20D Focusing

[RogersDA]

This is an RC car in a light tent. Shooting with a Canon 20D + 24-70L USM + Tripod + remote release. Camera set to autofocus - Center point only and targeted at the area just above the side view mirror (marked with the white square in image 1).

1. ISO 100, 1/160 sec., f/2.8, 58mm.
http://darphotos.smugmug.com/photos/165931285-L-LB.jpg

Image 2 - 100% crop of door.
http://darphotos.smugmug.com/photos/165931359-L-LB.jpg

Image 3 - zoom in on mirror and door handle.
http://darphotos.smugmug.com/photos/165931385-L-LB.jpg

What is odd here is that the autofocus point does not appear to be in focus. Rather, the door handle is the sharpest.


Images 4 and 5 - settings are the same, except shot at f/22.
http://darphotos.smugmug.com/photos/165931432-L-LB.jpg

http://darphotos.smugmug.com/photos/165931447-L-LB.jpg

At f/22 the area above the mirror has come into focus, somewhat.

So, I am thinking there's an issue here, but wanted other opinions. Anybody got some?

[Richard]

There do not appear to be any edges within the autofocus box. Could this be the reason it misbehaved? There are edges near by, but I don't really know what the radius of the autofocus is. Have you had focus issues on other shots?

Cheers,

[z_28]

Canon 20D + 24-70L IS USM + Tripod + remote release.



Sorry for correction, but such IS version don't exist yet :wink

[scwalter]

how close are you? According to the Canon lens chart the min focusing distance for the 24-70 is 1.25ft. Are you approaching that?

[RogersDA]

There do not appear to be any edges within the autofocus box. Could this be the reason it misbehaved? There are edges near by, but I don't really know what the radius of the autofocus is. Have you had focus issues on other shots?

Cheers,

Thanks for looking, Richard. I agree - there are no edges in the center of the box. However, you are right there is that one vertical edge. Apparently it is enough there for the camera to sucessfully autofocus:dunno. I don't have a clue how Canon actually does that.

Prior focus issues - not sure - that's why I was testing this. So many shots are not sharp with this lens, and my research showed my that this lens was very good.

[RogersDA]

Sorry for correction, but such IS version don't exist yet :wink

Me + my fingers + keyboard + midnight = :doh

Thanks for keeping me straight!

[RogersDA]

how close are you? According to the Canon lens chart the min focusing distance for the 24-70 is 1.25ft. Are you approaching that?
I may well be close. I need to get a ruler out and measure the distance. That should be from the focus point on the object to the sensor plane, correct?

A tape measure shows about 27" (2.25') from the focus point to the sensor plane, and about 20" (1.35') from the focus point to the front of the lens.

[gtc]

It may be too much of an ask for your autofocus-a focussing screen might be the best solution-such as Katzeye or Haoda-see Shay's thread.

As long as your viewfinder diopter is correctly set and you have a correctly installed focussing screen,your eye and brain will do the rest.

[kdog]

That's an inconclusive test, David. The sensitive area of focus points is actually about 3 times the diameter and width of the little indicator squares. Each AF point has multiple sensors that are arranged in rows and/or columns. The center AF point has both row and column sensors arranged in a plus-sign configuration. The rest of them are arranged in a line, either horizontally or vertically. So they lock onto either vertical or horizontal contrast lines best. If you use other AF points rather than the center one, it pays to know which are horizontal and which are vertical AF points. There are diagrams which show the layout, however I can't seem to find a reference to one right now. I do recall that the left and right sensors detect vertical contrast lines, while the top and bottom ones detect horizontal contrast. I forget what the inbetween ones use, but I rarely use them. I mainly stick to the center AF point which locks on to both horizontal and vertical. Your area of best focus in that shot is more level than the area which the square is over, so that's probably why the focus didn't end up where you thought it should.

So the results look normal for F2.8, since the DOF is so shallow. However, at F22, I would have expected the entire car to be in focus. You really need to calculate the DOF for your setup to be sure. There are online calculators for that. Here's one: http://www.dofmaster.com/dofjs.html

If you really want to characterize your camera/lens AF calibration, you need to set up a better controlled experiment. This site is useful: http://www.canon-dslr.com/Canon_Jan05/Canon_SLR_Focus_Test.htm

I have other tips too, but that should get you started.
-joel

[claudermilk]

Well, from the parameters in your origianl post it appears the f2.8 image is ever so slightly back-focussed. However as has been mentioned, this setup may be challenging the limitations of the lens minimum focus and the body AF system. Since at f2.8 the DOF here is paper-thin and that door handl cannot be very far from the intended plane of focus. My gut feeling is things are within spec.

My next thought is why shoot the lens wide open here? You are in control of the light, so stop down to at least f8 to put the lens in it's sweet spot and give a broader DOF to avoid the problem. Maybe back the lens away a bit and zoom in some more.

[RogersDA]

It may be too much of an ask for your autofocus-a focussing screen might be the best solution-such as Katzeye or Haoda-see Shay's thread.

As long as your viewfinder diopter is correctly set and you have a correctly installed focussing screen,your eye and brain will do the rest. I had been reading that thread, and I have missed the focusing screen from my previous film units. I was never clear why they did aways with those. I am seriously thinking of that purchase.


That's an inconclusive test, David. The sensitive area of focus points is actually about 3 times the diameter and width of the little indicator squares. Each AF point has multiple sensors that are arranged in rows and/or columns. The center AF point has both row and column sensors arranged in a plus-sign configuration. The rest of them are arranged in a line, either horizontally or vertically. So they lock onto either vertical or horizontal contrast lines best. If you use other AF points rather than the center one, it pays to know which are horizontal and which are vertical AF points. There are diagrams which show the layout, however I can't seem to find a reference to one right now. I do recall that the left and right sensors detect vertical contrast lines, while the top and bottom ones detect horizontal contrast. I forget what the inbetween ones use, but I rarely use them. I mainly stick to the center AF point which locks on to both horizontal and vertical. Your area of best focus in that shot is more level than the area which the square is over, so that's probably why the focus didn't end up where you thought it should.

So the results look normal for F2.8, since the DOF is so shallow. However, at F22, I would have expected the entire car to be in focus. You really need to calculate the DOF for your setup to be sure. There are online calculators for that. Here's one: http://www.dofmaster.com/dofjs.html

If you really want to characterize your camera/lens AF calibration, you need to set up a better controlled experiment. This site is useful: http://www.canon-dslr.com/Canon_Jan05/Canon_SLR_Focus_Test.htm

I have other tips too, but that should get you started.
-joel

Thanks for that information. I was not aware of the horizontal/vertical process for the different autofocus points. I knew that the center point was always stated to be more sensitve, and what you stated makes that more clear. I will certainly be going though the stuf you posted.

Well, from the parameters in your origianl post it appears the f2.8 image is ever so slightly back-focussed. However as has been mentioned, this setup may be challenging the limitations of the lens minimum focus and the body AF system. Since at f2.8 the DOF here is paper-thin and that door handl cannot be very far from the intended plane of focus. My gut feeling is things are within spec.

My next thought is why shoot the lens wide open here? You are in control of the light, so stop down to at least f8 to put the lens in it's sweet spot and give a broader DOF to avoid the problem. Maybe back the lens away a bit and zoom in some more.

f/2.8 was not for any reason other than testing the camera. Thanks, too, for the suggestions.

[claudermilk]

I can vouch for the KatzEye screen if you are thinking of getting one. I missed the split prism, too--so I did something about it. :deal It really is worth the investment if you do any manual focus work at all, or need to verify the AF.

[hs30540]

My 24-70L never focused properly until I sent it in to Canon to be calibrated.

Try this:

http://www.focustestchart.com/focus10.pdf