The Online Photographer

Check out our new site at!

Friday, January 05, 2007

Pentax K10D/21mmDA Autofocus

by Carl Weese

[Editorial Insertion:
Carl's Pentax K10D Thought Process to Date:
Part I: "On Buying a Camera"
Part II: "Less is More"
Part III: "On Testing a New (Digital) Camera, Steps A and B"
Part IV: "Weese on Anti-Shake"
Part V: "Anti-Shake Pt. II, or, Shake Reduction as Virtual Tripod for Nature Work"
End Editorial Insertion]

Two things to talk about today.

First, a clarification about what I’ve been up to here, which is not to do an in-depth product review in the mode of the several online sites, or ink-on-paper magazines, which do that sort of thing well. Nor to recommend any particular products to TOP's readers. The idea here is that some of TOP’s readers might enjoy following the thought process an experienced photographer goes through to choose and then learn to work with a new set of tools. There's no implication that what I chose will be right for anyone else, but rather that the process might help you find the right answer for your needs.

Second, a problem report with the equipment I’ve been talking about. Several readers who also have the Pentax K10D/21mm DA lens combination have written to ask whether I’ve encountered problems with the AF system. Yes, I have. I got in touch with the Pentax folks to tell them about the problems. I sent them JPEG files demonstrating the errors I was seeing with the lens I bought on the open market, and they sent me a fresh-from-the-warehouse second copy of the 21 to test. It worked exactly the same.

For years now I've been saying my wonderful 1989 Honda CRX would make it to 180,000 miles. It did, last week, and started coming apart at the seams. Mainly from being 17.5 years old rather than the miles. But it became hopeless and I had to get a new car. I lucked out and found a very hard-to-get Honda FIT (someone got tired of waiting and cancelled their order). It is a subject that might confuse an AF system if the sensor hits the plain metal instead of the wipers or other solid target.

Here is the problem. All AF lens/camera systems encounter situations—“focus problems” in the sense your math teachers used the word “problem”—that they can’t handle. Subjects with no detail or contrast, like the sky or the sky reflected in the paint on a shiny car, or a screen of winter shrubbery in front of a brick wall. In the first case an AF system can’t do anything at all, in the second, it can’t figure out whether you want the sticks or the wall, and depending on the distance, it may not be able to distinguish the sticks, period. In that case you get the wall, even if you wanted the sticks to be in best focus.

What I’ve found with the 21DA combined with the K10D camera—both quite new introductions—is that as long as the AF mechanism gets its act together the focus is correct. None of the back-focus/front-focus problems that afflict some samples of all lens/camera systems. When it’s right, it’s perfect, and that's with both samples of the lens on my camera.

What’s wrong is that simple focus problems throw the system for a loop. A six inch diameter tree trunk six or seven feet from the camera, placed squarely on the central focus spot, just isn’t seen by the AF system, which instead focuses well behind the target. Another problem, pointed out in an email from a reader who is working with the same equipment, is that at around 15 or 20 feet, the AF may miss the subject and jump not just to infinity, but actually past to the limit of the lens’s mechanical travel. I hadn’t found this before getting that email (thanks for the tip, Jim), but it did occur the next day when I used the K10D/21DA combination in very dim light. I’d like to test the lens on another compatible camera and I may do that soon, but I don’t have a different Pentax DSLR model handy at the moment.

Because I see the problems with both copies of the 21 and have reports of similar behavior from other peoples' camera/lens combinations, I don't think this is a sample defect issue. The 21 DA is too good a lens to give up on, so I’m continuing to use it, though playing it safe and using it only in manual focus. I’m neither a lens designer nor a software engineer, but my impression as a user of this gear is that what I’m seeing is a communication/compatibility issue that might be fixable, possibly even with a firmware update. I hope so. Maybe I’ll hear something about it from the elves in Golden, Colorado, as soon as they dig out from their second blizzard in less than a week.

Next time, we'll go on an Easter Egg hunt.

Posted by: CARL WEESE

Question from ShadZee: Are you having issues with your other lenses? I'm considering buying the 43mm Limited (and maybe the 31mm limited as well). I really hope the K10 would work flawlessly with these two.

Carl Responds: My 40mm (and a 70mm pancake Pentax sent me for testing) both work exactly as the AF should. No front/back issues, focus locks on quick and secure with the same sorts of subjects that fool the system with the 21 mounted.


Blogger Neils said...

But does the 40mm, you got one of those also right, does the 40mm lens focus accurately?

I'm a Nikon DSLR and a long long time Leica M shooter. This little Pentax still intrigues me.

5:19 PM  
Blogger ShadZee said...

Are you having issues with your other lenses? I'm considering buying the 43mm Limited (and maybe the 31mm limited as well). I really hope the K10 would work flawlessly with these two.

5:21 PM  
Blogger PatrickPerez said...

I was afraid I was going crazy(er). I thought you were describing a photo of your CRX, and the lens distorted it's shape. Glad to see it was a Honda Fit.


6:37 PM  
Blogger Ira Crummey said...

Just send me the 21mm and I'll check it out for you. It shouldn't take more than 8 or 10 months to be sure.

10:54 PM  
Blogger promenadeur said...

if you can read in german, read my post in The Public Eye Blog. If not, here are some first translations in English: "The pixel arrangement of the AF sensor is actually up to three times larger (linear extension), than the marking in the finder indicates us. A selected sensor has thus a clearly larger "field of view", than the appropriate AF marking in the finder shows. If a detail with higher contrast than lies in the direct surrounding field of the AF marks, the sensor will probably focus on it. Not on the face outline of our "victim", but on the higher-contrast structures of a wall, shrubs directly beside it for example.

Could that be a answer?

2:12 AM  
Blogger MartinC said...

A year or two before I bought my first DSLR I bought a Pentax MZS with one of the limited series lenses. I bought the body first and then sourced the lenses from another retailer on the other side of the country but when trying out the camera in shops in London I noticed immediately that there was something off with the AF - so much so that you could see through the viewfinder that manual focus and AF were not in agreement. I tried two bodies in shop #1, both duff and ended up trying a total of 6 in two or three other retailers before finding one that appeared just about ok. I then bought the 43 limited lens and was dismayed to find that 1) I simply could not get it to focus with anywhere near acceptably accuracy and 2) it would never focus to infinity. I tried one more sample of the lens before abandoning my brief experience with Pentax. My impression at the time was that during the film era these QC (or design) issues went largely unnoticed because the majority of users never scrutinised their results in the the way that people do now with the instant pixel peeping culture associated with the digital age. So it is with some interest and, I have to say no surprise, that I am reading all these reports now of duff AF in the K10 and the principle reason that despite its many obvious attractions and the notional appeal of the lovely limited series lenses, that I feel not a twinge of an impulse to divert from the Oly E system to Pentax.

3:22 AM  
Blogger Supasnapper said...

Sure auto focus is useful, but us olduns' remember the days when the Mk 1 eyeball was all we had to use for focussing, so what's wrong with
using manual? 'Use it or lose it'

4:49 AM  
Blogger Carl Weese said...


My 40mm (and a 70mm pancake Pentax sent me for testing) both work exactly as the AF should. No front/back issues, focus locks on quick and secure with the same sorts of subjects that fool the system with the 21 mounted.

6:42 AM  
Blogger Carl Weese said...


Your info could be related. In an inadvertant "test" of the AF problem, I was shooting strings of exposures, then for reference was going to hold up one finger and shoot, do the next string and hold up two fingers, etc. to track the tests. What happened was the camera couldn't focus on one finger, but snapped to focus instantly when I spread my hand out and gave the AF a broader target. I sent those files off to Pentax as well but have not yet heard back from them yet. One finger at 20 inches is about the same target as the six-inch diameter tree at 6-7 feet...

11:03 AM  
Blogger Michael said...

Carl, Mike,

This series is just the sort of thing I very much enjoy. Please keep it up! One by-product is that I may buy the Pentax, but that is immaterial. The point is looking over the shoulder of someone with wide experience and well honed judgment. Thanks!


11:10 AM  
Blogger Peter said...

Sure auto focus is useful, but us olduns' remember the days when the Mk 1 eyeball was all we had to use for focussing, so what's wrong with
using manual? 'Use it or lose it'

Not to be a wise-a$$ but have you ever tried focusing on the tiny, dim, ultra-smooth focusing screen that is in a reduced frame dSLR? With the better VFs it can be done (barely), but it's pretty tough. One of the reasons that I stuck went with (Konica) Minolta is because they were the one manufacturer who seemed to have a care about this sort of thing. (That plus the in-body AS.) As it is, I wasn't ultimately happy with the KM-5D for regular MF use (should've sprung for the 7D) and wound up replacing the screen with a 3rd party screen with split image w/micro prism collar.

Aside from the 5D there are times when I like to change up my style and pace and then I use a pair of all mechanical Bessa RF's. Essentially they are for those times when I want to manually focus, and do a host of other things manually. Once you learn how to use the RF patch properly it's heaven, I highly doubt any SLR can touch RFs in that department.

2:56 PM  
Blogger Joel said...

Weird! I think I'm seeing this with my K100D and the 21mm as well, though I'd not narrowed down the causes.
What I see is that in certain situations, the AF goes the wrong way - it decides to try focusing out, for example. Then, when it gets to infinity, it gives up. If I manually bring the focus in towards the plane of the object, then re-fire AF (I'm using the AF button, disabling the shutter button AF), it can then lock on.
So from my very limited experience (this has only happened to me twice), if the AF had picked the right direction, it seems like it would have locked on. But it stubbornly wants to go the wrong way (focus-in vs focus-out) and doesn't backtrack. Does this sound like what you see?


6:36 PM  
Blogger Joel said...

To be clear, with my post just above, I've only seen this 21mm focus problem twice, and I've otherwise found the 21mm fast to focus and incredibly spot-on. It's a gorgeous lens.

6:37 PM  
Blogger nextSibling said...

martin storz's suggestion earlier matches my experience with the K100D. The AF sensor seems substantially bigger than the LED patch visible in the viewfinder. The AF system seems to attempt to focus on the highest contrast edge within the sensor area. If that's not what you want, especially if the higher contrast edge is near the periphery of the AF sensor and a lot further away than where you want focus to be, mayhem ensues.

10:17 PM  
Blogger Morven said...

From everything I've read on Pentax forums considering the K10D (it's a possible purchase for me in the medium future), the lenses have no focus calibration and thus, short of horrid misalignment or brokenness, won't be the defective part if you experience focus issues.

With some AF systems, e.g. Canon, it seems that the lens does have some calibration for correct autofocus.

7:26 PM  
Blogger SuperJared said...

Just wanted to say thanks for the info on the 40mm...

and that my 1986 Honda CRX just pushed 211k miles :P

1:46 PM  
Blogger john91 said...

This "autofocus heading the wrong way" problem does not seem to be new. With my PZ1P, some difficult and rare focusing situations caused the AF mechanism to slam to one end or the other (near or far), and not recover. I attributed that to not having a contrasty edge to grab onto, and resorted to manual focus. My K10d seems to do a better job, -- it seems much harder to confuse -- but my experience is anecdotal, not systematic. The problem you describe is consistent with the AF Sensors not seeing the tree (or finger). Has anyone tried one of the other non-averaging AF modes? I don't have a 21mm. I, too, am willing to test your lense for you for a few months.

7:55 AM  
Blogger alan said...

Thanks for the clear descrption of the da21 focusing issue -- that's precisely the problem that I have had with it. My experience is that at short distance it has little trouble, but as soon as the subject gets beyond 7 feet or so, the lense focuses to infinity (or at least well behind the subject) while still indicating proper focus on the subject. When I do get the focus right, however -- especially on mid-distance subjects -- the results are great.

It's really frustrating, but it sounds like you're convinced it's endemic to the lens and I wouldn't have much success in sending it off for troubleshooting.

4:08 PM  

Post a Comment

<< Home