After trying numerous recommendations for setting Canon's AI Servo AF, I am getting only the rare in-focus picture of small birds in flight in their native environment. Based on this I have formed an opinion wrt the performance limits of Canon's "AI Servo AF". However, as this is a Q&A forum, and not a discussion forum, I'll pose my question as follows:


Does anyone have an objective test (or series of tests) that clearly show the limits of Canon's AI Servo AF to track and maintain focus on subjects in motion?

I do not have a great deal of experience in wildlife photography. Perhaps the "limits" are exclusively mine. When one searches online for information that documents "limitations" of Canon's AI Servo AF, there is virtually nothing critical of its performance. Yet, it's fairly simple to set up a test that shows the system consistently fails to track a moving subject. I've posted a short video on a makeshift "tracking test" I set up since originally posting this question.

A few points:

  1. My home-made "tracking test" is not intended to be the final word on the subject; it's simply a point of departure. And FWIW, I found if I zoomed my lens in on the "moving subject" in this video, the tracking improved. Could image size in the frame be a limiting variable?

  2. I chose a black subject because the wall was white, and I wanted to present the AF system with a high contrast subject against the background. I assumed that contrast was a limiting factor for AF performance, and so my intent here was to start with an "easy" subject to track.

  3. I'm aware that there are a lot of very sharp in-focus pictures of birds in flight posted on the Internet. However, it seems that most of these pictures are large birds against a contrasting background (e.g. blue sky).

  4. Based on some comments and answers received so far, my question seems to have caused some to feel that the objective of my question is defamation, or even a rant. This is not the case. I am asking this question here because I want to learn if my camera is capable of capturing the images I want. If not, I am prepared to try other approaches.

  • 1
    What kind of birds? Can you post examples? Could it be move blur? Are we talking about gliding seagulls or darting sparrows? And what lens are you using? – xenoid Dec 1 '19 at 21:14
  • @xenoid: Tits, finches & sparrows mostly; teal are another example, but easier to capture in flight. Wee birds that live in the brush, move very quickly, over short distances (5-10 meters is a long flight) and are drawn out mostly to feed. And other birds that won't show you the courtesy of flying high enough to provide a contrast-y sky background, slow enough to track, or even in flight for very long. I've tried 85mm f/1.4, 70-200mm f2.8, 500mm f/4 (all Canon) mounted an a 1DX Mk ii. Does that answer your question? – Seamus Dec 1 '19 at 21:26
  • @Seamus I'm not aware of any reviewers/testers that have an objective AF tracking test that can be qualitatively reduced to a xxx.xx (number) per yyyyy.yy (number). There are way too many variables regarding AF tracking to make such a thing practical. I have seen plenty of outstanding images of such small birds as you wish to photograph made with Canon cameras, particularly the 1D X Mark II and its predecessor. – Michael C Dec 1 '19 at 23:43
  • @MichaelC: Not looking for quantitative necessarily, but it should be empirical. I've seen a few outstanding pictures of small birds in flight also... but they didn't use Canon's AI Servo AF system; they pre-focused & used a "focus trap" device to trigger the shutter. If you've seen such images actually made w/ Canon AF, I'd like to see them, too - a link? – Seamus Dec 2 '19 at 1:39
  • 2
    I'm voting to close this question as off-topic because it isn't about how to achieve a specific photographic goal (taking photos of small, erratically moving, difficult subjects), it's about seeking validation that the equipment is to blame for not realizing that goal. – Michael C Dec 9 '19 at 3:53

I agree small birds in flight are extremely challenging, but I don't think the main challenge is camera's autofocus system. The main challenge is that it's hard to spot small birds (I once went to a natural park to photograph birds -- I heard a huge amount of birds all around me, but rarely saw them and on the rare occasion I saw one, it flied 10 meters and then hid somewhere I couldn't see it anymore), they fly away so quickly and so short distance that you can't get the camera ready and aimed. Furthermore, they're small so nailing focus is hard. Preferably you would like to use as many AF points as possible if the background is further away and you're not shooting birds among tree branches.

Here are some photographs of flying birds, taken with Canon's second cheapest DSLR camera (9-AF-point EOS 2000D) and one of the cheapest telephotos (55-250mm STM):

Flying seagull 1

Flying seagull 2

To me, the focus looks acceptable although this is not the sharpest lens or the highest-megapixel DSLR possible. Note shutter speed was 1/1000 s, so there could be some motion blur in these pictures.

If the Canon's cheapest autofocus system (9-AF-point system) can do this, what could you do with the better autofocus systems with more points?

Have you read these?

Does anyone have an objective test (or series of tests) that clearly show the limits (or the failings) of Canon's AI Servo AF to track and maintain focus on subjects in motion?

You could read the Tamron 100-400 lens review by Dustin Abbott. It uses Canon cameras with the Tamron 100-400 lens: https://dustinabbott.net/2017/12/tamron-100-400mm-f-4-5-6-3-vc-usd-review/

| improve this answer | |
  • The articles by R. Cicala are helpful, +1. They're on-point, though their "focus" (no pun intended) is on what happens after the subject has been acquired and "tracked". The acquisition & tracking is where Canon's AI Servo AF is struggling (ref) with small birds in their environment. When large "zones" are selected, the system simply won't recognize (acquire) the small subject bird. I'll post another video showing this behavior. In the meantime, still looking for something like Sicala's articles on acquisition and tracking :) – Seamus Dec 9 '19 at 16:43


I don't have a complete answer to my question yet, but I've turned up some things that are relevant; i.e. information relative to the limits of performance on Canon's AI Servo AF system. I'll post some of this information as I find it, rather than waiting. This seems a reasonable approach as it now appears likely there isn't one answer, but rather various situations and trials/tests that, in the aggregate, illustrate the limits of performance.

Some Findings:

  • This answer provided a link to a series of useful articles written by Roger Sicala, and published in the lensrentals.com blog. As of 20191209, Roger has posted 4 reports on this subject, and may be working on more? He clearly has some healthy skepticism of Canon's marketing claims, but he covers his subject objectively, and backs it up with real data.

    As interesting and relevant as his findings are, the results were published in 2012, and are a bit dated now. Also, his research dealt mostly with the electro-mechanical aspects of focusing - the closed loop servo system that manipulates the optical elements in the lens in an effort to drive the "error" in the phase detection sensors to zero. Of course this is a critical function of the AI Servo AF system, but it is not the entire system. There is another distinct and separate subsystem that handles Subject Acquisition and Tracking. This is the subsystem that my question concerned, and yes - my question was not clear on this distinction as I didn't understand it at the time.

  • I followed copious, well-polished advice from experts on photographing birds in flight, but failed to get good results for small birds in flight. After a couple of weeks, I had tried most of the suggestions that I had found. It became apparent that what I saw in my viewfinder was different than what the pundits saw and photographed. For example: Why did my AF points appear to hop randomly around the viewfinder when I tried to focus on a small bird at rest? Honestly, it seemed at times that the AI Servo AF system was trying to focus on everything except my subject!

    I decided that I would do a simple test indoors using a "moving subject" that I could control, and eliminate some variables (also is warmer indoors & my coffee machine is closer). Since AI Servo AF is not available in Live View mode, I needed another way to record the scene in my viewfinder. My method was to remove the eye cup from the viewfinder, and record the scene in the viewfinder with my phone's camera. This yielded a small, low quality video which I've posted on YouTube, but I hope it gets the point across. The central question raised by this video is, "Why does the tracking system leave the moving subject after a few seconds, and begin "tracking" the frame of my Velux window?! I'll have more results to post in 2-3 weeks after I return from travel.

  • Since performing my makeshift test, I've seen a similar test performed by DP Review on the Canon 5DS. The author opines that Canon's iTR (Intelligent Tracking and Recognition) is the subsystem responsible for tracking the subject in the viewfinder initially selected by the photographer. This opinion would seem to be correct given Canon's description of iTR in their AF Setting Guidebook (page 72):

    This is the algorithm used to determine AF point auto switching during [Auto selection AF], [Large Zone AF], and [Zone AF]. By detecting people's faces and subject's colors, extremely accurate subject detection and tracking is possible. EOS iTR AF does not operate outside of [Auto selection AF], [Large Zone AF], and [Zone AF].

    DP's assessment of iTR is that "performance is not up to par". The video would seem to bear this out. I am not sure at all what effect "close distances" has on the results, but: Canon appears to have engaged in a bit of hyperbole when they claim, "extremely accurate subject detection and tracking is possible".

    DP offers a theory for the iTR's limited performance:

    [iTR] struggles to offer pinpoint accuracy unless the subject is well isolated from other scene elements. Indeed, iTR subject tracking shows improved performance for subjects well isolated in depth, such as birds-in-flight, or a soccer player far out on a field, where there is a large disparity in subject distance between the subject and the background...

    Not to jump to conclusions, but this theory does offer rationale on why birds in flight are easy, and small birds in flight are difficult to impossible when the small birds are in their natural environment: The twigs, branches and leaves effectively form a cluttered background that the current iTR technology simply cannot deal with. At the least, this theory jibes with my experience.

    And for those that are interested, DP Review also conducted this test using another camera (Nikon D810). I found it interesting comparing these two videos on the number of views, and the "approval ratios" (i.e. thumbs up or thumbs down).


| improve this answer | |

Your Answer

By clicking “Post Your Answer”, you agree to our terms of service, privacy policy and cookie policy

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