2011, March 27.
Note: I updated the article to version 1.1.
I think, the feedback and experiments done after version 1.0 helped to understand the observed focus behaviour rather well and version 1.1 takes this into account.
An important (new) aspect is a description how low key studio photography, esp. when combined with an f/4 lens (or slower), can lead to inaccurate focus.
2011, March 07.
Fig.1: Accuracy of the Pentax K-5 phase detect AF vs. luminosity in EV. The above chart includes all measurements, i.e. various lenses, light colors, distances and apertures. The accuracy is measured as deviation of the focal plane from the sensor plane, in µm.
In modern times, each new release of a digital SLR camera seems to be accompanied by teething troubles. This applies to all makes across the board. Sometimes, they are fixed quickly by the vendor, like the “Green Line Syndrome” video issue in the Pentax K-7, the “Hot Pixel” video issue in the Nikon D7000 or the “String of Pearls” stain issue in the K-5. Sometimes, they aren't like the shutter-induced blur issue with the K-7 which LumoLabs succeeded to document in this blog.
Currently, there are two remaining teething troubles for the K-5 which are widely reported: wrong PTTL exposure with external flashes in some situations (confirmed in writing by 3rd party flash makers, even for Pentax' own flashes). And a systematic wrong lock of autofocus in low (tungsten) light.
Therefore, LumoLabs has decided to have a closer look at the issue. After careful evaluation and many hundred test shots we found the issue to be real. Pentax has unofficially reported to work on the issue. The pressure is on them to address the issue and the author hopes that our findings may contribute to their efforts. Buyers of the K-5 must be able to be confident that the issue is fixed sooner than later.
A preliminary copy of the paper was provided to Pentax earlier this week and the head of Pentax Europe officially receives a printed copy today. I have been told that Pentax engineering will receive a copy too.
The study
The results are too complex to be presented in the scope of a blog. Fig.1 above provides a first idea of our work.
Please refer to the complete paper for our findings:
- www.falklumo.com/lumolabs/articles/k5focus (HTML version)
- www.falklumo.com/lumolabs/articles/k5focus/K5Focus.pdf
(PDF version for download and offline reading)
In a nutshell, this is what we find:
- 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.
- 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).
- 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.
- 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.
- 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.
- 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. ;)
Move focal plane by 0xFF µm? Yes, do it so! :)
This may describe what's going on behind the curtain: a µm-valued variable becomes 0xFF (255) and causes a false shift of the focus plane by 255 µm. I call it the +0xFFµm hypothesis. :)
- Is it likely? No.
- Is it possible? Yes.
Please, read the full paper linked above if you have further questions and come back here to leave comments or questions. Thank You.
HI Falk, a truly great read..like always :)
ReplyDeleteI personaly have a K-5 with 18-55 WR (sold recently), DA* 16-50 f2.8 SDM and 35mm f2.4.
I noticed that only 18-55 WR and 35mm f2.4 have front focus in low light (house, appartment lightning) but 16-50 which has a SDM does not front focus...how come is that? Can you look into the matter? TNX.
Davor
Thanks a lot for your detailed analysis. I can tell you, that the Pentax K-r has the same issue. I use Pentax FA-50/1,4 and Tamron 28-75/f2,8, both lenses show the same front focus issue on low lights. I'm upset about this an honestly - I'm currently thinking of a Nikon D90 to replace my Pentax stuff!
ReplyDeleteoli
My K-5 front focuses in tungsten light level of 5-6 EV (metered with external lightmeter) even with fast lenses (50/1.4, 31/1.8, 77/1.8 etc.). With light level low enough to activate focus assist light it is always in spot (because of AF assist light of course). I don't agree with your findings, that it's a light level that causes the issue. I can properly focus with just a little glow of my (LED backlit) notebook and there is a significant front focus with 500W halogen modeling light of my studio flashes (with a subject/model close enough to be extremely well lit - say over 8 EV). So far I can override the issue by activating AF micro adjustment whenever I'm in a tungsten light environment.
ReplyDeleteHi Falk,
ReplyDeleteI really appreciate your effort. This is the most systematic analyze of the K5 front focus problem I have read so far.
Are you going to double check whether the front focus problem is Luminance dependent? (ie. add more Tunsten lights for more Luminance to see if it can compensate the problem)
TF
@jaad75: Thanks for your contribution. Basically, you find LED works and Halogen doesn't. Makes me think that there's more to it than rgb or color temperature, maybe a specific wavelength. You give reason for to carry out another experiment ;)
ReplyDelete@TF: Actually, the study contains the check for luminance vs. tungsten: you can check yourself if tungsten amplifies the problem. Just look at the dots (their color) and check more reddish/more yellowish ones have more blur. What I see is they have a focus shift earlier. But that could be explained by slight insensitivity for red.
Falk, I'm almost sure it's a wavelength or a range of it - that's why some people claim their tungsten micro adjustments don't work with some types of tungsten balanced compact fluorescents. BTW, have you seen this:
ReplyDeletehttp://www.youtube.com/watch?v=d2W5gwTtLqc
Hi Falk,
ReplyDeleteInteresting stuff, as usual. On my K10D, I have found my 31 Ltd consistently backfocuses in low light. This doesn't happen with other similar lenses (such as the 77 Ltd), which leads me to believe there is something dodgy about the 31 Ltd, which otherwise focuses fine in good light. With other lenses in low light the K10D will randomly misfocus either behind or in front, without any preference for either.
Hi, Falk. Great article. Thank you very much for tackling a topic that few reviewers would dare, but this is exactly what people need to know. Your shutter blur article was already most appreciated. Wish more reviewers would also do the right thing and think about their responsibility to their readers, and check or at least report these types of problem. It would be hard to think that those reviewers never heard of these highly publicised problems.
ReplyDelete@jaad75, yes, I've seen the youtube. While impressive, I hesitate to cite it because it is very much a macro situation.
ReplyDeleteBTW, before I settled for my three rgb bulb setup, I made some experiments with a flashlight which I could switch between tungsten and LED. While the problem did occur earlier with tungsten, the focus jump would occur with LED too if made dark enough.
If you look at Fig.4 of the study, you'll see that blueish light "jumps" focus at an about -2EV darker light level. Maybe this already covers the difference between tungsten and LED. However, halogen may be even worse than tungsten which can't be light color alone.
645D seems to be work better with its brighter lenses and may be little bit different XII+ implementation, may be they have copied the specs from 645D just but missed something ?
ReplyDeleteFalk, I guess -2EV by your measures or 0EV by external lightmeter (if I understand it correctly) is low enough for me...:)
ReplyDeleteAnyway with my K-5 front focusing problem occurs in relatively bright tungsten environment in my kitchen (let's put halogen modeling lights aside for a while). As I wrote before there is 5-6 EV (according to my Sekonic), so it still doesn't fit to your theory...
@jaad75, regarding your 6-8EV focus trouble. You say you can correct for it with AF microadjustment. Can't it be you see the "normal" defocus which can become as large as 140µm (cf. my EV8 shadow sample in the chart of this article).
ReplyDeleteI doubt the focus shift I talk about can be corrected with an AF micro adjustment. But maybe, I am wrong.
I've just briefly gone through your paper and think that your proposed formula for calculating the (absolute) "defocus amount" in um would not be reliable and exact enough as it is all based on the amount of edge transition blur which can be affected by various other factors in particular the optical quality/defects of a lens (which could then be affected by different aperture settings and sample variation, etc.).
ReplyDeleteTo obtain the true +/- um project image focus error at film plane, I believe that one should shoot a chart with inclined scale but which should have a centre FLAT contrasty object. The sharp point where it appears on the scale is measured of which this value (for the focusing errors) should be mapped into the value of defocus distance (no matter +ve or -ve) at the sensor plane by taking account into the focal length, aperture and so on. After all, this value will be real and exact.
Nevertheless, I must thank you for all your hard efforts spent and the good intention for doing this extensive test, which is surely very interesting and is meant to be helpful. Keep up your hard works, Falk, I am sure that I will always there to read your great tests. Finally, I just wish to say thank you also for your unreserved sharing and contribution to the Pentaxian community!
@Rice: Thanks for your kind comment and you are right.
ReplyDeleteMy defocus measurement is not as reliable and exact as it could be. I hope I said so in the paper too. But it isn't a flaw of my method.
I could calibrate each set of parameters (light color/luminance/distance/aperture/lens) for its value b0 using a manual focus bracket and using the best value. As one would do for a resolution test (not that test sites other than photozone.de would do so -- most just use a manual LV focus which wouldn't be good enough either).
However, my goal wasn't absolute defocus measurements but the general interdependency of defocus and other variables like brightness. So, I reduced the workload on me (I try to be efficient actually) and worked with a single value b0 for all measurements. I don't agree my method wasn't good /enough/ (for the given purpose). But others may disagree.
While your method (tilted test chart method) is used by many, I don't for a number of reasons. First, there is a risk that clutter enters the AF spot with wide angle lenses or larger distance. E.g., the area covered by the center AF spot with an FA31 in 1 m distance is large, maybe more than the width of A4 paper... Therefore, most tilted test chart results are either with too close distance or with too cluttered center. Then, there is a problem of field curvature which renders the method useless with, e.g., an DA*16-50. More importantly though, I need a method allowing automated numeric (and objective) extraction of accuracy, possibly to subpixel accuracies (although not required this time). Something seriously missing in most other tests. Without an automated procedure, you cannot run the large number of tests (and I only publish a subset) I do. But without the large testing base you cannot reach into the domain beyond the anecdotal test.
So, the progress with my method is that I actually abandoned the subjective evaluation of tilted chart test shots. And if necessary, it's precision can be increased. I actually should do so in order to study what's going on around EV 8 ...
Maybe, I should clarify a bit more. My method uses b0=2.19px.
ReplyDeleteIf the correct value for b0 would be 1.19 or 3.19px resp., then the corresponding value for 100µm (f/1.8) would shift to 103µm or 96µm which is a 4% error caused by a 50% error in b0. I hope you can see why I worked as I did.
I agree that the error is large for small values of defocus and indeed, there are no measurements resulting in a small defocus value for this reason. And I said so in the paper. But for the current study, I wasn't particularly interested in the region where the AF excels. It was more the region where it doesn't and my method was meant to be the best one to find it.
Can I get your advice please. My husband has just purchased a K-5 for me.
ReplyDeleteI had been hoping for a Nikon D90 after considering this and a Canon for a number of reasons, but my husband has been doing a lot of research behind the scenes wanting to surprise me and was hearing a lot of buzz about the Pentax K-5. My brother works for Pentax, so was able to get this significantly more expensive camera for a price closer to the D90 or at least close enough to consider it as an option. From a technical perspective, the K-5 sounds great - and likely a better purchase than the D90 which is no longer being made as far as I know.
After researching Canon and Nikon - and then holding them in my hands and considering the ergonomics of them and menus etc. - I had decided I wanted a Nikon D90. And, am hoping to get a 50mm 1.4f lens. The photographers whose style I like - family, lifestyle, stills, natural light etc. - all use this lens most of the time - Canon or Nikon.
I just want to ensure this IS a suitable investment long term as this is my first digital SLR and once you buy into a family you tend to stick with the line/investment.
I hadn't done any research on Pentax. I'm a graphic designer and all the photographers I know or have worked with use Canon or Nikon - so I'm a little weary and wondering if there's good reason for this that I should be aware of.
My hope and intention over the next few years is to get a lot more technically savvy and comfortable shooting in various natural light settings - family, friends, stills lifestyle etc. - indoors and out. Probably do some more artistic experiments as well for wall prints etc. I also find myself wanting to shoot in low light situations without a flash at parties etc., events or gatherings. Often at night. Which is why this information in this link concerns me with the K-5. Trouble focussing in low light.
http://falklumo.blogspot.com/2011/02/lumolabs-pentax-k-5-low-light-focus.html
But, to be honest some of these terms are a little over my head at this point, and I'm a bit overwhelmed trying to figure out (under some time pressure) if this IS a wise investment for me to keep.
I would appreciate any thoughts you have on this.
@Anonymous female graphic designer with a husband refusing to listen ;)
ReplyDeleteYou wanted a D90 and got a K-5. Good or bad?
Well, as you correctly realize, it is more a decision for a brand than a camera. Pentax is small (say 5% market rather than 40% market) because they had a bad start into the digital age and Hoya costed a few percent share when they took them over. Nevertheless, Pentax traditionally is a very strong and innovative brand and is currently growing quickly.
In general, Pentax currently is the brand of choice for photographers committed to APS-C, an SLR body as rugged and small as possible for street and field photography and outstanding image quality. Currently, there is no other APS-C camera with an equal image quality (be it low light or at sunshine) to the K-5 and with the Pentax Limited prime lenses, it can be exploited. It is however, a bad choice if you have firm plans to upgrade to 35mm full frame in the foreseeable future. Also, if you're an avid action or bird shooter, rather get a Canon 7D or Nikon 300Ds.
You ask about the low light focussing trouble described in my blog article. Pentax says (sometimes) they are going to fix it. If they do not, this is what it means: The K-5 may produce soft photos (with its phase autofocus) under very specific conditions: These include a lack of light (where a D90 probably would not focus at all) and/or a dark background or a slow lens. The main reason why people are complaining is that the K-5 won't tell you it missed focus then. Last time I saw the problem in the studio was where a Canon 5DmkII didn't focus w/o AF assist light at all.
In summary, I would not make my decision Nikon vs. Pentax depend on this little nasty problem which probably soon will be solved anyway.
You'll have to decide what camera you'll love more. If you don't fall in love with the K-5, with its small Limited prime lenses, the rugged and actually beautiful body, the silent shutter (which leaves Nikon shooters confused) and the fabulous image quality and dynamic range, then don't keep it. Photographers have to love the tool they use.
I hope this helps you.
Thanks so much for taking the time to provide helpful and informative advice.
ReplyDeleteLove your entrance by the way ;0). Gotta love him. His intentions were to get me the best we could afford - this is a big purchase for us. I think he did just that - and I hate to second guess his choice. But, I'd hate to invest so much on the wrong path for me at the same time.
I really like the feel, lack of sound and operations of this camera. It feels like a German car to me. Well-used space and planning etc. And, I think I'll probably really enjoy using it.
My only hesitation now is that all the photographers I currently network with are Canon or Nikon - and this is a big learning environment for me. However, in doing my research I'm learning there's a pretty strong Pentax community to learn from.
And, also - your comment about switching to full frame. I'm really not sure if it's something I'll move toward or need to. Is this something that Pentax doesn't currently provide or is simply not their strength?
The most action I'll likely be shooting is my constantly moving young boys and fun-loving families - so I don't think that's a concern.
I appreciate the perspective on the focussing issue. It's not my deciding factor any longer.
Thanks again. I'm sure I'll be back. The tech side of all of this is NOT my strength!!
please take a look at the new firmware released
ReplyDeletehttp://www.pentax.jp/english/support/digital/k-5_s.html
it says auto focus improved
@GfK. Thanks for the pointer. I plan to run a test with one lens over the weekend. And update my report accordingly.
ReplyDeleteGreat Work!
ReplyDeleteOne question however. I understand that It can be expected to see cameras front focus in tungsten. But why some of those new Nikons D7000 shift to back focus in tungsten? Do you care to investigate?
Just an observation about why there are so many Canon and Nikon (esp. DLSR) users. It is my sense the high rate of Canon and Nikon users primarily stems from the coattails of popularity. Many people assume something is good if it is popular. This notion stands to reason at many levels, but could also be attributed to a simple desire to own something that is perceived to be "better" than another brand. And a lot of people tend to follow this perception with the least amount of toil required. In other words, they would rather just buy a camera that is popular than look into any comparisons on a more detailed level. And if you look at Pentax there are plenty of reasons to buy their products instead, and while their sales are improving....they still lag behind the Canon/Nikon juggernaut. The aggressive marketing advantage Canon and Nikon enjoy (wisely) also contributes to their success....and this aligns perfectly with the popularity notions I mention. As Falk stated Pentax started off very slow and are trying to catch up....so it's ironic that Pentax was the leader in film SLR's for many years.
ReplyDeleteThat's not to say Canon and Nikon don't make great cameras. They do. And a lot of professionals won't use anything else for a variety of reasons having nothing to do with popularity. But Pentax is making strong inroads toward their relative monopoly (and Sony and Panasonic are bridging the gap with their strong mirrorless approaches) and I can foresee the day when there is more of a competitive balance (a parity that can actually be good for Canon and Nikon users)....though one would expect much of that hinging on Pentax and their marketing ideas (even though I like that they think about the cameras and the photographers who use them more than slick advertising).
In short, Pentax is the red-headed stepchild (as Seinfeld would say "not that there's anything wrong with that"), but the child is growing up and learning from their years of "therapy". Don't be surprised to see the kid become President someday.
When I was young Pentax were as popular as Canon (Nikon was then much more expensive than either of these brands), but not until the K-5 the Pentax brand has again gotten major interest from the consumers.
ReplyDeleteDisagree. The K-5 has certainly received a lot of well-deserved accolades, but as far as interest from consumers....the K-x, from the entry-level point of view, was a big success. The combination of great reviews, low price, excellent high ISO results (rivaling even some full frame cameras), and some radical color choices (esp. for DSLR) really helped Pentax become a viable option for beginner DSLR owners.
ReplyDeleteFor the first half of 2010 it was #6 in all DSLR sales. The Nikon D5000 with its articulating LCD was #3 and the omnipresent Canon 550D was #4). Considering Nikon and Canon dominate the advertising market, it's a pretty decent showing for a relatively niche company like Pentax.
For anybody watching the comments here: I published another blog article to look at the improvements brought by the v1.03 firmware.
ReplyDeleteAs for the comments above about market aspects: Let me make myself clear: I consider the K-5 to be the best APS-C camera today. All aspects combined. But Pentax really could make it a bigger hit if it reacted quicker to user feedback.
Hi
ReplyDelete"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."
Just to confirm after glancing at your blog entry and your report. Are the EV figures quoted above 'exposure EV values'?
We just bought a K5 a couple of days ago and immediately noticed frequent front focus with our fast f/1.4 50mm Pentax lens. Initially we played around with the AF fine adjustment and then came across your blog entry.
Just trying to correlate our experience in terms of front focus with our 'exposure EV' values with the values you quote.
Thanks
@Anonymous: The EV values I am quoting are explained in ample detail within the report. I don't consider EV values as obtained by a camera's meter to be accurate enough. They are accurate for daylight and an 18% gray target which turns out mid gray in sRGB. In all other cases one has to apply corrections.
ReplyDeleteThank you for your analisys of the problem, I have a K7 that also frontfocuses in low, tungsten light (but contrast-detect focus is spot-on).
ReplyDeleteI've always wondered why K7's AF-assist is green while those mounted on "legacy" flashlights and on camera from other manufactures is red/orange.
I came to the conclusion that my autofocus simply hates low, warm lights... and that's a pity since AW-balance is good.
Hello
ReplyDeleteThank you for your very interesting study
Now you have defined way to optically measure defocus, have you tried to compare these measures with the AFPredictor values (Defocus measure from AF sensor point of view?) present in Exif data?
I would be very interested to know about the correlation between these 2 values
@Anonymous, yes, I looked at correlation between defocus and AF integration time, AF predictor and AF defocus. But I found no strong correlation. All values go up if it gets darker and AF integration time maxes out at 255 ms or so. It is mostly at these max. AF integration times (but not always) that the front focus issue shows.
ReplyDeleteI'm very surprised that you found no correlation between AF Predictor and your optical measure of defocus
ReplyDeleteAF Predictor seems to give the amount of defocus (positive value for a BF, negative for FF) as measured by the AF sensor
You should at least observe, in normal conditions, a correlation between the sign of the AF Predictor and the sign of your df indicator
@Anonymous, you should take into account that the AF tries to make defocus zero. It accepts less perfect focus for low light but defocus will NOT measure the front focus issue. Otherwise, it wouldn't exist in the first place. Moreover, I didn't study the effect of small defocus or its sign.
ReplyDeleteThank you for your time to share this excellent and informative post. It 's my first time, during my visit here. I found a lot of material in your article. Keep it up. Thanks.
ReplyDeleteDigital Cameras Canon
I wanted to thank you for this great read!! I definitely enjoying every little bit of it I have you bookmarked to check out new stuff you post.
ReplyDeleteDigital SLR Cameras
I wanted to thank you for this great read!! I definitely enjoying every little bit of it I have you bookmarked to check out new stuff you post.
ReplyDeleteDigital SLR Cameras
I have acquired yesterday a k-5 and observed immediately the focus problem in low light, compared to a pentax k-x which works great. Now I'm going back to the shop to return the k-5. That focus problem is not acceptable in my opinion.
ReplyDeleteIs there any evidence that SR could improve autofocus accuracy in low light? I read about a patent somewhere stating that SR assists autofocus.
ReplyDelete@Andrew: I am unaware about such a thing and haven't seen any evidence for such an effect. Sensor shift SR and phase AF should be independent things. However, SR should improve contrast AF in low light a lot and w/o applying any patent. And there is a patent where the phase AF module sits on its own shift motor. The reason why SR helps with contrast AF is that contrast AF w/o SR would have to work on blurred images in low light (i.e., contrast would be random and never be high). Pentax SR is active during contrast AF :)
ReplyDeleteAny news, any hopes of improvement from Pentax?
ReplyDeleteNo news I am aware of.
DeleteDo you think the infra red could be rsponsabile for it (see also http://ricehigh.blogspot.com/2011/12/deepere-technical-investigation-into-k_16.html)
ReplyDeleteI don't think so. When the issue was first reported with the K-5, this was examined and it was found that IR cut filters don't alter the effect. However, I did not study this myself.
DeletePentax hasn't made yet any improvement in the AF module since its launch in the *ist camera back in 2003 (19 years with the same in the af module!!!!). You can see in the following link the patent http://www.freepatentsonline.com/6937818.html
DeleteThe AF sensor are designed to work accurately when f-number is 5,6 or smaller (8,11, etc.) but when we try to take a picture at, let's say 2.8, our troubles start, specially if the focus is close, because depth of field is really narrow.
The problem is really big if we are using large apertures and light with a warm color temperature.
Is there a solution? Well IMHO, it could be fixed with a better color temperature detection system , aka, auto white balance.
Pentax needs to design a new and better af module that works accurately at 2.8 apertures and more than 11 focus points. SAFOX VIII to IX+ are basically the same. There are only minimal software improvements between them.
SAFOX IX+ is IMHO obsolete for a long time, could be a decent af system for a entry level dslr, but not for a flagship model.
@Anonymous, your statements are basically wrong with only a small core of truth. E.g., you may look at various editions of FNAC (France) test magazine and see how much the Pentax AF performance (both AF.S and AF.C) improved over generations. In two more recent tests, the K-5 has beaten all competitor AF on accuracy. The K-5 AF is considered by most to be on par with the Nikon D7000 AF which means that it isn't yet in one league with Nikon D300 or D4.
DeleteI agree, there is room for improvement and this article deals with one: the problem with very low light when the light color isn't measured accurately anymore.
Your unspecific criticism really goes too far and I will not accept further comments from your part. Thanks for understanding.
I know about the fnac test. I'm a Pentax user for a long time, an I'm only saying that the improvements are no enough to compete with, for example, Nikon D300. There are lot of people asking for a brand new af system. Thats all what I said. Accept it or not, but thats is the reality. Thank you.
ReplyDelete