Autofocus Torture Test



Autofocus is one of the most important aspects of any camera. There has been much debate over the past few years about how well autofocus performs on the various camera models. We originally developed this AF torture test as part of an ongoing article about the autofocus problems that surrounded the original introduction of the Canon 1D MKIII. Using this test we were able to discern that there was most likely a hardware issue at the root of these AF problems. And, so there was. Anyway, people seemed highly interested in this particular autofocus test, so we continued to conduct this test during various camera reviews.

How we test:

The tests are fairly simple and straightforward. A camera is placed on a tripod and a runner runs directly at the camera at a steady pace. The tests are all performed using servo focus mode at f/2.8 using the best 300mm f/2.8 lenses available from the respective camera manufacturers. The camera exposure is set to ensure that a very high shutter speed (usually 1/2000 or higher) is achieved throughout the run. Normally, at least 5 test runs are conducted and then coded for the number of in focus images in each sequence. A percentage is then taken for each test run. These 5 percentages are then averaged to get an overall result. These tests are all done using the default camera settings and default custom function settings. Tests are done using the center focus point only, all available focus points, or both.

This is a specific Auto Focus torture test, not an overall Auto-Focus score:

Please note that these results do not necessarily reflect the overall autofocus system performance of any particular camera. They do provide a lot of insight into how well the AF systems stand up to one of the toughest tasks they can possibly face.


Most of the reviews containing the detailed results of each trial run can be found here and here.

So far, we have tested the Nikon D300, Canon 1D MKIII, Canon 40D, Nikon D3, Canon 1DS MKII and Canon 7D.

Drew Strickland

14 Responses to “Autofocus Torture Test”

Read below or add a comment...

  1. ecarr says:

    Would be really interested in the performance of the Canon 5D Mk2 with your testing procedure. It gets a lot of flack for it’s AF performance. But it seems to be pretty well in the center point department. Also it has the hidden focus points for AI servo mode (with custom function on). Thanks for the tests!

  2. Jim says:

    Anyway the 7D AF tech could be put in firmware update for 5DII

  3. darrin lingle says:

    no data for the nikons in all points focus? why not? i’ve had both and there are good and bad things about both.

  4. gipukan says:

    I would like to see the 450d xsi tested as well in your list please. Heard some rumours on the 7d not being the best in raw

  5. Gearsau says:

    I notice that the NIKON D3 test was conducted over a year ago. In my opinion, the only way to compare , is to run a test with all the cameras at the one time.

    The D3 has had a firmware update to what was used on your test.

    I would like to know the camera settings for the D3.

    In my opinion, I would never expect AF to work well on a low contrast subject like your test subject with a white tee shirt.

  6. Must say that i consider this test anything but trustworthy.
    I am a professional sports photographer.
    When using only the center AF-point i get more “keepers” than using all AF-points at the same time.
    Besides that, a runner who runs in a quite steady pace is anything but “torture” for the cameras AF-system.
    A pro camera should easily “nail” the focus in a test like this.

  7. Ellen says:

    My first question is whether the camera test began with you focusing on the runner at a distance or whether the subject was out of foucs. If he was out of focus, was the distance set to be too near or too far? This could have a significant impact on how many images in a trial were in focus. Another variable is what the runner was wearing and where the center AF sensor was pointed. For instance if it was set on a black area – or perhaps a white area, both lacking contrast, the results would likely be different than if it was aimed a t a wrinkle in the shirt where there is more contrast for the AF sensor to work.

    This is an interesting test, but it just scratches the surface unless you control – or specify – more of the variables.

  8. arbee88 says:

    You never state as to whether or not that these are current production models tested at the same time. If you are basing thins on test for caneras that may have been production in the past, rather than all tested together as current ‘fresh’ models, then the test would be worthless. You know how quickly things change and there are always firmware upgrades that tweak the performance of the camera.

  9. eric epie says:

    Do you have AF test for non-full frame cameras (i.e. Canon 450D 40D, 50D, Nikon D80, D60)?

  10. Axel says:

    Low contrast, featureless white T-shirt and backlit, GOOD JOB!!!

  11. Eduardo says:

    Like Axel said: Low contrast, backlit, featureless shirt. And, light conditions varying over time as the tests were done. This is the kind of “test” that would bring ridicule down on a well-meaning forum poster.


  1. [...] little camera bested every camera on our Auto Focus Torture Test when using the center focus point only. It came in second only to the revamped Canon 1D MKIII when [...]

  2. [...] the whole article here. AKPC_IDS += "654,";Popularity: unranked [...]

  3. [...] per comparare le prestazioni di vari corpi macchina. ProPhotoHome ha provato a condurre un piccolo test comparativo, con modalità discutibili, ma cercando di mantenere simili condizioni e presupposti, mettendo [...]