Canon Lens Auto Focus Accuracy – Part 2

Having set my mind at rest in Part 1 that my new lens wasn’t totally dodgy, I set about trying to make it even better.  In fact, I decided to repeat the process on all three of my main lenses.

Testing Process – attempt 2

After doing some more online research and making a few more “mistakes” first hand, I put together a few refinements to improve my calibration method.

Vertical focus target: I realised after I’d finished my first testing that the low-contrast details in the middle of the focus chart were bold enough that my focussing system was still capable of picking them up, so it was possible that some of the back focussing effect I saw in Part 1 was due to the focus point locking to something further back on the chart.  To eliminate this, I printed a simple focusing target (from the last image at the end of the page) and taped it to the front of a DVD box-set.  This target was then aligned with the zero point on the scale.  Canon say that a vertical target is better and based on my experience, I agree.  The lens calibration tools that are available commercially also use a vertical target, which is another good indicator.

Shooting tethered in Capture One Pro:  Having already installed all of the relevant Canon software and Capture One Pro (C1), I started C1 and set up a new tethered session.  With the camera switched off, I connected it to the PC with the USB cable provided and then switched the camera on.  In version 6.2 there isn’t support for Live View with my camera but there are other advantages to using C1 apart from the immediate large screen review of the shot.  With my first shot on each lens, I zoomed and panned to the relevant part of the scale in the C1 image, converted to black & white and enhanced the contrast.  Each new shot after that appeared on screen with the same settings rather than having to repeat this process each time, which made reviewing very quick.

Test at maximum focal length: I had initially intended to check each end of the zoom range but as you can only enter a single calibration value and the largest effect of focussing errors would be at the long end of the zoom, I cut this down.

Closer testing: Reducing the distance from camera to target reduces the depth of field.  I had thought that this would make the testing easier as it would be easier to see the effect of any error.  However, having tried to calibrate 200mm at 1.5m distance (7.5x focal length) and running out of microadjustment range, I realised that it wasn’t going to work!  Canon suggest testing at 50x focal length but I didn’t have room for this, so I had to be content with 25x.  For the 200mm I was at 5m and for the 100 / 105mm I was at 2.75m.  When I added the extender I should have moved further back but at 18x focal length I decided this was sufficient – and besides, I didn’t have space!  I could also have done the 100mm calibrations at 5m but the other limiting factor was that I found the size of the test chart in the image too small to make an accurate assessment of the focusing zone.

When the going gets tough: When I was struggling to decide where each zone of sharpness lay and which microadjustment setting was best, I exported a TIFF file out of Capture One and applied an Emboss filter in Photoshop Elements (there are a few references to this around – just Google “focus adjustment emboss filter”).

Keep notes: I quickly found that I forgot which images had which microadjustment settings, so I noted them in the metadata to make it easier to compare settings.

My final process and calibration

Disclaimers: this worked for me, it might not work for you!  There are probably better ways to do either individual aspects of these tests, or even the whole method.  Proceed at your own risk.

Other than the refinements above, I used the same principles as attempt 1.

  1. Set up camera and lens:
    1. Focal length of lens = maximum zoom
    2. Aperture = maximum (smallest number)
    3. File format = largest RAW files possible
    4. Aperture priority mode (Av)
    5. Focus = single-shot
    6. Focus point = centre only
    7. Image stabilisation = off
    8. Autofocus = on
    9. Mirror lock-up = enabled
  2. Set the testing scale at an angle and a focus target vertically, aligned with the zero on the test chart.
  3. Set the camera on a solid tripod with a remote release at least 25x the focal length away from the target.  Set the height of the tripod so the centre focus point is at same height as middle of chart.  Set the sensor plane of the camera parallel to the focus target.
  4. Align the centre focus point with the middle bar on the focus target.
  5. Manually set the lens to focus to infinity.
  6. Use the remote release to autofocus and release the shutter.
  7. Repeat steps 5 and 6 three times to allow for variations in the accuracy of the focusing system.
  8. Review the three shots.
  9. In steps of 5, try two sets of microadjustment and repeat steps 5 to 8 each time.  Determine where the likely “best” adjustment lies from your first 3 sets of pictures and try more settings.
  10. Repeat until best setting established.
  11. Use Photoshop Elements Emboss filter to emphasise the edges in the image if critical assessment is proving difficult.
  12. Once the final adjustment is determined, make sure it is set on the camera and take a “real-world” test shot to confirm the results.

Rinse and repeat for each lens…

If steps 9 and 10 don’t make sense here, they probably will if you are trying it.  For example, having established that each lens was back-focussing, I knew I would only need to adjust with a negative number.  I shot 3 images each at a microadjustment of 0, -5 and -10 and reviewed them on the PC screen.  Logic then pointed to which setting to try next – between 0 and -5, between -5 and -10 or less than -10.  For my 70-200mm lens, -5 wasn’t enough adjustment and -10 was too much so I tried -9 and -8. -8 looked just right at the time, although reviewing the shots later I decided this adjustment was slightly too strong and settled on -7.

Improvements

With hindsight, there are still refinements I’d make to improve this testing even more.

  • Get the scale even flatter on the book
  • Find a more scientific way of ensuring the sensor plane was parallel to the focus target and scale
  • Make sure that the scale is in completely even lighting

Results

All of my main lenses were back-focussing to some extent, even my trusty 24-105mm f4 L IS which I didn’t think I had any problems with.  Although this is only relevant to my camera / lens combinations and probably doesn’t help anyone else, the final adjustments I used were:

  • 70-200mm f4 L IS: -7
  • 70-200mm f4 L IS + 1.4x extender II: -6
  • 24-105mm f4 L IS: -4
  • 100mm f2.8 macro: -11

The negative numbers above make the back-focussing lenses focus more towards the front.

Here are two examples from the 70-200mm, click through to open them up to full size.  Compare the 9 behind the zero to the 9 in front, especially on the embossed version – this shows a slight front focus:

Micro adjustment focus test sample
Test scale at 200mm with an adjustment of -8 (100% crop)
Micro adjustment focus test sample - embossed
Embossed version of the test scale with an adjustment of -8 (100% crop)

And here are some of the real-world test shots.  The keyboard shot was taken on a tripod using the macro lens and the focus is spot-on.  The bar codes were taken hand-held with IS enabled at ISO3200 in burst mode and cropped to show the detail.  I checked at both 200mm and 70mm to make sure the micro adjustment hadn’t caused a problem at the other end of the zoom range.  I didn’t apply any sharpening to these; none of these are supposed to be “perfect” shots, just enough to convince me that the adjustments were OK.

Computer keyboard
"Real-world" test of macro lens adjustment - focus point was on the "a" of Caps Lock

 

Before and after adjustment - 200mm
Before and after adjustment at 200mm: before adjustment is on the left, after on the right

 

Before and after adjustment - 70mm
Before and after adjustment at 70mm: a subtle difference, but before adjustment is on the left, after on the right

Conclusion

I would love to test my lenses even more thoroughly, but I’d rather spend the time taking photos.  Besides, these two posts have been long enough!  Having finished the “technical” testing & calibration and trying some real-world examples, I’m now confident that any focus problems I see from here are purely down to the operator… 🙂