The Single Focus Calibration Penalty

Most camera manufacturers, with the notable exceptions of Sigma and now Tamron, only let you assign a single focus fine-tune calibration setting. Canon will let you calibrate zooms at two different focal lengths, which is slightly better than Nikon. What happens if you have a zoom lens that needs different focus calibration at various focal lengths or different distances? What happens is that you lose resolution on your intended focus target.

You would assume that this discussion doesn’t apply to mirror-less cameras or when you shoot using Live View. In those cases, the camera uses the image sensor to detect best focus, and focus calibration doesn’t apply, or does it?

Did you know that the Nikon Z mirror-less cameras actually still have a focus fine-tune option in their ‘Setup’ menu? Why in the world would they still offer it? The phase-detect pixels in the image sensor can still lead to focus errors when telling the lens how far to move focus, hence fine-tune is still there. Only the “Pinpoint” focus mode on the Z cameras uses contrast detect, which is, again, really slow. Sorry to burst your bubble.

Nikkor 24-70 f/2.8 AF-S E ED VR has terrible focus shift

I decided to use my Nikkor 24-70 f/2.8 AF-S zoom for some testing. This isn’t a cheap lens by any stretch of the imagination. It experiences a significant focus calibration shift when zooming it, but you can only calibrate it with a single “fine-tune” value.

I’m using the MTFMapper program from here to analyze some shots of a special focus target. First, I want to see how far focus calibration shifts after zooming, and secondly I want find out how much this focus error impacts the shot resolution on the target.

In my testing, I’m using phase-detect focus on my Nikon D610. I’m shooting unsharpened raw pictures on a sturdy tripod, and I have turned off vibration reduction. I triggered the shots using an infrared remote, and the mirror is flipped up for at least 3 seconds before tripping the shutter.

I take at least 10 shots at each zoom setting, to ensure I know what the “typical” focus distance is. There’s always some natural focus variation between shots, so it’s important to characterize what the range of focus distances actually is.

The focus target

The MTFMapper program uses a target like the one shown above to analyze focus. The red rectangle shown represents where you should point your camera focus sensor. The target itself is rotated at 45 degrees relative to the camera sensor, with the trailing (outer) edge of the big target rectangle rotated away from you. The trailing vertical edge of the big target rectangle is taller than its leading edge, so that the final photo makes it look more like a real rectangle instead of a trapezoid. This special target shape is intended to fix perspective distortion.

The Test

I shot the 24-70 lens at 70mm f/2.8 to verify correct camera focus calibration. I just so happens that 70mm on my D610 uses a focus fine-tune value of 0; it’s a non-zero value on my other cameras. The big rectangle vertical edge (underneath the red focus sensor shown above) should be in perfect focus. I took ten shots, where I manually de-focused the lens between shots, and then pressed my focus button to re-focus the target edge with phase-detect focus. I positioned the camera to have the test target just fill the frame.

After I verified correct focus calibration using the MTFMapper program to analyze my shots (at 70mm), I then zoomed out to 24mm and moved the camera/tripod until the test target again filled the frame. I didn’t alter the focus fine-tune setting, since the whole point of this test is to note the effect of the single focus fine-tune calibration at different zoom settings.

I shot another 10 shots at 24mm f/2.8, remembering to de-focus the lens before pressing the focus button to reestablish focus. I analyzed these photos in MTFMapper to find where the “typical” focus distance landed.

70mm f/2.8 focus results in MTFMapper

The 70mm typical results are shown above. The plane of best focus was verified to coincide with the leading edge of the big target rectangle (I drew a vertical red line to show where best-focus is). The average MTF50 resolution of the little square left-edges is about 33 lp/mm. The average MTF50 resolution of the little square right-edges is about 36 lp/mm. The peak resolution was 37.5 lp/mm.

24mm f/2.8 focus results in MTFMapper

The typical focus results at 24mm are shown above. The plane of best-focus missed the big vertical target edge by about 65mm (2.5 inches) at a focus distance of about 1 meter. To correct the missed focus, it would rquire a focus fine-tune calibration setting of about -6 on this camera. If I used this -6 fine-tune setting, then of course the 70mm focus calibration would be wrong. This is very, very irritating. Irritating enough for me to say a bad word.

The average MTF50 resolution in the plane of the target edge (where the camera focus point was) is about 38 lp/mm. The actual plane of focus (indicated by the vertical red line above) has an MTF50 resolution of about 50 lp/mm, or about 30 percent higher resolution. This means that I lost 30 percent of resolution where I actually told the camera to focus!

When I have the time and remember to do so, I change the focus fine-tune value to -6 when I am shooting in the vicinity of the 24mm focal length. Sometimes I'll try a small manual focus change instead. This is maddening to me, and fraught with error. The in-between focal lengths have varying degrees of focus error and their associated resolution loss.

Conclusion

The resolution loss your photos experience after zooming away from the calibrated focal length can be quite significant. Your very expensive lens may even be performing worse than a cheap ‘calibrated’ lens. You might even be blaming yourself for being a poor photographer who keeps missing focus.

It’s really unfortunate that Nikon cameras and lenses don’t let the user calibrate focus at multiple focal lengths and focus distances. I don’t consider switching to Live View a solution; not for action shots, for sure. And please don’t tell me I need to stop down to f/16. And mirror-less cameras aren’t the savior, either.

Sigma engineers noted this focus problem long ago, and fixed it by letting you calibrate their “global vision” lenses at 4 focal lengths and 4 distances per focal length. Tamron has since followed suit, so now their newer lenses have a similar calibration feature.

This problem may force photographers to switch to lenses with smart, programmable firmware like Sigma and Tamron. I don’t think that Nikon has any plans to address the issue. I personally fixed the issue by buying Sigma lenses; and no, they didn’t pay me to say that.