson850/7d-flash-problem

I went to a dealer, to see if an exchange of my camera would fix the problem. Every 7D camera in their shop we tried, from an early demo to the latest delivery from Canon, showed exactly the same problem.
--
Anders

Hi, I have the latest firmware yes. Once I have it working its very consistent. It reverted to "over expose in bounce / not respond to FEC" mode a couple of times I think. Working fine at the moment (bounce fine and responds to FEC) just not entirely sure what fixes it or causes it to go off again. I've been adjusting the FEC with camera rather than flash this time btw.

Andrew

Hi Anders, Your info & pdf are interesting. I'll have a proper think about your last paragraph. The second capture has to be within 16secs of FEL; correct? If you FEL'd on a 13or18% grey target, how would you summarise your expectations for 1) working normally & 2) faulty.

I am bouncing off the front of a softbox screen (both flash & screen well out of frame to one side) with no spectral or highly reflective surfaces anywhere near the captured scene. With ETTL/Autoflash this gives 2 to 3+ over exposure. (Shooting through the softbox - no up angulation on the flash-head - all is perfect.) (180 degrees rotate does not seem to induce the problem - so far anyway!) I'm going to check FEL (aimed at 13% grey) next. Thanks, Donald

Hi Andrew, Tx for that. I'll report further within next 9 hrs. Donald

ScratchDisk said:

Hi Anders, Your info & pdf are interesting. I'll have a proper think about your last paragraph. The second capture has to be within 16secs of FEL; correct? If you FEL'd on a 13or18% grey target, how would you summarise your expectations for 1) working normally & 2) faulty.

Click to expand...

My expectation is that I should get just about the same result if I have the internal flash of the 7D as a master, when using FEL, as I get if I put a 580 EX II on the camera and use that as a master. I find that reasonable.

It should be influenced by me choosing a good or bad target for the FEL to measure against. And it is, when I use the 580 EX II as a master, but fires the slave at 1/1 power all the time, when using the internal flash instead as the master.

Note that you have to follow my test procedure 100% to verify your camera. If you for example set ISO 100 instead, you may think it's OK, since then the flash needs to fire about at full power.
--
Anders

apersson850 said:

My expectation is that I should get just about the same result if I have the internal flash of the 7D as a master, when using FEL, as I get if I put a 580 EX II on the camera and use that as a master. I find that reasonable.

Click to expand...

Certainly! (I slightly misconstrued your earlier post.)

apersson850 said:

It should be influenced by me choosing a good or bad target for the FEL to measure against. And it is, when I use the 580 EX II as a master, but fires the slave at 1/1 power all the time, when using the internal flash instead as the master.

Click to expand...

Yes

apersson850 said:

Note that you have to follow my test procedure 100% to verify your camera. If you for example set ISO 100 instead, you may think it's OK, since then the flash needs to fire about at full power.

Click to expand...

Yes, I follow your point about the iso setting etc.

Thank you for your testing and detailed reports. From my testing last night, I would have concurred exactly with my 7D.

Bizarrely, on repeating my tests and others today the built-in is controlling perfectly. Today this includes (slave still in bounce orientation) ETTL & Canon Manual and back to ETTL. All first without then with the built-in contributing - & back to without! The relevant cam screen set FECs all work and FEL works off Dark, Light and Kodak grey giving over, under & neutral as it should. I'm still on 1.0.9 with fresh batteries all round.

This apparent recovery has really got me puzzled. I followed Andrew's sequence, but I have a very hard time believing that that has influenced events. I noticed that moderate use of the built-in takes a considerable toll on cam battery state - so I will be persuing that further.
Anyway thanks again for your findings, Donald

Andrew, Please see my post a minute or two ago - just above.

There is an apparent cure - I followed your sequence (I think) but I can't see how that would be responsible for the apparent improvement. When it works/controls as it should the built-in is great & accurate. But if it proves to be unreliable that will be all the more frustrating. Still testing.... Donald

I can't promise it is what I suggested. Just happy it is working. Till it stops working again we won't know.
Other thoughts are to test little things like:
1) does using FEC on flash cause a problem

2) does mounting flash on camera and unmounting it with it switched on cause a problem
Just trying to think what else I did.
Andrew

...well reversing the sequence, ie controller fully enabled before slave switched on, does not 'restore' the problem.

...and FEC set on the slave doesn't either! Do I sound as if I'm complaining that everything's perfect? I'm glad I haven't deleted my problem captures from earlier. It definitely did have a problem - I'm going to investigate battery charge a little further and also something I heard under 'ETTL 2' here;
http://cpn.canon-europe.com/content/masterclass/canon_flash.do

I wasn't imaging it either.
When its not working it is at least consistently bad for bounce.
Getting good results with bounce, a small brolly, grid, everything now.
Andrew

Hi Anders, I’ve been thinking about your set-up for investigating this issue. I agree that ETTL2 ought to be able to make a good attempt at your scenario and you have obviously found that ETTL2, as implemented in most 7D built-in flash units, does not! However, could I persuade you to try a faulty/failing 7D on a different, possibly slightly easier scenario, but one that still has the built-in set to ‘not contribute’ and a single slave angled up from the ‘straight-ahead’ position?

My concern centres about overlapping direct and bounced slave flash illumination. In the CPN link below, under ETTL2, the speaker talks about how ETTL2 distinguishes between ‘the subject’ and what it assumes to be glass or a mirror. If a 7D (wrongly) treats a portion of an image as if it were say a mirror (even if it obviously isn’t) it would allow that portion to ‘blow’ completely.

Would you consider testing a set-up with your slave comfortably out of frame to one side, with its head pointed up say 45 degrees, but completely away from your subject and towards a suitable bounce surface? This surface to be placed in such a way as to successfully bounce light back over your subject - giving soft side illumination. In your diagram – if I interpret it correctly – you are likely to have an overlap of both direct as well as bounced light falling on at least part of the subject. I accept that the direct portion will be weak because it will be outside the normal coverage cone. I would be intrigued to know whether or not a failing 7D might nevertheless pass this version of a ‘slave in bounce’ scenario test.
Cheers, Donald
http://cpn.canon-europe.com/content/masterclass/canon_flash.do

Hi Guys,

The problem seems to relate to exactly how a 7D under ETTL2 determines what it is going to call a ‘highly reflective surface’. When working normally, we are told that ETTL2 decides that ‘the subject’ is the principal area within the frame where the flash illumination exceeds the ambient by a moderate amount. Any area where there is a really great difference between ambient and flash illumination is assumed to be a ‘highly reflective surface’ such as glass or a mirror and is therefore ignored from the flash calculation and consequently ‘blown’ (i.e. the ‘+3 stops’ issue under discussion).

I’m beginning to think that the particular implementation of ETTL2 in a 7D has been set to accept too much area within the frame as being potentially a ‘highly reflective surface’. Allowing the built-in to contribute to the flash captured illumination may change the ‘settings’ under which the 7D ETTL2 works and make ‘+3 stops’ issue less likely.

I’m going to test further by comparing results under Ander’s set-up with results under mine – they both essentially use the built-in set to non-contributing and they both have the slave flash-head elevated above the straight-ahead position. Ander’s set-up may easily induce a ‘highly reflective surface’ assumption on the part of the 7D implementation of ETTL2, whereas mine is much less likely to elicit that assumption. Cheers

Interesting coincidence, but my guess is that it can't be anything more. Your logic does not hold up to close scrutiny.

First, you come up with a processor speed based on the time it takes light to go an arbitrary distance, then conclude that it is probably that speed, then conclude that that speed explains the arbitrary distance. Kind of self-fulfilling logic.

Also, it is highly unlikely that the metering system computes the results in one clock cycle - it would take several instructions (possibly hundreds or even more) to do the calculation. And the light sensor would never be tied exactly to the CPU speed like that - it is probably safe to assume the light sensor (being an A-D conversion) would most likely take much longer to even come up with a value, let alone pass that to the CPU in one clock cycle.

Note that the problem I see with my camera is not related to E-TTL II, as I'm using FEL (Flash Exposure Lock) when it happens. In the beginning, it didn't work with normal E-TTL II either, but that has "healed" by itself, for some reason. Not the problem remains when using FEL only. That bypasses all evaluative metering and processing the camera normally does (with E-TTL II), so it's not any reflective surfaces or anything that fools the camera.

The other main point is that it works when using the 580 EX II as a master. Hence it doesn't matter if some direct light hits the target, when I set the slave to bounce. The camera can handle that, when the 580 EX II is used as a master, so in my opinion, using the internal master it should be able to do the same.

Not sure if the comment about processing and such was directed to me, but I did once think about what kind of timing is needed to capture the flash illumination and read it at the proper time. Time of flight for light to pass 0.3 meters is one nanosecond. Ten nanoseconds is 3 meters. Up, bounce, down and then back to the camera could be 15 nanoseconds. That's the cycle time for a 67 MHz processor.

This reasoning was not to try to proove that there's a processing issue here, just to show that although light travels very fast, it's still figures within the same magnitude we are talking about. The burn time of the pre-flash is of course significantly longer, so it may be that this is no issue. But then, why does it work when the slave is aimed directly towards the target? The difference is only a few nanoseconds.
--
Anders

Mine appears to underexpose (under some conditions). I'm on the latest firmware (came loaded, so assume it's a newer body).

I do tend to shoot with an omnibounce, but whichever way, bounced, direct, I get underexposed results. I'm using a 420 but have a 580 on order.

Strangely, the first few shots I took were fine, it was only yesterday that I noticed it.

Actually, on wide angle, in brighter ambient light, it gives quite pleasing results, but in darker conditions, zoomed, it gets quite dark.

I'm using an old Sigma 18-200, though as a general carry round (with surprisingly good results), so I don;t know if it's a difficulty the system has with reading the focal length.
--
http://www.flickr.com/photos/philipsmith/

Hi, To reduce the variables, I would test initially with minimal ambient light and a Canon lens. You are giving Canon a let-out by using an off brand lens. An omnibouce isn't a problem except that it & bouncing generally eats power, so use a highish iso and only test at short range. The 420 won't do Manual Wireless, but that seems not to be a problem. Cheers

On further testing, the above hypothesis is certainly an insufficient answer.

Essentially I seem to have an intermittent problem. My central issue is that with the built-in not contributing to the capture and particularly if the camera is attempting to set a ratio between A:B (usually but not exclusively, if the slaves are in bounce orientation), I frequently get gross 3+ stops flash overexposure. But this does not happen all the time – sometimes the network works perfectly and responds to FEC exactly as it should.

Separately (?) FEL is unpredictable too. As designed, if you FEL off a Kodak grey card, the system should deliver a neutral flash exposure. If you FEL off a white card, the system should underexpose the remainder of the capture. If you FEL off a dark tone, the system should overexpose the remainder of the capture. Sometimes my 7D does exactly that over a Canon Wireless set-up. However, sometimes the network misfires completely, but more frequently the network delivers gross 3+ stops overexposure again.

If I use my 580(11) as a master, I have yet to have a single issue. It does multiflash ETTL/Autoflash, with the different forms of FEC exactly as it should. Similarly FEL so far, has been faultless, with the 580(11) as master. (And Canon Manual wireless too.)

Given the price of this camera – and for me, the potential attraction of having a built-in controller/master – this situation is extremely frustrating. You simply cannot rely on the built-in controller to deliver the goods under ETTL/Autoflash.

Andy0x1 said:

The new Firmware 1.1.0 appears to have resolved this for myself and others.

Thanks everyone.

Click to expand...

So there was problem.

Congratulation it solved

Resolved? Unfortunately possibly not. The basic problem is intermittent, so I'll reserve judgement on 1.1.0. Unfortunately the FEL problem remains. With an average tone scene, I may achieve a good ETTL Multiflash (both bounced) exposure. If I then FEL off a Kodak grey card, I get a 4+ stop gross overexposure. We're not there yet.