AF, the 7d and 5dII

[Frank Kolwicz]

I just completed a series of AF tests on the 7d (mostly), and the 5dII for comparison, after a very disappointing outing yesterday when none of the 200 or so frames I shot were critically sharp. The camera/lens were mounted on a decent Gitzo tripod, maybe not the biggest they make, and a Custom Brackets gimbal head and I used a remote release for all of the shots, except the hands-on with IS.

I took a lot of care to wait for the system to stop vibrating and I finally went outdoors and put the both tripods (target and camera) on the dirt when the weather got better in order to get off the bouncy house floor, which did have some effect, and to get higher shutter speeds (up to 1/2500 at ISO 1600) than my indoor light provided.

The important tests for me were at f/8, because that is visibly sharper than at f/4 and it's my standard setting. At f/4 it was hard to judge the narrow where best focus was, as it was all pretty fuzzy at 100% view. At f/8 the band of sharpness was easier to discern, even though it was larger.

After 5 hours of repeated tests with varying lens and camera settings and viewing the RAW files at 100% on a 24" monitor (600/4IS and 100-300L lenses) here are my conclusions for both cameras:

Best focus reliability: Live View at 10x and close focus eyeglasses. This was hard to do because at 960mm effective focal length with the 7d, the slightest touch on the system induces wild movements and keeping the high-contrast, well-lighted target image in sight to judge fine details for each of the repeated shots was tiring. There was a noticeable improvement in both absolute sharpness and repeatability compared to the other two ways I tested, but focus still varied visibly over 3 frames on the angled test chart, just not as much as the other methods.

Next best: AF and MLU. I waited about 10 seconds for the system to stabilize after allowing the lens to AF (changing focus between frames so AF would have to find the target) and trip the MLU. There were much wider variations in focus this way with 3 test frames than LiveView10x and overall the image was slightly softer even at 1/2500 sec.

Least sharp of the three methods: AF, IS and hands-on, the same as I would use in the field for birds (still at 1/2500). Neither of the latter two are all that easy to use except when the subject is stationary, which most small birds are not, so it's my default procedure and has been disappointing, to say the least.

I understand that Canon's 1 series cameras and maybe the 5dIII have better AF systems, ones that hold AF tolerances to +/- 1/3 stop versus 1 full stop for the 5dII and 7d, but the difference in absolute sharpness, not just the variability, is disappointing.

Please don't reply to tell me how good your system works unless you have a similar and equally demanding setup and subject - short lenses and large subjects just don't cut it when it comes to stressing a system!

[Bob Atkins]

The stricter the testing and the better the equipment, the more likely you are to find problems. AF isn't perfect. If you'd done these tests 20 years ago on a digital camera, or 30 years ago on a film camera, you might never have noticed the differences because the best results wouldn't have been as good as they are today.

If you'd done these tests on an optical bench with a 100MP sensor, you'd have found even more variability since images that looked equally sharp to you in your tests would start to look different.

At some point you have to stop worrying over absolute sharpness and concentrate on the photography. Nobody is ever going to look at your images with the same degree of analytic precision that you are doing. All photographers have the same issues. You may not see it in other's work because you don't have their original images to examine at the pixel level on a large monitor.

No system will ever be perfect. Even with a perfect sensor and a perfect lens held absolutely still, you're shooting through air, and with telephoto lenses you are shooting through a lot of air. Air isn't a static or uniform optical medium. That's why stars "twinkle". Air at ground level is even more unstable than the upper atmosphere and so shooting long distances along the ground will degrade the image. It's also why a small telescope in space will outperform a larger telescope on earth and why the best telescope sites are on the top of mountains.

Manual focus with live view will always be best because your brain and eyes are better than the camera's focus algorithms. However the camera will be faster than you and will be able to focus under conditions that you can't (handholding the lens for example, or on a moving subject). Both ways to focus have their place.

If you expect perfection you are doomed to be forever testing lenses and you will always be disappointed in the results.

[Frank Kolwicz]

Your remarks are important as a defense against the pixel-peepers, but as I said in my message, I was only forced to do the testing because of the failures I saw in my field images - I wouldn't waste most of the good daylight hours testing for the fun of it, even on a gray day. And, if you remember, this has been an on-going problem for me, one that still has no resolution. I keep double-checking my technique, equipment and shooting conditions to try to figure out why my image quality isn't what it once was.

I had been getting great images, then fewer and fewer were that good until none of them were as good as the best. The bad ones may have been +/- printable, but they deprived me of the flexibility of varying cropping and using images of birds that were rather small in the frame as I had been able to do when things were better.

I also hate to spend the money and lose the weeks or month in the field to send the lens to Canon only to have them do a routine clean and adjust and tell me nothing is wrong.

[Bob Atkins]

I used to do a lot of testing of my lenses. Like you I'd test and test my telephotos but eventually I stopped. I cane to realize that if I looked hard enough I could find fault with anything!

Why you should be seeing a decrease in image quality over time is a mystery unless something has changed (your technique, your subjects, your lenses). It's pretty unlikely that the quality of your lenses is degrading over time. It's possible that they might go out of AF calibration, but using Live View and manual focusing should take care of that, or at least show that there is some issue there. There are almost no camera issues that will result in reduced sharpness if you are using manual focus.

It's hard to know what to suggest. I suppose you could rent a new 600/4L IS II lens and a 5D MkIII body and see if they give you better results than you are getting with the equipment you own. At least that would exclude changes in your subject and technique. Of course this won't be a cheap test, and the whole point of owning a lens is that you don't have to pay to rent it!

With a lens like the 600/4L IS I'm surprised that you saw maximum sharpness at f8. I'd have expected sharpness to peak by f5.6 if not f4.5. That lens should not need to be stopped down to f8 to be sharp if accurately focused. If it does need to be stopped down then either focus is off (and the smaller aperture and larger DOF compensate) or there's something wrong with the lens. At f8 I'd expect to see a hint of diffraction softening with a really good lens.

[Johnny]

Frank, you could post some photos here if you like. Like Bob said, it is hard to understand what exactly is wrong. If your setup used to be better (sharper images) and you now have images that are not as sharp as they used to be it could be a calibration issue and most likely it is the camera unless something really bad happened to your lens. If you post some crops or a link to some of your raw files we could look at it and maybe we could help? If you can not get critically sharp images after the rigorous setup and testing you mentioned it is hard to know what else you could do but rent a new rig and see if that solves the problem.

[Frank Kolwicz]

I think my troubles are related to using the tripod versus my window mount. It's not universally true of my bad images, I get some either way, but with the old Gitzo 3-piece legs and Custom Brackets gimbal head I think I'm not steady enough on my feet and have enough body tremors to make using the longer effective focal lengths, 930 and 1344mm, too sensitive at shutter speeds I typically have to use under our almost perpetually cloudy winter (spring and fall!) skies.

With the window mount in the truck I'm in my steadiest position (short of full horizontal in bed), the mount is very short and my left arm is braced on the window, right arm braced against my body or the steering wheel and results are much more reliable, down to 1/100th sec. or less at times.

It may also be that I have good days and bad ones and don't really feel all that much different either way.

So thanks for your help and I'm going to try like hell to avoid any more postings on my focus problems - it's getting boring.

[Bob Atkins]

In the "old" days, photographers sometimes used two tripods when working with really long lenses. One under the lens and the other under the camera. Obviously this is only practical with static subjects. Not the rig for birds in flight!

There are/were also stabilizer arms available. The tripod goes under the lens as usual, but an adjustable length arm goes from the tripod socket on the camera to a clamp on one of the tripod legs. It's a bit more flexible than two tripods, but still a pain for moving subjects.

A beanbag on the roof of a car can be a very stable support, more stable than most tripods with a long heavy lens. Again though, not for birds in flight.

If you have to use a tripod and gimbal head, then you need the biggest, heaviest ones you can find. Of course there's a compromise to be made there. My most sturdy tripod is an old Bogen 5050, but it's too big, cumbersome and heavy to carry far. I compromise with a Gitzo 1348 CF leg set when I actually have to carry it far, but it's not as stable, despite costing about 3x what I paid for the Bogen.