A Site for Photographers by Photographers

Community > Forums > Canon EOS > EOS Digital SLR Bodies > Minimum aperture for AF to...

Featured Equipment Deals

A Brief History of Photography - Part II (Video Tutorial) Read More

A Brief History of Photography - Part II (Video Tutorial)

This video explores the second half of photography's history and development from the technological advances in the late 1800s through the beginnings of digital photography at the end of the 20th...

Latest Equipment Articles

The Ultimate Holiday Gift Guide for Photographers Read More

The Ultimate Holiday Gift Guide for Photographers

Find a little something for the photographers on your list with this ultimate guide.

Latest Learning Articles

Embracing Fall Read More

Embracing Fall

The weather may be getting colder but that doesn't stop photo.net photographers from enjoying the outdoors!

Minimum aperture for AF to work

Kasper Hettinga , Nov 12, 2008; 05:45 a.m.

I know most canon SLRs (except 1 series) have AF up to f/5.6. I was wondering (out of curiosity) what the science is behind it. On this forum, often reference is made to the amount of light, but I don't really believe that.

First, because f/5.6 to f/8 is just one stop of light whereas the AF system works of a range of 15-20 stops if I remember correctly, so this 1 stop should matter that much.

Also, I experienced difficulties with AF when using extension tubes (I calculated a reduced effective aperture of f/8), even though I had about 4-5EV of light (which should be more than enough for the canon AF system, as I also had good focus at f/2.8 with just 1EV of light) and I was trying to focus on a clear light/dark border, so contrast should be OK as well.

With searching p.net, I found a post by Bob Atkins referring to geometry of AF sensor, but I would like to hear some more details if that is the real reason :)


    1   |   2     Next    Last

James Colwell , Nov 12, 2008; 06:23 a.m.

You might correctly remember hearing or reading that "the AF system works of a range of 15-20 stops", but this is not relevant (and it's probably not correct). Autofocus is performed with the lens wide open. That's why you have AF problems with extension tubes. Also, your intuition is wrong - 1 stop is a large difference, especially for the AF system.

Kasper Hettinga , Nov 12, 2008; 07:01 a.m.

I know AF is performed with the lens open and that -1 stop is a halving of the amount of light.

Just some numbers which I wrote down yesterday:

I was outside, with a aperture/shutter combi of f/4 & 1/15 sec. This is 8EV of light. Very easy focusing. I then put a extension tube on the lens, reducing the effective aperture 2 stops, so I have 6EV of light. AF hunts and won't lock.

Going inside and removing the extension tube, I got a reading of f/4 & 1 sec. This is 4EV, so much less than outside with the tubes. However, the AF worked more or less fine in this situation, even though there were 2 stops less light than outside with tubes....

Bueh B. , Nov 12, 2008; 07:58 a.m.

Probably because the depth-of-field is much, much smaller with extension tubes, so the AF has difficulty getting a good locking. In macro mode it is usually better to use manual focusing anyway -- or prefocusing to the desired magnification and moving the camera back and worth until you get your framing.

Kasper Hettinga , Nov 12, 2008; 10:23 a.m.

Off course, I do the focusing manually for macro work (often by moving camera/lens). I was (am) just interested in more background on the AF sensor and why this limit of f/5.6 is quoted... :)

Dave T , Nov 12, 2008; 10:47 a.m.

I think it also depends on the amount of contrast in the scene you are working with. I can focus at f/8 (sometimes even f/11) on a dark, clear night on the moon, but then sometimes have problems focusing on a crescent moon with the same technique.

Bob Atkins , Nov 12, 2008; 12:04 p.m.

It's really nothing to do with the amount of light. You can get AF at f5.6 in really poor light, yet not get AF at f8 in bright sunlight. AF may be easier on brighter subejcets with higher contrast but that's a secondary effect. It is releated to geometry, but there's really no simple way to explain it. You can think of it as analagous to the aperture being equivaent to the baseline of a rangefinder camera, or you can think of it as being similar to the case of split screen focusing of manual cameras, where it's a common observation that the split screen section "blacks out" when you stop a lens down too far,

Most DSLRs user SIR (secondary image registration) TTL passive phase detection which is basically a beamsplitter scheme which uses the light from opposite sides of the aperture in a rangefinder type scheme. It's looking for the overlap of the images from both sides of the rangefinder system (both sides of the aperture) and if the angle between the paths of those images is too small (f stop too small), it's hard to tell when the images are in or out of registration (i.e. focus shift doesn't make much difference).

William W , Nov 12, 2008; 01:16 p.m.

The diagram is typical and not geometrically accurate. It is not detailed in regard to the workings of rangefinder type / shift AF system – but hopefully it is an aid to the explanation above, showing how a smaller max aperture lens, provides less differentiation, for the function of the AF system.

1. Lenses of two different max apertures – open

2. Main Mirror position during viewing

3. Main Mirror position during exposure

4. Focusing Screen

5. Pentaprism

6. Viewfinder

7. Digital Sensor

8. Sensor Filter

9. Shutter

10 AF Mirror


Typical Crossection View: DSLR with AF system

Peter Rowe , Nov 12, 2008; 01:29 p.m.

AF works like your eyes do - they view the scene from two different angles and can therefore judge distance and see in 3D. The AF system does the same by viewing from opposite edges of the lens viewing area and can judge distance by the difference in the images. Using a slower lens (high f number) the distance between the views gets narrower so the camera can no longer get the information it needs to judge distance - the error would be greater than the DOF of the shot so it would be hit and miss. It would be similar if your eyes came closer together - you would not be able to judge distance as well. So, it's really not related to the amount of light, it's the separation between the angles of view. This is why in low light you can focus at f5.6 but in bright light you can't at f8.

Robin Sibson , Nov 12, 2008; 01:33 p.m.

Bob's comments are - of course - exactly right, but there's a bit of supplementary information that may be of interest too. Back in the days of MF and the FD system, the professional bodies - the F-1 in its different versions - had optional microprism focusing screens for different max-aperture lenses. They differed in the angle at which the microprisms were cut. With the ones for fast lenses, you got a very strong out-of-focus effect with those lenses, so it was very easy to focus, but the screen blacked out with slow lenses making the microprisms useless. Conversely, the ones for slow lenses never gave much 'snap', but didn't black out even with slow lenses. Exactly the same principle applies with phase-detect AF, and many Canon bodies contain both standard-precision sensors that work down (that is, as slow as) to f/5.6 and high-precision sensors (at least at or near the centre) that work down to f/2.8. On a FF body it is possible to have HP down to f/4 and standard precision down to f/8, just at the centre. This was first provided in the EOS-3 and then EOS-1V film bodies, and is continued in the 1Ds series. I believe it is also provided in the 1D series but I don't know if it is 'tuned' for the smaller circle of confusion appropriate for the smaller format. It has never appeared on a 1.6-factor body, although maybe there the CofC requirement would be too demanding. It is a source of discontent among 5D users (well, me at least, and probably many more) that f/4HP and f/8SP AF are not provided, and it is a real disappointment that the 5DII does not overcome this. If it was possible for the EOS-3 prosumer body ten years ago, then why not now?

Nice diagram, Bill.

    1   |   2     Next    Last

Back to top

Notify me of Responses