[ytterbium]

Main question is, how big the size is and how it affects focusing on smaller size objects using wider angle lenses.

The full story. Bought a Sigma 30/1.4 (if not for corner sharpness and distortions, brilliant glass - vibrant colors, tack sharp in center, creamy bokeh, excellent contrast, flare resistant), while testing it after purchase (shallow dof closeups, focus test charts, isolated objects) didnt notice any constant fault, just some random missed shots.
But after more extensive real life usage, started to notice, that most of my shots had actual focus behind the main subject. First taught was that it is the lens which is faulty, because i had no noticeable problems with 18-55, sigma 70-300, F 50/1.7 and others. Wierd thing, that on close subjects (a portrait, filled object) the focus was always dead on.

After few trials i was able to isolate the problem. Testing showed, that on two K10D's the lens worked correctly. It misfocuses on k100ds in two cases. When shooting at angle (simple test: point camera at floor, 45 degrees, center focus point, single shot, focus is out of frame, or at its top - behind the subject). Other is when focusing/setting focus point on subject in similar/smaller size to spot metering area, having some relatively close background (human->wall, flower->bushes, vase->table). Semi-transparent objects (tree branches) and does not even need a close background, it goes straight trough them to infinity.
This (missed shots) starts to get very evident at subject to camera distances 3..4 meters and increases with distance (probably more due to subject relatively "getting smaller", than actually moving further away).
And the focus is always dead on, on something at background or infinity, not just some arbitrary distance further.

This happens with all focus points.

Since i have not noticed similar problems with other mentioned glass, i taught that this specific camera-lens combo makes this happen. Unfortunately i was wrong. After testing my camera at shop, with FA 35/2 (btw spectacular lens), found out it behaves exactly the same (floor angle shot missed, distant - medium sized subjects missed).

This rises two more questions.
What might be the solution. Since it is not actually a back focus, there is nothing to correct/adjust thus sending for repair might not help.. it feels more like camera starts to focus from infinity and locks on first subject it meets at selected focus point.
At this point my only solution seems to be K10D. Kind of too expensive, since in other ways k100d is fully functional and satisfying...
It still has warranty.

Is the camera actually faulty, or is it my shooting style and AF system specific issue (not an actual deffect).

Could this be a wide angle problem. Or it is easier to get/notice it such way.
I suspect, when using kit lens (18-55) it could be too hard to notice the problem because of its F-number at greater focal lengths and big DOF at 18mm (which have somewhat brighter F-number). As well as with longer lenses, when the subject usually is zoomed in, and shallow DOF makes you immediately notice error and seamlessly correct it. With wide'ish lenses, subjects are usually too small to notice in VF and refocus(try to..) while shooting, but the shallow DOF of bright lens makes it very apparent when viewing on PC or printing.

Have a look at the samples provided few posts further. They illustrate the problem much better.
Those posted here are confusing and not very obvios.
Since this is actually a photography forum, some samples to make it easier perceive the problem..
Sigma 30/1.4, nothing of a photographic value, just shots i took trying to isolate the problem:
Inbox Foto (foto.inbox.lv)
Pentax FA 35/2, same as sigma, backfocus'in around in shop:
Inbox Foto (foto.inbox.lv)
(BTW, you can see EXIF, right next to the lower right picture corner, click the link to extend)

Pardon my (non existent) English skills and the long story.

---

Last edited by ytterbium; 04-02-2009 at 09:18 AM. Reason: +

[Marc Sabatella]

It misfocuses on k100ds in two cases. When shooting at angle (simple test: point camera at floor, 45 degrees, center focus point, single shot, focus is out of frame, or at its top - behind the subject).

Is there an actual objec ton the floor that you are trying to focus on? If not, it shouldn't be surprising that it has trouble focusing. And if there is an object to focus on, are you closer to that object than the minimum focus distance for the lens?

Other is when focusing/setting focus point on subject in similar/smaller size to spot metering area, having some relatively close background (human->wall, flower->bushes, vase->table).

This much is normal, if I'm understanding you correctly. The focus "point" is actually much larger than it appears in the viewfinder, so if there is anything that looks close to your intended subject - even if it is not really close but just appears close, like a distant background object almost directly the subject - the camera might well focus on it instead. It's an age-old problem for AF systems, and one should switch to manual focus in these situations.

At this point my only solution seems to be K10D. Kind of too expensive, since in other ways k100d is fully functional and satisfying...

Again, if I am understanding you correctly this is perfectly normal for any camera with autofocus.

The link you posted is to a whole gallery, and it wasn't clear which ones exhibited problems. The ones I clicked on appeared to show the focus exactly where I'd expect. Is there a particular one you feel shows a problem?

I suspect, when using kit lens (18-55) it could be too hard to notice the problem because of its F-number at greater focal lengths and big DOF at 18mm (which have somewhat brighter F-number). As well as with longer lenses, when the subject usually is zoomed in, and shallow DOF makes you immediately notice error and seamlessly correct it. With wide'ish lenses, subjects are usually too small to notice in VF and refocus(try to..) while shooting, but the shallow DOF of bright lens makes it very apparent when viewing on PC or printing.

These are very good observations. I think you are completely correct here.

[ytterbium]

Thank you for your reply.

Is there an actual object on the floor that you are trying to focus on? If not, it shouldn't be surprising that it has trouble focusing. And if there is an object to focus on, are you closer to that object than the minimum focus distance for the lens?

1.) Plain floor, with pattern (contrast), way after minimum focusing distance (standing straight, camera to face, approx 2 meters). Surfaces at angle (especially floor) tends to "pull" focus away even from objects not at angle:
Inbox Foto (foto.inbox.lv)
(if there were no box, the focus wouldn't just be a little incorrect, most likely it would be somewhere at subject distance 3..5m, out of frame)
Wierd thing is that focus test chart, also in angle but much closer focuses correctly:
http://www.shutter.lv/index.php?name=Forums&file=download&id=1963

This much is normal, if I'm understanding you correctly. The focus "point" is actually much larger than it appears in the viewfinder, so if there is anything that looks close to your intended subject - even if it is not really close but just appears close, like a distant background object almost directly the subject - the camera might well focus on it instead. It's an age-old problem for AF systems, and one should switch to manual focus in these situations.

2.)It is not that "..it might focus instead..", its more like it will definitely lock on the darn thing behind unless you're very lucky. As for how big the subject must actually be, for AF not to fail i was not able to exactly determine (some images illustrating various sizes with incorrect focus):

Have a look at the samples provided fev posts furtherr. They illustrate the problem much better.
Those posted here are confusing and not very obvios.
http://www.shutter.lv/index.php?name=Forums&file=download&id=1957 ( those are the kind of shots i would most often get.. seemed easy target to me..)
Inbox Foto (foto.inbox.lv) (behind the lamp, 3x times in a row, mildest case, some times even the rod is OOF)
Inbox Foto (foto.inbox.lv) (somewhere at the fence, focus point was barrel front)
Inbox Foto (foto.inbox.lv) (back wall of a box, instead of front)
Inbox Foto (foto.inbox.lv) (well.. i guess its at infinity)
Inbox Foto (foto.inbox.lv) (at the subjects, just behin the magazine.. is it that small or complex?)
Inbox Foto (foto.inbox.lv) (dog, a little smaller than barrel, but correct)
Inbox Foto (foto.inbox.lv) (upper-right tree leaves, fail to infinity)
Inbox Foto (foto.inbox.lv) (not too sure, seems like fence instead of box, might be correct)
http://www.shutter.lv/index.php?name=Forums&file=download&id=1961 (Quite small subject and close background, virtually impossible to get shot like this[subject size, background distance] in focus)

Again, if I am understanding you correctly this is perfectly normal for any camera with auto focus.

3.)For me it seems too severe to be a general AF issue. Also, for me it seems logical, that between many possible targets (possible phase shifts in SIR AF case) camera should pick the closest one. Since the subject most often to be shot is in foreground, i assumed that focusing on foreground has to be primary AF goal. It feels like that the camera often just does not "see" the subject in front. I have not carried out any serious testing, but doing few shots with smaller (than 1.4) artificial (paper) outer aperture while focusing, made me feel that camera can see foreground subjects better.
What makes it worse, is that with such lenses (less than approx. 35mm) it becomes very hard to focus manually (k100d -> pentaMIRROR, factory focusing screen).
The samples aren't that informative because i usually delete the worst cases right after uploading from camera. So actually most of shots in that gallery have correct focus (they were to show when i did not observe this particular problem).
But in real life shots, for me its most noticeable when shooting a group of peoples indoors, with furniture and walls.. one might just guess where the focus will end up.

Someone with similar equipment (wide, bright lens, 30/1.4, 35/2, 31/1.8...+k100d/ds) could post their experience to shed some light, as i am about to sell the sigma ridiculously cheap. Why keep it if it cant make pictures.. but wouldn't like to regret, if the fix was simple and not that expensive.

These are very good observations. I think you are completely correct here.

4.) Well those are just logical conclusions after a brief testing and attempt to isolate the cause. But most of the time i had to observe many pictures to be deleted, with someones face all blurry and the shoulder just behind them in perfect focus.


I would really like to hear more similar lens(bright, wide/normal)-body(k100d/ds) combo user experiences.

Hope this all makes sense to you...
P.S. Btw, does that site (inbox.lv) loads fast, are you able to open the links at all? Im using it since its very fast for me, and i hope for internet users in my country, but i have no idea what speed and reliability it delivers for example to US users. Maybe i should use different hosting when sharing images for them.

[Marc Sabatella]

2.)It is not that "..it might focus instead..", its more like it will definitely lock on the darn thing behind unless you're very lucky.

I checked several of the examples you listed, and in most of them, I think it was extremely unrealistic to expect to any camera to have guessed the specific area you wanted to focus on - there's way too many potential targets close to the center point. In several cases, the target chosen was actually closer to the center than the thing you said you wanted it to focus on.

Does someone else see this differently? I'm having a hard time seeing a problem here - it looks like normal AF behavior to me.

The only shot that surprised me was the very first - the vase of flowers. Although this looked liked it might have been closer than the minimum focus distance? Possibly also that affected the last shot, although in that one, the fact that your subject was moving was perhaps the real problem - hard to lock on something that isn't sitting still, so no surprise the camera might choose something immobile if it is nearby.

[ytterbium]

Thank you again. In the first one i was well away from the subject.
Yes, it would be really good to hear others, until that i have nothing to compare against.
In the mean while ill just take few random shots to get some "real" samples.

[UnknownVT]

I know you have put a lot of effort into posting your problems -
but like Marc, I am having a hard time following and trying to figure out what the problem is.

Perhaps you can shoot a set that shows the problems more obviously -
rather than showing us your "good" shots - where we cannot see the problem.

Another way is to by-pass the problems you may be having entirely -
by switching to central point focus only -
that way you determine which point the camera is focussing on -
then hold that focus (light half-depression of shutter button)
and re-compose to take the shot.

[ytterbium]

Since I'm now indoors and the problem exhibits itself at greater distances, i may post them in here tomorrow when ill get out.

And your'e starting to convince me.
Usually i select the focus point manually, but iv'e tried using only center focus point and recompose - unusable with this lens.. because of the wide'ish fov, focal plane has greater "rotation" (opposite to almost parallel movement with longer mm) so its even more likely to get OOF image, and this time even without AF error.

Just noticed, that k100d does not like colorful, detailed targets, especially containing red and purple colors:
http://foto.inbox.lv/tamrac/Sigma2/IMGP7499.jpg
How big the center AF area has to be to see anything behind that bag and be able to get exact lock?

I highly appreciate your help, because i have limited possibilities to compare different gear (very few pentaxians nearby).

---

Last edited by ytterbium; 01-03-2009 at 07:52 AM.

[Wombat]

Just a thought - do you get the same problem whether you hand hold or use a tripod? Were your "problem" shots taken with SR on or off? Have you been able to observe the same problem at all shutter speeds and apertures that you normally use? I don't know what any of this would signify, but it just might be worth a check.

[UnknownVT]

Since I'm now indoors and the problem exhibits itself at greater distances, i may post them in here tomorrow when ill get out.

Usually i select the focus point manually, but iv'e tried using only center focus point and recompose - unusable with this lens.. because of the wide'ish fov, focal plane has greater "rotation" (opposite to almost parallel movement with longer mm) so its even more likely to get OOF image, and this time even without AF error.

Sorry maybe I am not understanding something -
as I am confused by what you are saying.

You just said the focussing error happens at greater distances -
greater than you can get indoors -
so can I assume this is likely to be >10metres?

At 10 metres for your 30mm f/1.4 lens the DOF is 4.05m to 18.75m (using the standard circle of confusion of focal length/1000) -

Even if we took the much, much more critical way to calculate the DoF using a circle of confusion of Focal length/2000 (ie: twice as critical) - at 10 metres the DoF is still 8.1m - 13m

if the point you are focussing on is out of focus then the focussing error is HUGE! even using the very critical 1/2000 circle of confusion the margin of error is still 2 metres in front and 3 metres behind ie: 5 metres -

and for the normal 1/1000 circle of confusion the margin of error is 6 metres in front and nearly 9 metres behind!!!!! that's a range of 15 metres - you could drive a BUS between that!

The Hyperfocal distance for a 30mm at f/1.4 is 21.4 metres - so focussing on that distance means everything from 10.7 metres to infinity will be in focus.

If one really wants to be hyper-critical use the 1/2000 circle of confusion
then the hyperfocal distance for the same 30mm f/1.4 len is then 42.8 metres - so focussing on that everything from 21.4m to infinity will be in focus/critically sharp.

Maybe I am missing something I cannot understand how one can see a mis- focus at longer distances, unless it is a huge gross misfocus - then there would be a VERY serious focussing problem with the camera.

[ytterbium]

Basically im almost convinced by you thats its me the error. In the mean time i've uploaded a little gallery and here is my comments on it. All shot with recompose, or focus point set at the subject, this time SR off, most of shots were fast enough(>1/500) to not need a tripod:
Inbox Foto (foto.inbox.lv) (The gallery itself)
Some of them have small diaphragm set, so they look fine. I uploaded them all, so the hit/error rate can be seen. In my opinion, some of them could fail due to subject, but in some the subject is too clear and "obvious". But as you requested here i have pointed out where it is most visible:
Inbox Foto (foto.inbox.lv) (lower right focus point, pointed at hump center, recomposed - a bus could drive.. just after the hump, not over it )
Inbox Foto (foto.inbox.lv) (Center AFp, then recompose, gray foreground wall, well this has to be some sort of fault, how a center point could see behind it?)
Inbox Foto (foto.inbox.lv) (the tree, right AFp - infinity i guess)
Inbox Foto (foto.inbox.lv) (pretty straight forward - fail)
Inbox Foto (foto.inbox.lv) (well, arguable.. seems easy target to me)
Inbox Foto (foto.inbox.lv) (tree, Center AFp, slight recompose)
Inbox Foto (foto.inbox.lv) (stubs in foreground, lower right AFp)
(click on the images for full size)

I hope this better illustrate, where and why i see the problem.
P.S. After reviewing the shots, it feels like there is some sort of defect after all.
Since there is no actual back focus, so it can not be fixed by front-back adjustments, one person hinted me, that maybe this could be caused by focus sensor tilt adjustments. Could this be true? If i sent a camera for repair, would they be able to replicate my problem and try to resolve this, or just assume that everything is fine and sent it back.

Problem is persistent, at any settings, at some it just becomes less noticeable (big DOF) or harder to achieve (no nearby background)

As for the amount of error, now i realize, that actually i never did took a look at the focus distance scale (because most of shots evem in camera looked acceptable for f1.4, and the exif shows only "DISTANT").
Most shots were taken, by putting a hand in front of my camera, activate AF to get subject out of focus, then focus to subject. In some cases, re-activating AF after getting first lock "on subject" gets the second lock correctly, but most of the time that doesn't help (it picks the same target).

[ytterbium]

So, should i send my camera to warranty or forget the bright wideangles?

[Marc Sabatella]

In my opinion, some of them could fail due to subject, but in some the subject is too clear and "obvious".

I'm just not seeing that. The samples that I looked at had several possible targets within range of the center "point" (large rectangle), and the camera chose one of those subject. I don't understand how you could possibly think the camera would read your mind here and happen to pick the subject you intended every time when there are other perfectly valid choices.

Note on some you appeared to closer than the minimum focus distance (like the tree that cuts directly across the frame and the camera focused on a plant behind it).

Once again, though, you've posted so many samples - including, some in which you appear to be agreeing that the camera is working fine - that it's kind of overwhelming. If you're really convinced there is a problem, could you post *just one* picture to clearly demonstrate this, with full explanation of what you focused on and how (eg, using center point & recomposing, etc).

In some cases, re-activating AF after getting first lock "on subject" gets the second lock correctly, but most of the time that doesn't help (it picks the same target).

Right. So you try again, moving the camera a little hoping it might make a different choice, but eventually, you should just switch to manual focus. Much easier with Pentax lenses and the "quick shift" feature. But again, this is how AF systems have worked for decades - from the evidence I see here, your camera is perfectly normal.

[ytterbium]

Ok, then briefly:
http://foto.inbox.lv/tamrac/Sigma2/IMGP7547.jpg

-Must be viewed at 100% (click on the image), since the DOF was getting quite big (picked because of good AF point coverage).
-Manually selected center focus point, no movement (without recompose).
-AF.S. SR off, Aperture priority.
-Handheld (1/3200 s @ F1.4 +0.7ev ISO200).
-Perfectly flat, opaque object, having perfect detail/contrast for focusing.
Covering not only, red center rectangle, but also spot metering area as well as af brackets showing surrounding AF points (just to be sure...).
-Camera put out of focus before shot (set closer than subject).

In addition:
To make it clear, i did not quite get what did you mean by large rectangle.. how big the center AF area actually is:
Inbox Foto (foto.inbox.lv)
(Pick this, or one of next two, which approximately matches the size - "spot, medium or wide").
If you think its the "wide", i still cannot see how camera could focus behind that wall (such AF area barely reaches over it), and how small space is left for surrounding AF areas.
Since i posted that image.. in similar shot, also with same sized dog , lower left focus point (more exactly over dog), after several tries resulted in sharp grass behind...too bad its already deleted.

If talking about distance, all subjects were at least 1 meter or more away, minimum distance for sigma is approx. 40 cm. You're saying that im too close.. no i just picked big enough subject to show, how obvious this error is, so with the addition of blur, it seems to be close.
Have you actually shot with similar configuration and did you ALWAYS had misfocus in such situations, or you used manual focus all the time?

I highly appreciate your help because i would be very happy to find a solution, since unless it is solved, basically i cant use any normal low light prime.

[UnknownVT]

This is perplexing you are still showing shots that are very hard to analyze in a simplistic way.

I'll take your word for your " Inbox Foto (foto.inbox.lv) (lower right focus point, pointed at hump center, recomposed - a bus could drive.. just after the hump, not over it ) " where the focus point locked - this would mean the actual physical focus was behind what the camera locked on.

Also " Inbox Foto (foto.inbox.lv) (pretty straight forward - fail) " - assuming the focus locked on the tree in the center foreground - this has got to be closer than 10 metres though

OK without trying to comment about your shots or shooting conditions I found this that might be of interest -

It's a test of the Sigma 30mm f/1.4 by Ken Rockwell but on Nikon bodies -
QUOTE:
" 2.) AF often focused slightly in front of my subject with my D200, potentially significant at the large apertures for which you buy this lens. I worked around this by pointing my AF sensors at something behind my subject if it was critical. This isn't a problem for real, live 3D subjects. This effect varies from sample to sample and camera to camera. It was fine on my D40. "

This appears to be the opposite to your focussing problem - but there obviously IS a problem with focussing with this lens.

He elborates further down the review -

QUOTE:
" Autofocus Accuracy
AF accuracy is fine for the things I would shoot with this lens, like people and kids in dim light.
The Sigma 30mm f/1.4 is very bad, depending on your camera, for photographing at long distances. The AF system, at least on my D200 at f/1.4 and large apertures, consistently wouldn't focus at infinity and instead focused too closely. I focused by scale instead for objects further away than about 50 feet (20m). If this is a lot of what you do, you probably won't be happy with this lens. It was perfect with my D40. This will vary from sample to sample and camera to camera.
AF accuracy is always a problem for shooting test charts with f/1.4 lenses. This has driven me crazy since the 1970s. For hand-held shots in dim light of wiggly subjects (the whole point of this lens) the AF accuracy is fine. For shooting flat test charts it's poor, but no one buys this lens for shooting test charts. For test charts and astronomical use, get the discontinued Nikon 28mm f/1.4 AF or the insanely good new Canon 85mm f/1.2L II.
At f/1.4 this is obvious if you're dumb enough to be shooting at f/1.4 in daylight. In real use it's invisible at f/4, and in the dark it's not much of an issue since most things worth shooting are deeper than the depth of field anyway. If you worry about these issues, don't get this lens, or at least try a different sample. "

No insult intended - those are his words about "dumb" - and I am sure not directed at anyone in particular

So after struggling with your long and arduous analysis -
perhaps there may be a problem/incompatibility with the Sigma 30mm f/1.4 and your sample of the Pentax K100D Super.

BTW - the EXIF data seems amiss reporting the lens to be 300.0 mm(!) -
a f/1.4 300mm would be incredible

[jeffkrol]

Ok, then briefly:
Inbox Foto (foto.inbox.lv)

-Must be viewed at 100% (click on the image), since the DOF was getting quite big (picked because of good AF point coverage).
-Manually selected center focus point, no movement (without recompose).
-AF.S. SR off, Aperture priority.
-Handheld (1/3200 s @ F1.4 +0.7ev ISO200).
-Perfectly flat, opaque object, having perfect detail/contrast for focusing.
Covering not only, red center rectangle, but also spot metering area as well as af brackets showing surrounding AF points (just to be sure...).
-Camera put out of focus before shot (set closer than subject).

In addition:
To make it clear, i did not quite get what did you mean by large rectangle.. how big the center AF area actually is:
Inbox Foto (foto.inbox.lv)
(Pick this, or one of next two, which approximately matches the size - "spot, medium or wide").
If you think its the "wide", i still cannot see how camera could focus behind that wall (such AF area barely reaches over it), and how small space is left for surrounding AF areas.
Since i posted that image.. in similar shot, also with same sized dog , lower left focus point (more exactly over dog), after several tries resulted in sharp grass behind...too bad its already deleted.

If talking about distance, all subjects were at least 1 meter or more away, minimum distance for sigma is approx. 40 cm. You're saying that im too close.. no i just picked big enough subject to show, how obvious this error is, so with the addition of blur, it seems to be close.
Have you actually shot with similar configuration and did you ALWAYS had misfocus in such situations, or you used manual focus all the time?

I highly appreciate your help because i would be very happy to find a solution, since unless it is solved, basically i cant use any normal low light prime.

Send the camera in w/ the lens to be calibrated together.