Where do you get that stuff about a look up table from?
The usual: I read it somewhere. We know lenses are "chipped" and the
stored values can be changed by a calibration procedure. But I don't
insist upon a lookup table; it could also be just a set of parameters
to describe a translation formula.
Do you realize what a ridiculously huge look up table one would need
with a zoom lens?
No, how huge? Certainly you would not waste table space where the
performance was linear and simple interpolation would suffice.
Simple interpolation? We are talking 1980's electronics here. You do NOT have computing power in lenses, just simple logic.
The camera NEVER can say move x amount of
units, when the camera does NOT know what units the particular lens
uses,
That's the whole point of the lens doing the translation from camera
units: the camera does not have to know the lens units, instead the
lens knows how to translate from camera units to it's units. Think
of it as like mechanical gearing: a fixed input can result in
different outputs.
The camera DOES NOT KNOW DISTANCE. Why? Because it does NOT KNOW focal length. And to know distance it would need to know an EXACT focal length. This really is not hard to grasp.
So... if the camera does not know distance (it just does not know, do
you understand that the camera can not know distance... a very
important point).
You've never explained this: given that the phase detect prisms are
essentially a split image rangefinder, why doesn't the camera know
the distance? Calculating distance based on parallax with a fixed
baseline predates cameras.
Same story: To calculate anything, one must know exactly what optics one is dealing with. And the camera just can NOT know the exact optics characteristics, it can not know distance. The only way for the camera to know distance (or needed movement to achieve focus) is by having two readings and a lens movement in between, so the camera can tell the lens how much more (most probably in a percentage instead of "units") the lens has to move to achieve focus.
And you really will NEED to know the exact focal length to know what
amount once would move to reach focus.
Focal length is not that important: what's important is how far a
particular lens group needs to move to achieve focus. Even for the
same focal length, and internal or rear focus lens will require
different amounts of movement than a conventional front element or
all element focus movement.
Focal length is VERY important, without knowing focal length one can NOT judge the phase difference at ALL. Use your brain, and look through a tele lens and a wide angle lens. You will see the huge depth of field difference , and this is exactly what the phase detect system has to deal with too.
know, how does it tell the lens how many units to move? Only the lens
would know what an unit means.
Exactly.
So why do you keep on proposing that the camera tells the lens to move a certain amount of units after just one reading when the camera does not know distance and does not know units.
If you will want to keep contending it, then explain just that. How
can the camera see distance. It can not.
Tell that to Canon. Look at the bottom of this page:
http://www.canon.com/bctv/faq/aft.html
While it's a video camera, it says that it's "TTL-Secondary Image
Registration Phase-detection System" was developed for SLRs. Note
where it says "determines the positional relationship between the two
images to detect the amount and direction of defocusing. " Amount AND
direction!
Note three very simple things. One, this is NOT a DSLR with exchangeable lenses, the system can know more about the lens than your DSLR does. Two, note the very elaborate lens system BEHIND the prism (equivalent of the mirror in your DSLR), which also makes this a very different system than your DSLR.
Three: There is NO explanation on how "distance" is calculated.
So again, why does this example contradict anything I am saying?
I will lay it out again in clear steps how the AF will and has to work. The phase detect system ONLY can see anything when the image is relatively in focus, it needs edges/contrast to compare. So, it takes several readings in between which the camera tells the lens to move, till edges get detected. If the steps taken by the lens overshoot the relatively in focus area, the lens will hunt. If the camera gets close, it takes note of the phase difference. If tells the lens to take a small step. If the phase detect detects improvement, it tells the lens to move a calculated percentage of last step to achieve focus. If it detects a deterioration, it moves the wrong way, so it tells the lens to move the other way and takes a reading again.
And it took him no time to implement the change.
...