Great thread, folks.
My experience with AFMA was similar to the one of Hawks. I actually wound up adjusting only my 1.4 and 1.8 primes, as it turns out, likely because they are my most precise lenses and I did not exceed 1-2 frames per each AFMA setting.
Thanks Neuro for methodical explanation.
Great post, Neuro. That illustration is great for describing the difference between precision and accuracy.
Originally Posted by neuroanatomist
Chuck misspoke here slightly. The maximum aperture of the taking lens doesn't matter at all -- all that matters is the f-number rating of the activated AF sensor. (Aside from the obvious fact that which AF sensor will be activated does depend on the f-number of the taking lens, of course.) His statement implies that since an f/1.0 lens has much thinner depth of focus than an f/2.8 lens that it will have higher autofocus precision. In fact, the precision is exactly the same as the f/2.8 lens, because the autofocus sensor only sees a tiny sub-aperture from inside the annular ring corresponding to f/2.8. That's only the center point of course; most autofocus sensors on non-1D bodies are f/5.6, so the precision is much lower.
Thanks, Daniel. In fact, if he misspoke, it was more than slightly! I specifically asked that question of him in the same email: "Is the relevant ‘depth of field’ for those statement the DoF at the maximum aperture for the lens currently mounted? If so, how are lenses with apertures wider than the AF point sensitivities handled – e.g. is an f/4 lens accurate within its f/4 DoF or its f/5.6 DoF, and likewise, what is the case for an f/1.2 lens at either the center f/2.8 or an off-center f/5.6 AF point?"
His response was: "The fact that the AF points are functional with apertures as small as f/5.6, f/4 or f/2.8 respectively depending on the camera model and AF point under discussion does not imply that their measuring precision is limited to the depth of focus at those apertures. The AF detection system has the capability of calculating depth of focus based on the maximum aperture of the lens, whatever it happens to be."
By the way, depth of focus is different from depth of field.Depth of field deals with object space, such as the person that is 10 feet away with 5 inches DOF, while depth of focus deals with image space, such as the tiny fractions of a millimeter in front of and behind the sensor. The difference is important when it comes to autofocus for several reasons. One is because imperfect collimation of the autofocus has a much stronger effect on wide angle(e.g. 24mm f/1.4)than telephoto (e.g. 100mm f/5.6), even if the depth of field is the same. There are a few calculators around the web that will give you the depth of focus numbers, such as Bob Atkins'.
<meta content="text/html; charset=utf-8" http-equiv="content-type" />
<meta content="text/html; charset=utf-8" http-equiv="content-type" />
Very interesting. Wouldn't be the first time he was wrong, but I'm surprised he's so specific and adamant about this. I would suggest that he reads Canon's 1D3 white paper. Page 17 clearly illustrates why f-numbers faster than f/2.8 have no effect.In fact, you could literally cut off the outer parts of an f/1.0 lens, remove the inner parts (e.g. the f/5.6, f/8, etc. annular rings), and even remove the non-sub-aperture parts of the f/2.8 ring and the f/2.8 autofocus sensor would still function exactly the same. The extra glass that makes an f/2.8 lens into an f/2 or faster lens has absolutely no effect on the f/2.8 autofocus sensor.
Here is an analogy. Imagine you are looking at the world through a 3cm cardboard tube. There a 1cm hole in a wall, so when you look through the tube your view is restricted by the hole, not the tube. If the hole is expanded to 3cm, you'll have the maximum 3cm view possible. If the hole is expanded to 10 cm, the view is restricted by the tube, not the hole, so you can still only see 3cm.
The high precision f/2.8 autofocus sensor can only look at the world through an f/2.8 tube. If there is an f/5.6 hole in the wall, AF wont work. If the hole is expanded to f/2.8, then it will have an unobstructed view and work normally. If the hole is expanded to f/1.0, it makes no difference because the view is restricted by the f/2.8 tube. However, unlike a cardboard tube put up to your eye, the camera autofocus sensor can never move to look at different parts of the hole -- it is always firmly fixed in the exact same direction.
<meta content="text/html; charset=utf-8" http-equiv="content-type" />
<meta content="text/html; charset=utf-8" http-equiv="content-type" />
Daniel and Neuro
Ok, I have a meeting to go to and limited time to digest all this. But I read Neuro's post on the fly this morning.
So possibly to break this down to laymans terms that maybe easier to understand tell me if this sounds correct.
That the AF point, will vary each time we engage the AF function of the camera. The amount it varies will be within the 1/3 to 1/2 depth of focus. Depth of Focus would be releated to Depth of Field but less depth because of the proximity it is measured from the camera. Because it is not accruate to the point, it produces a group of shots similar to say a rifle and scope would. But by adjusting this "group" foward and back with the AF microadjustment we are looking for more of an average of shots that are "Spot On" rather than pin point accuracy each shot.
(This would explain why it seemed harder to tune the 24mm and 35mm than the tele's. It would also expain why as you go further from your target, it might appear less adjustment is neccessary because your Depth of Field increase andit wouldincreases the area of Auto Focus ).
But isn't the margin for error lower on a fast lens?
This is my understanding is well. With respect to Chuck, I don't believe him.
Yep - I knew this, but when I asked the question of Chuck I didn't think through to the fact that any specification related to the AF sensors would have to be based on depth of focus, not depth of field. Chuck corrected me as well.
Me, too. I asked the question because your explanation and excellent analogy seemed logical to me at the time.
Of course, this has major implications for very fast (e.g. f/1.4-1.2) lenses. With off-center AF points, the depth of f/5.6 focusing precision is going to be much wider than the depth of field at f/1.2, and even when using an f/2.8 high-precision center AF point with 1/3 depth of focus precision, that zone offocusing precision will still be wider than the actual depth of field (depending on subject distance).
@<span class="user-name"]<span>HDNitehawk, I think you've got it...
I wanted to resurrect this thread for an instance to see if I could get some feed back.
I have read how to do a mfa on the camera, many of them are long winded time consuming ways. I have done several of the methods, after giving it some thought I tried another way.
I was doing this with the telephoto on a tripod but not sure why it wouldn
I
I think I will experiment using live viewwith the 24-70 lens tonight on the 5D Mark II. The lens has only been used once by my wife since I got it back from Canonafter I sent it in for service about 8 months ago. See if an average would work, then check it taking sets of pictures and do a comparison that way.
What got me thinking on this was that I wanted to do a mfa last week with the 7D in the field, because it seemed like it was back focusing on the 500mm F4L. ( I didn't have time to check the new camera before I left on my trip). I think in reality it was focusing just fine and the probelms I thought I was having were actualy my technique and learning the new AF system. It was dead on off the tripod at home.
It would seem logical to do it this way, in the field you could just set up on any point, do a focus and then a 10x check just to make sure you were hitting your target.
Posting Permissions
Reply With Quote