Upload & Sell: On
One thing that hasn't been mentioned is we're still a month away from Sigma's AF correction widget, which this lens is designed to use. I haven't gotten one yet, but I've been able to get some background. Unlike camera MA adjustment, the Sigma will give us a lot more options.
I don't want to write a book, but some things to consider:
1) Focus groups don't move linearly. Moving from 4 to 8 feet may be 800 steps, from 8 to 12 650 steps, from 12 to 16 414 steps. You can actually see how nonlinear it is by following the path of the AF helicoids.
2) Ultrasonic focus motors estimate how far they're moving for a given current burst. There are tension adjustment rings in them that, even when properly calibrated, still have a 'range of acceptable' movement per current.
3) The mechanical resistance of the lens changes at different focal distances and is slightly different for each lens.
4) The in-camera algorithms in a Canon camera are not designed for the motor, focusing rings, and distances in a Sigma lens. Third-party lenses 'borrow' the best-fit algorithm from an existing Canon lens, the chip in the third-party lens is converting that instruction.
So it makes sense that some copies may AF accurately at certain distances but not at others. They may have slightly tighter collars in the helicoids (those only come in 2 or 3 sizes, so in some lenses they're slightly more resistant than others because of a snugger fit). Or that AF motor my be moving fewer or more steps than the current tells it to.The resistance to movement is different at different portions of the focusing curve.
You also have to consider that the motor - focusing system of a third-party lens has to be able to mimic the motors Canon uses and the motors Nikon uses (which are very different, trust me, I take them apart for a living). That has to complicate things further. For example, the Canon 35 L has some 3200 steps of focus but the Nikon a very different amount (don't have the exact number). The Sigma insides has to be able to do either one depending on which mount (and electronics) are attached to it.
Unlike MFA, the Sigma lens programming thing will let us set adjustment at 4 to 16 focus distances if I read correctly, calibrate the actual motor function to the algorithms, etc. I think it will also allow us to compensate for focus shift at different apertures. It will be a much more powerful tool than what we get with the cameras MFA.
Will it work better? I don't have a clue. But it certainly seems possible. In fact, now that I see a little bit about the tool it seems amazing that third-party lenses could work without it.
I've spent the last 2 months reading more stuff about how phase detection AF works than I ever, ever wanted to. Everything from patents to academic journals to white papers. The more I read, the more amazed I am that it ever works at all.
Edited on Jan 11, 2013 at 03:25 AM · View previous versions