The K20D, K-7, and K-5 support fine adjustment of the autofocus (AF) system per lens. Some other cameras (K10D, K100D) have a (hidden) service menu that allows access to a single AF adjustment.

Here are some hints to consider when attempting to optimise the AF performance with such fine adjustments:

• Put the camera on a tripod to avoid hand-shake.
• Don't shoot a ruler or other objects where the AF sensor has multiple targets to lock on to. AF sensors cover a bit of area and may be slightly off-centre. Get a focus chart or another target which helps you to know where the camera focused on.
• Make sure that the AF sensor you are using (typically the centre one) only "sees" the target. Move it around the target initially, making sure that the camera can not obtain focus. If still obtains focus then you either have not sufficiently pointed away from the target yet, or the AF sensor "sees" parts of the measurement print that it should not see.
• Don't shoot test targets under Tungsten light if you want to fine-tune AF for daylight shooting. Tungsten light will cause the AF system to slightly front focus. This holds for every Pentax DSLR using a SAFOX AF module without a "+" in its designation, i.e., in particular for every DSLR before the K-7.
• Try to shoot a flat rather than a tilted target. The Pentax service manual for the K10D shows a setup with a flat target and a tilted ruler. Here's a corresponding image including instructions (courtesy by rawr).
• Do a series of tests with the lens starting from infinity and its minimum focus distance respectively a number of times. Results will not be a 100% consistent with each other across the series but you'll see a trend. Doing a single shot only may let you end up on a non-optimal setting. This is also true for calibrating a focusing screen.
• Doing a series of shots with focus bracketing may yield a more precise result than the method that tries to place the centre of focus on a "zero line" of some ruler.
• Try to read the focus point near to the point where you focused. If the ruler is way off the focus point used, you may optimise for a particular field curvature of your lens.
• The distribution of the DOF before and beyond the focus point approaches 1:1 the closer you get to the subject. The often quoted 1:3 ratio only applies for a subject at a third of the hyperfocal distance.
• Make the test using a focal distance you expect to use in real world shooting. For different focal distances, the AF adjustment may be slightly different. A rule of thumb for a general adjustment is that the distance should be roughly 25-50 times the focal length of the lens. Canon recommends the factor 50 but most lenses have quite a deep DOF at this distance; not much use for fine-tuning. If you get it right at 25 times the focal length (Lens Align recommendation) then the 50 times focal length distance should be fine as well.
• The widest possible aperture (lowest f-ratio) will make it the easiest to check where the focus is, but in particular very fast lenses (lowest f-ratio <1.8) typically introduce a slight focus shift with aperture change. So again, you may want to calibrate, say a useful f/2.8 rather than an extreme f/1.4. This way you can achieve a compromise that works for a wide range of aperture settings rather than just a narrow range.
• Evaluate real world results and let these govern your adjustments, rather than relying on test setups.

Note that it may not be possible to find one adjustment value that will work equally well for all focal lengths of a zoom. You have a number of choices for calibrating a zoom:

• use the longest focal length (recommendation by Canon)
• use the focal length you use most
• use the focal length where sharpness accuracy will be most critical
• choose a compromise setting that gives you the best comprise for all focal lengths

The focus chart by Tim Jackson assumes that tilted targets are not problematic. I don't like that its focus target is ambiguous, i.e. a rather wide line (bar) in the middle.

The focus chart by Yvon Bourque also uses a tilted target. It nicely avoids unwanted focus targets and has a precise point (line) of focus, but I'm not sure how well it works if the lens requires a different focus for its centre vs its perimeter.

Jeffrey Friedl's chart allows read out of focus near the target but one has to make sure to use a low enough grey level version so that the AF does not pick up the read out pattern. Also, the AF target is not as unambiguous as it could be.

The chart featured in Bob Atkin's Focus Testing article seems to be a good compromise.

If your camera supports Live View you can also try this chart which is based on provoking moiré patterns. It avoids any potential issues that tilted targets may have. However, make sure that the AF module isn't fooled by the colour temperature of the screen you are using. Validate your results using natural light. You may also use Live View in order to obtain accurate Live View focus first and then check whether refocussing with phase AF causes any lens movements. Of course the latter approach will work with any target, but take caution to nevertheless obey all of the aforementioned caveats.

A commercial solution is the LensAlign tool. It uses a flat target. Not sure how far away the ruler is from the centre area which is used to obtain focus. The LensAlign website features a "LensAlign User Guide" which you may find helpful and a "LensAlign Distance Tool" which makes a suggestion at what distance you should measure a particular lens (25 x its focal length) and what focus distribution (percentages of the areas that should be in front and behind the target point for a given distance) you should expect. The "Mk II" version of the LensAlign tool seems to be made of relatively flexible material (to support disassembly into a small flat parcel for transport). A user reports "They look ok, but there could well be some mis-alignment in all of these parts. I would be shocked if there wasn't.". If they user is right then such mis-alignment could have an impact on the precision obtainable with this particular (cheaper) version of the LensAlign tool.

There is a less expensive version called Spyder LensCal. It lacks the alignment aids of the LensAlign tool, though.

You may build your own LensAlign tool, by following the instructions for how to build a GhettoCAL.

P.S.: If you would like give me reputation points for compiling this article, please use the corresponding button at the top right of this post or a corresponding button in any of my posts below.

[jimH]

How do you access the "hidden" service menu on a K10?
Thanks

jimH

Edit: Firmware 1.30

[photog]

Thank you for this explination. Does anyone know how large each division is?
Ron McDermott

[Class A]

How do you access the "hidden" service menu on a K10?

Please have a look at this K10D AF Adjust thread. With your firmware version (1.30) you'll need software. Prior to that one could use a key combination used after start up. There are also ways to downgrade the firmware, involving a modified firmware which has a patched release date.

---

Last edited by Class A; 08-07-2009 at 11:00 PM.

[Class A]

Does anyone know how large each division is?

I'm not aware of a corresponding diagram.

EDIT: I found this diagram showing the AF areas.

In any event, you can perform the "black dot" test: Choose some uniform background and place a rather small object in front of it. Focus on that small object. Switch to MF or use some other method of avoiding refocusing and move the camera so that the small subject moves around the focus area that interests you. If metering is activated (tap on the shutter button to activate it) the camera will tell you with the green hexagon in the viewfinder when the selected AF sensor sees something in focus. By approaching the subject from all directions, you can find out how large it is and where it is placed.

[photog]

I have an addition to my recent post. Testing with a number of lenses fron my Sigma 17-70 to a couple of old F lenses, I discovered by experimentally that I actually needed -10 (full scale correction)! The magnification facility was very helpful as checking on computer from test to test becomes extremely tedious.
What sort of corrections have other members found necessary.
Ron McDermott

[opiedog]

ron,

just curious, at 17mm, does your sigma 17-70 shift focus when changing from f2.8 to say f7.1?


thanks!

jordan

[photog]

I did not specifivally test for effect of aperture changes. But I sincerely hope not! I will make tests as opportunity offers.
Ron McDermott

[Arpe]

So if one's lens is Back Focusing, does one apply a negative correction in one's camera, or positive.

[Class A]

So if one's lens is Back Focusing, does one apply a negative correction in one's camera, or positive.

The adjustment value should be positive.

[jgredline]

The adjustment value should be positive.

Yep, agreed!

[mrdfred]

I have used the debug mode on my K10d after downloading the English version of the usb program to turn the debug menue on. I was worriied about using it as I didn't want to permenantly screw up my camera. Prior to using it, I tested all my lenses, DA 18-55, DA*16-50, DA 50-200, Kiron 100 f2/8 macro, at max zoom and max apprature. I tripod mounted the camera and used a shutter release to take the test shots. I moved the camera to 10X the focal length of the lens and shot 3 times at the same focal length and apprature at different targets, both tilted and flat. I took 3 shots of each target with each lens, both tilted and flat. All my lenses were back focusing at about the same amount. I found that the DA 18-50 kit lens was completely out of focus at wide open and at the widest zoom, I'm suprised the camera even achieved a focus lock. I used the focus targets from Tim Jackson, Focus Test Chart V2.00, (both tilted and flat), Jeffery's Test Chart V1.0, (both tilted and flat), and a BAR code label from my cell phone box, multiple verticle lines of varieing thickness, numbers, and text of varieing sizes on very white paper. The tilted targets showed me the direction (Back or Front focusing) my lenses were off. The flat targets were used to make the correction once I accessed the Debug menu. The Bar code target proved the most helpful in fine tuneing the adjustment. l also checked if I could return my camera to the default settings, and was pleased to find that the debug adjustments can be undone. Adjust to the right to bring the focus point foward and to the left to move it back.

[philbaum]

To answer the OP's question, one of our club members brought in his LensAlign equipment and let some of us try it out.

I first used it on my DA-300 and ran into the problem of getting too close to the minimum focus distance of the lens where results got squirrely, backed up and it seemed to work fine. Then took out my DA 50-135, which has been a reliable focuser for me, and results were all over the place. Now looking back on it, i didn't understand the need not to focus in the center of the flat target and not near the slanted ruler, so i may have well picked it up in the focusing some of the time. The guy who owned it didn't explain where one needed to place the focus.

Short Story of micro adjustment in the field: I do a lot of camera work for a local playhouse. I knew that with Tungsten light, the focus shift is towards front focusing. So at the beginning of a recent session, i noticed some difficulty getting sharp focus, looked at the shots and noticed that the farther one got back on the face, the less focused it was, dialed in a negative 5 micro-inches, and it worked better. The point is - once you get used to how to adjust micro-adjusting, its possible to adjust one lens for use indoors on the fly using real shooting indicators - i think :-)

Here's another idea for a DIY focus setup. Take a tripod with a head and set it up 15 feet away or whatever distance one wants to use. Focus on the tripod head using the AF button with half-shutter focusing disabled. Take a metal ruler like the one suggested above, drill a 1/4" hole through it, install a QR plate on it using a 1/4" nut, and then mount it on the same tripod head. take a pic of the ruler and micro adjust as needed. remove the QR plate and ruler, focus again on the tripod head only, reinstall the slanted ruler and check the previous adjustment. Adjust as needed, etc.

A little cumbersome but it eliminates any possibility that the AF will pick some other item to focus on.

Thanks for this discussion, VERY helpful

[Class A]

Hi Phil,

I'm glad you found this discussion helpful.

Now looking back on it, i didn't understand the need not to focus in the center of the flat target and not near the slanted ruler, so i may have well picked it up in the focusing some of the time.

It is absolutely vital that you don't focus on the ruler, but on the flat target. Prior to that you should align the target with the aiming aids. If you focused on the ruler, it is no wonder that your results where all over the place. You could use "PhotoME" to check where you focus point was for a certain image (unless you used the focus-recompose method).

Here's another idea for a DIY focus setup. ... Focus on the tripod head using the AF button with half-shutter focusing disabled.

The good thing about your setup is that you measure at the same place where you focused. Most other methods (including LensAlign) assume that focus is the same for the target (centre) and measurement area (typically towards the image edge). I think this suboptimal as most lenses do not have a completely flat field of focus. Not sure about the magnitude of the error but why not eliminate it by putting the ruler exactly where one focused?
With your proposed solution, I would just suggest that you provide something less ambiguous than a tripod head which has too many features the AF could latch on. Using a flat target card on top of the tripod head should work.

[Class A]

You may build your own LensAlign tool, by following the instructions for how to build a GhettoCAL.