p.25 #1 · ['NEW Fix' UPDATE!] - MkIII AF still broken
I did the following experiment to check the hypothesis that the “motion blur”/ Ghost in the OOF parts may be caused by the AF part still focusing while the exposure takes place.
On my 1d MkII I set PF 16. With this function, the exposure is made when the AF detects the subject is in focus. The camera was on a tripod end I used a steady object (the lens trunk) put on a table in an angle in order to have something in focus on slightly different distances.
The first picture is with AF in one shot AF for comparison and to set a base.
Next I put the lens on Manual focus and manually changed the focus distance on the lens from way back focus to way front focus. While doing so the camera detected momentarily something in focus and triggered the exposure. You can see it was too late and the focus point had shifted to the front. This was to be expected because focus by the lens was still moving towards the front.
I repeated the scenario by moving from manually front focus to back focus.
Did the same in portrait orientation.
The result is on www.marcelknaapen.nl/AFDYN
All images IS800, aperture f1.8 shutter 1/5000 sharpness +1. A little noise reduction was added to get rid of some distracting chroma noise.
The OOF blur seems quite natural and the same in all directions, nothing like the type of blur (ghosting / motion) that was shown in the RG images.
p.25 #3 · ['NEW Fix' UPDATE!] - MkIII AF still broken
Marcel Knaapen wrote:
The OOF blur seems quite natural and the same in all directions, nothing like the type of blur (ghosting / motion) that was shown in the RG images.
Indeed it does, thanks for taking the time to do this (I've wanted to do something similar, but haven't had the time).
p.25 #4 · ['NEW Fix' UPDATE!] - MkIII AF still broken
FYI, I used centre focus point + surrounding points. Thats's what I use standard (although for some shots I choose an other focus point). I didn't bother changing it for this little experiment. Center point was aimed at the middle n of Canon.
The reason for the test was not about front or backfocus but to examine what the focus blur itself looks like when the AF does not stop moving during the exposure.
p.25 #5 · ['NEW Fix' UPDATE!] - MkIII AF still broken
Kind of an interesting data point for the 1Ds MkIII AI Servo performance, quoted from another thread:
veroman wrote:
Here's a review that compares 100% crops of the 1Ds III and a 40D. http://www.juzaphoto.com/eng/articles/canon_eos_1ds_mark3_review.htm
To be expected, the 1Ds III crop is considerably larger. But is it any more crisp and detailed than the 40D crop? Not sure, but I don't think so. Perhaps when printed the IQ difference between the two would be more obvious.
— Steve
Check out the AI Servo sequence; he equivocates 17 out of 24 images in focus, and attriubtes some of the errors to user technique. But, he uses the same 'target point' in the image, despite that part of the subject moving to different parts of the AF array as the runner nears the camera. A 70% hit rate, but I'd sure like to see which AF point was used, because looking at his sequence, without that info his results could either be better or worse than he has stipulated. He's certainly got some nice OOF images tossed in there...
p.25 #6 · ['NEW Fix' UPDATE!] - MkIII AF still broken
Wow interesting test. As Jeff says without more data (focus points) it's hard to judge. A few of those he says are sharp don't look very sharp to me. There also seems to quite a variation in the sharpness of some of the sharp ones. Looks to me like the 1D3 and the 1Ds3 do share the same AF system.
p.25 #8 · ['NEW Fix' UPDATE!] - MkIII AF still broken
I thought I had seen somewhere before the introduction of the 1Ds3 (and before all the noise about 1D3 focus issues) that the 1D3 and 1Ds3 were going to share the same focus mechanism. This test would certainly seem to agree.
p.25 #9 · ['NEW Fix' UPDATE!] - MkIII AF still broken
That is absolutely what we all heard, and considering the timing, it really couldn't be anything but. The only differences one could rationally expect are firmware in regard to predictive AF algorithms, since the frame rates are different.
However, the one thing I might wonder about is the physical design of the viewfinder (since it has to be different), and how the AF module is stuffed in there, and potentially how the submirror interacts with the new cross-type sensor(s). Who knows, maybe the entire AF assembly was designed for the Ds-Series, and was hacked into the 1D with the assumption that it would just work? We may never know until somebody dissects the MkIII and MkIV versions and compares them...
p.25 #10 · ['NEW Fix' UPDATE!] - MkIII AF still broken
Jeff wrote:
...The only differences one could rationally expect are firmware in regard to predictive AF algorithms, since the frame rates are different.
...
Wouldn't predictive AF be designed to utilize the delay time between mirror-up and shutter-open, rather than time between successive frames (i.e. frame rate)? Since the D and Ds have the same shutter delay after pressing the release (I think), the rest of the timing could be the same, except for the dead time (mirror down for viewing) while the image is being processed. I think it's possible that they use the same AF predictive algorithm no matter the fps. For that matter, the fps of the 1D can be set to lower speeds, including 5 fps and slower.
p.25 #11 · ['NEW Fix' UPDATE!] - MkIII AF still broken
That's probably true, Jess, which also makes you wonder whether these AF systems are or are not 'optimized' for these fast shooting rates. Changing speeds with the MkIII certainly doesn't affect its accuracy in any way.
It's a funny thing that this camera tends to almost always err on the front-focusing side, and you can (quite) often find something within the AF array that is perfectly focused. I guess statistically you might expect that, but it's quite consistent. It really makes me wonder --despite choosing only a single AF point--if the camera is still using the non-selectable AF-assist points in some way, whether intentional or not.
p.25 #13 · ['NEW Fix' UPDATE!] - MkIII AF still broken
Jeff wrote:
It's a funny thing that this camera tends to almost always err on the front-focusing side, and you can (quite) often find something within the AF array that is perfectly focused.
Tha'ts almost as if the predictive focus algorithm is incorrect. But then what would cause OOF images in single shot mode? It would be weird to think that there are two different problems within the same system.
p.25 #15 · ['NEW Fix' UPDATE!] - MkIII AF still broken
You guys are gonna love this:
I just got off the phone with Customer Relations, and they told me that my camera is in shipping, ready to come back to me. They said:
"AF mirror mechanism adjusted, 'stopper, submirror' replaced. All adjustments and corresponding tests have been completed to Canon factory standards."
Again.
After yet another re-explanation of the current predicament, I told them to immediately stop shipment of the camera to me until I can speak with someone who actually knows something about my particular camera, and that if it was shipped and I hadn't had such contact, I would refuse to sign for it and it would come right back to them.
p.25 #17 · ['NEW Fix' UPDATE!] - MkIII AF still broken
Jeff, holy crap! this is what i said to you in the PM i sent you. you should still ask them what they do to validate the camera from a functional standpoint. that is, if you get a hold of anyone who actually knows something... don't know if anyone like that exists at canon.
p.25 #19 · ['NEW Fix' UPDATE!] - MkIII AF still broken
My main contact is out until Tuesday, and the guy today finally admitted that most were at CES, therefore fairly unavailable for anything detailed. Who knows...
A 70% hit rate, but I'd sure like to see which AF point was used, because looking at his sequence, without that info his results could either be better or worse than he has stipulated. He's certainly got some nice OOF images tossed in there...
