1. Ridiculously large means: knowing from every very very small steps
in focus. If you expect a lens to focus up to 3 mm precise at 3
meters distance, and 1 mm at for instance 30 cm distance, imagine
what a ridiculous amount of "x amount of units" need to be stored.
And that is just for one focal length. Now add an about infinite
amount of zoom steps to it.
Just not going to happen.
2. So the lens also needs to know all that? What kind of computing
power do you imagine a lens to have? And when were AF lenses
introduced? 1987 for Canon EOS?
What kind of computer did you have? What did eeproms look like back
then? How would all this intelligence be achieved in lenses back then?
3. Canon Extenders do tell the camera (not the lens) they are there.
The camera reacts by refusing to use AF when the resulting f-stop
exceeds f5.6.
With other tele converters, like the cheap ones from Soligor, Kenko,
Tamron, all the extenders do is route through the electrical
contacts. The camera does NOT know the new focus length, does not
know at what distance the focussed point is, does not know the
changed f-stop. The lens does not know of the existence of the TC
either.
Put your lens in manual focus. Turn as the focus manually. Notice how
the phase detection lets you know when it sees you achieve focus. Now
tell me the AF system only looked once, then told you how much to
turn and just let you be. It just does not happen. Phase detection
does not calculate the distance, in an SLR. If tells the lens to
move, checks again, then tells the lens the percentage of last
movement to move to sort of achieve focus lock. Nothing to do with
look up tables, or advanced computing power in lenses.
Just:
Move till the phase detect gets close to focus (else it just sees
blur). To know how when that is, focus keeps getting checked. If the
steps a lens makes overshoots the close to focus that the phase
detect needs, you have a hunting lens.
Now if the phase detect gets two readings where it actually CAN
detect edges/contrast (it is close to focus already and checked
that), it then calculates from the last movement the percentage of
the last step the final step has to make.
And this final step is the source of most back and front focus cases.
And about aperture f2.8 mentioned above, the only reason f2.8 gets to
use a more precise sensor is because f2.8 sort of guarantees a
certain amount of light. It is an arbitrary choice, just like f5.6
for AF or no AF is an arbitrary choice.
Hi
So having thought about it for a bit longer:
1. You're quoting an accuracy value for the lens at 30cm, 3m etc. and saying that there are lots of other distances so that's a big look up table. Why would the lens need a big look up table of accuracy values? That's got nothing to do with being in focus. Does it even need a look up table. The camera could tells the lens that it is so many units out of phase, and the lens converts that phase difference into degrees to rotate by using some constant scaling factor. I imagine there is a look up table though.
With the EF(S) lenses I've used, focus doesn't change with zoom and you can demonstrate that one for yourself pretty easily on the MF setting and lighting up the focus indicator the whole way through the zoom range without touching teh focus dial. Therefore, if the look up table really exists then it doesn't need to consider zoom, and your look up table is suddenly looking a lot less infinite. If the look up table exists, then given that something the kit lens focuses through about 50 degrees, 200 points is more than enough to calibrate it. Not that big really is it? Not really that intelligent is it.
2. Again what exactly is it that you think the lens needs to know. All it needs is a conversion of phase difference units as measured by the camera into the degrees rotation used by the lens. Even in 1987, a lens would have been capable of multiplication if it was just a scaling factor. I don't think anyone ever claimed there was a dual core Pentium in there. if it's a look up table approach then that's a pretty stupid lens too.
3. If the zoom doesn't impact the focus, then neither do teleconverters.
The manual focus indication is interesting and it's the only point you make that suggests the possibility of a closed loop in AF mode. The use of closed loop in MF mode doesn't prove it, but suggests the possibility.
The claim that the f2.8 increased AF sensitivity is arbitrary doesn't stack up either. You get the increased sensitivity whatever the ambient light, so the physical explanation of the longer baseline sensor operating with the bigger aperture is more plausible. Otherwise why not activate the increased sensitivity when ever the light is good enough, whatever the minimum aperture? Equally an aperture greater than f5.6 may black out parts of the regular sensor, hence the reason why AF stops altogether?
I'm still not convinced one way or another by the open loop versus closed loops, but the arguments of look table complexity and supercomputing lenses just don't cut it with me. Sorry.
If there isn't a look up table, then what do all these "re-calibrated" lenses really undergo when they get sent to the service centre?
