Focusing: are all bodies/lens compatible?

[short-circuit]

I just experienced strange things with AF:
A friend of mine asked me for help about problems he had with his brand new EOS 20D: very poor focusing with his Canon 28/70 F2.8 L.
I first made hand calculation to know what AF precision should be for each combination of sensor/aperture/length to have the focusing error below 1/6 of the depth of field. (first sharp point is at DOF/3 before focused point and last sharp point 2*DOF/3 after so DOF/6 seems for me the maximum error we can allow).
FOR EOS 20D and 28mm at F2.8. DOF=36.7cm at 2m
FOR EOS 20D and 70mm at F2.8. DOF=5.7cm at 2m
FOR 1dmk2 and 28mm at F2.8. DOF=46.4cm at 2m
FOR 1dmk2 and 70mm at F2.8. DOF=7.2cm at 2m
First remark: due to pixel size EOS20D is more demanding for lens AND for AF. Ok, we already knew that...

I tested my lens, knowing I could allow about 1cm error at 70mm at 2m and 6cm at 28mm. I tested my mk2 with my Tamron 28/75 F2.8. +0.8cm error at 70mm and +2cm error at 28mm...nice it works as expected.
Then I tested his body with his lens: -20cm error at 2m!!:eek:
Then my body with his Canon 28/70L lens: -14cm error. (and his lens seems to work fine one analog body)
He made oher tests and his body is good with some lens and very poor with others.
For me it sounds like a compatibility problem. I read that lens are 'calibrated' and can have 'tolerance errors' but I don't understand this point as for we the body has an AF sensor and use a correlator to maximise the sharpness...for me the lens is just following the body control signals to find the best possible contrast: you could have 2mm distance between body and lens if you have the electrical contacts and inside focusing range it should work! (Ex: macro ring works with AF). Ok, there is the AF motor precision but nobody already complained about the L series right?

The Canon vendor seems to be completely lost with those problemes and body and lens are going back to Canon for the second time.
Any Idea? Any experience?
Thanks!

[ScottE]

The 1D has a larger sensor than the 20D. There for at any given focal length a lens will appear to be more wide angle on the 1D than on the 20D. As your table illustrates, a wider angle lens will always have more depth of field than a more telephoto lens.

If you shot the 1D at 70 mm you would have to shoot the 20D at about 57 mm to frame the shot the same. I wouldn't be surprised to see that the DOF on the 1D at 70 mm and the 20D at 57 mm were closer in DOF.

[tim]

Don't forget the 20D has a 1.6X crop factor, not a 1.6X magnification factor. The DOF isn't changed just because the sensor ignores the outside of the image.

[short-circuit]

You are right, but I just tried to compare, so you need to keep one thing constant: either you take the same 'vision angle' or the same length. (I define the DOF relatively to the pixel size. In analog, it was hard to define, but now in digital, we can do that!)

But for same 'vision angle' the D20 has a longer DOF due to the smaller sensor:
(lens length is squared somewhere in the DOF formula)
FOR 1dmk2 and 70mm at F2.8. DOF=7.2cm at 2m
FOR EOS 20D and 70mm at F2.8. DOF=8.7cm at 2m

My goal was not to evaluate DOF but AF precision for a given lens. Some internet pages tells that AF precision should be about 1cm at 2m, but giving an absolute value is not for me so good, I prefer to say:
AF precision should be 8cm for a 28mm F2.8 on a 1Dmk2 (DOF/6)
AF precision should be 1.2cm for a 28mm F2.8 on a 1Dmk2 (DOF/6)
AF precision should be 0.13cm for a 28mm F2.8 on a 1Dmk2 (DOF/6) (!!)

[pierrot]

But for same 'vision angle' the D20 has a longer DOF due to the smaller sensor
I don't agree: the DOF has nothing to do with the size of the sensor at all. But it is related to the actual distance between the (optical) center of the lens and the focusing plan (the film, sensor, or whatever you can imagine).

But anyway, the autofocus sharpness is not dependent of this distance. It's a feedback system. As far as the autofocusing system refers to contrast-measuring on the sensor and that the lens movement is driven by these measurements, there should be no problem: the "body" should move the lens in order to achieve the best contrast, and then stop it.
The autofocus doesn't rely on an "information" sent by the lens which could be biased by a wrong lens-sensor distance for instance, it relies on the actual contrast perceived by the sensor. Thus it's ability to (theoretically) achieve perfect focusing whatever the lens

[short-circuit]

DOF has to do with focal length and pizel size. (and pixel size with sensor size and px number)
approx: DOF=3*H*distance/(H-length)
with H=length^2/(2*px_size * Aperture)
with length in m. 2*px_size is the max allowed confusion circle...
...anyway I have the impression that we agree but that's more a language problem. To agree on DOF we must agree on the 'sharpness criteria'. I took 2*px_size because I want an AF precise enougth for the expensive 8MP of my body! :)
ok..I close the debate...
I agree on your explanation on AF.
I just finisnet test on all my lenses. Sigma 12-24 15-30 28-70F2.8 Tamron 28-75 Canon 27-300 and 50F1.8
- No big problem detected except from the Canon 50F1.8: almost any possible error between -10 and +10cm. This lens has a good optical system but a very poor mechanical one: you feel that it is a very cheap design.
- New lenses shows better AF percision (probably more mechanical precision)

[S230]

I just experienced strange things with AF:
A friend of mine asked me for help about problems he had with his brand new EOS 20D: very poor focusing with his Canon 28/70 F2.8 L.

This was the very same problem that my friend experienced and was also one of my original questions posted.
http://photography-on-the.net/forum/showthread.php?t=50304
Even my 300D is a bit out of focus and I think it's alignment with the points. I now shift slightly upwards and seems to be better but need to try more.

[short-circuit]

I found a great site:

http://www.hkdotnet.com/FrancisPhotographyChannel/AF_Test/index.htm
http://www.hkdotnet.com/FrancisPhotographyChannel/AF_Test/AF_focus_test_result.htm
http://www.hkdotnet.com/FrancisPhotographyChannel/equipment.htm

But still don't know what can be 'AF calibrated' inside a lens...