[rkt]

And here is one with the Tamron at 4.5 too... just to make an apples to apples comparison...

if you see the text below the central focus marker that says "Focus test chart v2.00" ... it is more "in focus" in the Tamron snap vs. the Pentax ... which would probably be the expected result if the Tamron was front focusing ... the Pentax focus seems better centered around the central line ...

Can it be now concluded that this lens is front focusing?

---

Last edited by rkt; 08-23-2010 at 12:44 PM.

[Marc Sabatella]

Sorry, I don't mean to be a @#$%^ about this, but I'm wondering - did you really shoot these vertically like that, or are these crops?

If they are crops, then it *might* be the case that you were shooting from too far away, and the "Focus test chart V2.00" might have been in range of the central AF sensor, thus stealing the attention of the camera's AF system and invalidating the test. It's important to shoot from close enough that only the focus target is in range of the sensor. Meaning nothing else should be within, or even close to, the center brackets etched into the viewfinder, since the AF sensor is about that big, maybe a bit bigger.

If you did shoot these vertical and that's the whole picture, then I think those shots probably do the trick, although I have to admit I've never thought through what effect shooting vertically rather than horizontally might have. Assuming it was done on tripod, I don't think it should invalidate the results. But handheld tests at small apertures are also always a little suspect, as just moving a couple of centimeters between the time focus is achieved and when the picture is taken totally throws off the results at wide apertures.

I like that you tested both lenses at f/4.5, to see if DOF was simply masking any camera focus error when you tested the 18-250. And I suspect your analysis is correct - the fact that the "Focus test chart V2.00" is more in focus with the 17-50 than the 18-250 is good evidence that the 17-50 is front focusing. I do still wonder, though, if it gets better as you repeat the test, as it still seems to me that "many" if not "most" of the things that would lead a lens (as opposed to a camera) to FF would also cause it get better with each successive attempt. Also you might want to see what happens if you manually focus correctly.

Again, I don't say all this to be a @#$%. I say it because I've been there, done that - made every possible testing mistake in the book chasing down a phantom problem that turned out to be my own error all along. Wasted a lot of time on the effort, sent a perfectly functioning camera in for repair, returned two other cameras I bought to replace the first, etc. It took quite a while for me to understand all the ways I was screwing up my tests, but once I sorted it all out and learned to perform tests in a more controlled manner, the problems I had thought I saw all went away. So I know well how easy it is to be misled.

I'm not saying *all* reported problems turn out to be phantoms, but when I see people possibly repeating the same mistakes I know all too well from my own experience, I do like to try to help sort through them.

[rkt]

Any feedback is good feedback for me... that's why I decided to do this test as well ... since I know 'paranoia' can make you see things ... and trust me, I have been paranoid ... since I expected this lens to be 'perfect' based on all that I read about it and the fact that I thought I would avoid all the issues that go with the DA*

did you really shoot these vertically like that, or are these crops?

These were shot with the target flat on a table (therefore, horizontal on the table surface), and the camera on a tripod (IR remote to shoot) ... at approx. 45 degrees... I tried to follow the testing instructions totally... then I cropped out approx. the middle half (in terms of width, the height is as shot) ... and then pasted the three crops into a new photo approx 150% the original width to get all three together in the same pic ...

I do still wonder, though, if it gets better as you repeat the test, as it still seems to me that "many" if not "most" of the things that would lead a lens (as opposed to a camera) to FF would also cause it get better with each successive attempt.

I didn't get this point completely... you mean if I keep trying out the lens then the FF may reduce (even disappear)? Or that I may isolate the exact issue better by testing the lens more accurately?

From what can be seen in the test pics (and I am still open to more tests if suggested, since I would rather prove myself wrong on this issue than go back to deciding between DA* & Sigma) is it conclusive either ways ?

[rkt]

Here, at 50mm, I have tried to get as close to possible with AF still working ... (i corrected the WB so the shot looks different... rest of the setup is the same... flat table, tripod, remote)

Left, shot with AF... right, shot with MF ... ISO 100, 2.8 at 1/800 ... both shots ...

---

Last edited by rkt; 08-23-2010 at 12:44 PM.

[Ash]

The manual focusing seems to be inacurrate from this example. Back focused by about 6mm.
The AF trial looks more accurate - if anything, BF by 2mm.

Do you get similar focusing troubles in real life images, both close and far focusing?

[WheresWaldo]

Sorry, I don't mean to be a @#$%^ about this, but I'm wondering - did you really shoot these vertically like that, or are these crops?

I am not going to respond to the above sentence.

If they are crops, then it *might* be the case that you were shooting from too far away, and the "Focus test chart V2.00" might have been in range of the central AF sensor, thus stealing the attention of the camera's AF system and invalidating the test. It's important to shoot from close enough that only the focus target is in range of the sensor. Meaning nothing else should be within, or even close to, the center brackets etched into the viewfinder, since the AF sensor is about that big, maybe a bit bigger.

This was my point exactly about the usefulness of these printed charts but you refused to acknowledge it! This is also the reason why the LensAlign and other homebrew concepts work much better, even at distances that are too far for these printed charts the LensAlign still works properly. I have read your other posts and know that you are an intelligent guy, I just don't see how you can cling to certain beliefs about the efficacy of these printed focus charts.

If you did shoot these vertical and that's the whole picture, then I think those shots probably do the trick, although I have to admit I've never thought through what effect shooting vertically rather than horizontally might have. Assuming it was done on tripod, I don't think it should invalidate the results. But handheld tests at small apertures are also always a little suspect, as just moving a couple of centimeters between the time focus is achieved and when the picture is taken totally throws off the results at wide apertures.

Well the center auto focus point is a cross point and should work either way.

At this stage I have nothing more to add to this conversation so I will leave this thread alone.

[Marc Sabatella]

I cropped out approx. the middle half (in terms of width, the height is as shot)

OK, in that case, it would appear the "Focus test chart" is out of range of the focus sensor, so the test is good.

I didn't get this point completely... you mean if I keep trying out the lens then the FF may reduce (even disappear)?

Temporarily, yes. One of the possible causes of FF (or BF) would be if the gearing in the lens is off. The camera figures out how far out of focus the picture is when it starts, figures out much to turn the lens to get it in focus, and turns the screw the appropriate number of times - but if the gearing within the lens is off, the lens might not get all the way in focus (or might overshoot). Sure, the csmera supposedly checks again, but it doesn't keep check and keep checking until its perfect. So after the first try, it gets "close", and lets you take the shot. Now that it's close, if you immediately try repeating the test, the camera will see that it's close already, and won't have to turn the screw very far to get it in focus - and the result should be closer than last time. Repeat again and it should get closer still. At least, *if* the problem is in the gearing. But I'm still pretty fuzzy on what else could psosibly be wrong in a lens to cause FF (or BF). Some have hypothesized that it has to do with the fact that focus might shift due to "residual spherical aberrations", but I can't say I've sene a really clear and convincing explanation as to how this could cause the AF system to see a picture that is already in focus and deliberately throw it out. So it would be worth trying that focus it manually, then see if AF actually throws it out or not.

[causey]

I did numerous AF and MF tests with 'my' two copies of Tamron 17-50mm--perhaps more than a hundred fifty shots (among which at least 20 careful tests with tripod and charts)... I did MF tests for comparison with the Tammies and other lenses, and noticed that, while there was some 'decentering', at least the object I had focused on was always in focus, whereas the AF shots with the Tammies were slightly out of focus (on the borderline between focus and blur). I also noticed that whether the MF was slightly front or back (at f/2.8) depended on how I rotated the ring: more 'back' effect, if rotated counter-clockwise (or 'up'), and more 'front' if rotated clockwise ('down'). A kind of inertia.
And I think if you take numerous real life--and/or causal test--pictures and remark the same front focusing issue over and again, those numerous shots count for something. If not directly relevant, they can at least back up the results of the more serious tests.
Also, Tamron 17-50mm, unlike a lot of other lenses, has a history of front-focusing problems; admittedly, not so many with K20D.

[Marc Sabatella]

If they are crops, then it *might* be the case that you were shooting from too far away, and the "Focus test chart V2.00" might have been in range of the central AF sensor, thus stealing the attention of the camera's AF system and invalidating the test. It's important to shoot from close enough that only the focus target is in range of the sensor. Meaning nothing else should be within, or even close to, the center brackets etched into the viewfinder, since the AF sensor is about that big, maybe a bit bigger.

This was my point exactly about the usefulness of these printed charts but you refused to acknowledge it!

That's because I'm saying something quite different from you. You were saying the thinness of the line was a problem. I'm saying that's not true at all, as I see it. The *only* problem I am identifying here is that if you shoot from too far away, or at too extreme an angle (too close to parallel to the chart), some part of the box *around* that line might end up in range of the sensor. Again, the problem has *nothing* to do with the thinness of the line it's solely a question of performing the test incorrectly, such that the box ends up in range. if you perform the test correctly, that *won't* be a problem.

So no, we aren't saying the same thing at all. I still don't understand how the thinness of the line could possibly cause problems.

I have read your other posts and know that you are an intelligent guy, I just don't see how you can cling to certain beliefs about the efficacy of these printed focus charts.

I'm be glad to give them up if you can explain what the problem is. So far, I just don't see it. If the thin line is the only thing within range of the AF sensor - and it *is* if you perform the test correctly - then there is nothing else for the AF sensor to lock onto. it *will* lock on to the thin line. the fact that the line is thin shouldn't - as far as I can tell - cause any problems, nor should the fact that this thin line is on an angle. I just don't see how either of those facts should present even the slightest difficulty. Which is to say, it isn't *obvious*. there might be some obscure thing going on that requires advanced knowledge of optics or physics to understand, but I don't have that kind of background. Could you explain what you think the problem actually is with thin lines or angles?

[causey]

But I'm still pretty fuzzy on what else could psosibly be wrong in a lens to cause FF (or BF). Some have hypothesized that it has to do with the fact that focus might shift due to "residual spherical aberrations", but I can't say I've sene a really clear and convincing explanation as to how this could cause the AF system to see a picture that is already in focus and deliberately throw it out.

My absolutely uneducated guess is that this issue is characteristic to certain combinations of cameras and lenses. It might have to do with how quickly responsive the AF system is--and K200D's system is pretty sluggish--and with some optical quirk of the lens. I have no idea. In any case, phenomena come first, explanation comes second. (Unless we can't safely discount the hypothesis of mass hysteria.)

[crwl]

I have a K10D + Tamron 17-50 and I must say autofocusing with the Tamron is somewhat temperamental. Especially at the wide end. And focusing to infinity, it's always better done manually (it sometimes fails to lock focus at all when pointing somewhere far away). I quickly tried the same lens on a K100D and didn't notice any problems in a quick test; neither have I noticed any troubles with AF on the same K10D with the FA 35/2 AL or DA 18-55 3.5-5.6 WR.

I even sent the lens to warranty repair, but as far as I could see, it came back in the same condition I sent it in. And it makes terrific photos when correctly focused and is sharp all across so there's definitely no decentering problems. I'm thinking of upgrading to K20D with its per-lens AF adjustment when I find one inexpensive enough... (offers welcome).

[rkt]

This is turning into an interesting discussion... thank you, everyone, for participating...

A couple of questions...

(1) are the tests (and the results above) conclusive either ways? What is your conclusion if you got these results? Would you conclude that the lens needs to go back, or that what can be seen is within the 'tolerable' range and that it can be lived with... (I mean, nothing is ever totally perfect, as long as it is not totally outright unacceptable ?

(2) if the answer to the first part of (1) is NO, then can you propose some alternate tests / compositions that I can take pictures of to reach a 'conclusion' of some sort ... since I have just a few days before the 'exchange' option turns into a 'send for repair' option ...

Now that it's close, if you immediately try repeating the test, the camera will see that it's close already, and won't have to turn the screw very far to get it in focus - and the result should be closer than last time.

I have tried to remove this effect by throwing the lens off-focus before each shot... by focusing on something different and then refocusing on the target line...

Causey, what did you finally do? Did you totally get rid of this lens and get another substitute (if yes, then which one?), or did you decide to 'live with it' till you got a better body (like K20 or something with AF adj.) or resorted frequently to MF when in doubt?

[causey]

Causey, what did you finally do? Did you totally get rid of this lens and get another substitute (if yes, then which one?), or did you decide to 'live with it' till you got a better body (like K20 or something with AF adj.) or resorted frequently to MF when in doubt?

Fortunately, I was able to return them... I am still thinking of buying the Pentax 16-45mm, but I think I'll wait some more, because the DxO software really does miracles on the kit lens' output. Sharpness-wise, I'm pretty satisfied, and the kit lens has its advantage: it's very little. I've learned to appreciate the convenience of small lenses, so I'm trying to build a modest set of primes in addition to the travel zooms (kit lens and DA 50-200mm). The Tamron experience was so bad, I might have become wiser about lenses I might try to get a Tamron 17-50 again after I buy a K-x body, but that's not going to happen until prices go very-very low. Who knows? Maybe the Tamron will be a bit cheaper, too, although I doubt it.

I PM-ed the guy who wrote a bad (grade 4) review about Tamron 17-50mm on account of its front focusing problem with a K10D. It appears to be the lens, not the camera, nor the camera/lens combination. He has just received the lens back from servicing, and now he says it focuses wonderfully at all focal lengths. I asked for his permission to post his detailed reply here; now waiting for an affirmative signal from him.

[WheresWaldo]

That's because I'm saying something quite different from you. You were saying the thinness of the line was a problem. I'm saying that's not true at all, as I see it. The *only* problem I am identifying here is that if you shoot from too far away, or at too extreme an angle (too close to parallel to the chart), some part of the box *around* that line might end up in range of the sensor. Again, the problem has *nothing* to do with the thinness of the line it's solely a question of performing the test incorrectly, such that the box ends up in range. if you perform the test correctly, that *won't* be a problem.

So no, we aren't saying the same thing at all. I still don't understand how the thinness of the line could possibly cause problems.



I'm be glad to give them up if you can explain what the problem is. So far, I just don't see it. If the thin line is the only thing within range of the AF sensor - and it *is* if you perform the test correctly - then there is nothing else for the AF sensor to lock onto. it *will* lock on to the thin line. the fact that the line is thin shouldn't - as far as I can tell - cause any problems, nor should the fact that this thin line is on an angle. I just don't see how either of those facts should present even the slightest difficulty. Which is to say, it isn't *obvious*. there might be some obscure thing going on that requires advanced knowledge of optics or physics to understand, but I don't have that kind of background. Could you explain what you think the problem actually is with thin lines or angles?

Now I understand where we are different. I did not mean to imply that the thin line was the issue, and I don't think I used those exact words. What I said was the target is too small. The target becomes even smaller the more oblique your angle is and also becomes smaller the further away you are. The smaller the target area the less likely that you will know where you are focused. It was a word interpretation difference and my inability to find the precise ones to explain my point.

There is one point in favor of the printed charts; Pentax uses a printed chart when calibrating cameras! The problem lies in all the other variables that we humans introduce when doing the same test. If you read the Pentax K series service manuals, I have, Pentax uses a precise set up. They only test with one lens and at one distance and with a precise method of exposure and they use a jig. Every camera that they test for service uses the same set up. We introduce variable distances, different lenses, different maximum apertures as well as different lighting.

One more note I read somewhere but don't have a direct link (might have been the same service manuals) that focusing from infinity and focusing from minimum focusing distances could produce different results. I have confirmed that my 2 K20Ds act this way. I believe I read that this is due to the cross sensor used in the center. From infinity it one axis is dominant and from close the other predominates, that would indicate to me that after you assure that the test scenario is accurate you need to verify focusing from both directions and compare the results.

[causey]

Mathias (Kameraten) kindly agreed to my posting his reply:

"It so happens that my 17-50 just a few days ago returned from servicing. In addition to the faulty focusing it had also developed a habit of reporting the focal length incorrectly to the camera which was upsetting both shake reduction and flash.
The thing had some chip repaired and electronically calibrated and wow, now it's quite different! It actually hits focus both near and far, with zero adjustment. Earlier it had to be adjusted as far back (-10) as possible in order to hit even sometimes.
Since I originally extensively tested three different brand new copies of the lens, with three different bodies, I'm bound to believe Tamron has made some kind of product development that has gone into the repair kits too. Or the original lenses were all just incredibly badly adjusted. On the K10D it was useless.
There sure are a lot of weird and even wrong things with the Pentax focusing system, but thats mostly a different story."