Focus chart to check all focusing points

My camera (7D mark II) seems to have a problem ...

I want to see if that is the case or I am doing something wrong - the issue: center focusing point is accurate, right hand side focusing points are accurate ... left hand side - especially the top ones (the ones I tend to favor) not so accurate ... (this is not necessarily related to a full frame or crop camera - some 5D mark IV users have reported the same issue - and reading those reports I got the idea to check mine).

I sent my camera to Canon 2 times so far for focus inconsistencies - according to them nothing is wrong - but I know for a fact that they checked using the center point only.

The issue is apparent with all my lenses so I think that might be too much of a coincidence to blame de lenses!

The idea is to print a 100x66 cm chart, mount it on something hard and square the camera as best as possible to it - am I right in thinking that if the chart perfectly fits in live view the camera will be aligned correctly? - That is question 1.

Question 2 - what will be the best pattern for it?

At the moment I put together this https://drive.google.c​om …-B1EuTKy46ZTBFb0UzaXVQU​jA (full size link)

Resized version

First improvement ... to help aligning the chart!

Interesting experiment. I would be interested in learning the results.

If you could rig a mirror to the center (temporarily), that would help to get the lens axis exactly perpendicular to your test chart.

I guess it would not be that surprising if the lens mount is slightly off parallel to the sensor, but the question is whether the discrepancy is significant.

I've used a mirror to align other 3D rigs ... but I was thinking that if all the white triangles will touch the side of the screen in live view this way the alignment will be more precise ... am I right or not?

xseven wrote in post #18185817
I've used a mirror to align other 3D rigs ... but I was thinking that if all the white triangles will touch the side of the screen in live view this way the alignment will be more precise ... am I right or not?

I guess you would be correct, but I suspect in practice it will be very hard to do an exact registration on the LCD.

Canon warns that repeating patterns can fool AF. That might not be an issue if you use Live View... but you could also stick something like a postage stamp under each AF point location.

Archibald wrote in post #18185819
Canon warns that repeating patterns can fool AF.

I was unaware of this fact ... in what way would this be an impediment?

xseven wrote in post #18185820
I was unaware of this fact ... in what way would this be an impediment?

The AF works by merging two images, with one part coming from the upper part of the lens and the other from the lower part. Naturally the parts need to be the same part of the subject. It works like a rangefinder. With repeating patterns it could try to merge neighboring parts of the image.

And thinking some more about registering the triangles around the periphery to the LCD - this would also depend on how symmetrical the lens is. And all lenses can be expected to have some asymmetry due to manufacturing tolerances.

You are doing challenging work, measuring a small error in the alignment of the flange or sensor or AF system! Your technique needs to be very good, probably better than Canon's tolerances.

Indeed ... there are distortions ... I will add another "inner rectangle" as a reference point so this way I can calibrate each lens around the middle of the range where the distortion is minimal ... even with distortion the corner triangles should be enough to insure the "calibration"

A single line passing the AF point is better than this kind of pattern. That kind of pattern will always give inconsistent results. It's like trying to focus on a venetian blind.

If all your AF points are sensitive to horizontal lines (like they are in a 1DX, for example), then you could print as many lines as there are rows of AF points. Nothing else is needed to test all of them. Having only one line per AF point virtually eliminates the risk for aliasing among the lines.

Interesting experiment but I'm not sure how you are going to differeniate between focus sensor aligment and lens alignment or defects.
I suppose it might be best if you use a macro lens that has a flat field and is known to be perfectly centred.

I think this sort of test might call for a far higher precision setup than anyone could manage outside of optical laboratory conditions which is probably why the average camera service centre cannot find the fault.

Ray

You are WAY over thinking this. All the focus chart you will ever need can be made in less than a minute with:

1.) 1 sheet of 8.5 x 11 paper
2.) 12" ruler or straightedge
3.) 1 black pen
4.) 1" of scotch tape

Lay paper out in landscape orientation. Make two marks, each one inch long, one parallel to the short edge, one parallel to the long edge. Affix to wall.

Shoot from 50X focal length of lens mounted. Cross points will see either/both of the marks. Each linear point will see the line perpendicular to itself.

Most of the time I just shoot a brick wall. My wife complains when I stick things to her walls.

If multiple lenses will give similar results ... I guess that lowers the chances that I have a problem with the glass
A set of horizontal lines can work only for a certain distance and focal length ....
What about a simple square pattern? Would that be better?

A small paper will not give me the results I want ... center point is accurate as well as the right side ... the left side is playing me ....
So I need to test the camera on both sides, without touching it ... at a "normal" distance .... not in the macro realm ... because another thing I noticed ... micro-adjustment is different at different distances ... so I will consider at least 2-3 meters to the target .

xseven wrote in post #18185957
If multiple lenses will give similar results ... I guess that lowers the chances that I have a problem with the glass
A set of horizontal lines can work only for a certain distance and focal length ....
What about a simple square pattern? Would that be better?

I have no solid theory here, just my own ideas, which have been known to be wildly amiss.

My thinking is as follows. Say I am unsure if this camera/lens is focusing where I want it to focus. If I only give it one thing that it CAN focus on (each individual line), I know EXACTLY where it was focusing. If you have two things it COULD have focused on, how do you know which one was the CAMERA's subject. I used my method to determine that the top row of AF points on my 60D was just above the little boxes in the viewfinder. That explained a lot.

Your line of thinking is correct if you are dealing with a 3D scene ... but in this case all the lines are at the same distance ... so, I think, there is no confusion

But checking the "corect" correlation between the focus squares in the viewfinder and the actual focusing points on sensor is definitely something I need to do.