[falconeye]

Is your set of criteria a positive list only? In other words, I guess that I cannot conclude from it that there is a significant improvement if the background is not bright, right?

Correct. I didn't anticipate a limited effect and didn't run enough tests.

In order not to overload the discussion with too many results, let me append ONE (of many) test where I did the following:

FA31 f/1.8, 1m, white wall background, black&white subject, WB at daylight, AF assist light off, phase AF center spot.

The light is a normal 40W tungsten bulb dimmed between 0% and 100% (orange dots) and a red 40W tungsten bulb (red dots). All other explanations as described in the study.

The attachments are as following:
1. from firmware version 1.02.21.07.
2. from 1.03.22.11 (current).
3. a separate test for 1.03 (not available for 1.02) where WB was manually set to reddish or auto.

I think WB set to Auto or Daylight didn't make a difference (I received a separate PM pointing into the same direction). I tend to believe that the 1.03 fix may only have a positive impact with a darker background or a slower lens. But this remains to be confirmed.

Attached Images

     

---

Last edited by falconeye; 03-18-2011 at 02:53 AM.

[RonHendriks1966]

Thanks Falk,
this does confirm my findings yesterday. I was in a bar (ST.Patricks Day) and had to adjust my camerasettings to get good focus with my DA*55mm lens. True not the best settings, but there was a clear difference in focussing between the stage and behind the bar where the light was of different colourtemperature.
Gr Ron.

[dlacouture]

Well, this is disturbing...

I'm gradually finding such shortcomings to this v1.03 update. I have good results with light targets against dark backgrounds, but the other way around it's less successful... It's almost as if they reversed the behavior...

I'll give the manual WB selection a try...

[planedriver]

Hi all, Guten Tag Falk! I've also updated my K-5 with the 1.03 version and my first impression was that it made things better under simple situations and the AF became usable down to EV level 3 or even 2. However under "more complex" lighting situations it is still very unreliable.

Here are my pictures. Airplane cockpit, Sun has just gone down in front..not completely dark, and all sorts of light sources from the LCD displays, map-lights and bulbs in the panels. Focus point was on my collegaue's face and as you can see the panel on the overhead panel framed by a yellow frame is in focus-not his face.

K-5 da12-24 iso3200 12mm f4 1/10





now with the DA35 2.4: iso2500 1/25 f2.4 EV-0.7

focus was on his eyes..but the glasses came out in focus.





Also interesting that the same da12-24 under similar conditions has the tendency to back-focus on my K-x.

Now thinking about my previous experiences I always wondered how is it possible that pictures taken in similar conditions (and within a few moments) come out with different color temperature? I have the impression that the ones which are perfectly exposed are the ones with correct WB..now I have to check my files but I think that those are the ones which are also focused correctly. ..or ?

Nothing scientific just impressions..


Andras

[Class A]

The attachments are as following:
1. from firmware version 1.02.21.07.
2. from 1.03.22.11 (current).
3. a separate test for 1.03 (not available for 1.02) where WB was manually set to reddish or auto.

Maybe your lens requires a new AF calibration after the firmware update? It seems that the new firmware introduced a slight overall FF bias for your lens. Someone posted that resetting their AF adjustments and then recalibrating made a difference, but I personally wouldn't give the latter story too much weight. It might be that the AF adjustment for bright daylight also changed a bit. If so, it could mean a global improvement for your V1.03 results. Still leaving a nasty jump...

It seems that V1.03 has better "red" performance. In the V1.02 shots one can observe red dots that despite being located at higher EV have higher defocus. This never happens with V1.03. It seems that a linear fit through the orange and red dots respectively shows a higher negative slope for the V1.03 firmware. In other words, the performance in very low light is worse, but it gets better faster than the V1.03 version (at least for the red dots). So I'd say one can see an effect (not sure if it is statistically significant, though, and the "slope" observation might be just an artefact of too few samples).

[starbase218]

I would say the second example can be attributed to the AF areas being much larger than the point you see in the viewfinder. I wouldn't be surprised at all if an area would span his left eye and the frame of his glasses. The sharp contrast of the bright frame against a dark background makes for a very appealing focusing target.

[areidjr]

Correct. I didn't anticipate a limited effect and didn't run enough tests.

Thanks for your efforts, Falk.

I can't say that I'm surprised by your results. After 1.03 was released I immediately bought a K-5 (I already had and returned 3) so I could use it for an event (indoor wedding) last Sunday. The results were unimpressive. Maybe 15-20 properly focused shots in challenging lighting conditions and maybe an 5-10% totally OOF. The rest had just enough FF to make render them useful, but quite disappointing. OTOH, I did manage to capture the event without missing any shots. When using my K100DS, the shots I got were all good and properly focused, but I missed a lot due to the AF not locking.

The other observation was that the wider the FL, the worse the problem. So, for me, the new FW was, perhaps, an improvement, but not a fix. I still find the K-5 AF to be unpredictable and unimpressive under artificial light.

[jolepp]

That cockpit is a useful scene in that it allows one to see the front focus in a real-life picture, usually there FF plane is in thin air and all you really know that the focus is not good. Otoh it might be a difficult scene for color temperature measurement as there is light from the (tungsten?) bulbs mixing that coming in the windows. Then again the cameras (viewfinders) and objectives (focus rings, feel and throw) are apparently built with the assumption that AF works, so it should, in all circumstances where the sensor can take a resonably nice snap as such, or then we should have the option of a Pentax split prism / microprism focus screen at least (the lack of such a beast is also telling as for assuming AF to work and/or most using it). That AF does not work well in dimmish light is to be expected I guess, then again it would be most useful just there as MF gets an (especially ) iffy business (especially ) with the large aperture lenses one gets to push available light photography to the new frontier made possible by sensor (low-ISO) improvements.

My test scene experience with the DFA 100mm 1:2.8 is that AF remains useful to somewhere between EV 2 - 1 which is enough for me as such. However, the Tamron 17-50 1:2.8 I would use in real life shots seems to fail between EV 3 - 2 which is so, so OTOH this is requires -10 even in daylight and is not a Pentax lens so maybe it is not fair to expect it work and/or I should be cross to Tamron about that ... it seemed to work pretty well with the k-x though when I first got it. The primes I have are MF so testing them would be testing my own techique as well as the accuracy of the focus indicator so the results would not mean much either way .

I guess I'll eventually slap in the El-Cheapo split prism focus screen currently sitting in a drawer and see how well I can compete with the AF system as it is. Unfortunately, mucking with even that is not such a good idea when I still need to send the body in for stains (and was holding on doing that in case dealing with the AF issue would need sending in so that the two issues might be combined in one service round trip). At this point getting separated from the K-5 might be good for both the owner and the gadget in case the AF woes in fact continue ... at times I find myself wondering whether the thing would make a satisfying sound when dropped from a height, or what amount of TNT might completely vaporize it ...

[starbase218]

@Falk: Some people say that Pentax DSLRs are not the only ones suffering from focus issues in badly lit situations. Do you know if this is the case? Perhaps it is an idea to test some other DSLRs as well and see how they behave?

[Class A]

...at times I find myself wondering whether the thing would make a satisfying sound when dropped from a height, or what amount of TNT might completely vaporize it ...

Having some fun, eh?

Best to not obsess for a while. I feel your frustration and boy I wished myself the firmware fix had been successful. Maybe it is good news for me. I get to sit out the K-5 as well, after having resisted the K-7. My K100D is still going strong and maybe my next upgrade (K-3?) will be a full-frame and/or without issues (hey, one can dream, right?).

Or, if the Sigma SD1 will be affordable, I might ask Sigma whether they want to swap my three Sigma K-mount lenses for the same models in Sigma mount. If not, I've read that the Sigma mounts is a K-mount mechanically and a Canon EOS mount electronically so some conversion should do the trick.

Come on, Pentax, bring me home to the flock!

[falconeye]

Maybe your lens requires a new AF calibration after the firmware update? It seems that the new firmware introduced a slight overall FF bias for your lens.

This isn't what data tells.
My methodology needs fine gauging (focus bracket series calibration for every color temperature, lens, aperture and distance) to properly cover the area below 30µm. A step I skipped because I care for the area above 100µm only.

My FA31 needs no calibration and actually focusses pretty well in full daylight. Results around 70µm shall be considered "good" and it may be either front or back focus then. This is within the "Canon 1/3 DoF region" spec too. Just look at the results worse than 70µm.

@Falk: Some people say that Pentax DSLRs are not the only ones suffering from focus issues in badly lit situations. Do you know if this is the case? Perhaps it is an idea to test some other DSLRs as well and see how they behave?

This is true but I won't test it.

The K-5 manages to focus dark enough for it's light meter to fail (with an f/5.6 lens). It leaves every other dSLR in the dust AFAIK. The point is that the K-5 would lock focus in a way where it fools you believe it aquired focus while in fact it was a false focus. This is much less a problem with the K-7 or with other brands, from what I hear. I think this is the downside of the K-5 AF module's spectacular light sensitivity.

[Pål Jensen]

@Falk: Some people say that Pentax DSLRs are not the only ones suffering from focus issues in badly lit situations. Do you know if this is the case? Perhaps it is an idea to test some other DSLRs as well and see how they behave?

Autofocus isn't very precise anyway and all AF systems give a ballpark figure. Why not use manual focus and use AF when it is needed in order to get the shot?

[Smeggypants]

Hi all, Guten Tag Falk! I've also updated my K-5 with the 1.03 version and my first impression was that it made things better under simple situations and the AF became usable down to EV level 3 or even 2. However under "more complex" lighting situations it is still very unreliable.

Here are my pictures. Airplane cockpit, Sun has just gone down in front..not completely dark, and all sorts of light sources from the LCD displays, map-lights and bulbs in the panels. Focus point was on my collegaue's face and as you can see the panel on the overhead panel framed by a yellow frame is in focus-not his face.

hey there planedriver, this pretty much sums up my findings as well.

I found my K-5 was front focussing much more with more complex lighting. A combination of light from my computer monitors and tungsten light made it front focus much more than just tungsten light

Also interesting that the same da12-24 under similar conditions has the tendency to back-focus on my K-x.


Andras

I foudn that when I got to EV0 or undet my K20D started to back focus btu he K05 was stil front focussing.

[Smeggypants]

Hi there Falk

My tests on 1.03 were done with both FA50/1.4 and sigma 30/1.4 both lenses at f1.4 and at a subject distance of 30 inches ( which is near enough a metre )

My white balance was set to Auto and the back ground was not bright in most tests/ Light colour was tungsten in the tests that showed front focus, although a more complex lighting condition ( computer monitor+tungsten ) made the FF far worse

Hi friends,

I am through the analysis of the v1.03 data and I am a bit confused. Maybe, somebody can help before I write my update.

According to my measurements, v1.03 brings no significant progress IF the following conditions are met:

1. The FA 31 lens is used at 1m distance and f/1.8 aperture.
2. The background is bright.
3. The camera white balance is manually set to daylight (maybe doesn't matter).
4. The light color is tungsten (maybe doesn't matter).

Wrt (3.), I see some positive effect if the camera white balance is set to automatic or tungsten. However, I don't have enough data points (I have 21 measurements for this case) to call this effect significant, i.e., these 21 data points all lie within the cloud of 94 data points obtained with daylight white balance. I even see some negative effect of v1.03 if the subject has a red color and the lamp has not.

OTOH, I trust the numerous positive reports and I must assume that I must miss this effect for some reason.

Maybe, the white balance effect is significant. Maybe, the effect is only noticeable with lenses slower than f/1.8. Or maybe, only at higher EV values with darker backgrounds.

Unfortunately, I am far away from my lab and can't do further measurements for the next three weeks.


Is there anybody wanting to check if the positive v1.03 effect disappears with a wide aperture and/or a non-matching camera white balance?


Thanks everybody.

[starbase218]

This isn't what data tells.
My methodology needs fine gauging (focus bracket series calibration for every color temperature, lens, aperture and distance) to properly cover the area below 30µm. A step I skipped because I care for the area above 100µm only.

My FA31 needs no calibration and actually focusses pretty well in full daylight. Results around 70µm shall be considered "good" and it may be either front or back focus then. This is within the "Canon 1/3 DoF region" spec too. Just look at the results worse than 70µm.


This is true but I won't test it.

The K-5 manages to focus dark enough for it's light meter to fail (with an f/5.6 lens). It leaves every other dSLR in the dust AFAIK. The point is that the K-5 would lock focus in a way where it fools you believe it aquired focus while in fact it was a false focus. This is much less a problem with the K-7 or with other brands, from what I hear. I think this is the downside of the K-5 AF module's spectacular light sensitivity.

So you're basically saying that although the K-5 may under certain conditions frontfocus, other cameras are likely to not be able to focus at all in those situations?