This picture turned out really well:

enter image description here

and here Lila is running towards the camera, and the aperture is wide open. I know the AF system uses the phase of light and diffraction to work out when it is in focus, but the time it takes to raise the mirror is surely at least 50ms the subject can move in this time.

So my question is this, does the AF system understand velocity and assume it is linear at the instant it starts to raise the mirror?

I've deliberately not mentioned my camera, because I am curious about AF as a system rather than specific to one camera.

The camera raises the mirror and sets the aperture when it takes a picture, and with the mirror raising or raised, the AF system no longer works, so it can't know it is in focus.

  • 2
    \$\begingroup\$ Great photo, BTW. \$\endgroup\$
    – JDługosz
    Jan 26, 2015 at 12:24
  • \$\begingroup\$ "uses the phase of light and diffraction to work out when it is in focus" <-- this is not correct. The word phase in "phase detect autofocus" refers to something completely different (and yes, I do think it's a stupid name for it, but that's what engineers chose to call it...) It has absolutely nothing to do with the wave nature of light. If you're curious how it works, the illustration in the Wikipedia article is very helpful \$\endgroup\$
    – Szabolcs
    Jan 26, 2015 at 16:58
  • \$\begingroup\$ Please do not edit the title, this isn't about moving subjects, it's about the camera dealing with "blind-ness" of the camera while it raises the mirror. This applies to stationary targets, moving ones, camera moving, camera rotated. \$\endgroup\$
    – Alec Teal
    Jan 26, 2015 at 17:14
  • \$\begingroup\$ @jdlugosz thanks! i.imgur.com/KGONWcP.jpg she's very sweet really (I see that picture and see "love incoming" - we'd shouted "treat" - not everyone sees this) \$\endgroup\$
    – Alec Teal
    Jan 26, 2015 at 17:20
  • \$\begingroup\$ @Alec I am sorry, I should have not appeared so argumentative. I deleted irrelevant comments. "Phase detect AF" works the same way as split-prism focusing screens, which you can google for if you are interested. \$\endgroup\$
    – Szabolcs
    Jan 26, 2015 at 18:11

1 Answer 1


Most AF systems offer a tracking mode, usually called something like "servo AF" or "continuous AF". In these modes the camera does indeed attempt to calculate the velocity of the object being tracked and account for the time to raise the mirror and open the shutter.

In the top of the line bodies the calculations are quite sophisticated and will take acceleration / deceleration, curved trajectories into account.

  • 1
    \$\begingroup\$ I read about details recently, and recall that for Canon EOS it is the AI Servo setting that does exactly this. (I think liner extrapolation is good enough for such a small time interval) \$\endgroup\$
    – JDługosz
    Jan 26, 2015 at 12:22
  • \$\begingroup\$ I would have thought linear too. Curve fitting beyond this becomes iffy. Also this explains how I was able to go so long without ever using continuous AF (I left it in auto - and it chooses) \$\endgroup\$
    – Alec Teal
    Jan 26, 2015 at 17:03
  • \$\begingroup\$ Had some thoughts while in the shower. The AF system doesn't really have access to distance, it's not linear. Curve fitting with this would be difficult - so I think it is linear. \$\endgroup\$
    – Alec Teal
    Jan 26, 2015 at 17:16
  • \$\begingroup\$ It does know the distance. Phase detection focus does. Video focus just has to fiddle until it's sharpest. A SLR will snap immediately to the correct focus with one look. \$\endgroup\$
    – JDługosz
    Jan 26, 2015 at 19:05
  • 2
    \$\begingroup\$ overview of the canon system \$\endgroup\$
    – db9dreamer
    Jan 27, 2015 at 3:19

Your Answer

By clicking “Post Your Answer”, you agree to our terms of service and acknowledge you have read our privacy policy.

Not the answer you're looking for? Browse other questions tagged or ask your own question.