FROM: "Jim Williams" 
SUBJECT: Re: Can AF "see" better with a fast lens?
DATE: Mon, 23 Nov 1998 04:38:32 GMT
ORGANIZATION: @Home Network
NEWSGROUPS: rec.photo.equipment.35mm

>> I'm not sure that I follow. From the these two posts, I surmise that
>> AF is reliant upon sufficient light reaching some kind of AF sensor in
>> the camera through the lens. Yet a fast lens lets through more light
>> than a slow lens. So does that mean that during daylight that it has
>> no bearing because there is more than sufficient light but in low
>> light that it does?
>
>
>As I understand it, it is totally based on contrast and not light until a
>certain point.  Nikon is 5.6 (at least in mine) and the instructions state
>this.  After that, there is not enough light to differentiate the contrast.
>


Okay, this is going to be a little complicated, but I hope it will clear up
some things. The way the AF in SLRs works is by "phase detection." The
basic principle is kind of like the split-image rangefinder focusing aid in
a manual-focus SLR. This focusing aid consists of two shallow prisms, which
angle your eye's view so it sees light rays coming from the two opposite
edges of the lens. When the lens is correctly focused, these edge rays (by
definition) must cross at the plane of the focusing screen; that means
objects seen by the left edge of the lens and those seen by the right edge
of the lens will line up with each other as seen through the split-image
prisms. If the lens is incorrectly focused, the edge rays will cross either
ahead of or behind the focusing screen; that means the rays from the left
edge and right edge will be displaced relative to each other, and lines
will appear "split" through the prisms. (If that doesnt' seem clear, try
sketching the lens, focusing screen, and edge rays on graph paper; you'll
see how the rays merge when the lens is in focus, but are separated when
it's out of focus.)

The AF system works the same way, except that instead of your eye it uses
two (or more) pairs of CCD arrays. Optics in the AF system work the same
way as the split-image prisms, directing light from the left side of the
lens to one CCD, and from the right side of the lens to the other CCD. The
patterns of light and dark in the subject cause the individual elements of
the CCD segments to put out different values, so that the total output of
each CCD could be graphed as a wiggly, square-edged waveform corresponding
to the light and dark patterns in the subject. The AF system's CPU compares
the waveforms from the two CCDs to see whether or not they're "in phase" --
analogous to your eye looking at the two halves of the split-image
rangefinder prism to see whether or not they're lined up. If the waves are
in phase, the system knows the lens is in focus. If they're out of phase,
it can determine the amount and direction of the error based on the
direction and displacement of the two waves relative to each other, and it
uses this information to drive the AF motor to focus the lens.

You can see from this description that the whole thing works kind of like
the two-window rangefinder in a traditional rangefinder camera such as a
Leica... except that instead of two windows, the two rangefinder images
come from the opposite edges of the lens. This type of rangefinder is more
accurate the farther apart the two windows are, since this longer "base
length" provides more displacement when the lens is out of focus. With the
AF system, a faster lens (wider diameter) gives the equivalent of the
longer base length, since the left and right edges are farther apart.
That's why, all other things being equal, a faster lens gives better AF
performance -- the phase differences in the waves are more dramatic and
easier for the CPU to detect.

This also explains why AF systems won't work with some teleconverters or
with lenses of slow maximum aperture. It isn't just that the light hitting
the sensors is dim -- otherwise, it should still work on a very bright day
or with very bright subjects. The problem is that if the lens' aperture is
too small, the AF sensors' view of the right and left edges is cut off.
It's the same kind of problem you see with a split-image prism when you
stop down the lens to preview DOF; usually one half of the prism or the
other will "black out," because its view of the edge of the lens is blocked
by the diaphragm blades.

Sorry this has taken so long to explain, but I hope it will make a lot of
AF quirks somewhat clearer.

If you have any question, remark, comment, want to share some philosophy or just want to express your opinion about these pages, feel free to send email to: w.j.markerink @ a1.nl