Galbraith is finally done testing Canon's AF

[ron203]

http://robgalbraith.com/bins/content_page....cid=7-9314-9545

[ericisaac]

wow, couldn't they have just said yes, no or maybe?

[macgyver]

wow, couldn't they have just said yes, no or maybe?
[a href=\"index.php?act=findpost&pid=212486\"][{POST_SNAPBACK}][/a]

The last paragraph of the last page:

Canon has put considerable effort into correcting and improving EOS-1D Mark III autofocus in the 14 months since the cameras was introduced, but they have not yet made it, or its high-resolution counterpart the EOS-1Ds Mark III, deliver reliable autofocus of subjects in motion.

[Mort54]

Regardless of the brand loyalties we all have to some extent, I think we can all sit back and admire what Rob has done here. He deserves a grateful round of applause, in my opinion. He clearly loves the files he gets from the 1DIII and 1DsIII (and deservedly so), tho he clearly is still disappointed with some aspects of the AF in these bodies. I think the 1DIII and 1DsIII owners have been well served by Rob and have better cameras as a result of his efforts.

[francois]

Regardless of the brand loyalties we all have to some extent, I think we can all sit back and admire what Rob has done here. He deserves a grateful round of applause, in my opinion. He clearly loves the files he gets from the 1DIII and 1DsIII (and deservedly so), tho he clearly is still disappointed with some aspects of the AF in these bodies. I think the 1DIII and 1DsIII owners have been well served by Rob and have better cameras as a result of his efforts.
[a href=\"index.php?act=findpost&pid=212497\"][{POST_SNAPBACK}][/a]

I agree, Rob has put an extraordinary amount of work to dissect the Mk 3 autofocus system issues. I'm curious to see Canon's take on his findings.

[Eric Myrvaagnes]

Rob's results make me glad that most of my photography is of things that don't usually do much moving while my camera is aimed at them (mountains, trees, ets.)  

[Khun_K]

http://robgalbraith.com/bins/content_page....cid=7-9314-9545
[a href=\"index.php?act=findpost&pid=212481\"][{POST_SNAPBACK}][/a]

I did not use the camera tested by them but I found my 1Ds MK3 to be perfectly OK.

[francois]

I did not use the camera tested by them but I found my 1Ds MK3 to be perfectly OK.
[a href=\"index.php?act=findpost&pid=212683\"][{POST_SNAPBACK}][/a]

Me too, but I don't shoot much fast action with a 1Ds Mk3. So far, I've been very satisfied with its autofocus performance.

[michael]

It's important to note that only photographers who are shooting subjects in rapid motion at larger apertures who are going to encounter focus issues with the new Canon One series cameras. The other 95% will likely never see it.

But, having said that, it's a big black eye for Canon.

Michael

[Slobodan Blagojevic]

It's important to note that only photographers who are shooting subjects in rapid motion at larger apertures who are going to encounter focus issues with the new Canon One series cameras. The other 95% will likely never see it.

But, having said that, it's a big black eye for Canon.

Michael
[a href=\"index.php?act=findpost&pid=212738\"][{POST_SNAPBACK}][/a]

Michael, not that I disagree with you... everything you said above is correct... just wanted to add my angle:

Just because a Ferrari can take you to the nearest supermarket to buy groceries, does not mean people buy Ferraris based on that ability. I would also argue that 95% of Ferraris will never be fully driven the way they are intended to be. So, if something would be inherently wrong with Ferraris, but  would only show up at higher speeds, 95% of owners would 'likely never see it'. However, if it would ever turn out that Ferraris could not actually deliver on their basic promise (i..e, superior performance under conditions most owners would never take it to), most people would not be buying it and saying "Oh, well... at least it can take me to the supermarket".

Back to cameras: using your numbers, it is plausible that 95% of owners of 1D buy it not because they need it for 'subjects in rapid motion at larger apertures', but because precisely that use is what is supposed to be the main selling point for that camera, its sheer ability to do it. Without it, it loses its main marketing appeal, and that is the 'big black eye for Canon', as you rightly said. Most high-end products generate their profit not from a few professionals they are designed for, but from mass sales to those impressed by the pro performance.

[Mark D Segal]

The most disturbing thing Rob Galbraith's article underscores is the existence of problems with corporate culture at Canon. I say this fully recognizing the huge technological contribution they have made to digital imaging, but in this particular instance and others, there is a demonstrable insufficiency of product testing and quality control before these expensive high-end products leave the factory, combined with delay admitting issues and in some cases less than satiasfactory performance servicing the consequences. That they had the nerve to cut their collaboration with Rob Galbraith is one example of behaviour which underscores what I am saying here. Readers may also remember the issues with the IPF 5000 printer.

One should also bear in mind that this company manufactures a range of medical diagnostic equipment (opthamology, radiological, etc.) on which patients' health may depend. One may wonder whether they do a better job of QC and error correction on that stuff.

[budjames]

I've been a Canon shooter since high school. My first SLR was the Canon FTb in 1972. I just purchased the Canon 1DsMkIII 2 months ago and I've had a 40D for about 6 months. They work fine, but I'm not shooting fast action sports.

It is still a bummer that there is no clear acknowledgement by Canon to address the issue if only to clear the air. Now that Nikon has a FF DSLR at 40% less than the 1DsMkIII, I think that they might start to gain market share at the expense of Canon.

I have too much invested in L-glass and accessories to consider a switch. I'm very happy with my images, but I still want Canon to fix the problems.

Like previous posters, I agree that Ron deserves major kudos for his testing and research efforts.

Cheers.
Bud James

[Mark D Segal]

Canon has acknowledged the focus issue. It took them a good while, but that said, they have been working on improvements to firmware and for the 1DMk3 the hardware. Rob thinks they haven't fully solved it yet, which means they have more work to do - a different story from saying they haven't acknowledged it. The additional - and serious - issue which Rob raises in his article is whether they will devote more attention to a full fix, or will they refocus their sight on the next models to the detriment of customers owning current models.

[Mark D Segal]

It's important to note that only photographers who are shooting subjects in rapid motion at larger apertures who are going to encounter focus issues with the new Canon One series cameras. The other 95% will likely never see it.

But, having said that, it's a big black eye for Canon.

Michael
[a href=\"index.php?act=findpost&pid=212738\"][{POST_SNAPBACK}][/a]

Michael, unfortunately the issue isn't confined to subjects in rapid motion according to Galbraith, viz:

"Every word about the EOS-1D Mark III and firmware v1.1.3 applies equally to the EOS-1Ds Mark III with firmware v1.0.6 installed. It shares the same tendency to imagine or overestimate forward motion and the same difficulty holding focus on a stationary subject. "

"With either the EOS-1D Mark III and firmware v1.1.3 or the EOS-1Ds Mark III with firmware v1.0.6, AI Servo autofocus won't hold still on a static subject. "

"Just enough testing was done to reveal a few things. Whereas the older firmware can introduce a change in focus distance with each One Shot activation, even if the AF point is aimed at the same subject at the same distance each time, the newer firmware seems to give more repeatable and accurate focus results under the same conditions. But if the test is done slightly differently - setting the focus to some other distance, then focusing one time on the subject using One Shot, autofocus variability seems about the same as before: there's too much."

"To give yourself a reasonable chance with One Shot, it looks like you'll want to pump the autofocus two or three times with it pointed at the same area of the subject. Alternatively, switch the camera to AI Servo, configure it to autofocus only when the rear AF-ON button is pressed, then "lock" the focus by releasing AF-ON. Not only did AI Servo produce more consistent stationary subject autofocus than One Shot (albeit in limited testing), it also means you don't have to switch AF settings when the subject starts to move. Just press AF-ON.
I Servo autofocus won't hold still on a static subject. "

"Static focus stability - which was a real source of telephoto shooting troubles with the EOS-1Ds Mark III through the early months of 2008 here - is much improved, though some twitchiness remains."

"When these cameras are set to AI Servo, the newest firmware is loaded and the subject being photographed is stationary or moving slowly, the results are much better than before: autofocus is now acceptably stable and acceptably accurate. The cameras' tendency to shift the focus is still lurking inside, it's just mostly dormant."

These quotes would seem to indicate that the focus issue is more general, but the firmware up-grades help (less than 100%) with both cameras.

Where you say the other 95% will likely never see it - maybe so. The problem here is the real difficulty of sorting out cause and effect when there is also an anti-aliasing filter in the path softening the image regardless of how well focused it is. So what we really need to do when we pixel-peep raw files for focus is adjust for background anti-aliasing. This means we are looking at second-order differences of sharpness between a correct and a slightly incorrect focus. Perhaps not easy except under well-controlled conditions designed to parse this dual causality.

Cheers,

Mark

Edited 10:10 PM EDT to unscramble two quotes.

[zlatko-b]

I agree -- they did acknowledge the focus issue -- and numerous cameras were recalled for a sub-mirror fix -- and there was a firmware update that made improvements.  This is not a company that has ignored the issue.  If they've failed to address it fully, it's not for lack of trying.

Rob G. still uses the cameras for his day-to-day work (and he explains why).  This is remarkable when you consider the demands that his sports work places on the AF system.

I've had pretty good success with the 1D3, but my subject matter is entirely different (not fast moving) and I almost always use One Shot.

[Tim Gray]

Yes, Canon acknowledged a problem.  (note past tense)  But I suspect their position is that it's been fixed.   One of RG's important points is that the 1DMKII has better performing AF than the 1DMKIII (at least in certain circumstances).  Canon has not acknowledged that, and probably never will.

At some point in the future, I hope RG will take an equally rigorous look at the 1DMKIV.

[Mark D Segal]

At some point in the future, I hope RG will take an equally rigorous look at the 1DMKIV.
[a href=\"index.php?act=findpost&pid=212856\"][{POST_SNAPBACK}][/a]

If they let him, it's best done before the camera hits the market!

[Kagetsu]

Personally I was more interested in a point mentioned on the second page, and a little further on the third page of his latest update, in reference to the low light performance of the AF. This is a problem I've also noticed, and to me just seems very poor that either it's not been addressed or is insignificant to many people who use this camera (and the 1D III).

Personally I found my old 5D to focus significantly better in low light situations.

[Mark D Segal]

Personally I was more interested in a point mentioned on the second page, and a little further on the third page of his latest update, in reference to the low light performance of the AF. This is a problem I've also noticed, and to me just seems very poor that either it's not been addressed or is insignificant to many people who use this camera (and the 1D III).

Personally I found my old 5D to focus significantly better in low light situations.
[a href=\"index.php?act=findpost&pid=212898\"][{POST_SNAPBACK}][/a]

Kagetsu,

I have to say there are aspects of Galbraith's article which leave me perplexed as thorough a job as he appears to have done, and it is this quandary which caused me to comment on Michael's contribution from yesterday - because when Michael tells us 95% of us won't see an issue with focusing, he's drawing on a wealth of experience. Yet based on what he said, unless I've misunderstood something, Galbraith seems to be seeing things that many other people don't see. To wit, I own a 1DsMk3 and I know three other people here in Toronto who own one also. All of us use this camera primarily in one shot mode and none of us are sports photographers. All of us have marveled at the quality of resolution this camera displays, but at the same time (and now just speaking for myself) these raw files do need judicious sharpening to fully reveal this quality - a factor which must relate to the A-A filter, which of course is there for good reason.

Now turning to your point about focusing in low light situations, this issue has come up before - in this Forum if I'm not mistaken. When things like this arise, the first thing I do is get worried and ask myself - O.M.G. does this mean I bought a lemon? The next thing I do is settle down and test it. So late at night when all was dark except for a low voltage ceiling light on the second floor landing of our house, I pointed the camera into a dark room and took a succession of shots with the door to that room incrementally closed further so less and less light could enter the room. That camera kept on focusing decisively until it was so dark in there I could hardly see what the camera was focusing on - at a distance of about ten feet lens to subject. So in the case of my sample of this camera, there is no low light focusing problem. But that doesn't mean there isn't a problem - you say you have one and I have no reason to doubt you. So from this limited sample of two, one may begin to infer that perhaps there isn't a GENERIC technical problem with low-light focusing, but another - and in a sense more elusive - problem of variable performance between samples of this camera, and that may boil down to a question of standards maintenance/quality control.

Cheers,

Mark

[Mark D Segal]

Yes, Canon acknowledged a problem.  (note past tense)  But I suspect their position is that it's been fixed.   One of RG's important points is that the 1DMKII has better performing AF than the 1DMKIII (at least in certain circumstances).  Canon has not acknowledged that, and probably never will.

At some point in the future, I hope RG will take an equally rigorous look at the 1DMKIV.
[a href=\"index.php?act=findpost&pid=212856\"][{POST_SNAPBACK}][/a]

Tim,

I'm coming back to your post once again, because you've indirectly raised another issue which forcibly comes to mind in this context. You will recall that the 1DMk3 hit the market well before the 1DsMk3. You may also recall that customers observed the focus problem soon afterward and also well before the release of the 1DsMk3. I also recall seeing speculation that maybe the release of the 1DsMk3 was being held back till Canon verified that none of these focus problems would affect this model. Contrary to that speculation, we now we have Galbraith saying that the focusing problems discussed in relation to the 1DMk3 also apply to the 1DsMk3. If this were correct, it is really distressing to contemplate that the 1Ds3 was released knowing the likelihood of technical flaws, and tends to reinforce one's impression that perhaps Canon has some product testing and QC issues to deal with. Given the amount of money we invest in what is supposed to be the pinnacle of professional equipment, and eventhough I'm using mine I believe quite successfully, I find all of this really discomforting, because I'd like to believe that at this price point we wouldn't need to be concerned about either embedded technical defects or sample to sample variability of important performance characteristics.

Cheers,

Mark