[Guide] How to test a lens for de-centering

[puddleduck]

Premise: Wide angle lens are very prone to "de-centering" - a manufacturing defect (or caused by drop damage!) whereby optical elements within a lens are no longer aligned correctly. This generally manifests itself as the right or left hand side being noticably softer than its opposite side. Lens elements move back and forth within a lens its when focusing - if the elements do not move "as one" then softness can manifest itself on the frame.

This is where this photographic cliche - the dreaded brickwall shot! - becomes handy.

To do this properly, you need go decent light, and a wall that is pefectly flat.

The first task is to take a control shot - this is made with the camera in a landscape orientation.

Here's one I made earlier!

Now we have our control shot above - without re-focusing the camera - rotate the camera 180 degrees so it is upside down - and re-take the shot. This is our comparison shot.

If a lens is decentered, when the camera is rotated upside down, any uneven sharpness with the framer will move from right to left or vice versa between the control shot and the comparison shot.

The final stage is the take the comparison shot, and rotate it back to the correct orientation in Photoshop etc.

This comparison can be very revealing:

This quick test allows us to clearly compare left and right side (and top / bottom) performance - in the case of this example, the lens in question has a severe de-centering problem on the right hand side.

This test is easy to do, and takes 5 minutes at most, and I'd definately recommend spending the time - better to find out quickly, rather than have image rejections down the line.

HTH

 

[malla1962]

Why can't you just put up 2 targets and look at both sides?
Your upside down shot does not look square to the wall.
What lens was you testing?

 

[puddleduck]

Why can't you just put up 2 targets and look at both sides?

Could do - but the rotation is more definitive.

Your upside down shot does not look square to the wall.

Its fine.

What lens was you testing?

I was testing a Sigma 12-24.

This is that top corner in the real world...

...the brickwall test confirmed the issue. Its very useful - if the softness flips right to left or vice versa after rotation, its a 101% confirmation of de-centering.

 

[coldpenguin]

How do you ensure that the camera is
a) steady
b) the same distance
when upside down (assuming that most wide-angle lenses don't have tripod mounts)?

 

[Tyke Tiler]

How do you ensure that the camera is
a) steady

Use a shutter speed that your confident will eliminate your vibrations.
Even without IS or VR, it's fairly simple to take a crisp frame without a tripod, it's all about shutter speeds and how you hold your camera.

b) the same distance?

when upside down (assuming that most wide-angle lenses don't have tripod mounts)?

It seems to me that distance precision isn't that much of a significant variable in this test really, if your lens is de-centred it's going to have roughly the same results despite similar distances.

T.

 

[Bruce120]

Thanks Andy, easy to follow and just tested my Tokina 12-24.

 

[HoppyUK]

You don't need a brick wall, nor do you need to turn the camera upside down. Any target with some detail will do, preferably at some distance to minimise any potential focusing variance, eg car number plate from a few yards. Be sure to shoot at maximum aperture (lowest f/number) where any problems will be most obvious.

Focus carefully on target using centre AF spot, ensuring a shake-free shutter speed. Lock focus and exposure. Position target in the corner of the frame and take a pic. Repeat with the target in the other three corners. Compare the four images, and if one or more is significantly worse, you have a problem. Very slight variances are quite common, and usually acceptably "within tolereance" - faults are obvious.

Just remember that you are looking for consistency, not ultimate quality, which is often quite poor with a severe test like this, ie the extreme corner of the frame at lowest f/number will show the lens at its weakest point. Wide zooms are especially bad at this.

You don't need a PC - just blow them up on the camera screen. It takes less than a minute. It's a very reliable 'good copy' test or an 'is it damaged' test.

 

[classcams]

Not enough info given at the start.
1/ what was the focal length
2/ Is it a full frame or cropped shot.
Looking at the blow ups of the corner objects one seems a clean shot considering it`s at fully open aperture.

 

[HoppyUK]

Not enough info given at the start.
1/ what was the focal length
2/ Is it a full frame or cropped shot.
Looking at the blow ups of the corner objects one seems a clean shot considering it`s at fully open aperture.

Not sure what difference the focal length setting makes, or whether it matters if it's full frame or not. If you're testing for centring, which is far and away the most common problem with a faulty lens, and it's out (either because the lens has been poorly assembled, or damaged) then four identical photos of an object positioned in the corners, shot at lowest f/number, will show it.

If nothing shows up in such a test, there can't be much wrong