[bjan]

Dear Ash, when taking pictures with wide open aperture like 2.8 or faster under artificial light wich is a very common situation especially when you think of the capabilities of the camera and the available lenses you expect your target to be in focus .
In normal day to day ( or better evening to evening ) situations the Front Focus is very obvious when shooting with fast set up and no test conditions or special skills are required to have this experience. In case you only shoot with smaller apertures and under daylight conditions NO PROBLEM sir ! But he... where would you buy a K5 for ?
So Falk did not do the testing because it is difficult to see the problem in everyones shooting life but to find out where it comes from and maybe even how it could be solved.
Again Pentax should hire those kind of people in order to get their act together.
The best DSLR in the world becomes a very unreliable partner for a more serious photographer with these kind of quirks the K5 had and still has after 1.03 !

[Ash]

Dear Ash, when taking pictures with wide open aperture like 2.8 or faster under artificial light wich is a very common situation especially when you think of the capabilities of the camera and the available lenses you expect your target to be in focus .

If I understand you correctly you want for me to appreciate the utility of large aperture shooting in low lighting situations. I can say I'm well aware of that, having been shooting low-light events quite often myself with cameras such as the K10D, *ist D and K20D that struggle in high ISO situations.

In normal day to day ( or better evening to evening ) situations the Front Focus is very obvious when shooting with fast set up and no test conditions or special skills are required to have this experience. In case you only shoot with smaller apertures and under daylight conditions NO PROBLEM sir ! But he... where would you buy a K5 for ?

I have never decried the importance of accurate focusing in low light conditions. Nor did I ever mean to say that the low light FF issue was not obvious in some K-5 models. All I was saying was that my own K-5 seemed to have passed the test before and after the firmware update. This does not negate the experiences of others who may be having major FF problems both before and after the firmware update.

So Falk did not do the testing because it is difficult to see the problem in everyones shooting life but to find out where it comes from and maybe even how it could be solved.
Again Pentax should hire those kind of people in order to get their act together.

Which is what I suggested when I mentioned such a scientific study of the issue should not go unremunerated...

The best DSLR in the world becomes a very unreliable partner for a more serious photographer with these kind of quirks the K5 had and still has after 1.03 !

Not every K-5 owner has this problem, so as in the other long-winded thread in the K-5 forum... the K-5 is *not* a failure. Pentax has been and is still rectifying issues with its current models, which is what every other camera company does as well...

[Vaikis_]

problem that new light color sensor doesn't work corectly as expected, it can be sensor errors from Chromatic aberration.

[bjan]

I have never decried the importance of accurate focusing in low light conditions. Nor did I ever mean to say that the low light FF issue was not obvious in some K-5 models. All I was saying was that my own K-5 seemed to have passed the test before and after the firmware update. This does not negate the experiences of others who may be having major FF problems both before and after the firmware update.

You make me crazy . I tend to believe that there are different versions on the market with your story in mind. Who else is under this impression ?
(I secretely hope so)

[Ash]

You make me crazy . I tend to believe that there are different versions on the market with your story in mind. Who else is under this impression ?
(I secretely hope so)

I'm sorry, I do not understand this time.
Different versions of what? K-5s? I'm not sure of that.
But I'm so far convinced that the FF problem is manifesting itself in varying levels among the users reporting in this forum alone. If there is evidence of the contrary, I'd be interested to see it.
In the meantime, I'll just keep shooting with my accurately focusing K-5.

[falconeye]

But have you tested more than one K-5?

Is it possible that the difference between the positive experience I have had with the 1.03 firmware and yours is down to a a faulty camera body or lens on your end?

One K-5 only. But its AF performance (in absolute terms) still is pretty impressive. So, I doubt my unit is any worse than yours.

As said before, I'd like to hear about the circumstances your positive experience applies to.

I may ask for a replacement camera and if Pentax agrees, I would have a second unit to test.

The meter is needed in certain lighting situations but there should be a range of colour temperatures where the AF should obtain (almost) correct focus even in very low levels.

It would thus be a shame to lose the AF sensitivity in all situations just because it produces incorrect focus in some situations.

One of the new charts (DA*50-250) shows good focus accuracy below the focus shift transition for daylight. It was true daylight from an outside window made narrow enough to obtain low enough light values.

I think it really just is that the AF module falls back to a daylight hypothesis when in reality, no data is available.

Falk, I'd like remind you that the idea about Pentax AF problems because of CA in AF sensor is idea of GordonBGood and oleg_v.
And it's very reasonable idea.

The (longitudinal) CA dependency of the AF-measured phase is well described in articles and patents 10 years old.

So, this was the first possible assumption one could make.

But it was wrong. CA in the AF sensor isn't the problem as it is ubiquitous with phase detect AF sensors. A too low sensitivity of the accompanying colorimetric sensor is the culprit. And not even that is true. It is the lack of treatment in the firmware that both sensors have different light sensitivity which is the culprit. I'd say that in the K-7, both sensors had about equal sensitivity but they only upgraded one of them for the K-5 without adding a corresponding firmware handler.

And Pentax may still not understand this. Or otherwise speaking, why did they hack some special (yet to be identified) cases in the v1.03 firmware without addressing the above root cause? Of course, the other option is I made some errors in my analysis (or my K-5 is different) and Pentax did what was possible. Both is likely.

[Class A]

One of the new charts (DA*50-250) shows good focus accuracy below the focus shift transition for daylight.

I didn't see this chart when I first read the revised version and didn't find it when I looked for it now (in the HTML version). Where is it?

I think it really just is that the AF module falls back to a daylight hypothesis when in reality, no data is available.

Sure, but why then be "be bothered" about the camera not refusing to lock focus? In daylight (moonlight) lighting, trying to lock is the right thing to do and when the meter fails the camera has no way of knowing what light it should compensate for (except by guessing by using recent measurements or recent white balance results from recent images). Would you rather see the K-5 refuse to lock focus in any light and thus lose potentially spectacular locks in good conditions just to avoid mis-focus in adverse conditions?

Ideally the camera should warn the user that a mis-focus may have occurred due to a daylight assumption. Or better, it could check the image to see whether the daylight assumption was correct. It could even perform a contrast detect on the image taken, to check whether focus had been obtained (would be a slower "portrait mode" ).

A warning could prompt the user to check focus and/or use explicit white balance settings to provide information to the camera or ask the camera to obtain a colorimetric reading from the last image taken.

But it was wrong. CA in the AF sensor isn't the problem as it is ubiquitous with phase detect AF sensors.

Well, isn't the CA in the AF sensor the ultimate cause? If it weren't there then there wouldn't be any need for a colorimetric sensor (and the latter's failure to provide information in low-light or with slow lenses wouldn't matter).

A too low sensitivity of the accompanying colorimetric sensor is the culprit. And not even that is true. It is the lack of treatment in the firmware that both sensors have different light sensitivity which is the culprit.

I don't think I agree because if the colorimetric sensor really has pretty low sensitivity then fixing the firmware would produce a K-5 that struggles to obtain focus lock in relatively good light (e.g., dim modeling lights) and slow lenses. Is that really what we should expect from a camera in that price range? Was the K-7 really not any better? Why didn't we hear about K-7 front-focus reports or "My K-7 doesn't lock focus with modeling lights" reports then?

I'd say that in the K-7, both sensors had about equal sensitivity but they only upgraded one of them for the K-5 without adding a corresponding firmware handler.

Possible, but if it were true I wonder why your initial report didn't state that the K-5 has no FF problem in the range that it can be expected to focus (e.g., K-7 territory)? I'm not complaining about your initial report, I just wonder what to expect from the K-5.

I'm still not sure whether the "K-5 FF problem" is a result of exaggerated expectations (doesn't look like it because the f/1.4, EV -1 specification is not met, is it? The specification doesn't mention a specific light requirement) or a real failing to perform in conditions where one should expect a camera of this caliber to perform (i.e., at least reach K20D / K-7 levels).

Whether the V1.03 fix does everything what can be done given the hardware used is very hard to tell.

---

Last edited by Class A; 03-28-2011 at 06:56 PM.

[falconeye]

I didn't see this chart when I first read the revised version and didn't find it when I looked for it now (in the HTML version). Where is it?

You're right. The HTML export crippled some of the side-by-side images (again). I refer to the right side of Fig.16 in the PDF version, available here as well:
http://www.falklumo.com/lumolabs/articles/k5focus/images/results03/60PhaseWhite03-b0_18.png

There are some acceptable (EV -4) and one excellent (EV -2) focussing result with daylight with a lens showing problems with tungsten at EV +2.

Sure, but why then be "be bothered" about the camera not refusing to lock focus?

I won't contradict you. It may be a mere matter of philosophy.

It is like the difference between release and sharpness priority. Given that the K-5 supports both modes (meant for AF.C though), why not extend the meaning to this problem domain too?

Personally, I prefer my tool to not hide false assumptions. Even with metering, I can have under and overexposure warnings. But no AF fail warning. The least I can ask for is an AF fail warning after the image is taken. Because then, the AF spot color can be determined and the assumption be asserted. In the 1.03 update: nothing of this sort. This is what disappoints me. I am disappointed with respect to the easily feasible. Not nenessarily wrt the competition.

Well, isn't the CA in the AF sensor the ultimate cause? If it weren't there then there wouldn't be any need for a colorimetric sensor (and the latter's failure to provide information in low-light or with slow lenses wouldn't matter).

Ok, then phase AF is the ultimate cause because it can never be as accurate as contrast AF. I prefer the system view and there, the achromatic AF relay lens and the colorimetric sensor form a system not inferior to an apochromatic AF relay lens.

I don't think I agree because if the colorimetric sensor really has pretty low sensitivity then fixing the firmware would produce a K-5 that struggles to obtain focus lock in relatively good light (e.g., dim modeling lights) and slow lenses. Is that really what we should expect from a camera in that price range? Was the K-7 really not any better? Why didn't we hear about K-7 front-focus reports or "My K-7 doesn't lock focus with modeling lights" reports then?

I've seen 5DmkIIs engage their AF assist light with studio modelling lights ...

If a camera doesn't lock focus immediately in the studio, it normally is close enough and just using AF assist light or another subject feature does the job. Still better than a soft result after the session is over... Note that a K-5 still would be very fast and would still outperform a K-7 because it wouldn't need to hunt.

Possible, but if it were true I wonder why your initial report didn't state that the K-5 has no FF problem in the range that it can be expected to focus (e.g., K-7 territory)? I'm not complaining about your initial report, I just wonder what to expect from the K-5.

When the initial report appeared, I didn't yet understand the nature of the problem (the two sensors of different sensitivity). And then, I didn't study the K-7 to be sure about the relative performances.

But I insist in one thing: if my tool says "ready" it must BE ready ... That's the difference between tools and toys. And you know this. You're originally from a toolmaker's country

I'm still not sure whether the "K-5 FF problem" is a result of exaggerated expectations

Again, my problem is not where the K-5 cannot focus. It is where the K-5 fools you and says it can while it can't.

[WerTicus]

my K-5 failed me under modeling lights... but i noticed the issue and turned up their brightness thus resolving the front focus. This was with 1:02 though.

[Class A]

I refer to the right side of Fig.16 in the PDF version, available here as well:
http://www.falklumo.com/lumolabs/articles/k5focus/images/results03/60PhaseWhite03-b0_18.png

Thanks.

Personally, I prefer my tool to not hide false assumptions.

I fully see your point and I agree. On the other hand, there should be a way to exploit the AF module's sensitivity in cases where it is safe to do so.

The least I can ask for is an AF fail warning after the image is taken.

Yes, it wouldn't be hard to do, except perhaps for high-frequency shooting modes.
One could even imagine a self-calibrating system that adjusts the AF compensation as long as there are hints that the same scene is still being shot (time stamps, metering). Could be a mode that could be turned on and off.

Ok, then phase AF is the ultimate cause because it can never be as accurate as contrast AF.

Touché, but phase AF is fine in most situations and we just want it to perform in more situations.

I prefer the system view and there, the achromatic AF relay lens and the colorimetric sensor form a system not inferior to an apochromatic AF relay lens.

Hmmh, I prefer the full sensitivity of an AF module with an APO relay lens over an additional colorimetric sensor with a dependency on the widest aperture of a lens, potential alignment issues, different sensitivity for different colours, etc. Having the metering chip double up as a colorimetric sensor probably makes it less sensitive for its metering purpose. It appears to me that the colorimetric sensor solution used by Pentax is primarily motivated by cost reduction considerations. Nothing wrong about that as long as it works.

I've seen 5DmkIIs engage their AF assist light with studio modelling lights ...

I trust the owner knew what they were doing and it wasn't in an "always on" mode or something similar.

That's the difference between tools and toys. And you know this.

Sure. Admittedly, my expectations towards cameras/lenses in this price bracket aren't as high as they used to be. Lenses without QC issues seem to be the exception rather than the norm and cameras have issues like tilted viewfinders, mis-calibrated electronic horizons, mis-aligned AF point overlays, general FF/BF of the body, firmware bugs, etc. Maybe I'm giving all these reports too much weight but it seems one needs a bit of luck to get copies (of lenses or cameras) that live up to their "tool" specification and are not let down by "toy tolerance" issues. I'm probably being unfair here but I'm a bit annoyed that I cannot order a camera with confidence but rather would expect to find out that I'll have to send it back for one reason or the other.

Again, my problem is not where the K-5 cannot focus. It is where the K-5 fools you and says it can while it can't.

Would you say that the K-5's AF performance is at least as good as that of its predecessors?
If so, I'd be thinking of a K-5 purchase more favourably again. However, why were there FF-complaints from users? If they were unfounded, why did Pentax release a "fix"? Did Pentax believe there was no real problem but did some small optimisation just to be seen to react? Why are some reporting significant improvements?

The whole story still doesn't quite add up for me and at the moment I don't feel confident that I could get a K-5 that performs in reasonable conditions. I would not want to resort to my K100D for some shots, that would be silly wouldn't it? Maybe the K-5 is fine in all situations in which the K100D would lock focus but at the moment I don't have a good feel for whether or not that's the case.

P.S.: It appears you are of the opinion that the K-5 doesn't have an FF problem but rather a "false focus lock" problem. I'm fine with that view as long as the K-5 doesn't FF where a K-7 / K20D would not either. However, I'd still find it a bit sad that the apparent sensitivity of the AF module is let down by a comparatively insensitive colorimetric sensor.

[Ash]

Falk, you're right. The more aggressive approach the K-5 AF module has with its focus locking mechanism may therefore lead to a false assurance that the AF has accurately locked onto its subject, and allow the photographer to take the shot in any case. The K20D would keep trying under low lighting and either lock on well or give up. The K-5 I've found gives up less easily, but in low non-tungsten lighting it does try reasonably hard to fine focus and hit the spot accurately.

As I mentioned in the other thread comparing the K20D with the K-5, I had to try really hard to get my own K-5 to front focus, but that was only in EV1-2 lighting. I haven't yet tried candlelight or distant street lighting to know how it would fare in those conditions, but come to think of it, I don't know when the next time I'd be faced with shooting under those conditions.

[falconeye]

Class A and Ash: full agreement.

my K-5 failed me under modeling lights... but i noticed the issue and turned up their brightness thus resolving the front focus. This was with 1:02 though.

Do you remember the shooting conditions? Probably a low key situation? Any chance you replicate the situation with 1.03?

I still hope that 1.03 brought improvements in situations with a gray background. A situation I didn't test in my study.

Because if it now performs in low key studio situations and just fails under the conditions as laid out by my study (i.e., rather dark really), then I aggre with Class A that the remaining complaint is the missing warning rather than a failing focus.

[dlacouture]

Class A : having owned a K7 for a full year before the K5, and shooting my kids in relatively unchanging environments, the K5's FF problem leaped at me, so I'd say there is definitively a gap in the K5 low-light AF capabilities compared to the K7...

Regarding the v1.03, in casual use, I've seen an improvement, but the exact situations where it does occur are quite unknown to me... Seems to me that a white target surrounded by dark is now better handled, but I can still see nasty cases of FF with faces in relatively well-lit situations...

Yesterday I took snap of my newborn baby in some low-light settings (some at 1/20, f/1.4, 3200 iso, so that's "EV" 0.3), and they all came out perfectly focused, even given the hair-thin DoF at this aperture...

And sometimes, the same lens will repeatedly miss focus in EV4 situations! Quite stressful...

[areidjr]

And sometimes, the same lens will repeatedly miss focus in EV4 situations! Quite stressful...

That's been my biggest frustration with the K-5. The circumstances under which it will FF seems to be a bit unpredictable. If I knew exactly where it would fail I could work around it.

[deejjjaaaa]

Yesterday I took snap of my newborn baby in some low-light settings
.

you are not supposed to use any tungsten bulbs in EU !