[jolepp]

The above mentioned Thread seems to be stopped. Will the discussion be somewhere else?

Earl

https://www.pentaxforums.com/forums/pentax-k-5-forum/134259-k-5-auto-focus-tr...catch-all.html

[Bramela]

The above mentioned Thread seems to be stopped. Will the discussion be somewhere else?

Earl

Can be, but only if the personal slap-fest does not recur.
More and more, Moderators are becoming sick and tired of the insults, arguing and petty picking over K5 issues.
It is fine for K5 issues to be discussed at length, but, as there seems to be widely varying experiences with K5 , certain posters seem to think that it is ok to get stuck into others with differing points of view/experiences.
It will not be tolerated.

[Aristophanes]

PentaxImaging of the US advertises the K-5 as a"Professional" camera and if you care to look at the specifications for the K-5, you might ask yourself who is being duped or harmed. Compared to how certain previous generations of Pentax cameras performed, it is quite reasonable to expect the K-5 to meet its specifications. What is the point of buying something, where the superlative advantages of the product are crippled by either design, manufacturing, assembly, hardware or software issue.
Add to that the various other issues that have received widespread attention, it is only prudent for "informed" consumers to wait instead of going thru the hassles that others have experienced so far. Some consumers won't wait, buying into another system they perceive as more reliable and functional.

This forum and others, and forum members like "Falconeye" provide exactly the kind of information that gives us a chance to make a somewhat educated decision when buying specific camera products. I am the one parting with my money, expecting a product that works as advertised.

Once again, products fail and have faults, but when these faults are systemic of the product line, then it may be time to put on the brakes. There really is no fine line between "Marketing" and reality. Meet your stated specifications or lose me as a customer.

Can be, but only if the personal slap-fest does not recur.
More and more, Moderators are becoming sick and tired of the insults, arguing and petty picking over K5 issues.
It is fine for K5 issues to be discussed at length, but, as there seems to be widely varying experiences with K5 , certain posters seem to think that it is ok to get stuck into others with differing points of view/experiences.
It will not be tolerated.

It's hardly "petty" if there is empirical evidence as falk provides. How some people take bad news is petty, but not the effort itself. The whole point of falk's paper is to cut through the "varying" experiences. His data set is not large, but his methodology validates what hundreds of posters here and on DPR have noticed. The balance of facts point to a design/manufacturer's error or outlier tolerance that may not be in line with consumers'. This is rapidly becoming fact, not conjecture. The last time falk did this the Moderator arbitrarily posted his shutter slap analysis to another forum, despite the fact this was "news" and had been passed along as such to Pentax officially. The public release of an empirical study qualifies as news in this industry.

[Smeggypants]

Can be, but only if the personal slap-fest does not recur.
More and more, Moderators are becoming sick and tired of the insults, arguing and petty picking over K5 issues.
It is fine for K5 issues to be discussed at length, but, as there seems to be widely varying experiences with K5 , certain posters seem to think that it is ok to get stuck into others with differing points of view/experiences.
It will not be tolerated.

But it's OK for some posters to shout User Error at those of us who actually have a faulty K-5 and have posted images to prove it. I found that pretty personally insulting yet when I respond I get infracted.

I take responsibility for getting embroiled in the slap fest. I'm sure the moderation has been even handed.

[Christine Tham]

Yes, the hypothesis needs more data. There seems to be some lens dependency indeed.

I don't have the time to do tests with my lens collection as thoroughly as you, but I can't seem to reproduce the problem at all with some lenses.

A thought that just occurred to me: perhaps the problem may be stray light from a wide open lens entering the AF sensors?

The binary behaviour can then be explained by stray light (which is a kind of "noise") overwhelming target light data at low EV. This would also explain why the behaviour may change with different lenses.

I have noticed for example the presence of a bright light source well away from the sensor lines may cause misfocus. Since the AF mechanism works when the lens is wide open it may well be that there is an issue with the AF mechanism not rejecting stray light.

May be interesting to do a test under varying conditions - ie. with lens hood attached and detached, also with presence of stray light.

Moreover, I think that maybe the K10D struggles in low light as well. Unfortunately, I could not repeat the test for other cameras or brands. Nevertheless, I would be surprised to learn that focus for the K10D jumps in such a binary way when the light is dimmed.

With my very limited testing, I think the behaviour of my K10D is "binary" at around the same light levels as for my K-5. For the same lens.

[dlacouture]

[*]The paper clarifies conditions to hit or avoid the issue. White light (halogen is not white enough though) and a wide lens stopped down help to work around the problem. AF assist light typically doesn't help though. But an LED flash light does.

First, big thanks for your paper...

Falk, I confirm that on my K5 (with the last known firmware), the AF assist lamp goes off several times while focusing (I was one of the two persons that reported this in the other thread, Chris Tham and me).

With my FA50/1.4 mounted, in dark situations, the AF assist lamp will fire once before moving, then, if it's really dark, for nearly each micro-adjustment (which is to say, usually, only one). If I voluntarily mess with the AF module by moving the camera while it focuses (panning or traveling), the AF assist lamp blinks up to 4 times (3 micro-adjustments) before AF fails...

The resulting pics seem to be focused, contrary to my other, AF-Assist-free low-light pics...

I'll have to check with the 18-55, if the behavior is the same...

EDIT:
I've done tests with my 18-55WR and my Tamron 70-200/2.8... In both cases, the behavior is identical to the FA50 : the AF assist lamp will blink as long as needed to lock AF (up to 4 times).
Focus accuracy is good in each instance (took about 20 pics with both lenses), while without AF lamp the K5 will FF quite a bit.
So in my case the AF assist lamp does work.

An interesting point I've found is that the camera will behave identically (obvious FF) in the following situations:
- AF assist lamp disabled.
- AF assist lamp enabled but blocked by the hand.

So maybe the problem lies in the fact that the camera may not actually compensate for low-light (I don't know if this is indeed needed?), as it thinks it will have the AF assist lamp anyway... So, about at the same time the assist lamp should kick in (even when disabled), the '+' sensor is not considered anymore (as it would mess with the green beam focus, I think).

This could be partially dismissed by Falk's findings that the light color has no real impact, but I think this is not really proved yet, as using a blue filter in front of a tungsten bulb is different from having a true daylight (or pure blue) source.

EDIT 2:
Going through tests again, I've found out that the AF assist lamp will come on more easily with the FA50 than with the 18-55!!!
I mean that, in the same lighting situations (where the K5+FA50 will start to use the AF assist lamp for low-contrast targets), the K5+18-55 will hardly ever use it (and will end with blurred pics)...
This is quite the opposite to what I'd have thought...

[Christine Tham]

Okay - I have done a simple test to test the "stray light" hypothesis.

Using a single battery, and using the light from a low powered torch to partially illuminate the battery in a dark room ...

I am able to confirm that the K-5 achieves perfect focus on the battery, even under very low amounts of light (f1.9, ISO100, close to minimum focusing distance, Exposure time 4sec, AF assist light did not come on).

Falk, are you able to do a series of tests with only spot lighting directed at the target area to be focused, and confirm that there is no FF under these conditions?

If you are able to replicate the results, it goes a long way in explaining the wide variance in reported results - from unable to replicate problem, to minor FF, to severe FF to completely blurred pictures.

If this hypothesis is true, it would mean that the issue is hardware rather than software, and probably unlikely to be fixed on the K-5, unless Pentax completely changes the design of the AF mechanism to eliminate/control stray light better.

Sorry to be the bringer of bad news for people hoping that there will be a firmware fix soon.

Attached Images

 

[Christine Tham]

Okay, not so simple.

I have now tried my simple experiment with different torches at different intensities.

The colour of the light still makes a difference.

Anything close to white light is not a problem, a very yellowish light will still cause the camera to misfocus.

So the stray light hypothesis does not completely account for the problem, but it can help explain some of the variance in the results.

So there still may be hope for a firmware fix.

But my personal feeling is that a fix is unlikely, since we are talking about a problem that has persisted through several generations of cameras (potentially at different magnitude from camera to camera).

So at this stage my advice would be: if you have a camera and you are happy with it, then keep it (my case). If you have a camera and you are not happy with it, return for a refund whilst you can and wait for news from Pentax.

[Christine Tham]

Going through tests again, I've found out that the AF assist lamp will come on more easily with the FA50 than with the 18-55!!!
I mean that, in the same lighting situations (where the K5+FA50 will start to use the AF assist lamp for low-contrast targets), the K5+18-55 will hardly ever use it (and will end with blurred pics)...
This is quite the opposite to what I'd have thought...

This is a really interesting discovery, and may confirm why I also get different results with different lenses.

I'm starting to think that the FF problem may be related to spherical aberration - somehow what the AF sensor detects is different from the viewfinder or main sensor, and the issue is related to lens curvature, and somehow the function of the AF assist is also tied in to this, which is why the AF assist behaves differently on different lenses.

[areidjr]

Thanks for the effort you put into this study. I found it to be both interesting and informative.

If there is any weakness, it would be the fact that, although your findings mirror those of many other users, your sample size of one is really not enough to extrapolate your findings across the entire population of K-5's. That's not to say that your findings are not 100% correct, but just that a sample size of one is not statistically significant.

Again, thanks for the great effort and report.

[Christine Tham]

Okay, I have now graduated from shooting batteries to shooting the output from the torch itself (the bare bulb).

The results are even weirder.

Shooting directly at the light source (a single incandescent bulb) still causes the the camera to misfocus, so it now appears the issue (for my camera at least) is not related to low light at all - since the bulb filament is providing quite strong light - more than an AF assist light would provide.

It appears the K-5 will misfocus if the predominant light source in the frame is very yellowish in colour. Even if the target subject of focus is the light source itself.

I think this eliminates the stray light hypothesis, and certainly raises the prospect that it's spherical aberration related to lens curvature - which is why the misfocus is related to light frequency more than light level. This explains why some people are reporting misfocus even at quite high light levels.

I am not sure this sort of problem is easy to fix in firmware, but it may be possible to fix it. It would require a lookup table based on the lens used, the white balance of the light source (not sure this is something the AF sensor can detect - the fix may require the camera take 2 shots - a test shot and a follow up shot with the focus adjusted to compensate).

[Christine Tham]

Here is an attached photo showing the K-5 misfocusing on an incandescent torch - hopefully you can see on this photo the camera is FF by quite a large amount - surprising since the light source is relatively strong so this is not an issue relating to light level at all.

[deejjjaaaa]

I agree with Quicksand. I tested my K20D and K100DS and found they FF in tungsten light and focussed properly in daylight, no matter if it was dim or bright. I didn't bother to test my K-x when I got it, but I know it does the same thing.

Kx is the same - w/ FA50/1.4 (ok MY kx and MY FA50/1.4) you can have the target to be brightly lit w/ tungsten lamp of 100wt (or 150wt) withing ~1' from the target to able to shoot handheld @ ISO200 (metering off the target itself which is a sheet of white paper - so it does not matter how dark is the room itself) and get different focusing vs the daylight... so the dimness has nothing to do w/ that.

[Ubuntu_user]

Isn't this something that has been/is a problem in past Pentax cameras? Because it seems that I could replicate the FF on a torch with my K-7. However, this is not enough of an issue to pose problems normally under regular shooting.

[deejjjaaaa]

Isn't this something that has been/is a problem in past Pentax cameras?

I believe that it is the same thing and no "+" in SAFOX actually changed anything dramatically... may be made it a little less pronounced when coupled w/ the newer DA lenses, especially darker ones vs FA lenses