Reading mode: Light Dark

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

Go to last post
Errors seem to be very systematic depending on light level, i.e. it should be possible to fix. It is also strange that focus assist lamp is not used correctly. There is various things that can help to come over this problem in software. The module and algorithm are just developed for 645D and K5 and 645D has much more light to use from its large size MF lenses and worked accurately in halogen modeling light. If you fit the same thing into APS-C it needs some adjustment.
 
I question how K5 stacks up to it peers, good bad or indifferent. Do you plan on testing other manufactures models?
--
Dave
 
First of all, good work, thank you for the research but I do have a few questions....
falconeyes said:
LumoLabs: Pentax K-5 low light focus study is published

1. The K-5 as it presently ships indeed has a flaw in its phase detect autofocus module or software which causes it to front focus in low light below a lens-dependent threshold in EV.
Click to expand...
I do not really understand lens threshold.... Is that at optimal exposure and does it change when , for example in night shootingI stop down to get the picture dark enough to be realistic.

For example next pic:



1/30s f/2.0 at 24.0mm iso1600

If I use "green button it wold get much brighter but in order to refelct teh situation I stopped down or used faster exposure to get a more real life view.
falconeyes said:
2. If it does, it seems to consistently focus ≈ 255 µm behind the sensor plane (although with a significant ± 75 µm scatter of results which is about twice as large as the normal scatter of result).
Click to expand...
The 75 µm scatter is indeed significant, more that 1/4th. But that 255 µm, how much FF would that really mean in an image (ofcourse depending on FL but someone must know how much that is on a 31, 50, 100 or whatever mm FL.
Then we can use a DOf calculator etc etc....
falconeyes said:
3. Faster lenses seem to keep working in lower light but of course, are prone to more blur when the front focus does eventually happen. Slower lenses can already start to front focus at light levels metering as 4 EV or 6 EV even. A fast lens may work down to 0 EV in white light.
Click to expand...
weird....
falconeyes said:
4. Light sources other than daylight emphasize this problem as they simply appear darker to the AF module. Moreover, it seems to be moderately color blind for red which further emphasizes the effect in deep tungsten light.
Click to expand...
Mayb that explain that K-7 and K20D also hade FF or BF probs in certain sitiations I know???? (Actually in there my K-5 (stained version) did better than the others...)
falconeyes said:
5. The effect is real and can negatively impact the daily work of a photographer. On the other hand, it is possible to run into a low light tungsten situation without the problem.
Click to expand...
shouldn;t that be "nightly" work?
falconeyes said:
Please, refer to the full study and accompanying material before discussion. I am available to answer questions. Enjoy the read :)
Click to expand...
Oh man, I have enough reports of meetings and decisions to read now already....
--




falconeyes said:
The difference between genius and LBA is that genius has its
limits.
  • Janneman ( adaptation of the Kings quote from Albert Einstein)
Click to expand...
 
kadarpik said:
Errors seem to be very systematic depending on light level, i.e. it should be possible to fix. It is also strange that focus assist lamp is not used correctly. There is various things that can help to come over this problem in software. The module and algorithm are just developed for 645D and K5 and 645D has much more light to use from its large size MF lenses and worked accurately in halogen modeling light. If you fit the same thing into APS-C it needs some adjustment.
Click to expand...
If the 645D doesn't suffer it is likely to be the K-5's software as hypothesised in this paper.

I hope it is a software problem otherwise I may skip the K-5 too.

K10/20D were mostly fine, AF is slow and inaccurate in brighter light than the newer cameras though.. K-7 shutter blur and K-5 AF error.

--
Regards,
Dan

My Photos:
http://www.pentaxphotogallery.com/danielfranco
http://www.flickr.com/photos/honourabletyr/
 
Eric,
I've enjoyed reading your comments before and so respect your opinions.

There are 2 women in my photoclub that work together on weddings, both have bought the D700. seems to be popular for the wedding pro's.

I'm curious why the D7000 is being passed by. It too seems to be troubled by some AF problems, and i don't have enough information to determine if Nikon is able to fix these problems for their customers. There are some limitations to the D7000 that i'd rather not buy into, besides my having a number of Pentax lenses at the moment.

But thats my question - why did you pass over the D7000?
--
Phil B
K20D, K10D
 
janneman02 said:
I do not really understand lens threshold.... Is that at optimal exposure and does it change when , for example in night shootingI stop down to get the picture dark enough to be realistic.
Click to expand...
Well, if you didn't read the answer in the paper ...

In a nutshell: EV is EV if an 18% gray target under white light exposes as 18% gray (50% in sRGB). Otherwise, you have to apply corrections which I did.
janneman02 said:
If I use "green button it wold get much brighter but in order to refelct teh situation I stopped down or used faster exposure to get a more real life view.
Click to expand...
As I said, image and target brightness must be cinsidered, not just EXIF.
janneman02 said:
Then we can use a DOf calculator etc etc....
Click to expand...
The paper contains an example. IIRC, it says something like 25cm on 1m distance at 31mm f/1.8.
janneman02 said:
janneman02 said:
3. Faster lenses seem to keep working in lower light but of course, are prone to > > Please, refer to the full study and accompanying material before discussion. I am available to answer questions. Enjoy the read :)
Click to expand...
Oh man, I have enough reports of meetings and decisions to read now already....
Click to expand...
Drop those other reports. I am sure they are boring to death ;)
 
Frankly, I find all this a little troubling (not the tests but, the results). I guess people writing about this were not crazy after all (sometimes it is user error you know).

I really don't want to buy into this until I feel it's addressed. Sometimes you only have one chance to get a shot. Until I know the outcome of all this, or if there will be a fix through software - I am probably going to hold off on purchasing a K5 at this time - keep on with my trusty 2 K20s that I have.

Will you keep us posted? Will you let us know if you hear back from Pentax engineers about this? Maybe there is a simple fix for it.

I still think the K5 is a great camera but, for what I need it for AF needs to be accurate.

Thanks for doing the study.
 
Hi Falk:

Thanks for the study. In real world use I find the K5 has accurate low light focus about 80% of the time and misses when you least expect it to. However, my copy seems to be much better than the Kr. Are there any plans to do a study on the Kr as there is likely a larger issue on this model?

As others have stated it would be useful to compare to D7000 and 7D to get a across brand perspective. My sense is that no brands are trouble free.

Dale
--
http://www.pbase.com/abundant108

http://www.pentaxphotogallery.com/home#section=ARTIST&subSection=272176&subSubSection=1787360&language=EN
 
Dale108 said:
As others have stated it would be useful to compare to D7000 and 7D to get a across brand perspective. My sense is that no brands are trouble free.
Click to expand...
I keep seeing these "other brands also have the tungsten problem" posts lately, and they seem a little aside the point here...

Well, this is not even an inter-brand problem, but an intra-brand one, and a big one at that!!!

The K7 was particularly successful at tackling those low-light/tungsten lighting problems (at least, mine was, when compared to my K10 and K20 before it, and my Kx alongside it). I've run quite some tests on this problem with each of these bodies (I've discovered this problem while shooting with my K10), and never could push my K7 to a fault : it either focused spot-on or failed to lock.

Now, Pentax specifically advertised the K5's SAFOX IX+ as being an improvement over the K7 in low-light focusing, which is true in term of speed, but glaringly false in term of accuracy (for me, at least).

We are not complaining about a tungsten focus shift here, which plagued AF modules since day one, way back in the 80's... Such a focus shift is actually small and often results in a mere softness of the image, not a blatantly misfocused shot as I can often see with my K5...
 
moving_comfort said:
Falk, can you elaborate on this statement a bit? (from section 4.9, item 4)

4. EV specifications are typically according to external light meter readings and then the transition would probably appear to be at 2 EV rather than 0.5 EV.

What EV specification frameworks are at play here? It seems you refer to more than one.

(Thanks, btw, huge effort on your part, exceptional work.)
Click to expand...
I think I explained a bit about the problem in the paper too.

To compute EV from shutter, aperture and ISO is straightforward and I called the number LV1 in my paper.

Most will stop here.

However, this value only makes sense if an 18% target is metered and its photo turns out 18% gray too. In daylight white and with no levels applied. This is the standard situation which is almost never met.

The easy part is to cope with sRGB (18% appears to be about 50% then) and to switch off levels in Lightroom (requires setting contrast and brightness and black level and levels to zero)

However, the target is white and the light is colored. This will lead to 2-4 lower values for EV after proper correction. Compared to LV1 alone. Where one has some freedom how to correct for light color. My version leads to lower values than a simple color temperature method. But my colors were too extreme to compute a temperature.

In short, I look into the image and compute a 2nd LV2 from the deviation between the white target and the mid-bright and colored image. This 2nd value is negative. Then I add both values.
 
Suse B said:
Will you keep us posted? Will you let us know if you hear back from Pentax engineers about this? Maybe there is a simple fix for it.

I still think the K5 is a great camera but, for what I need it for AF needs to be accurate.

Thanks for doing the study.
Click to expand...
I don't know how this thread will be going. But I will certainly update my blog as I learn more about the situation.
 
What he said.

And as much as I would like to do a comparison this other brands, it isn't on my short list. First, I'd like Pentax to sort this out.

However, I think phase AF focus accuracy is an underetimated subject, by testers at least. It is work to test it out, agreed. And maybe, my methodoligy is new actualky motivating others to do similiar tests. I don't know.

With increasing resolution, I think AF sensors, esp. with APS-C, run into limits. Read the image understanding white paper (vited in the study) to understand why it is a bigger problem with APS-C than full frame. IMHO, this should become a standard testing procedure for cameras, any brand.
 
I get confused.

I read your reports which intimates categorically that the problem is systematic and affects ALL k5's.

Then I take a k5 turn the lights down really low and take shot after shot pin sharp.
How do you explain the discrepancy of real world and your lab results.?

Examples below









I have gone through 3 k5's (stain) the images are from the second, the first behaved the same and I haven't tested the third for low light accuracy.

I've yet to find a k5 that exhibits these alleged traits , I can show that AF varies with colour temperature as expected but not a systematic AF failure under low lighting conditions.
--
My PPG

http://www.pentaxphotogallery.com/home#section=ARTIST&subSection=1471087&subSubSection=0&language=EN
My Photo Stream
http://www.flickr.com/photos/awaldram/
 
falconeyes said:
falconeyes said:
Oh man, I have enough reports of meetings and decisions to read now already....
Click to expand...
Drop those other reports. I am sure they are boring to death ;)
Click to expand...
Boring they are but also about people loosing their jobs and budget cuts in the firedepartment so that actually makes any FF issue rather irrelevant...
falconeyes said:
falconeyes said:
Then we can use a DOf calculator etc etc....
Click to expand...
The paper contains an example. IIRC, it says something like 25cm on 1m distance at 31mm f/1.8.
Click to expand...
This is to me the confusing part, 25 cm on a distance of 1 m???

Are u sure??? I have heard people complain about "out of focus" or "oft". Some people sa they really had to look into it but this is HUGE...

Odd thing, I do cheat on DPR a bit and do visit other forums, not just the one that shall not be named here but a few local Pentax forums (NL, BE) an I have not ever heard on one of those of this issue.... Even the thread I started("Do you have the FF issue?") gt just a few "No"'s

--




falconeyes said:
The difference between genius and LBA is that genius has its
limits.
  • Janneman ( adaptation of the Kings quote from Albert Einstein)
Click to expand...
 
Gotta agree with you there.

The other thing I find strange is the statement we often see that contrast detect is more accurate as if this is to be expected.

Now when I went to school admittedly many years ago phase detect AF is many times more accurate if pre-calibrated correctly

The advantage of Contrast detect is it doesn't require additional hardware to function lowering manufacturing costs and is more fool proof not requiring pre-calibration.

The down side is contrast detect requires bags of light.

My test seem to indicate the status quo has not altered and Phase detect will continue to operate reliably long after contrast can no longer function, But this again seem to go against the popular opinion here (and seems only here).
--
My PPG

http://www.pentaxphotogallery.com/home#section=ARTIST&subSection=1471087&subSubSection=0&language=EN
My Photo Stream
http://www.flickr.com/photos/awaldram/
 
awaldram said:
The other thing I find strange is the statement we often see that contrast detect is more accurate as if this is to be expected.
Click to expand...
Funnily, both my Tamron 28-75 and 70-200 focus very badly in CDAF, even in full daylight... The 70-200 will misfocus by about 20cm on a 4m target! Direction of misfocus depends on the direction the lens was traveling : BF if the lens was going toward infinity, FF in the other direction... It's as if the camera could not stop the lens fast enough...

I cannot even start to find an explanation to this behavior...
 
awaldram said:
... Then I take a k5 turn the lights down really low and take shot after shot pin sharp.
How do you explain the discrepancy of real world and your lab results.?
Click to expand...
In your case, it's a typical example of a systematic user error.
Sorry I couldn't help it :).

--
Cheers,
Alex
 
il_alexk said:
awaldram said:
... Then I take a k5 turn the lights down really low and take shot after shot pin sharp.
How do you explain the discrepancy of real world and your lab results.?
Click to expand...
In your case, it's a typical example of a systematic user error.
Sorry I couldn't help it :).
Click to expand...
That fine if my 'user error' causes my camera to behave impeccably in low light but correct usage results in miss-focused shots then all power to my 'user error'. :)

Mind you "An 'error' is a deviation from accuracy or correctness" so I guess its not my error after all.

http://en.wikipedia.org/wiki/Error
il_alexk said:
--
Cheers,
Alex
Click to expand...
--
My PPG

http://www.pentaxphotogallery.com/home#section=ARTIST&subSection=1471087&subSubSection=0&language=EN
My Photo Stream
http://www.flickr.com/photos/awaldram/
 
If you can see a mistake in my approach to reproducing this 'malaise' the k5 suffers from then please enlighten me.

We all make mistakes and me more than others :)

I mean I feel really left out what with lab results an all, If a Spanish shopkeeper can test 100 k5's and the're all faulty whats wrong with mine ?

Maybe your correct and I'm to stupid to see it.

I had the stain (twice) I can reproduce quite a few P-ttl bugs so I want to miss-focus with everybody else as well.

I mean 25% inaccuracy should really be easy to spot shouldn't it?

--
My PPG

http://www.pentaxphotogallery.com/home#section=ARTIST&subSection=1471087&subSubSection=0&language=EN
My Photo Stream
http://www.flickr.com/photos/awaldram/
 
You must log in or register to reply here.

Keyboard shortcuts

f
Forum
m
My threads
Mobile site
About
Editorial content
Cameras & Lenses
Community
All content, design, and layout are Copyright © 1998–2025 Digital Photography Review All Rights Reserved.
Reproduction in whole or part in any form or medium without specific written permission is prohibited.
When you use DPReview links to buy products, the site may earn a commission.
©GPS Media - Guides, Products, Services.
Back
Top