AF Micro-Adjustment tool

[Bart_van_der_Wolf]

Hi folks,

Having acquired my 1Ds3 (wow, what an amazing tool), and having spend some time getting used to the controls and menu layout, I decided to calibrate my lenses using the AF micro-adjustment feature. I can tell you, it makes a huge difference in focus accuracy and repeatability.

I've made a simple tool to assist in the procedure, and you can download it here. It should be displayed on an LCD screen a 100% zoom setting (so at native size).

It works by exploiting the interference patterns or moiré between the R/G/B LCD elements and the camera's LCD elements when directly viewed with Life View. With good optics and perfect focus, the moiré is maximized. Depending on shooting distance, choose the Life View zoom setting (full, 5x 10x) that shows it best. When the situation permits, you will see larger R/G/B aliasing of the LCD elements. You'll also notice how narrow the field of focus is, especially with lenses that react strongly to small focus ring rotations.

This is my first version of the target, therefore I may update it with some other features, also depending on feedback. So let me know if it works for you, or what might help to make it better.

Bart

[Doug Kerr]

Hi, Bart,

That sounds like a very interesting approach.

Originally Posted by Bart_van_der_Wolf

It works by exploiting the interference patterns or moiré between the R/G/B LCD elements and the camera's LCD elements when directly viewed with Life View.

I'm having trouble visualizing the principle here. Are you saying that the moiré phenomenon lets the viewer better discern the sharpness of the image of the "focus target", or what? I'd enjoy being able to grasp a little better what chain of effects you are exploiting here. It sounds fascinating.

Thanks.

[Bart_van_der_Wolf]

Originally Posted by Doug_Kerr

I'm having trouble visualizing the principle here. Are you saying that the moiré phenomenon lets the viewer better discern the sharpness of the image of the "focus target", or what?

Hi Doug,

What happens as one manually adjusts the focus is that at the exact optimal focus setting the background will change from uniform gray into larger (colored) aliased dots when viewed on the camera's LCD. The circles and cross hair will allow to acquire AF easily, and when calibration is optimal, moiré will be maximized on the camera's LCD.

A procedure that works for me:
- I switch to LifeView, which in its current implementation will only allow manual focus.
- Optimize manual focus by searching for maximum aliasing. This will only occur a best focus (and assuming a decent enough lens is used), otherwise the defocus will act as a low-pass filter and prevent the aliasing.
- Switch off LifeView, and watch the lens barrel's focus indicator for the next step.
- Use AutoFocus (single AF spot) to focus on the (laptop) LCD screen, and watch the direction of adjustment. That will show whether the current AF calibration setting will front or back-focus.
- Apply an adjustment via the camera menu, and repeat the procedure. Once the adjustment is optimal, there will be no difference between manual and auto-focus.

The flat computer screen will prevent misinterpretation of the focus distance, because the AF system cannot react to phase effects from subjects at other distances. That makes it quite easy to get repeatable results. And because the computer LCD emits light, it can be easily done indoors, at common shooting distances for the lens to be calibrated for.

For super tele lenses it is probably easier (for distance reasons) to use a Zoneplate type of chart outdoors.

Bart

[Doug Kerr]

Hi, Bart,

Originally Posted by Bart_van_der_Wolf

What happens as one manually adjusts the focus is that at the exact optimal focus setting the background will change from uniform gray into larger (colored) aliased dots when viewed on the camera's LCD. The circles and cross hair will allow to acquire AF easily, and when calibration is optimal, moiré will be maximized on the camera's LCD.

How neat!

. . . otherwise the defocus will act as a low-pass filter and prevent the aliasing.

Of course! What the moiré does is allow us to visualize the higher-frequency components in the image (just as many so-called "contrast detection" AF systems actually do), and of course they increase as the "low pass" effect of the defocus spread function declines.

This of course makes perfect sense as moiré is a visible manifestation of aliasing, which occurs from high-frequency components beyond the Nyquist limit for the system. Less HF content - less moiré.

The flat computer screen will prevent misinterpretation of the focus distance, because the AF system cannot react to phase effects from subjects at other distances. That makes it quite easy to get repeatable results. And because the computer LCD emits light, it can be easily done indoors, at common shooting distances for the lens to be calibrated for.

Very clever!

Brilliant, my friend.

Best regards,

Doug

[Ben Rubinstein]

Sounds incredible, all I need now is for my 5D's to have had this feature! :-)

[Nicolas Claris]

ABSOLUTELY brilliant Bart!

I sure will give it a try tomorrow…
For zoom lens like the Canon 24-70 what lenght would yoy choose for calibrating? 50 mm?

[Bart_van_der_Wolf]

Originally Posted by Nicolas Claris

ABSOLUTELY brilliant Bart!

I sure will give it a try tomorrow…
For zoom lens like the Canon 24-70 what lenght would yoy choose for calibrating? 50 mm?

Canon suggests to use the longest focal length of the zoom range, 70mm in your example. Of course you can check afterwards how well that works for the shorter ones, but in my experience they are harder to calibrate because the DOF is already a lot more. The DOF of the shorter focal lengths will probably exceed the adjustment range anyway.

Afterall, what the AF micro adjustment does is add an offset to the lens calibration data to change the average focus position within the DOF limits of the widest aperture. That is done more easily at longer focal lengths.

Bart

[Nicolas Claris]

Thanks Bart
I'll post my findings tomorrow…

BTW, nice Santa this year, hé!

[Bart_van_der_Wolf]

Originally Posted by Nicolas Claris

BTW, nice Santa this year, hé!

Yes, Santa Mark III !

Bart

[Nicolas Claris]

Once you realize that the tests should be run wide open, it becomes really easy…

So, thanks Bart (and Canon…)

My Sigma 12-24 comes back to life… micro adjustement settings = +15 !!! (and I thought it was a good copy!)
Canon 24-70 = +3
Canon 70-200 ƒ2.8 = 0
haven't check the 500 yet… nor the 6mm

[Bart_van_der_Wolf]

Originally Posted by Nicolas Claris

Once you realize that the tests should be run wide open, it becomes really easy…

Yes, sorry for not making that more clear. LifeView should be used with aperture wide open (most shallow DOF). So, how did you like using the test target? Was it easy enough to see the moiré? Should I add some detail to allow inspection of an actual shot, although there's no real need to shoot anything?

So, thanks Bart (and Canon…)

My Sigma 12-24 comes back to life… micro adjustement settings = +15 !!! (and I thought it was a good copy!)

It doesn't mean the lens is bad or anything, just that the combined calibration of body and lens produces consistent front or back-focused results, which can now be self-adjusted on some models.

I know what you mean with coming back to life, my 200mm f/2.8 required +19, the other lenses much less. From average to pin sharp, without the need to inspect focus bracketed sequences or even take a single shot, what a difference. I've found the adjustment in actual shooting conditions to bring about a much better average focus accuracy. There can still be slight differences due to subject contrast on 3D subjects, but the average is better.

It can also reveal a bit of a desillusion on some zoom lenses, pinsharp on one end of the range, and soft/defocused on the other. An alternative would be to use a sort of average adjustment, or just optimize at the long end, and MF on the short end. I'm not sure if CPS lens calibration will affect differences between short and long focal lengths.

Well, it saved me from having to let CPS calibrate the lot.

Bart

[Nicolas Claris]

Originally Posted by Bart_van_der_Wolf

So, how did you like using the test target? Was it easy enough to see the moiré? Should I add some detail to allow inspection of an actual shot, although there's no real need to shoot anything?

Hi Bart
It was not so easy at first to determine the 'optimum' moiré on the LCD, but once you've found, the rest is very easy.
I combined the results while shooting with different adjustment settings and open unsharp raw at 100% in CS3.
They did confirm with the same results than your method does :-)

It is very important to do the tests wide open otherwise you get very unconstent results…

[Bart_van_der_Wolf]

Originally Posted by Nicolas Claris

Hi Bart
It was not so easy at first to determine the 'optimum' moiré on the LCD, but once you've found, the rest is very easy.
I combined the results while shooting with different adjustment settings and open unsharp raw at 100% in CS3.
They did confirm with the same results than your method does :-)

It is very important to do the tests wide open otherwise you get very unconstent results…

Okay, thanks for the feedback. I've noticed that some LCDs are much easier to use for this purpose than others. My LaCie 321 turns into in colored aliasing (because it is very sharp), whereas my laptop shows more modest monochrome dots.

I did mention that the optimal focus plane is very thin, so it requires a slow rotation of the focus ring on the lens. I'll think about it to see if things can be made even more sensitive.

Something that can help is using a different distance. I've built in 2 levels of detail, to accomodate for different distances, maybe a third is needed. Let's see how others react.

Bart

[Nicolas Claris]

Well, my NEC 2180UX is quite Sharp but the range length (specially with the 24-70) is quite important from might moiré to vibrating moiré.
Yes distance was important but the X10 view did help a lot for accuracy, making the center ring sharp… while still keeping in the moiré range (not sure I'm explaining clearly!)

[Herbert Dorfan]

Hi Bart,
Thank you for the benefit of your elegant thinking. Two qustions that come from my inexperience: Is it important that plane of the digital sensor of my 1dMk3 be parallel and centered with my lcd monitor, as I am able to make things work with my <200mm lenses, but not with my 300mm f2,8 where I don't see colors? (I have lens opened to 2.8). Do you have any suggestions how to accomplish this, if it is necessary?
Herb

[Bart_van_der_Wolf]

Originally Posted by Herbert Dorfan

Hi Bart,
Thank you for the benefit of your elegant thinking. Two qustions that come from my inexperience: Is it important that plane of the digital sensor of my 1dMk3 be parallel and centered with my lcd monitor, as I am able to make things work with my <200mm lenses, but not with my 300mm f2,8 where I don't see colors? (I have lens opened to 2.8). Do you have any suggestions how to accomplish this, if it is necessary?
Herb

Hi Herb,

It is not absolutely necessary to have the target and the sensor plane parallel. If they aren't then you will probably see a zone of moiré parallel to the in-focus plane. However, it is beneficial for repeatable focus to do have things approximately parallel.

Reasons for not achieving the moiré can be;
- too far out of focus, but that's also what the crosshairs are for, a visual clue and as well as an AF sensor aid
- distance too close, or magnification to large (which is the same), which prevents small enough detail to cause moiré
- distance much too far, which will cause the moiré to have too low an amplitude to see
- the lens has too low resolution
- or the manual focus is too reactive to nail the exact spot, try slower manual focus adjustments

You may have to vary the shooting distance a bit with your 300 mm. It should be plenty sharp enough to achieve moiré (color or monochrome), but requires shooting from a bit further away. It can also help to not use the maximum Life View zoom. Sometimes x5 is good enough or even better to see moiré clearly, try toggling the different magnifictions at some different distances.

I am actually working on an improved target version right now, one that's less distance critical. So your feedback on the final condition that you needed to see the moiré can help me a lot (I don't have the 300mm to test myself).

You might want to try this target which has a little resemblance to the direction I'm thinking of for the new target. The moiré from that target will manifest itself as concentric smaller rings.

Bart

[Herbert Dorfan]

Thanks for you response, Bart,
I set up at a distance of 3.5 meters. The moire lines seem more predictably obtainable on the original target (with cross hairs) than with the concentric rings. In the later case the concentric rings resulting from moire seeem to dance over the screen and are, to me, less obvious.
Herb

[Bart_van_der_Wolf]

Originally Posted by Herbert Dorfan

I set up at a distance of 3.5 meters.

Was that for your 300mm, or for the shorter lenses?

The moire lines seem more predictably obtainable on the original target (with cross hairs) than with the concentric rings.

Yes, I did try to make the chance and amplitude of large area moiré as nasty as possible in the original.

In the later case the concentric rings resulting from moire seeem to dance over the screen and are, to me, less obvious.

Okay Herb, thanks for the feedback. It helps me to concentrate on the variations of the basic concept that are more likely to succeed.

Bart

[Herbert Dorfan]

Hi Bart,
The 3.5 meter setup distance was used for my 300mm lens, only.
Herb

[Bart_van_der_Wolf]

Originally Posted by Herbert Dorfan

Hi Bart,
The 3.5 meter setup distance was used for my 300mm lens, only.
Herb

Good to know you managed to get it done. The distance can be helpful for others to use as a startingpoint. Of course things will need to be verified at actual shooting distances, initially coming from infinity focus and from close focus, but I expect the adjustment you found this way to be very close.

Bart

[Wil Andrews]

Here's some very good news. This test with a 1MIII with the focus fix from Canon shows tack sharp auto focus with the three lenses I have with me (16-35L, 24-105L, and 70-200L). Earlier tests with other means showed some adjustment needed before the camera was sent in for repair.

Conduction the test as described produced such fantastic results that I felt a more rigorous test should be done. I devised a two step test:

Step 1: Do exactly what has been described here.

Step 2: Do not move camera.

2.1: While not in Live View, but in Auto Focus mode, manually rotate focus ring to de-focus the lens.

2.3: Press shutter half way down to actuate auto focus.

2.4 Without moving anything, enter Live View.

If the moire lines are still there, the auto focus returned the lens to tack sharp focus.

This part of the test should only be done after the user is satisfied with any adjustments made.

I added this test to see if the lens motor and or auto focus circuit has any hysteresis (slop in returning to same point). I am delighted to report that all works fine.

The only question I have is that Canon recomends that the tests be made at a "distance similar to actual subject matter you shoot . . ." But the argument against that is that Canon doesn't and can't do it that way themselves in their service center.

Time will tell how all this works out.

Wil

[Bart_van_der_Wolf]

Originally Posted by Wil Andrews

Here's some very good news. This test with a 1MIII with the focus fix from Canon shows tack sharp auto focus with the three lenses I have with me (16-35L, 24-105L, and 70-200L). Earlier tests with other means showed some adjustment needed before the camera was sent in for repair.

Conduction the test as described produced such fantastic results ...

Hi Wil, that's great news. I can imagine the anxiety that was caused by the stupid handling of the AF crisis by Canon. And they also missed an opportunity to turn it into a great comeback. Anyway, I'm glad my test target helped you to verify that things are now the way they should have been from the start.

... that I felt a more rigorous test should be done. I devised a two step test:

Step 1: Do exactly what has been described here.

Step 2: Do not move camera.

2.1: While not in Live View, but in Auto Focus mode, manually rotate focus ring to de-focus the lens.

2.3: Press shutter half way down to actuate auto focus.

2.4 Without moving anything, enter Live View.

If the moire lines are still there, the auto focus returned the lens to tack sharp focus.

This part of the test should only be done after the user is satisfied with any adjustments made.

Correct. I've noticed some irregular AF behavior when the microadjustment is not done yet, but much less (if any) when things are perfectly adjusted.

I added this test to see if the lens motor and or auto focus circuit has any hysteresis (slop in returning to same point). I am delighted to report that all works fine.

Yes, it's a good test to let the AF approach focus from infinity and from close focus positions, and verify accurate focus with Life View.

The only question I have is that Canon recomends that the tests be made at a "distance similar to actual subject matter you shoot . . ." But the argument against that is that Canon doesn't and can't do it that way themselves in their service center.

Well, the recommendation is kind of obvious. One shouldn't test at the closest focusing distance and assume everything is fine, when infinity focus is more important for ones shooting. However, I think there is not too much difference between medium distance calibration and using the lens at other distances. Exceptions may be the super-tele lenses and Macro lenses, because their optimal performance is probably at the extreme distances they were designed for. The great thing about the AF microadjustment target is that one is required to adjust the testing distance to something that allows the moiré to be generated, and that automatically covers most lenses, with the possible exception of Macros (although my 100mm seems fine at all distances). Macros can still be tested by looking at how well they focus on the LCD's sub-pixels.

Bart

[Andy Bennetts]

Bart - thanks for an ideal quick method to apply the microadjustment. Works a treat (well almost - see later) and I've set up my 70-200 f2.8 at +13 and my 24-70 f2.8 at +20.
This last lens still shows a very slight focus shift even at +20, and the lens is brand new. It doesn't seem to affect the pictures, but to be fair, I haven't had much experience with the lens yet. I'm surprised Canon's quality control has resulted it a lens that it at least in pure technical terms, a bit "off".
Any comments from anyone - should the lens go back to Canon to be set up properly?

[Bart_van_der_Wolf]

Originally Posted by Andy Bennetts

Bart - thanks for an ideal quick method to apply the microadjustment. Works a treat (well almost - see later) and I've set up my 70-200 f2.8 at +13 and my 24-70 f2.8 at +20.
This last lens still shows a very slight focus shift even at +20, and the lens is brand new. It doesn't seem to affect the pictures, but to be fair, I haven't had much experience with the lens yet. I'm surprised Canon's quality control has resulted it a lens that it at least in pure technical terms, a bit "off".
Any comments from anyone - should the lens go back to Canon to be set up properly?

Hi Andy,

I also have one lens with an 'extreme' AF-microadjustment setting (+19 if I remember correctly). What it means is that for the specific body/lens combination the tolerances are perhaps a bit skewed to different directions, which results in a cumulative adjustment that's that far off to one direction.

The trouble with many zoom lenses however is that the short and the long end require different amounts of correction. It therefore depends on how you use that lens, whether it makes sense to send it in for re-calibration. When you usually shoot wide open, things are much more critical than when you often use average apertures. When you shoot mainly at the longest focal length, things are again more critical due to less DOF.

In general, if most lenses are close to zero (say +/- 10) and one lens is much more out of range (say + or - 20), and if it is an issue because it's not enough, and due to your type of shooting you need to get it better, then try sending it in and see if it helps. When all lenses are skewed to one extreme side, then the body might need some recalibration as well and it is best to send in both/all, but only if you can't fix it within the +/- 20 adjustment range. As long as it can be adjusted with the AF-microadjustment, there's no real issue.

The only question I personally have is; does zoomlens re-calibration allow to get rid of the differences between the short and long end, as that's the most anoying thing left? It's maybe too soon, but it would be helpful if someone (with AF-microadjustment settings before and after) could answer that from experience.

Bart

[Steve Febbraro]

Originally Posted by Bart_van_der_Wolf

- Use AutoFocus (single AF spot) to focus on the (laptop) LCD screen, and watch the direction of adjustment. That will show whether the current AF calibration setting will front or back-focus.

So glad I found this forum, and the information on the micro-adust LCD chart and procedure.

I do have question however..When watching the direction of adjustment, looking down at the lens, does a leftward move mean the lens is backfocusing?

Thanks.

[Bart_van_der_Wolf]

Originally Posted by Steve Febbraro

So glad I found this forum, and the information on the micro-adust LCD chart and procedure.

I do have question however..When watching the direction of adjustment, looking down at the lens, does a leftward move mean the lens is backfocusing?

Hi Steve, welcome to OPF.

When you first let the camera autofocus, and then find that you need to focus the lens further away (rotate the lens anti-clockwise as seen from behind the camera) to achieve maximum moiré it means that the camera was front-focusing.

To get a sense of how much you need to adjust, I've used 2 small strips from a post-it note when I didn't have any other barrel markings I could use, and it made things easier. Try and stick one on the rotating focus ring part and one on the fixed part of the barrel, the pieces of paper just touching each other. After you use the AF Microadjustment tool to get perfect focus, draw a thin matching line on both pieces of paper. When you now use AF you'll see how far the rotation is off. Now iteratively change the adjustment setting until AF produces matching lines again. That works quite fast and all you are left with is some tests to see if the AF is consistent, independent of the focus approach direction (starting from closeup or infinity), and at other distances.

Bart

[Steve Febbraro]

Originally Posted by Bart_van_der_Wolf

Hi Steve, welcome to OPF.

When you first let the camera autofocus, and then find that you need to focus the lens further away (rotate the lens anti-clockwise as seen from behind the camera) to achieve maximum moiré it means that the camera was front-focusing.

To get a sense of how much you need to adjust, I've used 2 small strips from a post-it note when I didn't have any other barrel markings I could use, and it made things easier. Try and stick one on the rotating focus ring part and one on the fixed part of the barrel, the pieces of paper just touching each other. After you use the AF Microadjustment tool to get perfect focus, draw a thin matching line on both pieces of paper. When you now use AF you'll see how far the rotation is off. Now iteratively change the adjustment setting until AF produces matching lines again. That works quite fast and all you are left with is some tests to see if the AF is consistent, independent of the focus approach direction (starting from closeup or infinity), and at other distances.

Bart

Excellent.

Thank you Bart for the explanation and for the welcome. I will apply your instructions to the process.

[Steve Febbraro]

Bart - Thank you.

Using your method I adjusted all my lenses in about 1 1/2 hours, which is a relatively short time considering the amount of lenses.

For my zooms, I used the focal length I use most, as a guide.

My results are below. These are all Canon lenses.

35mm f/1.4 - +15 (This was the most adjustment of any lens. I went back to make sure and the settings remained the same).
135mm f/2 - -3
70-200mm f/2.8 - 0
24-70mm f/2.8 - +10
24-105mm f/4.0 - -3
17-40mm f/4.0 - 0
100-400mm +3
300mm f/2.8 - +3

Using the LCD was much easier for me, as my eyes have difficulty picking up the fine nuances of front and back focusing. Using the LCD really made the job so much easier.

Thanks so much for sharing your work.

[Bart_van_der_Wolf]

Originally Posted by Steve Febbraro

Using the LCD was much easier for me, as my eyes have difficulty picking up the fine nuances of front and back focusing. Using the LCD really made the job so much easier.

Thanks so much for sharing your work.

You're welcome. It is indeed not all that easy to pinpoint the optimal focus setting when shooting a test chart e.g. at a 45 degree angle, including the fact that the DOF is approx. 1/3rd in front and 2/3rd behind the focus plane. Also, angled test charts leave room for variation.

By using moiré to our advantage (for a change), it has become straightforward to find an extremely fine region of optimal focus, no guessing involved, and no need to shoot images and load them on a computer, and shoot some more, and ..., etc.
As you have found, not only is it accurate without room for doubt (moiré is an optical interference that's maximized by sharp lenses and optimal focus), it also saves a lot of time.

I like sharing stuff like that, because it will hopefully stimulate others to share as well. We can only benefit from joining forces.

Bart

[John_Nevill]

Bart,

I'll add my name to the thankyou list, sterling work.

I've since set-up two modified 1DMkIII's with an array of lenses upto and including the EF600 with a 1.4TC, your chart does an excellent job.

I just wish they were my 1DMkIII's, they seem to be much better at low light AF than my 1DNs.