Forgot Password
Pentax Camera Forums Home
 

Closed Thread
Show Printable Version Search this Thread
12-29-2010, 07:03 AM   #286
Veteran Member
filorp's Avatar

Join Date: Jun 2008
Location: Aberdeen Scotland
Photos: Gallery | Albums
Posts: 389
QuoteOriginally posted by Wheatfield Quote
The focus error appears to be independent of colour temperature.
In my case i does related directly to color temperature - iv showed it already in this post https://www.pentaxforums.com/forums/pentax-k-5-forum/119808-k-5-flash-11.html#post1304633. And if Kr and K5 exhibits the same problem and assuming that both are the constructs of the same time period maybe the same designer even, thay can share the same design hardware bug - i hope not but like monthy pythons singed "always look at the dark side of life" .....

12-29-2010, 07:13 AM - 1 Like   #287
Pentaxian
falconeye's Avatar

Join Date: Jan 2008
Location: Munich, Alps, Germany
Photos: Gallery
Posts: 6,862
QuoteOriginally posted by Wheatfield Quote
The focus error appears to be independent of colour temperature.
That's not true. It depends but in a complex way.

I did one more test to rule out or confirm another possible explaination. The tests have been around 1EV which is above the threshold AF assist light is triggered. ISO100, 2s, f/1.8 to be precise.

First with tungsten, a bit brighter: focus almost ok.
Next with halogen: focus miss.

So I thought, it may be the flicker from artificial AC light. So, I tried my battery-powered headlight positioned to give indirect light, even a stop less:

So, next with 3LEDs: focus hit!

I thought this is it, it is the flicker the battery doesn't cause. My battery-powered headlight has two more settings: mini tungsten bulb and 1LED. And it rules it out:

Next with battery tungsten: focus miss (luminosity between the 3 and 1 LED cases)
Next with 1LED: focus hit!

At the last position, the AF assist light wanted to come on 3/4 of times. I report the result without it.

The difference between battery powered tungsten and LED is spectrum only, ruling out the difference in luminosity and flicker.

So, daylight and LED are best, tungsten is worse and halogen is worst. A non-monotone dependency on color temperature.

My best guess still is a simple software error where correction factors are running off. Reminds me of the flash situation actually where the distance-dependent correction factor seems to run off too (following some discussion in Germany). I wouldn't be surprised to learn it is the same engineer to blame here...

To Pentax, I advise a code review of the critical sections of the firmware. One of the advantages of open source actually...


And anybody who likes to think it is a hardware bug: Maybe. But the likelyhood of software is much larger comparing the design complexities involved. Always stick to Occam's razor...

Last edited by falconeye; 12-29-2010 at 07:24 AM.
12-29-2010, 07:20 AM   #288
Veteran Member
filorp's Avatar

Join Date: Jun 2008
Location: Aberdeen Scotland
Photos: Gallery | Albums
Posts: 389
QuoteOriginally posted by falconeye Quote
My battery-powered headlight has two more settings: mini tungsten bulb and 1LED.
WOW!!! Boy and his toy!!!
12-29-2010, 07:51 AM   #289
Pentaxian




Join Date: Mar 2009
Location: Perth Australia
Photos: Albums
Posts: 1,511
I've noticed front focus issues under bad light too now... unless the light is so bad that the green assist comes on, which then allows it to be perfect again.

At least the front focus seems consistent and id expect a firmware update soon.

12-29-2010, 07:53 AM   #290
Senior Member




Join Date: Feb 2010
Posts: 116
Original Poster
Dear Falk,
Always good to have you very pragmatic and technical approach on this forum. Thanks.
I would like to ask you to shine your light to 2 questions.

1)
Would you know if a permanent exposure of the AF sensor to the main sensor like on the translucent mirror system of SONY would improve AF accuracy by the nature of the difference of system ?

2)
Do you foresee a faster contrast detect AF in the near future that rules out issues like the one we discuss in this thread ? With faster I mean fast enough to satisfy DSLR users in speed and accuracy. I expect the Fuji X100 to be quit fast with contrast detect and maybe without the possible flaws we see amongst phase detect cams ?
12-29-2010, 08:07 AM   #291
Loyal Site Supporter




Join Date: Apr 2010
Location: Tennessee
Posts: 5,413
This is one of the reasons I am not a first adapter. Hopefully this is not similar to the Canon 1DIII AF issue that Canon struggled with for almost a year. The 1DIII got such a bad reputation that Canon had to accelerate the 1DIV. A lot of sports guys went to Nikon D3 for its better high-ISO and better AF.

We know Pentax reads the websites and they address the issues that are serious even if they don't acknowledge the problems publicly.
12-29-2010, 08:22 AM   #292
Site Supporter
jpzk's Avatar

Join Date: Jan 2009
Location: Québec
Photos: Gallery | Albums
Posts: 6,969
QuoteOriginally posted by jolepp Quote
This would seem like a neat, pocketable method:

https://www.pentaxforums.com/forums/pentax-slr-lens-discussion/126249-how-int...ml#post1307972

The best part of the above IMO is that you not only see if there is a problem but whether it is back/front focus. The DA 16-50 should show the symptoms (if any) clearly at 50mm, f=2.8 and a short distance (~0.5m). In general you'll want minimum DOF for testing, and that comes with large aperture/long FL/short subject distance.

I would assume that the issue this thread is about, that is, different focusing under low artifical light vs. daylight affects all K-5s, but you never know. Testing for FF/BF with the fast lenses you have makes sense in any case. This way, you might avoid sending in the body for adjustment and/or getting a lens body combo where you run out of adjustement range. If you have other fast lenses you could use in the circumstances consider bringing them too.

It seems that my Tamron 17-50 1:2.8 front focuses with the K-5 in any light and that I would need a bit more adjustment for best results: knowing what I know now I would have done well bringing that in when I went shopping. Luckily I also have a stained sensor, which should mean a replacement anyway ;-)
Thanks jolepp!

So, I'll take the DA*16-50 along and perhaps a faster prime lens, such as the FA 31 or 43 Ltd.

The store I am going to has large windows, letting plenty of natural light in and they also have tons of fluorescent lighting ... not the best to test "tungsten" light ... plainly not tungsten.
But I suppose they must have some area in the store where there is tungsten lighting. I'll soon find out. I'll do a comparison between both natural and artificial light.

For the stains, that'll be easy enough to find out.

Cheers.

JP
12-29-2010, 10:47 AM   #293
Veteran Member
deejjjaaaa's Avatar

Join Date: May 2007
Location: steel city / rust belt
Posts: 2,046
QuoteOriginally posted by falconeye Quote
Can we please keep the church in the village?
(old German proverb. Sorry if it doesn't make sense in another language )

This is a simple software bug. Software bugs happen and are easy enough to fix.

One way to see it's a software bug (read firmware) is that the K-5 doesn't even activate AF assist light and is pretty confident when locking wrong focus.
did you see how close was ___100w___ lamp to the target ? that is enough light not to activate any AF assist light... but again - those were not K-5 cameras

12-29-2010, 03:54 PM   #294
Pentaxian
Wheatfield's Avatar

Join Date: Apr 2008
Location: The wheatfields of Canada
Posts: 10,042
QuoteOriginally posted by filorp Quote
In my case i does related directly to color temperature
QuoteOriginally posted by falconeye Quote
That's not true. It depends but in a complex way.

K. I was referencing to a couple of posts I read where the (day) light was very low and there was front focus issues, though filorp, you didn't mention what EV you were exposing at, so to my mind, in your instance, the jury is still out.
12-29-2010, 05:10 PM   #295
Pentaxian
falconeye's Avatar

Join Date: Jan 2008
Location: Munich, Alps, Germany
Photos: Gallery
Posts: 6,862
QuoteOriginally posted by bjan Quote
Dear Falk,
Always good to have you very pragmatic and technical approach on this forum. Thanks.
I would like to ask you to shine your light to 2 questions.

1)
Would you know if a permanent exposure of the AF sensor to the main sensor like on the translucent mirror system of SONY would improve AF accuracy by the nature of the difference of system ?

2)
Do you foresee a faster contrast detect AF in the near future that rules out issues like the one we discuss in this thread ? With faster I mean fast enough to satisfy DSLR users in speed and accuracy. I expect the Fuji X100 to be quit fast with contrast detect and maybe without the possible flaws we see amongst phase detect cams ?
1)
That should not make a difference. Assuming that both systems divert about 1/3 of light to the AF module. That's because AF is locked prior to the exposure.

2)
Yes. From what I know I would say it is already feasible today but isn't deployed yet for consumer electronics. Or the engineers lack some genius

A sensor like in the K-5 (10fps full frame, 100fps when windowed) should allow for a contrast AF system outperforming phase AF, let alone accuracy. I know that phase AF requires fewer measurements (ideally only one). But that's not necessarily the limiting factor if the imaging sensor captures 10x the light of the AF module.

I carefully watch the current contrast AF algorithms at work. And having programmed one myself, let me say that they still did not discover most of the tricks I had to invent to solve the problem 15 years ago E.g., I managed to eliminate hunting with contrast AF.

Nevertheless, as it isn't doing me doing the work, I expect it to take 5 more years before contrast AF will beat phase AF.


P.S.
Sorry if the above statement sounds a bit bold. But I really thought from my professional work experience that contrast AF is a solved problem in the industry only to discover that consumer electronic products still suck in this respect.

Last edited by falconeye; 12-29-2010 at 05:27 PM.
12-29-2010, 06:53 PM   #296
Veteran Member




Join Date: Jan 2007
Location: Northern California
Posts: 475
Not So Fast

QuoteOriginally posted by falconeye Quote
That's not true. It depends but in a complex way.

I did one more test to rule out or confirm another possible explaination. The tests have been around 1EV which is above the threshold AF assist light is triggered. ISO100, 2s, f/1.8 to be precise.

First with tungsten, a bit brighter: focus almost ok.
Next with halogen: focus miss.

So I thought, it may be the flicker from artificial AC light. So, I tried my battery-powered headlight positioned to give indirect light, even a stop less:

So, next with 3LEDs: focus hit!

I thought this is it, it is the flicker the battery doesn't cause. My battery-powered headlight has two more settings: mini tungsten bulb and 1LED. And it rules it out:

Next with battery tungsten: focus miss (luminosity between the 3 and 1 LED cases)
Next with 1LED: focus hit!

At the last position, the AF assist light wanted to come on 3/4 of times. I report the result without it.

The difference between battery powered tungsten and LED is spectrum only, ruling out the difference in luminosity and flicker.

So, daylight and LED are best, tungsten is worse and halogen is worst. A non-monotone dependency on color temperature.

My best guess still is a simple software error where correction factors are running off. Reminds me of the flash situation actually where the distance-dependent correction factor seems to run off too (following some discussion in Germany). I wouldn't be surprised to learn it is the same engineer to blame here...

To Pentax, I advise a code review of the critical sections of the firmware. One of the advantages of open source actually...


And anybody who likes to think it is a hardware bug: Maybe. But the likelyhood of software is much larger comparing the design complexities involved. Always stick to Occam's razor...

I also think that this is a software problem, but I am not so convinced anymore that it is wavelength-dependent, or mainly wavelength-dependent.

Here is why:

Today I used a set of ND filters on and off my FA 50 f1.4 and I could see a definite focus shift with the the exact same light source when the light was dimmer. In fact, I could add ND filters and see the focus shift more and more to the front as the light through the lens dimmed.

Try this with your led light source. Simply light the subject with it and get a focus lock that is accurate and take the image.

Add ND filters to reduce the amount of light through the lens significantly but keep the led source on the subject. Obtain focus lock with the filters in place and take the image. I think you will see a front focus shift.

In my case, the WB temp selected by the camera with and without ND filters was only 200 units different (my light source was daylight window light, btw), which is certainly far smaller than the difference in color temperature between any white led and any halogen light source.

Of course, none of us can perfectly control any of these experiments to remove all possible variables except for the one we are interested in, but I think that it is just coincidence that we are normally shooting in lower light that is also tungsten or halogen.

Ray
12-29-2010, 07:22 PM - 1 Like   #297
Pentaxian
falconeye's Avatar

Join Date: Jan 2008
Location: Munich, Alps, Germany
Photos: Gallery
Posts: 6,862
Ray, I don't think we disagree.

I agree the problem emerges when light fades. I only wanted to illustrate that the spectrum plays a role large enough to shift the point of problem by several stops.

A more complete study would be a 3D plot plotting defocus against luminosity and color temperature.
12-29-2010, 08:01 PM   #298
Veteran Member




Join Date: Jan 2007
Location: Northern California
Posts: 475
Yes - We Agree

QuoteOriginally posted by falconeye Quote
Ray, I don't think we disagree.

I agree the problem emerges when light fades. I only wanted to illustrate that the spectrum plays a role large enough to shift the point of problem by several stops.

A more complete study would be a 3D plot plotting defocus against luminosity and color temperature.
Falk, the point is really moot as we cannot fix the problem, but controlling the light source color temperature is the hard part.

One would need a neutral grey shooting area and a calibrated light source that can be varied in intensity and color temp. There are such lights, as I have seen some made up of RGB led arrays, but I certainly do not own one. I do have access to a colorimeter device, however, but even using that properly is a bit tricky in my experience.

This is the beauty of ND filters on the lens. Good ones do not affect the color temperature so we can isolate to one variable as much as possible by using a single light source (DC power preferred to eliminate color and intensity swings with AC cycle).

I think it would be pretty conclusive if we can get the same shift with a daylight source as well as with a lower temperature source. This would be pretty good proof that the issue is somehow related to light level/contrast and maybe not so much to the wavelength.

I know this much:

Using the same light source, I have added/removed the NDs from several of my lenses (I have a good selection ) and the focus shift is easily noted without even taking an image. You can see the movement of the lens focusing element from the focus lock position with the filters and the new focus lock without the filters.

It is pretty convincing even if more than a bit puzzling. Frankly, my money was on IR shift and/or lens aberration issues prior to my ND experiment today.

Ray
12-30-2010, 02:03 AM   #299
Senior Member




Join Date: Feb 2010
Posts: 116
Original Poster
QuoteOriginally posted by falconeye Quote
A sensor like in the K-5 (10fps full frame, 100fps when windowed) should allow for a contrast AF system outperforming phase AF, let alone accuracy. I know that phase AF requires fewer measurements (ideally only one). But that's not necessarily the limiting factor if the imaging sensor captures 10x the light of the AF module.
\very interesting . THanks for this info. I look forward to the new Olympus XZ1 announced yesterday and the FUJI X100 in respect of fast contrast detect. I do see compacts getting more and more serious for may of the DSLR shooters. Shallow Dof`s are never going to be achieved like in larger sensor cams but general results at consumer sized prints are getting hard to distinguish . Are you still shooting with the DP1 ? However a perfectly funcioning K5 would be the cam to beat as a allround all purpose camera giving us the best of all worlds.
12-30-2010, 02:22 AM   #300
Forum Member




Join Date: Dec 2010
Posts: 53
QuoteOriginally posted by falconeye Quote
1)
That should not make a difference. Assuming that both systems divert about 1/3 of light to the AF module. That's because AF is locked prior to the exposure.
Though, if one is tracking a subject and shooting continously, there should be some benefit because of constantly availablle information for the AF unit.
Closed Thread

Bookmarks
  • Submit Thread to Facebook Facebook
  • Submit Thread to Twitter Twitter
  • Submit Thread to Digg Digg
Tags - Make this thread easier to find by adding keywords to it!
camera, conditions, dslr, focus, k-5, k-5 ii, k-5 iis, k5, lens, light, pentax k-5, tungsten
Thread Tools Search this Thread
Search this Thread:

Advanced Search


Similar Threads
Thread Thread Starter Forum Replies Last Post
Front Focus and tungsten light question Big G Pentax DSLR Discussion 8 01-18-2010 12:25 PM
released Kx firmware 1.01 anyone cares? Mystic Pentax News and Rumors 4 12-05-2009 09:19 PM
Can K20 be adjusted for indoor Tungsten light? philbaum Pentax DSLR Discussion 12 07-22-2009 05:40 AM
A ship in bad light mingdie Post Your Photos! 11 08-07-2008 03:52 AM
Tungsten light option k100d super problem? demoleman Pentax DSLR Discussion 17 02-18-2008 04:58 PM



All times are GMT -7. The time now is 01:57 PM. | See also: NikonForums.com, part of our network of photo forums!
  • Red (Default)
  • Green
  • Gray
  • Dark
  • Dark Yellow
  • Dark Blue
  • Old Red
  • Old Green
  • Old Gray
  • Dial-Up Style
Hello! It's great to see you back on the forum! Have you considered joining the community?
register
Creating a FREE ACCOUNT takes under a minute, removes ads, and lets you post! [Dismiss]
Top