I thought it was time to risk revealing my utter and complete ignorance again, now - in reference to the recent speculation about, shall we say, in-lens step-motor support on the new Pentax(es).
The thing is, I've always been wondering if the placing of the motor is really such a big deal. It seems like everybody take it for granted is that the reason why Pentax have slower auto-focus than e.g. Canon (I haven't even verified that this is the case, but let's assume it's true), is that the motor controlling it is to weak, and that it is hard to find one that's powerful enough because the movement is slowed down by all the clockwork between the motor and the moving parts. I must admit that before I noticed people shouting here about the lack of "USM", I never saw these mechanical components as much of an issue, at least not for anything but very large and heavy lenses, and always assumed the main limiting factors on focusing speed were the quality of the AF algorithm and the processing power of the camera electronics. According to Bojidar, the Pentax AF system will based on data from a dedicated CCD, calculate the direction and number of rounds to turn the AF screw to reach the point of focus, and then use a simple electric motor and feedback from an encoder to move the correct amount. A 2nd pass for fine-tuning is also mentioned (possibly by someone else.) Now, it seems like this calculation of rounds is a not entirely trivial and requires a certain amount of raw processing power. Not much by today's standard, I would assume, but think back to the time that AF was first introduced. Then there is the encoder/counter mechanism. There is no direct way to control how far a normal electric motor moves - all you can tell it is what speed to run at, and not very accurately. So you have to measure the movement somehow. This is apparently done via a standard shaft-encoder setup, which essentially sends signals that are encodings of angles of rotation. The AF chip will "count" these signals so that it knows how many turns of movement there has been. Now, in order to stop at the right moment, you have to have a unit that keeps checking the count pretty often - which means some more processing power requirements. And not only that - as Bojidar also points out, you can't just stop the motor instantaneously as the correct amount is reached, but need to slow down gradually - so the routine doesn't merely have to check the count, but also needs to calculate what the speed is and what it wants it to be, and work out what the input voltage the motor must get in order to reach the desired speed. Depending on how close control is needed, these may be fairly complex calculations... And (this is the real point), the faster you move, the more often you have to check the count and do the maths, and the quicker you need the actual calculations to complete, as everything has to be done before you're supposed to check and update again. Or conversely, if you can't do the computations fast enough, you have to slow down the movement instead. Now, that's a longer explanation than the one I intended to write, but the point is that the AF chip has work to do while the motor is moving. Are you saying that this is now negligible? Another side of the story is that the USM motors are of stepper-type, which essentially means that all this processing is not needed; a stepper can be told more directly how far to move. It probably needs to be clocked at a much higher rate than the one at which you do the checks and updates mentioned above, though. In any case, I've also wondered how a setup with a stepper in the body would work out. Any comments? (OK that's the stupidest question so far.) - Toralf -- PDML Pentax-Discuss Mail List [email protected] http://pdml.net/mailman/listinfo/pdml_pdml.net
