LumoLabs: Pentax K-5 low light focus study is published

dlacouture · 03-02-2011, 02:11 PM

Christine, I think that the ample variation of reports is mainly because the target has nearly more importance than the lens and light levels used...

I went nuts trying to figure out why my K5 focused properly on focus charts at lower light levels than when I had obvious FF doing portraits of my kids... It's simply because their lovely pinkish faces actually register 2-3IL lower than a B&W target...

Christine Tham · 03-02-2011, 02:21 PM

Originally posted by dlacouture

Christine, I think that the ample variation of reports are mainly because the target has nearly more importance than the lens and light levels used...

I went nuts trying to figure out why my K5 focused properly on focus charts at lower light levels than when I had obvious FF doing portraits of my kids... It's simply because their lovely pinkish faces actually register 2-3IL lower than a B&W target...

According to Falk's theory it's all based on the amount of light reaching the "calorimetric" sensor (whatever that is).

But, like you, I am also finding that the brightness of the target relative to background may have a bearing on the focus.

However, if you remember, my attempts to quantify the issue have not been successful - your experiments re a white target against a black background and vice versa seem to be more successful.

Of course, my experiments are further complicated by some of them done on a standard focusing screen, and some done on a 3rd party focusing screen, and I suspect the difference had an impact on observed behaviour.

So my fear is that whilst Falk's theory explains a lot, it may not explain everything.

Class A · 03-02-2011, 02:28 PM

Originally posted by dlacouture

Using a tungsten flashlight into the viewfinder, my camera focused properly!!!

Thanks for the test. Very illuminating.

Originally posted by dlacouture

So, under a given light level threshold, the camera will revert to "daylight focus"...

If this happens "as designed" than this doesn't bode well for a fix since it may mean that the colorimetric sensor is too insensitive. Maybe it isn't the same as in the K-7? Let's hope that there is something else going on and/or that the user will have a manual way (be it through a WB setting, yuck) to let the AF sensor know what the colour temperature of the target is.

Originally posted by dlacouture

I guess the reports of misfocuses in "low daylight" situations are actually based upon the false assumption that when the sun just went down, the WB is still "daylight", whereas in reality there is an heavy shift toward blue...

Bluish light should cause back-focus.

Gimbal · 03-02-2011, 02:45 PM

Originally posted by dlacouture

BINGO!!!!

Using a tungsten flashlight into the viewfinder, my camera focused properly!!!

Tried to repeat the test but couldn't get anything useful out of it. Both LED and tungsten light seemed to either do nothing or if bright enough confuse it so much it just couldn't find focus at all.

A couple of times the camera desperately turned on the assist light (even though it wasn't THAT dark) when trying to fight the back light through the eyepiece.

dlacouture · 03-02-2011, 02:48 PM

Originally posted by Class A

If this happens "as designed" than this doesn't bode well for a fix since it may mean that the colorimetric sensor is too insensitive. Maybe it isn't the same as in the K-7? Let's hope that there is something else going on and/or that the user will have a manual way (be it through a WB setting, yuck) to let the AF sensor know what the colour temperature of the target is.

Yeah, that's what I think, too... And this explains the behavior in my previous experiment, where I used a black object against a white background (proper focus as the light temp can be easily determined), then a white object against a black background (FF, surely not enough light to determine the light temp).

Maybe it's just a case of "trusting" the '+' sensor results a little while longer?

Originally posted by Class A

Bluish light should cause back-focus.

Reading my post again, I found it quite obscure! So what I said (using complicated and useless phrasing!) is exactly what you said : the misfocuses occurring in low "daylight" are surely caused by the bluish cast of evenings, thus causing a back-focus...
Will have to wait tomorrow for this test!

Another thing just came to my mind : some may remember that I claimed I had severe back-focusing under fluorescent lighting? I was not able to reproduce it without reproducing the exact situation, so I let it go...
But come to think of it, it occurred in the exact situation where I had a small target surrounded by darkness.
So I guess the camera reverted to "daylight focus" again...

Catalana · 03-02-2011, 02:49 PM

Time to get busy designing a "third party" Tungsten Light viewfinder attachment for the K-5.

We can call it the "Front Focus Correction Adapter".

Christine Tham · 03-02-2011, 02:56 PM

Okay - based on the report that resetting the camera to "factory settings" "cured" the FF problem, I suspected that AF fine adjustment may skew the AF correction.

So I did two tests.

AF Fine adjustment turned on but with the adjustment parameter set to 0:

AF Fine adjustment turned off:

So it seems turning the AF fine adjustment on and off does impact the issue.

PS - I will cross post the results on the K-5 thread as well. Interesting!

Quicksand · 03-02-2011, 03:15 PM

Arg! So many variables!

dlacouture · 03-02-2011, 03:41 PM - **2 Likes**

Originally posted by Gimbal

Tried to repeat the test but couldn't get anything useful out of it. Both LED and tungsten light seemed to either do nothing or if bright enough confuse it so much it just couldn't find focus at all.

A couple of times the camera desperately turned on the assist light (even though it wasn't THAT dark) when trying to fight the back light through the eyepiece.

Sorry, I should have said "toward the viewfinder", not "into", as I actually hold the flashlight at quite an angle to the viewfinder (I hold it sideways at something like 60° from the viewfinder axe, toward the viewfinder). Its light did not really shine into the viewfinder, but more like "grazed" it from the side.

Otherwise, the light could travel all the way through the lens and mess with the AF sensor by lighting the target itself!

EDIT :
Just done another dumb test: still using my "100% FF" test scene, I added a small tungsten light at the edge of the scene, but not as a focus point, and taking care that it would not light the AF target. So, theoretically, it should have no impact upon focusing, except by giving a good light temp reading to the colorimetric sensor...

Guess what : the camera focuses spot-on!!! It surely identifies the tungsten light and applies the proper focus shift...

falconeye · 03-02-2011, 04:51 PM

Originally posted by dlacouture

BINGO!!!!
...
So, under a given light level threshold, the camera will revert to "daylight focus"...

Originally posted by dlacouture

...
Just done another dumb test: still using my "100% FF" test scene, I added a small tungsten light at the edge of the scene, but not as a focus point, and taking care that it would not light the AF target. So, theoretically, it should have no impact upon focusing, except by giving a good light temp reading to the colorimetric sensor...

Guess what : the camera focuses spot-on!!! It surely identifies the tungsten light and applies the proper focus shift...

dlacouture, you are my hero!

You did both tests I didn't find time to run and found results as predicted by the theory.

Moreover, you found that the colorimetric sensor falls back to daylight if it can't take a light temperature measure. So, the 250µm shift I see probably is the correction value between daylight and warm tungsten which is a hefty value.

Probably, the corresponding shift for the K-7 is smaller (if it exists) and maybe it is so because the AF aperture is smaller (less sensitive to light).

So, after all, we may see the downside to the increased light sensitivity of the new AF module. I've been told as a child that there are two sides to the same coin

Originally posted by Christine Tham

Are we sure the K-7 does not have a problem?

Now that you ask: No, actually not. But maybe, it is less pronounced.

Originally posted by Christine Tham

So my fear is that whilst Falk's theory explains a lot, it may not explain everything.

That's my fear too. Esp. I have some doubts how it would explain what is going on below the Planck scale (sorry, not everybody will get the joke ...

).

Originally posted by Christine Tham

So it seems turning the AF fine adjustment on and off does impact the issue.

Ok, you just added another dimension to the problem

Anybody wanting to repeat Christine's test? If confirmed, we should leave AF adjustment alone.

Just to be sure:
Christine, you entered "0" for both, per lens AND global adjustment, when using "0", right?

dgaies · 03-02-2011, 05:00 PM

Originally posted by falconeye

Anybody wanting to repeat Christine's test? If confirmed, we should leave AF adjustment alone.

I'll take a stab at repeating the test.

Class A · 03-02-2011, 05:22 PM

Originally posted by dlacouture

Just done another dumb test:...

Not a dumb test at all! Kudos for making all these useful tests and observations. (I'll add a rep in a minute).

Originally posted by falconeye

Esp. I have some doubts how it would explain what is going on below the Planck scale...

I also find the theory a bit lacking in terms of furthering the state of the art regarding quantum gravity.

Joking aside, Dlacouture's tests seem to indicate that the colorimetric sensor has a global view of the scene only and hence it won't be possible to have a proper solution. I had hopes that the colorimetric sensor might at least resolve the 77 metering segments and thus could focus on the area associated with a focus area, but that doesn't seem to be the case. Unless, of course, there is a stupid bug that causes the colorimetric sensor data to be ignored based on a global illumination value, rather than on the relevant local one.

Either Pentax has used capable hardware but needs to fix the software, or the K-5's AF might never be (almost) perfectly independent of the colour temperature of the target.

Christine Tham · 03-02-2011, 05:56 PM

Originally posted by falconeye

Anybody wanting to repeat Christine's test? If confirmed, we should leave AF adjustment alone.

Just to be sure:
Christine, you entered "0" for both, per lens AND global adjustment, when using "0", right?

Yes, I can confirm both adjustments were set to 0 for the out of focus shot.

I would like to caution that I don't believe turning off AF fine adjustment completely eliminates the problem. So it's not a magical cure all.

If you recall, when I first purchased my K-5 (just under a month ago, how time flies) I said I was able to replicate the problem, but only at very low light levels and the misfocus was so slight I said I did not believe it will significantly impact normal shooting. These tests were all conducted with AF fine adjustment turned off.

I have a suspicion that we are dealing with two problems here - or two slightly different manifestations of the same problem.

One is a "micro" problem - which occurs around EV -1 to EV 2 which is what you measured and results in a slight focus shift.

One is a "macro" problem - which can occur even at relatively high EV levels and can cause completely out of focus pictures.

With all due respect, I am not sure you have quantified the "macro" problem - and I don't believe anyone else has either.

It may well be that the "macro" problem is associated with turning on AF fine adjustment. I seem to recall that people who have reported significant out of focus issues were also people who have fiddled with AF adjustment and people who claim not to have experienced the problem were those that said they have left the AF fine adjustment at factory settings.

I am not sure, I have not cross correlated all the posts, justing putting the thought out to the ether.

dgaies · 03-02-2011, 06:29 PM

Originally posted by dgaies

I'll take a stab at repeating the test.

After running several tests, I've concluded that turning off the AF adjustment doesn't appear to have an impact on mitigating the FF issue. There were certainly pairs of test shots where the shot with the focus adjustment turned off was in better focus. However, there were also shots where they were equal and shots where the opposite was true; where turning off the focus adjustment made the shot worse. While I can't with 100% certainty conclude that there is no effect of turning the AF adjustment off, it seems more likely to me that differences shown in individual pairs of tests shots is just the normal variation in the amount of focus error at these low EV levels.

**Michaelina2** · 03-02-2011, 06:51 PM

Originally posted by falconeye

Christine, you entered "0" for both, per lens AND global adjustment, when using "0", right?

No...

Do a global AF reset for your lenses...

Menu > C26 AF Fine Adjustment > 2 On > Reset > OK > Menu > OK > Menu > Camera ready state

The read out after AF Reset should be:

Apply All.... +/- 0
Apply One.... Unset

--------

After the camera reset (Manual pp 335), I set the Program Line to MTF, selected "P" mode, then the C26 reset (above).

It may not be 'science,' but I see a HUGE improvement in AF accuracy. In fact, I no longer need to AF fine tune any lens... load, lock, and Go!

Cheers...

03-02-2011, 03:41 PM - 2 Likes	#264
dlacouture Veteran Member Join Date: Jul 2008 Location: Var, South of France Photos: Albums Posts: 1,074	Originally posted by Gimbal Tried to repeat the test but couldn't get anything useful out of it. Both LED and tungsten light seemed to either do nothing or if bright enough confuse it so much it just couldn't find focus at all. A couple of times the camera desperately turned on the assist light (even though it wasn't THAT dark) when trying to fight the back light through the eyepiece. Sorry, I should have said "toward the viewfinder", not "into", as I actually hold the flashlight at quite an angle to the viewfinder (I hold it sideways at something like 60° from the viewfinder axe, toward the viewfinder). Its light did not really shine into the viewfinder, but more like "grazed" it from the side. Otherwise, the light could travel all the way through the lens and mess with the AF sensor by lighting the target itself! EDIT : Just done another dumb test: still using my "100% FF" test scene, I added a small tungsten light at the edge of the scene, but not as a focus point, and taking care that it would not light the AF target. So, theoretically, it should have no impact upon focusing, except by giving a good light temp reading to the colorimetric sensor... Guess what : the camera focuses spot-on!!! It surely identifies the tungsten light and applies the proper focus shift... Last edited by dlacouture; 03-02-2011 at 03:58 PM.

03-02-2011, 05:22 PM	#267
Class A Pentaxian Join Date: Aug 2008 Location: Wellington, New Zealand Posts: 11,251	Originally posted by dlacouture Just done another dumb test:... Not a dumb test at all! Kudos for making all these useful tests and observations. (I'll add a rep in a minute). Originally posted by falconeye Esp. I have some doubts how it would explain what is going on below the Planck scale... I also find the theory a bit lacking in terms of furthering the state of the art regarding quantum gravity. Joking aside, Dlacouture's tests seem to indicate that the colorimetric sensor has a global view of the scene only and hence it won't be possible to have a proper solution. I had hopes that the colorimetric sensor might at least resolve the 77 metering segments and thus could focus on the area associated with a focus area, but that doesn't seem to be the case. Unless, of course, there is a stupid bug that causes the colorimetric sensor data to be ignored based on a global illumination value, rather than on the relevant local one. Either Pentax has used capable hardware but needs to fix the software, or the K-5's AF might never be (almost) perfectly independent of the colour temperature of the target. Last edited by Class A; 03-02-2011 at 05:59 PM.

Similar Threads
Thread	Thread Starter	Forum	Replies	Last Post
Any tips for low light focus with K-5	designinme_1976	Pentax K-5 & K-5 II	5	11-21-2010 08:38 PM
focus hunting in low light	sorin	Pentax DSLR Discussion	12	07-17-2010 02:20 PM
Low Light auto focus	JohnKSA	Troubleshooting and Beginner Help	6	03-10-2010 04:19 AM
Pentax Low Light Focus	indy1984	Pentax DSLR Discussion	15	02-17-2010 09:42 AM
EV low light focus question	tarsus	Photographic Technique	3	06-26-2008 08:22 AM

03-02-2011, 02:11 PM	#256
dlacouture Veteran Member Join Date: Jul 2008 Location: Var, South of France Photos: Albums Posts: 1,074	Christine, I think that the ample variation of reports is mainly because the target has nearly more importance than the lens and light levels used... I went nuts trying to figure out why my K5 focused properly on focus charts at lower light levels than when I had obvious FF doing portraits of my kids... It's simply because their lovely pinkish faces actually register 2-3IL lower than a B&W target...

03-02-2011, 02:21 PM	#257
Christine Tham Veteran Member Join Date: Nov 2008 Location: Sydney, Australia Photos: Gallery Posts: 2,269	Originally posted by dlacouture Christine, I think that the ample variation of reports are mainly because the target has nearly more importance than the lens and light levels used... I went nuts trying to figure out why my K5 focused properly on focus charts at lower light levels than when I had obvious FF doing portraits of my kids... It's simply because their lovely pinkish faces actually register 2-3IL lower than a B&W target... According to Falk's theory it's all based on the amount of light reaching the "calorimetric" sensor (whatever that is). But, like you, I am also finding that the brightness of the target relative to background may have a bearing on the focus. However, if you remember, my attempts to quantify the issue have not been successful - your experiments re a white target against a black background and vice versa seem to be more successful. Of course, my experiments are further complicated by some of them done on a standard focusing screen, and some done on a 3rd party focusing screen, and I suspect the difference had an impact on observed behaviour. So my fear is that whilst Falk's theory explains a lot, it may not explain everything.

03-02-2011, 02:28 PM	#258
Class A Pentaxian Join Date: Aug 2008 Location: Wellington, New Zealand Posts: 11,251	Originally posted by dlacouture Using a tungsten flashlight into the viewfinder, my camera focused properly!!! Thanks for the test. Very illuminating. Originally posted by dlacouture So, under a given light level threshold, the camera will revert to "daylight focus"... If this happens "as designed" than this doesn't bode well for a fix since it may mean that the colorimetric sensor is too insensitive. Maybe it isn't the same as in the K-7? Let's hope that there is something else going on and/or that the user will have a manual way (be it through a WB setting, yuck) to let the AF sensor know what the colour temperature of the target is. Originally posted by dlacouture I guess the reports of misfocuses in "low daylight" situations are actually based upon the false assumption that when the sun just went down, the WB is still "daylight", whereas in reality there is an heavy shift toward blue... Bluish light should cause back-focus.

03-02-2011, 02:45 PM	#259
Gimbal Pentaxian Join Date: Apr 2007 Location: Sweden Posts: 2,106	Originally posted by dlacouture BINGO!!!! Using a tungsten flashlight into the viewfinder, my camera focused properly!!! Tried to repeat the test but couldn't get anything useful out of it. Both LED and tungsten light seemed to either do nothing or if bright enough confuse it so much it just couldn't find focus at all. A couple of times the camera desperately turned on the assist light (even though it wasn't THAT dark) when trying to fight the back light through the eyepiece.

03-02-2011, 02:48 PM	#260
dlacouture Veteran Member Join Date: Jul 2008 Location: Var, South of France Photos: Albums Posts: 1,074	Originally posted by Class A If this happens "as designed" than this doesn't bode well for a fix since it may mean that the colorimetric sensor is too insensitive. Maybe it isn't the same as in the K-7? Let's hope that there is something else going on and/or that the user will have a manual way (be it through a WB setting, yuck) to let the AF sensor know what the colour temperature of the target is. Yeah, that's what I think, too... And this explains the behavior in my previous experiment, where I used a black object against a white background (proper focus as the light temp can be easily determined), then a white object against a black background (FF, surely not enough light to determine the light temp). Maybe it's just a case of "trusting" the '+' sensor results a little while longer? Originally posted by Class A Bluish light should cause back-focus. Reading my post again, I found it quite obscure! So what I said (using complicated and useless phrasing!) is exactly what you said : the misfocuses occurring in low "daylight" are surely caused by the bluish cast of evenings, thus causing a back-focus... Will have to wait tomorrow for this test! Another thing just came to my mind : some may remember that I claimed I had severe back-focusing under fluorescent lighting? I was not able to reproduce it without reproducing the exact situation, so I let it go... But come to think of it, it occurred in the exact situation where I had a small target surrounded by darkness. So I guess the camera reverted to "daylight focus" again...

03-02-2011, 02:49 PM	#261
Catalana Senior Member Join Date: Feb 2010 Location: California Posts: 190	Time to get busy designing a "third party" Tungsten Light viewfinder attachment for the K-5. We can call it the "Front Focus Correction Adapter".

03-02-2011, 02:56 PM	#262
Christine Tham Veteran Member Join Date: Nov 2008 Location: Sydney, Australia Photos: Gallery Posts: 2,269	Okay - based on the report that resetting the camera to "factory settings" "cured" the FF problem, I suspected that AF fine adjustment may skew the AF correction. So I did two tests. AF Fine adjustment turned on but with the adjustment parameter set to 0: AF Fine adjustment turned off: So it seems turning the AF fine adjustment on and off does impact the issue. PS - I will cross post the results on the K-5 thread as well. Interesting!

03-02-2011, 03:15 PM	#263
Quicksand Pentaxian Join Date: Apr 2008 Location: Central Ohio (formerly SF Bay Area) Posts: 1,519	Arg! So many variables!

03-02-2011, 04:51 PM	#265
falconeye Veteran Member Join Date: Jan 2008 Location: Munich, Alps, Germany Photos: Gallery Posts: 6,871 Original Poster	Originally posted by dlacouture BINGO!!!! ... So, under a given light level threshold, the camera will revert to "daylight focus"... Originally posted by dlacouture ... Just done another dumb test: still using my "100% FF" test scene, I added a small tungsten light at the edge of the scene, but not as a focus point, and taking care that it would not light the AF target. So, theoretically, it should have no impact upon focusing, except by giving a good light temp reading to the colorimetric sensor... Guess what : the camera focuses spot-on!!! It surely identifies the tungsten light and applies the proper focus shift... dlacouture, you are my hero! You did both tests I didn't find time to run and found results as predicted by the theory. Moreover, you found that the colorimetric sensor falls back to daylight if it can't take a light temperature measure. So, the 250µm shift I see probably is the correction value between daylight and warm tungsten which is a hefty value. Probably, the corresponding shift for the K-7 is smaller (if it exists) and maybe it is so because the AF aperture is smaller (less sensitive to light). So, after all, we may see the downside to the increased light sensitivity of the new AF module. I've been told as a child that there are two sides to the same coin Originally posted by Christine Tham Are we sure the K-7 does not have a problem? Now that you ask: No, actually not. But maybe, it is less pronounced. Originally posted by Christine Tham So my fear is that whilst Falk's theory explains a lot, it may not explain everything. That's my fear too. Esp. I have some doubts how it would explain what is going on below the Planck scale (sorry, not everybody will get the joke ... ). Originally posted by Christine Tham So it seems turning the AF fine adjustment on and off does impact the issue. Ok, you just added another dimension to the problem Anybody wanting to repeat Christine's test? If confirmed, we should leave AF adjustment alone. Just to be sure: Christine, you entered "0" for both, per lens AND global adjustment, when using "0", right?

03-02-2011, 05:00 PM	#266
dgaies Veteran Member Join Date: Oct 2009 Location: Maryland / Washington DC Posts: 3,917	Originally posted by falconeye Anybody wanting to repeat Christine's test? If confirmed, we should leave AF adjustment alone. I'll take a stab at repeating the test.

Thread Tools	Search this Thread
	Search this Thread: Advanced Search

03-02-2011, 05:56 PM	#268
Christine Tham Veteran Member Join Date: Nov 2008 Location: Sydney, Australia Photos: Gallery Posts: 2,269	Originally posted by falconeye Anybody wanting to repeat Christine's test? If confirmed, we should leave AF adjustment alone. Just to be sure: Christine, you entered "0" for both, per lens AND global adjustment, when using "0", right? Yes, I can confirm both adjustments were set to 0 for the out of focus shot. I would like to caution that I don't believe turning off AF fine adjustment completely eliminates the problem. So it's not a magical cure all. If you recall, when I first purchased my K-5 (just under a month ago, how time flies) I said I was able to replicate the problem, but only at very low light levels and the misfocus was so slight I said I did not believe it will significantly impact normal shooting. These tests were all conducted with AF fine adjustment turned off. I have a suspicion that we are dealing with two problems here - or two slightly different manifestations of the same problem. One is a "micro" problem - which occurs around EV -1 to EV 2 which is what you measured and results in a slight focus shift. One is a "macro" problem - which can occur even at relatively high EV levels and can cause completely out of focus pictures. With all due respect, I am not sure you have quantified the "macro" problem - and I don't believe anyone else has either. It may well be that the "macro" problem is associated with turning on AF fine adjustment. I seem to recall that people who have reported significant out of focus issues were also people who have fiddled with AF adjustment and people who claim not to have experienced the problem were those that said they have left the AF fine adjustment at factory settings. I am not sure, I have not cross correlated all the posts, justing putting the thought out to the ether.

03-02-2011, 06:29 PM	#269
dgaies Veteran Member Join Date: Oct 2009 Location: Maryland / Washington DC Posts: 3,917	Originally posted by dgaies I'll take a stab at repeating the test. After running several tests, I've concluded that turning off the AF adjustment doesn't appear to have an impact on mitigating the FF issue. There were certainly pairs of test shots where the shot with the focus adjustment turned off was in better focus. However, there were also shots where they were equal and shots where the opposite was true; where turning off the focus adjustment made the shot worse. While I can't with 100% certainty conclude that there is no effect of turning the AF adjustment off, it seems more likely to me that differences shown in individual pairs of tests shots is just the normal variation in the amount of focus error at these low EV levels.

03-02-2011, 06:51 PM	#270
Michaelina2 Site Supporter Join Date: Jan 2008 Location: SE Michigan USA Photos: Gallery Posts: 1,300	Originally posted by falconeye Christine, you entered "0" for both, per lens AND global adjustment, when using "0", right? No... Do a global AF reset for your lenses... Menu > C26 AF Fine Adjustment > 2 On > Reset > OK > Menu > OK > Menu > Camera ready state The read out after AF Reset should be: Apply All.... +/- 0 Apply One.... Unset -------- After the camera reset (Manual pp 335), I set the Program Line to MTF, selected "P" mode, then the C26 reset (above). It may not be 'science,' but I see a HUGE improvement in AF accuracy. In fact, I no longer need to AF fine tune any lens... load, lock, and Go! Cheers...