Erratic Auto Focus With Ultra Wide Angle Lens

[dje]

I’ve just purchased a Sigma 10-20mm lens for use with my Canon 600D (T3i). When I tried it out initially, I noticed that the AF was somewhat erratic at the 10mm end for objects more than about 1m away. By that I mean that the focus ring would not be positioned consistently in the one position when focusing on the same object on more than one occasion.

This was disconcerting initially as I thought the lens might be faulty. So I took it back to the shop with my camera to discuss it with the sales staff. They did a focus check test and said it was OK but I’m not sure they really understood the issue. I tried another lens of the same type and it was the same. Then I tried a similar (more expensive) Canon lens and it also exhibited similar behavior. I then came to the conclusion that the problem was a limitation of the AF system in the camera when used with an Ultra Wide Angle lens. So I will learn to live with it. I will be using manual focus on a tripod (for landscapes) a fair bit of the time anyway.

I would be particularly interested to hear if others have experienced this behaviour and also whether anyone has an explanation for it. A Google search revealed others have experienced this but I haven’t seen any convincing explanation as to why it occurs.

My thoughts are that with a focal length of 10mm,

The focus ring has to be moved a fair bit for the image to go significantly out of focus. This means that the sensitivity of the focus adjustment is low and as a result the loop gain of the AF phase comparison control circuit is also low (causing inaccuracy)

The detail in the focus area is very compressed due to the low focal length and hence the AF system has difficulty in detecting contrast changes.

The AF focus area (I’m using centre point focus) covers a relatively large part of the image in which different objects are at different distances and it is difficult to know which object will be chosen for focus adjustment

But I’m only speculating !!!

If anyone can shed some light I’d appreciate it.

Dave

[jcuknz]

The CiC calculator tells me that if you focus at 1.5m you have DoF from 0.68m to infinity so perhaps you should do that and give AF a rest? The lens seems to have a focus scale on it so it is possible from a photo at dpreview.
If it is the Sigma 10-20mm F3.5 EX DC HSM I imagine the Sigma 10-20mm F4-5.6 EX DC HSM would be similar

I think the same way as you with regard to large target areas for AF. AF doesn't know what we are thinking

[casper]

hi Dave,

I have just (2 days ago) purchased the same lens for my Canon 7D, so thanks for the information. I will check mine and get back to you. Your thoughts on the possible cause are more complex than my understanding, but I may be able to confirm your suspicion that the issue is generic.
regards,
-Noel

[dje]

The CiC calculator tells me that if you focus at 1.5m you have DoF from 0.68m to infinity so perhaps you should do that and give AF a rest? The lens seems to have a focus scale on it so it is possible from a photo at dpreview.
If it is the Sigma 10-20mm F3.5 EX DC HSM I imagine the Sigma 10-20mm F4-5.6 EX DC HSM would be similar

I think the same way as you with regard to large target areas for AF. AF doesn't know what we are thinking

It's the f/4-f/5.6. Yes the dof is large so focus is not super critical.

[dje]

hi Dave,

I have just (2 days ago) purchased the same lens for my Canon 7D, so thanks for the information. I will check mine and get back to you. Your thoughts on the possible cause are more complex than my understanding, but I may be able to confirm your suspicion that the issue is generic.
regards,
-Noel

Thanks Noel, I'd be most interested to hear how you get on. The 7D may have a better AF system than the 600D.

Did you get the lens on special from Ted's ?

Dave

[Shadowman]

I’ve just purchased a Sigma 10-20mm lens for use with my Canon 600D (T3i). When I tried it out initially, I noticed that the AF was somewhat erratic at the 10mm end for objects more than about 1m away. By

The focus ring has to be moved a fair bit for the image to go significantly out of focus. This means that the sensitivity of the focus adjustment is low and as a result the loop gain of the AF phase comparison control circuit is also low (causing inaccuracy)

The detail in the focus area is very compressed due to the low focal length and hence the AF system has difficulty in detecting contrast changes.

The AF focus area (I’m using centre point focus) covers a relatively large part of the image in which different objects are at different distances and it is difficult to know which object will be chosen for focus adjustment

But I’m only speculating !!!

If anyone can shed some light I’d appreciate it.

Dave

Are you using the viewfinder or Live View?
What exactly is the object you are trying to focus on? Could you post an image?

[Inkanyezi]

There is a reasonable explanation to the fact that the focus ring lands at different position.

All adjustment systems have some kind of hysteresis, and this will be more noticeable for focusing when the focal length is very short. Thus it might stop in various positions, although you aim at the same subject from the same distance. At longer focal length, the setting may also vary somewhat, but not as noticeable as when the focal length is very short.

[dje]

Are you using the viewfinder or Live View?
What exactly is the object you are trying to focus on? Could you post an image?

John I'm using the viewfinder so it's phase detection with the dedicated Af sensors.I've been focusing on a variety of subjects - trees, fence, house etc

Dave

[dje]

There is a reasonable explanation to the fact that the focus ring lands at different position.

All adjustment systems have some kind of hysteresis, and this will be more noticeable for focusing when the focal length is very short. Thus it might stop in various positions, although you aim at the same subject from the same distance. At longer focal length, the setting may also vary somewhat, but not as noticeable as when the focal length is very short.

Thanks Urban

If I remember my control system theory correctly (it was a long time ago),hysteresis is an esential part of a control loop to ensure stability. This relates to my point about control loop gain and precision of focus adjustment.

Dave

[casper]

Thanks Noel, I'd be most interested to hear how you get on. The 7D may have a better AF system than the 600D.

Did you get the lens on special from Ted's ?

Dave

hi Dave,
I see you have the f4-f5.6. I got the f3.5, so it will be interesting to see what happens. I picked mine up via ebay, Australian stock, 2 year warranty, $525, seller name dcexpert (in Brighton, Victoria). I was going to get the Canon 10-22, but best available price went from $598 to over $700 when the AUD:USD rate dropped over the last few weeks.
- Noel

[dje]

hi Dave,
I see you have the f4-f5.6. I got the f3.5, so it will be interesting to see what happens. I picked mine up via ebay, Australian stock, 2 year warranty, $525, seller name dcexpert (in Brighton, Victoria). I was going to get the Canon 10-22, but best available price went from $598 to over $700 when the AUD:USD rate dropped over the last few weeks.
- Noel

That's a good price for the f/3.5 Noel.

[DanK]

Have you tried center point only AF? With a wide angle lens, the locations found by the AF points are far apart, and they may be different in distance.

I don't know this, but it would seem logical that the large DOF at very wide angle might introduce more variability in AF results, as the contrast gradient with changing focus would be shallow. Anyone know if this is correct? In any event, with the huge DOF, it shouldn't matter.

[dje]

Have you tried center point only AF? With a wide angle lens, the locations found by the AF points are far apart, and they may be different in distance.

I don't know this, but it would seem logical that the large DOF at very wide angle might introduce more variability in AF results, as the contrast gradient with changing focus would be shallow. Anyone know if this is correct? In any event, with the huge DOF, it shouldn't matter.

Dan I only ever use centre point focus. I actually tried auto af point selection with this as well but it made no difference.

Yes the large depth of focus means it's not critical.

Dave

[casper]

Thanks Noel, I'd be most interested to hear how you get on. The 7D may have a better AF system than the 600D.

Did you get the lens on special from Ted's ?

Dave

hi Dave,
Miserable day in Sydney. Tried indoors with 7D and f3.5 Sigma 10-20, tripod, spot metering, 19 point AF. I set up a variety of objects from 1-4 metres, and took a bunch of shots without changing any settings. Sometimes it did not refocus, sometimes it refocussed but returned to a similar position, and sometimes to a completely different position. Checked AF points, and they changed from shot to shot. It does seem that the more potential focal points there are with a wide angle and large DoF, the more confused AF becomes. We may need to focus manually where this is a problem.
- Noel

[dje]

hi Dave,
Miserable day in Sydney. Tried indoors with 7D and f3.5 Sigma 10-20, tripod, spot metering, 19 point AF. I set up a variety of objects from 1-4 metres, and took a bunch of shots without changing any settings. Sometimes it did not refocus, sometimes it refocussed but returned to a similar position, and sometimes to a completely different position. Checked AF points, and they changed from shot to shot. It does seem that the more potential focal points there are with a wide angle and large DoF, the more confused AF becomes. We may need to focus manually where this is a problem.
- Noel

Hi Noel

Thanks very much for doing those tests and reporting your results. It sounds like you've had a similar experience to me. I will be manually focusing most of the time I think.

I've been out this afternoon trying out my 10-20. I need to get a feel for what you can do with 10mm to get the most out of it. There are some articles around. Should be interesting.

Dave

[Stagecoach]

I’ve just purchased a Sigma 10-20mm lens for use with my Canon 600D (T3i). When I tried it out initially, I noticed that the AF was somewhat erratic at the 10mm end for objects more than about 1m away. By that I mean that the focus ring would not be positioned consistently in the one position when focusing on the same object on more than one occasion.

I have the Sigma 10-20mm 4.5/5.6 so got it out tonight and put it on the D300 to see if I could understand your concerns/observations.

Firstly, In a large room where the lighting changed from very bright at one end to very dim at the other I found the the lens focused immediately with no hunting with the centre focus point over an area of good change in contrast (target) in all of 10 instances in the dim area. Although the focus assist illumination was on speed and repeatability of focus was good.

Out of curiosity I then tested in Live view, something I never use normally, and found it erratic and hunting with the dim lighting but this was only to be expected.

Secondly, I placed the rig on a tripod and focused it on a point (good target) some 4 mtrs away that was not very well lit to try and understand your comment that ""By that I mean that the focus ring would not be positioned consistently in the one position when focusing on the same object on more than one occasion""

I'm now confused as with this lens the 'focus ring' (the rubber serated grip) does not rotate when focusing in AF. The focus ring manual rotational movement is approx 130 degrees between near and far focus BUT the focus ring can be turned manually past these points continuously due to its clutch mechanism.

So I fail to see how it can be possible to have the focus ring being positioned consistently when it does not move in AF mode. In addition if you were to focus manually, you could only expect consistency in the rings positioning if you were assured that you did not rotate the ring in excess of its 130 degrees between near and far datums.

Perhaps I am reading your concerns incorrectly ?

Grahame

[dje]

I have the Sigma 10-20mm 4.5/5.6 so got it out tonight and put it on the D300 to see if I could understand your concerns/observations.

Firstly, In a large room where the lighting changed from very bright at one end to very dim at the other I found the the lens focused immediately with no hunting with the centre focus point over an area of good change in contrast (target) in all of 10 instances in the dim area. Although the focus assist illumination was on speed and repeatability of focus was good.

Out of curiosity I then tested in Live view, something I never use normally, and found it erratic and hunting with the dim lighting but this was only to be expected.

Secondly, I placed the rig on a tripod and focused it on a point (good target) some 4 mtrs away that was not very well lit to try and understand your comment that ""By that I mean that the focus ring would not be positioned consistently in the one position when focusing on the same object on more than one occasion""

I'm now confused as with this lens the 'focus ring' (the rubber serated grip) does not rotate when focusing in AF. The focus ring manual rotational movement is approx 130 degrees between near and far focus BUT the focus ring can be turned manually past these points continuously due to its clutch mechanism.

So I fail to see how it can be possible to have the focus ring being positioned consistently when it does not move in AF mode. In addition if you were to focus manually, you could only expect consistency in the rings positioning if you were assured that you did not rotate the ring in excess of its 130 degrees between near and far datums.

Perhaps I am reading your concerns incorrectly ?

Grahame

Grahame

My apologies, you are correct. The focus ring doesn't move with AF. I should have been referring to the movement of the distance scale !! My concern was that the distance scale doesn't position itself consistently when the lens is used with my 600D for distances greater than about 1m.

Nikon cameras may behave differently to Canon with this lens too. Don't they use the focus motor in the camera ?

Dave

[Stagecoach]

Hi Dave,

After posting I wondered if you were referring to the distance scale. Did some tests noting the repeatability of that and yes there is some slight variations at farther distances that I would simply put down to the 'automatics'.

As previously mentioned every electro mechanical control system is affected by hysteresis, dead band and to an extent friction. The dead band is one of the most significant in that it is an inbuilt function variable in size where no corrective control is undertaken. This is necessary to stop continual hunting, in other words it causes the system to say OK that's near enough.

[pnodrog]

The auto focusing system to avoid hunting is only looking for an aceptable focus not perfect focus. Depending on the amount and from which direction the focusing motor(system) is seeking an acceptable result the final focus position will vary much more for a wide angle lens than it would for a telephoto lens.

Try manual focus and you will probably find your own variations will be just as great. If you stop down and use depth of feild preview you will find it even more dificult to reproduce the focus ring's postion unless the focus point is extremely close.

The world is full of acceptable compromises.

[tclune]

My Tokina 12-24 f/4 lens often fails to focus at all in good light on my D5000. I have assumed that it is a matter of just about any focus being good, so the camera's algorithm just can't figure out which one is "best." When the system refuses to focus, I just set it to manual and eyeball it, because as others have said the DOF is so large. FWIW