Calibrating a lens...is auto focus less reliable when zoomed?

Question

So, I'm calibrating my lenses to a new camera body, a 5D Mark IV. When calibrating my 70-200 ƒ2.8 (version 1), I found that the focus was accurate at 70mm. However, when I calibrated it at the 200mm end, it was inaccurate.

The problem is, the inaccuracy is erratic. Rather than being just back focused or just front focused, which I can adjust for, the 5D would sometimes be back focused, sometimes front focused, and sometimes almost accurate.

Being that it is a long lens, I figured maybe camera shake might be causing it, so I retested using mirror lock-up to minimize shake. But I'm still seeing the same result.

I found the same behavior on my 17-40 ƒ4.0. Accurate at 17, erratic at 40.

I didn't see this behavior a couple years ago when I calibrated the same lenses using my 6D Mark II.

Ideas anyone?

What kind of testing protocol are you using? Improperly aligned targets almost always return false results. If you're aiming at a sloped target, you're never going to get consistent results. The target should be almost perfectly perpendicular to your camera's sensor. Any sloped scale should be far enough to the side that there is no danger the camera is trying to focus on it (because most AF "points" are much larger than those little squares in the VF and you can't be sure exactly where the camera is aiming. Also, are you testing under fairly bright, full spectrum lighting? — Michael C, Feb 06 '20 at 07:34
Related: Which offers better results: FoCal or LensAlign Pro? and What methods can be used to micro-adjust autofocus of a camera body? and Why does my DSLR focus on the background instead of my subject when taking shallow DoF photos? — Michael C, Feb 06 '20 at 07:46
For a more comprehensive overview of how to diagnose AF issues, including AFMA related issues, as well as a plethora of links to related questions and answers here at Photo.SE, please see: this answer to How do I diagnose the source of focus problem in a camera? — Michael C, Feb 06 '20 at 07:48
I'm using the SpyderLENSCAL device and procedures, so aperture wide open, proper distance away for the focal length based on their min/max distance charts, device level with camera, etc. The one variable that may be an issue is the lighting as I'm under tungsten lighting. I have a daylight balanced LED I'll try out. — Darrin, Feb 06 '20 at 18:10

score 1 · Answer 1 · answered Feb 06 '20 at 23:58

Ok, for future people searching for this issue, I think this is resolved.

The issue (for me, at least) was not enough light on the target. I was doing the testing indoors, beside a window mid-day (curtains open) and below a tungsten dining room light. But this was not enough light on the target for the auto focus to perform it's best. IIRC, my shutter was about 1.5 seconds in those conditions.

So today I lit it with a daylight balanced LED panel and re-tested. I always expect some variability, so I take 5 shots at each focal range I test, refocusing between each.

This time there was still some variability, but instead of + or - 12 units (on whatever scale they use), it was more like + or - 3. And there was only one shot that was 3 off, so I'll assume that it was an outlier.

So thanks everyone for your comments.

score 0 · Answer 2 · answered Feb 06 '20 at 15:32

In a way you could say that autofocus is less reliable when a lens is zoomed... if you do not correspondingly increase the focus distance.

More accurately; what happens is that the depth of focus decreases, which makes any focus errors more apparent (there is always an error tolerance; there has to be).

Also, many zoom lenses have more optical errors at the extremes (one or both). That certainly can impact autofocus accuracy/reliability.

But IME, the main issue with autofocus calibration is poor/improper methods.

Calibrating a lens...is auto focus less reliable when zoomed?

2 Answers2