Please help me interpret my focus tests

[Vertigo1]

After feeling that some of my pictures were softer than they should be, I've just done some quick focus tests using the ubiquitous chart.

The results are below and somewhat confusing so I'd appreciate it if anyone could offer their opinions. I have a 400D with two lenses, a Sigma 17-70 and a Canon 70-200 F4L IS. I've taken a test pic at both extremes of focal length on each lens. The problem is I'm trying to work out what's at fault here, one or both of the lenses and/or the body itself. Here we go:

Sigma @ 70mm:
Looks like about 5-10mm of front focus on this one

Sigma @ 17mm:
This is the worst, with a fair degree of front focus

Canon @ 70mm:
This one looks like it has a fractional amount of front focus, maybe 2mm

Canon @ 200mm:
This one looks like about 5-10mm of back focus to me

 

[CT]

I think the first thing you need to do is do these tests again and keep the test chart absolutely square to the camera horizontally. It's easy to be out with your results when the chart is skewed as much as it is in some of these shots. Take your time with the lining up of the chart.

You also need to keep a constant angle throughout the tests, the angle seems to vary a lot. Set everything up square and use a tripod, then when you change lenses everything should remain square.

 

[Vertigo1]

Unfortunately I don't have a tripod - will have to try to borrow one.

As for keeping the chart square, it was as square as I could get it, the Canon @ 70mm is a bit off whilst the Sigma @ 17 is purely due to the wide angle - the other side of the chart was symmetrical, it's just the crop makes it look wonky

 

[Steep]

Rest the camera on a pile of books/table top or something and hold it still while you make any changes. Set the chart a known distance away (there's a little mark on your camera to show where the sensor position is), use a metre stick or a tape to measure it out. As CT has said the results will only be good if the set up is perfect.

 

[Vertigo1]

Ok I'll give it a go - might be able to borrow a tripod. In the meantime, can nothing be deduced from the results above? The big problem is obviously the Sigma at 17mm, which seems to be front-focusing horrendously.

 

[mho]

Thats one of the worst type of tests there is to determine focus accuracy.

there is one out there which is much better, as you focus on a subject which is in the same plane as the sensor and the focus points and the measurement is taken at 45 degrees, this gives a much more consistant set of tests, I cant find the link to the charts at the moment. but will look later

Some other points to note

always use natural light or a well lit room
focus to infinity before focusing on the chart and taking the picture (take 3 pics at each setting etc) cuts down user error.
repeat the tests but don't focus to infinity between each shot.
you will be surprised how the results may vary
print the chart out at the highest resolution and its darkest black setting to make sure to get a deep black to help with focusing.

good luck with the tests

 

[Vertigo1]

Be interested to hear more about this other test (or any other useful tests) if you can find the details

 

[mho]

this is the sort of thing you need, where your focusing on a subject in the same plane as the sensor and lens.

Photo borrowed from POTN - all rights are theirs

this takes out any errors when you focus on a subject which is at 45 degrees to the plane of the lens and sensor / sensor points.

I think canon recommends the tests be carried out at MAX aperture and 10 times the focal length of the lens. with mirror lock up tripod supported, remote release. Centre point only

I should say as well is that the target you focus on is designed like that to give the AF system the best chance of atainig focus as it is using the cross sensor in the centre af point with highly differing levels of black and white which the AF system will love.

Another thing to note is that not all Dslr's focus points actually show the focus points actual sensitive area, the 10D had focus points larger than the area that lit up, i'm not sure about other cameras but will try to find out.

 

[StewartR]

In the meantime, can nothing be deduced from the results above?

No. Simple as that. Without a really solid test setup, it's not safe to draw conclusions about a few mm of front- or back-focus.