AF out or user error

[Equilibrium8]

Hi

I got a Canon 17-55mm lens a while back. I love the lens, but in low light I keep getting pics that are out of focus (close enough to look OK in viewfinder, but focus on eyes is out when I see on PC monitor). I thought it may be backfocusing, but I did a test in semi-controlled environment at home with low light and test charts were fine...

I'm shooting on a 50D and I don't seem to get the problem on other lenses.[IMG][/IMG]

Here is a shot of focus test at home. I think I have deleted shots I mean, but I will check and upload if I find any.

Does this appear to be user error or lens issue?

F/6.3
Exposure 1/200
ISO100
Focal length unknown (+- 10cm from object)

Thanks.

[Inkanyezi]

As EXIF data is not present in the image, it is not possible to see what focal length and what aperture that were used. It is also difficult to see whether there is a focusing issue, as depth of field is rather large. The lens has probably been stopped down somewhat when the test was shot. Try doing the test at the largest aperture, lowest number, and do it at more than one focal length. Use Av and set compensation to +1 for the test chart, and move in as necessary for wide angle and out for tele. Three shots should be sufficient, one at very wide, one at mid-length and one at longest.

It is very common that a lens will need calibration to focus correctly, so it is worth testing.

[Equilibrium8]

As EXIF data is not present in the image, it is not possible to see what focal length and what aperture that were used. It is also difficult to see whether there is a focusing issue, as depth of field is rather large. The lens has probably been stopped down somewhat when the test was shot. Try doing the test at the largest aperture, lowest number, and do it at more than one focal length. Use Av and set compensation to +1 for the test chart, and move in as necessary for wide angle and out for tele. Three shots should be sufficient, one at very wide, one at mid-length and one at longest.

It is very common that a lens will need calibration to focus correctly, so it is worth testing.

I forgot to add EXIF. I updated now, but only have the readouts from Win7 folder view which doesn't include focal length. I will do another test following your guidelines. Thanks!

[Geoff F]

Those charts give a rough guide but the method isn't totally accurate.

See this site as well http://photographylife.com/how-to-qu...tofocus-issues

[jcuknz]

This article may be of interest to you http://photographylife.com/what-is-focus-shift

I gather most cameras focus wide open and then close to the required aperture the way we worked with manual lenses in the past [ and some still do ]

[dje]

Hi Kenny

I'd suggest you try the method Geoff references in his post. You may find this easier and more reliable.

Just a couple of comments on your test shot :

• The shot hasn't been taken square on to the chart and may not have been close to 45 deg in elevation
• Did you use a tripod ?
• The centre focus area is quite blurry - something seems wrong.

Dave

[revi]

In addition, for this test to be as reliable as possible, make sure that the line on which you focus is perfectly parallel to the sensor.
In the posted image that bar seems to be tilted to the right, and that can mean that when you focus with the centre focus spot on the centre of the bar, the edges are not in the same plane. That would give an error in your estimation of the true point of focus. You can check that by comparing the right and left scales, any difference in what you see indicates a possible error in alignment

[Colin Southern]

Keep in mind too that AF systems definitely can struggle in low light.

[Equilibrium8]

Those charts give a rough guide but the method isn't totally accurate.

See this site as well http://photographylife.com/how-to-qu...tofocus-issues

Thanks! I'll giver this a try. I tried Urban's advice, but I couldn't find anything wrong with the shots.

[Equilibrium8]

This article may be of interest to you http://photographylife.com/what-is-focus-shift

I gather most cameras focus wide open and then close to the required aperture the way we worked with manual lenses in the past [ and some still do ]

Thanks. Very interesting article. I'm surprised this hasn't come up in discussions or books I've read.

[Equilibrium8]

In addition, for this test to be as reliable as possible, make sure that the line on which you focus is perfectly parallel to the sensor.
In the posted image that bar seems to be tilted to the right, and that can mean that when you focus with the centre focus spot on the centre of the bar, the edges are not in the same plane. That would give an error in your estimation of the true point of focus. You can check that by comparing the right and left scales, any difference in what you see indicates a possible error in alignment

Yes. Keeping it straight seems tricky. I struggled to keep the different focal lengths the same height when I did the other test last night.

[Equilibrium8]

Hi Kenny

I'd suggest you try the method Geoff references in his post. You may find this easier and more reliable.

Just a couple of comments on your test shot :

• The shot hasn't been taken square on to the chart and may not have been close to 45 deg in elevation
• Did you use a tripod ?
• The centre focus area is quite blurry - something seems wrong.

Dave

Hi

It wasn't quite square. I used a tripod. The blurry may be what I am seeing in other pics.

[Equilibrium8]

I will try Geoff's idea out later. For now, I uploaded some of the test at different focal lengths I took last night. I don't see anything helpful, but I thought maybe some here with better eyes and who knows what to look for could comment.

1/5 sec at f 2.8 17 mm
[IMG][/IMG]

1/6 sec at f 2.8 40 mm
[IMG][/IMG]

1 5 sec at f 2.8 55 mm


These are still not quite straight on. They were all taken on tripod. I did notice the ballhead I thought was super sturdy was slipping a bit with the narrow line.

[Stagecoach]

Hi Kenny, there's reams written about this test method and it's questionable accuracy on the web, much of it conflicting and much of it generated recently due to a known auto focus problem on a certain new camera model a while ago.

I'm no expert and have also never had the need, or felt the need, to undertake testing of my auto focus. From looking at the three at 55 and 40 mm focus looks fairly even (assuming chart at 45 deg) and errors are more than likely due to the chart 'alignment'. I will not comment on 17mm FL as without knowing the camera subject distance as the DoF is so wide here.

One useful tip I did read is that whilst undertaking these autofocus tests undertake a 'manual' focus one (without moving the camera) at the same time for each scenario preferably using liveview if not fully confident using the viewfinder. This then gives a good comparrison.

Grahame

[Equilibrium8]

Hi Kenny, there's reams written about this test method and it's questionable accuracy on the web, much of it conflicting and much of it generated recently due to a known auto focus problem on a certain new camera model a while ago.

I'm no expert and have also never had the need, or felt the need, to undertake testing of my auto focus. From looking at the three at 55 and 40 mm focus looks fairly even (assuming chart at 45 deg) and errors are more than likely due to the chart 'alignment'. I will not comment on 17mm FL as without knowing the camera subject distance as the DoF is so wide here.

One useful tip I did read is that whilst undertaking these autofocus tests undertake a 'manual' focus one (without moving the camera) at the same time for each scenario preferably using liveview if not fully confident using the viewfinder. This then gives a good comparrison.

Grahame

Thanks Grahame. One of the links above mentioned using Live View, too. It seems Live View uses contrast detection AF and not the phase detect normally used.

[Stagecoach]

Kenny, I have just finished reading that article this minute and Live view was used for one scenario but not with Manual Focus. The tip I read previousy only advised the use of Liveview to 'assist' with manually focussing.

An interesting article but omits any reference to hysteresis that can cause differing errors dependent upon focus direction (from near or far) due to the dead band incorporated in all control circuits to alleviate continous hunting. Just one of the many 'possibilities' when delving deeply into possible causes.

[dje]

Thanks Grahame. One of the links above mentioned using Live View, too. It seems Live View uses contrast detection AF and not the phase detect normally used.

Kenny with Live View focussing, you are using Contrast Detection and this measurement is done on the main sensor. With normal AF focus, you are using Phase detection on separate auxiliary sensors which receive their light through a different optical path. This means that proper calibration of the Phase detect setup is essential for Phase Detect AF to work properly. So the "photography life" method potentially gives you an idea how well it is calibrated.

The latest three images do appear to be as you would expect - with the centre focus point the sharpest part of the image and then the front and back gradually going out of focus. However the centre point still doesn't look terribly sharp to me (but is much better than the first shot you posted). How does it compare to your other lenses.

Dave

[dje]

Kenny, I have just finished reading that article this minute and Live view was used for one scenario but not with Manual Focus. The tip I read previousy only advised the use of Liveview to 'assist' with manually focussing.

Graham I have done this test using Manual focus (in magnified Live View mode) as I also think this is the better way to go than LV mode AF.

Dave

[Equilibrium8]

Kenny, I have just finished reading that article this minute and Live view was used for one scenario but not with Manual Focus. The tip I read previousy only advised the use of Liveview to 'assist' with manually focussing.

An interesting article but omits any reference to hysteresis that can cause differing errors dependent upon focus direction (from near or far) due to the dead band incorporated in all control circuits to alleviate continous hunting. Just one of the many 'possibilities' when delving deeply into possible causes.

Sorry, I misread that. It makes sense to try manual focusing for reference. Thanks!

[Equilibrium8]

Kenny with Live View focussing, you are using Contrast Detection and this measurement is done on the main sensor. With normal AF focus, you are using Phase detection on separate auxiliary sensors which receive their light through a different optical path. This means that proper calibration of the Phase detect setup is essential for Phase Detect AF to work properly. So the "photography life" method potentially gives you an idea how well it is calibrated.

The latest three images do appear to be as you would expect - with the centre focus point the sharpest part of the image and then the front and back gradually going out of focus. However the centre point still doesn't look terribly sharp to me (but is much better than the first shot you posted). How does it compare to your other lenses.

Dave

Thanks for explanation! I will print those other charts out and try them tomorrow. I haven't compared with other lenses yet, as I haven't had issues with them, so I will do a few comparison shots at the same time.