Forgot Password
Pentax Camera Forums Home
 

Reply
Show Printable Version Search this Thread
04-10-2009, 05:36 PM   #1
Veteran Member




Join Date: Nov 2007
Posts: 1,553
Real World AF Speed Data. K20D and K10D

I used exiftool to pull three tags out of my directory of photo files. The tags were; Software, AFIntegrationTime, and EffectiveLV. The software tag reported either, K20D Ver 1.01 or K10D Ver 1.30, so it was easy to separate the data. I opened the data in a spread sheet, massaged the data into something workable, and did a few calculations. The graphs are below.

My assumption is that the AFIntegrationTime tag denotes the time it takes for the camera to lock focus, and that EffectiveLV is the overall light level of the metering in the photo. If someone has knowledge that these assumption are wrong, please speak up.

First graph is the number of photos by EffectiveLV so that an idea can be formed of the sample. Probably 98% of the photos where shot with a screw driven zoom lens





Next is the chart of the average AFIntegrationTime per EffectiveLV.




I'm pretty sure the higher AFIntegrationTime at the higher EffectiveLV levels is because I am using a lens I didn't have before and it is known for its slower focus, the Tokina 80-400mm. The lower LV photos will probably have been taken with Sigma 18-50mm f2.8 macro on both the K10D and K20D, so I think that is where the real meat of this exercise lays.

I probably should break the data out further by lens, but I happy enough with this.

Thank you
Russell


Last edited by Russell-Evans; 04-10-2009 at 11:36 PM. Reason: fix tag name
04-10-2009, 06:06 PM   #2
Veteran Member




Join Date: Mar 2007
Location: Boston, PRofMA
Photos: Albums
Posts: 3,053
<scratches head>
So the K20D takes longer to focus? Do you have the graphs backwards? :-)
04-10-2009, 06:33 PM   #3
Veteran Member




Join Date: Nov 2007
Posts: 1,553
Original Poster
QuoteOriginally posted by kenyee Quote
<scratches head>
So the K20D takes longer to focus? Do you have the graphs backwards? :-)
Like Santa, I checked it twice. I have made some lens changes over time, so take it as it is. I'm mostly posting this so others will try it out on their photos. If in doubt about your AF speed, why not take the time to quantify your findings?

Thank you
Russell
04-10-2009, 07:33 PM   #4
Veteran Member
OregonJim's Avatar

Join Date: Mar 2009
Location: Willamette Valley, Oregon
Posts: 1,329
I like the idea, but you can draw no meaningful conclusion as there are too many variables. Separate the data by lens, and track battery levels - then you'll be getting closer.

04-10-2009, 10:38 PM   #5
Forum Member




Join Date: Sep 2008
Posts: 75
Are these all done with the same lens? I think that the lens might have to be the same for both cameras since different lenses have different distances they need to travel in order to focus. For example, a 50mm f/1.4 might be a lot quicker at focusing compared to a DA 50-200mm. F-stops might also play a role because of the differences in depth of field.
04-11-2009, 12:06 AM   #6
Veteran Member




Join Date: Nov 2007
Posts: 1,553
Original Poster
QuoteOriginally posted by turbosaturn Quote
Are these all done with the same lens? I think that the lens might have to be the same for both cameras since different lenses have different distances they need to travel in order to focus. For example, a 50mm f/1.4 might be a lot quicker at focusing compared to a DA 50-200mm. F-stops might also play a role because of the differences in depth of field.
This is a mix of all my lenses. The graphs aren't meant to show that the K20D is faster or slower than the K10D. I wanted to know the average time it takes for the AF to lock for my lens set for a given light level. The understanding I get from the graphs is that anything above EV 6 and my lenses focus fast enough. Under EV 6 and my lenses will take more than a second to lock and probably will hunt. EV 6 is a brightly lit home.

I also got an idea of where I tend to shoot in EV which is interesting in itself to me. Knowing that I get into EV 5, 4, and 3 means I maybe need to start carry a fast focusing prime as well. I need to do focal length to EV graph to determine if I can find an optimal focal length to carry. After carrying it and using it for a while I can go back and see if it really helped. I'll be able to see if the lens really focuses faster in the light level used.

I'm not really interested in knowing exactly if my K20D is faster or slower than my K10D for each lens. I didn't do it for that because it doesn't matter to me. I still own and use my K10D, but I've spent the money on the K20D so what is the point?

All the above is about me, but my reason for sharing is that it is interesting that the exif data contains enough information to answer questions about AF speed at light levels, AF speed for a lens at light levels, ... The idea is if you are interested, there is a way to actually quantify a lot of the general arguments about AF speed posted here.

Wouldn't it be nice to really know that the DA 40mm or the FA 35mm were the fastest focusing lenses instead of have to guess? You could have the entire prime lens line up ranked by AF speed if you wanted. Would you choose the DA 16-50mm, Tamron 17-50mm, or Sigma 18-50mm, over the other, if you knew one was much faster focusing in low light than the others? Heck, you could even answer the question: "Is the SDM or screw drive focusing faster on the 16-50mm or 50-135mm". That's what I'm trying to communicate, a way to do something.


And all with real world data!


Thank you
Russell

Last edited by Russell-Evans; 04-11-2009 at 12:15 AM.
04-11-2009, 12:10 AM   #7
Veteran Member
res3567's Avatar

Join Date: Sep 2008
Location: Houston Tx.
Photos: Gallery
Posts: 1,876
So you must have read that long post that was left earlier about the AF complaints.
04-11-2009, 12:14 AM   #8
Veteran Member




Join Date: Nov 2007
Posts: 1,553
Original Poster
QuoteOriginally posted by res3567 Quote
So you must have read that long post that was left earlier about the AF complaints.
Yes

Thank you
Russell

04-11-2009, 12:18 AM   #9
Veteran Member
res3567's Avatar

Join Date: Sep 2008
Location: Houston Tx.
Photos: Gallery
Posts: 1,876
My take on that post is "GOODNESS GRACOUS" Thanks for your input. You went and did your own research and started a fresh post to give a different perpestive. Admireable.
04-11-2009, 12:40 AM   #10
Pentaxian
dosdan's Avatar

Join Date: Dec 2007
Location: Brisbane, Australia
Photos: Gallery | Albums
Posts: 1,697
QuoteOriginally posted by Russell-Evans Quote
Under EV 6 and my lenses will take more than a second to lock and probably will hunt. EV 6 is a brightly lit home.
Russell, the max. AFIntegrationTime shown on your graph is 1/4s with EV6 being 1/10s.

Dan.
04-11-2009, 04:08 AM   #11
Veteran Member




Join Date: Nov 2006
Location: Hong Kong
Posts: 1,934
Good Job! But, there is something wrong..

Interesting analysis and actually you're clever enough to extract those EXIF data and plot them.

However, I doubt about the true meaning of the "AF Integration Time", which doesn't look like the *real* and *actual* AF time at all. Just imagine how the camera can focus with tens of millisecond and I don't believe at low EV near 0EV, the K10D and K20D could focus in 0.25 second, which contradicts all other measured AF timings for both cameras.

According to the Imaging Resource, even the AF motor don't move, the AF thinking time is 147 milliseconds (0.147s) for the K10D:-

RiceHigh's Pentax Blog: Is the K10D Really Sluggish in Shutter Lag?

And, PopPhoto measurement shows that at EV2 or below, the K20 requires more than 1 second to focus:-

Sony - Camera Reviews from Photography Experts

Moreover, the best AF focusing timing is just 0.35 second even at EV12 - a bright environment.

So, afterall, I guess the "AF Integration Time" means something else, not the AF time from start to completion. Maybe it just meant the "AF Calculation Time" by the camera's processor?
04-11-2009, 07:05 AM   #12
Veteran Member




Join Date: Nov 2007
Posts: 1,553
Original Poster
QuoteOriginally posted by RiceHigh Quote
However, I doubt about the true meaning of the "AF Integration Time", which doesn't look like the *real* and *actual* AF time at all. Just imagine how the camera can focus with tens of millisecond and I don't believe at low EV near 0EV, the K10D and K20D could focus in 0.25 second, which contradicts all other measured AF timings for both cameras.

The unit isn't part of the raw data of the tag. exiftool adds the "ms" to the output just like it adds the human readable lens names. exiftool has an -n option that will only report values.

from the exiftool man page.
-n Read and write values as numbers instead of words. This option
disables the print conversion that is applied when extracting
values to make them more readable, and the inverse print conversion
when writing. For example:

So maybe the data means what the data means, but the units are off by a power of ten in the formatted output units?

The other thing to consider is this is real world data. My lens isn't probably going from infinity to its closest focus as it would need to in a test environment. The lens focus is going to be some random distance from perfect focus, so unless the lens hunts from end to end, I would expect the focus speed to be better in real life than in the any test environment. The lenssimply isn't going to be moving as far.

Thank you
Russell

Last edited by Russell-Evans; 04-11-2009 at 07:50 AM. Reason: Last paragraph.
04-11-2009, 07:10 AM   #13
Veteran Member




Join Date: Nov 2007
Posts: 1,553
Original Poster
QuoteOriginally posted by dosdan Quote
Russell, the max. AFIntegrationTime shown on your graph is 1/4s with EV6 being 1/10s.
I'm thinking the units are being reported wrong as they are not part of the exif data. My bad for not catching it, however.

Run exiftool -n -AFIntergrationTime *.pef

from the exiftool man page.
-n Read and write values as numbers instead of words. This option
disables the print conversion that is applied when extracting
values to make them more readable, and the inverse print conversion
when writing. For example:

Thank you
Russell
04-11-2009, 08:42 AM   #14
Veteran Member




Join Date: Nov 2006
Location: Hong Kong
Posts: 1,934
I don't think so..

QuoteOriginally posted by Russell-Evans Quote
So maybe the data means what the data means, but the units are off by a power of ten in the formatted output units?
Not an unit problem, I think.

QuoteQuote:
The other thing to consider is this is real world data. My lens isn't probably going from infinity to its closest focus as it would need to in a test environment. The lens focus is going to be some random distance from perfect focus, so unless the lens hunts from end to end, I would expect the focus speed to be better in real life than in the any test environment. The lenssimply isn't going to be moving as far.
I thought about that from the very beginning the first time I read your post. But, however, as I mentioned, the IR found that the "calculation time" for Single AF even the lens *is not moving* at all is 147ms. So, how come the camera (say the K10D) can have a 7.5ms AF timing?

Besides, as I have also said, it is just *impossible* for the camera to have an *average* AF time of 0.25s at EV0 as the *best* timing of K10/20's AF at even EV 12 is 0.35s, which resemble my reallife experience. If Pentax could make a DSLR that could focus in 0.25s at EV0, then they had a winner!

In short, owing to the unrealistic figures that are obtained, I believe the "AF Integration Time" does mean something else (as I have also guessed in my last reply post - my first reply) and I am quite sure that it does NOT mean the *whole* AF time that is required for achieving the AF, completely.
04-11-2009, 02:46 PM   #15
Veteran Member




Join Date: Nov 2007
Posts: 1,553
Original Poster
Information from the author, I'm wrong.

On Sat, Apr 11, 2009 at 3:25 PM, Phil Harvey wrote:
> Hi Russel,
>
> My understanding is that the AFIntegrationTIme is essentially the exposure
> time required by the AF sensor. Actual focus time will be greater because
> the lens doesn't start focusting until after the AFIntegrationTime ends. I
> believe the milliseconds units come from the K10D debugging mode output, but
> I can't remember for sure. I tested this for the K10D and gives reasonable
> values, but hasn't been thoroughly tested for other models.
>
> I hope this make sense.
>
> In one of your posts you mention "Under EV 6 and my lenses will take more
> than a second", but you must have realized this is incorrect. 100ms = 0.1s.

This makes sense to me once explained; the AF sensor needs to collect enough light to be able to measure the focus, so in low light it is going to take longer in the same way the main sensor is.

Thank you
Russell
Reply

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, data, dslr, focus, k10d, k20d, lens, photography, photos, tag
Thread Tools Search this Thread
Search this Thread:

Advanced Search


Similar Threads
Thread Thread Starter Forum Replies Last Post
The real world difference f/1.4 -vs- f/2 Oscar1 Pentax SLR Lens Discussion 13 02-06-2010 10:33 PM
Real world ISO 3200 and beyond! K20D codiac2600 Post Your Photos! 16 03-01-2008 05:53 PM
PopPhoto admit using K10D AF speed data in K20D review Peter Fang Pentax News and Rumors 31 02-24-2008 04:48 PM
CA in the DA*16-50 vs Tamron 28-75 in real world 123K10D Pentax SLR Lens Discussion 10 02-23-2008 02:44 PM
Real world use. Rob.K Pentax DSLR Discussion 29 02-12-2007 01:26 PM



All times are GMT -7. The time now is 12:00 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