Why do lenses need AF calibration?

Started Oct 15, 2003 | Discussions thread
Ben Strasser Senior Member • Posts: 1,763
Re: Wrong - Re: Closed lopp: Proof:

Maybe our terminology or my understanding of an "open" and "closed" loop servo system is at issue here. My understanding is that an "open" loop system looks at AF sensor data and moves the lens stepper motor a set distance to what is expected to be correct focus. My understanding of a "closed" loop system is that the AF sensor data causes the lens stepper motor to move and continually adjusts for the new data, such as overshoot or change in AF sensor information. IMHO, if the AF system in the 10D were "open" loop, you would never account for changes seen by the AF sensor unless you initiated the half-shutter press again to start the process. Also, again IMHO, AI Servo would not be possible because focus lock would indicate the process was complete.

My experience with servo motors, open and closed, has been that a "closed" system uses feedback to insure the servo has reached the point it was sent to. Feedback can be via a comparator counter where the amount of "steps" the motor is told to move is compared to the number of "steps" the motor actually moved or by looking at the data for the initial command, ie. AF sensor data and resampling.

It is also possible, Canon's focusing algorithm is such that the focusing process is either timed or terminates with the focus lock indication. In AI Servo mode you never get a focus lock indication and the lens focusing is always active. Focusing appears to stop after focus lock in One Shot mode and also AI Focus mode if it has been decided the target is not moving. IMHO, a closed loop system which terminates with focus lock indication.

processor wrote:
AF first analyses the curves on half-pair sensors, finds the shift
between them, estimates the correct focal plane position, and
begins to move the lens. Close to focused position, it gathers the
data again and refines the focal plane position. Sometimes - but
rarely - additional points of data are gathered, if the previous
data was not of good quality. Therefore, AF on Canon SLRs is OPEN
loop and strongly depends on perfect calibration of all components.

It's like if it you drive from A to B and it's 100 miles, and you
see on the map that A to B is half of A to C, you know A to C is
approximately 200 miles. You don't need to make measurement EACH
mile - that would be more accurate, but also slow. But you need
your odometer and map be calibrated precisely for "open loop" to
give correct result.

The curve "lens shift vs. focal plane offset" is a second-order
polynomial and can be drawn (and the root - precise focus - can be
found) by knowing only 2 points (the focal length is the third
parameter which is reported by the lens).

What you observed in your experiment, you moved the diopter in
before AF gathered the last data point (which happens very shortly
before AF declares focus lock). When AF made the measurement, the
new data was completely out of line, so the camera repeated the AF

Open loop.

Doug Warner wrote:

Derek Hawkins wrote:

for the system to be open loop the camera would have to initially
calculate in advance the amount of pulses to send to the lens
stepper motor based on just an OOF image.....Heck, without focus
error feedback, the camera wouldn't even know whether it was

I agree, and to prove it, I set my 10D on a tripod with my 24-70L.

1. Aimed it at a page on the wall 2 feet away.
2. Manually defocus lens to infinity.
3. Press * to focus. When the focus mechanism started to rotate, I
moved a +1 diopter close-up lens in front of the camera lens.

The focus scale actually reverses and backs up to compensate for
the +1 diopter.

IF it was open loop and took a reading at the start of the focus
cycle and moved the elements a fixed distance based on that
measurement, it would be OOF with the +1 in place. Instead, it
compensates for the change, on the fly..

Now that I think of it, the same text can be done by setting the
focus at minimun distance while aiming at a distant scene. As the
focus cycle begins, quickly move a test target in front of the lens
at it's minumum focusing distance and watch what happens. It'll be
heading for infinity focus, and will instantly reverse and zero in
on the new, closer focus point.

Closed loop, closed issue.

Post (hide subjects) Posted by
Keyboard shortcuts:
FForum PPrevious NNext WNext unread UUpvote SSubscribe RReply QQuote BBookmark MMy threads
Color scheme? Blue / Yellow