Forgot Password
Pentax Camera Forums Home
 

Reply
Show Printable Version Search this Thread
02-20-2012, 07:28 AM   #31
Inactive Account




Join Date: Jan 2012
Posts: 202
QuoteQuote:
+10 on the lens + 10 on the adjust all...
The AF adjustment values are not culmulative. Further, the [ALL] or [ONE] value used is the one entered ("OK"d) last. This is easily tested by checking the AF Adjustment value in the EXIF info for pictures.

For instance, if I set [ALL] to +5, then set [ONE] to +8, the camera shoots with +8. If I then go back in and OK [ALL] at +5, the camera shoots with +5. If I set [ALL] to 0 after setting [ONE] to +8, the camera shooots with 0. AF Off shoots with 0.

02-20-2012, 07:30 AM   #32
Veteran Member




Join Date: Jan 2009
Location: London
Posts: 466
QuoteQuote:
Please see Understanding Camera Autofocus or Canon EOS Autofocus - BFCV and micro-adjust (MA) - Open Photography Forums . Simply put there are many variables which come into play and if they don't work in unison, problems occur
I have read these but still cannot see why lens manufacturing variations should result in poor focus, because the AF system uses feedback to maximise contrast.

Is the issue caused by the AF sensors being located in a different plane to the CMOS/CCD image sensor? Where are the AF sensors? Presumably they cannot be within the image sensor otherwise you would get blind spots there
02-20-2012, 07:52 AM   #33
fuf
Forum Member




Join Date: Jan 2012
Posts: 52
QuoteOriginally posted by peterh337 Quote
I have read these but still cannot see why lens manufacturing variations should result in poor focus, because the AF system uses feedback to maximise contrast.

Is the issue caused by the AF sensors being located in a different plane to the CMOS/CCD image sensor? Where are the AF sensors? Presumably they cannot be within the image sensor otherwise you would get blind spots there
As you write, the AF sensors are located in a different plane: please see the first image at Pentax Auto Focus Systems . Basically the mirror which you see when you unmount a lens lets a fraction of light through to another mirror which directs the light to the AF sensors.
02-20-2012, 07:52 AM   #34
Veteran Member
octavmandru's Avatar

Join Date: Jan 2009
Location: somewhere around
Posts: 615
Original Poster
QuoteOriginally posted by jmg257 Quote
The AF adjustment values are not culmulative. Further, the [ALL] or [ONE] value used is the one entered ("OK"d) last. This is easily tested by checking the AF Adjustment value in the EXIF info for pictures.

For instance, if I set [ALL] to +5, then set [ONE] to +8, the camera shoots with +8. If I then go back in and OK [ALL] at +5, the camera shoots with +5. If I set [ALL] to 0 after setting [ONE] to +8, the camera shooots with 0. AF Off shoots with 0.
That is something new!

In my understanding ALL is the "camera offset", ONE is the individual lens adjust. They should be additive, in the menu both values show in the same time. Otherwise one or the other should grey out, isn't it?

02-20-2012, 08:00 AM - 1 Like   #35
Inactive Account




Join Date: Jan 2012
Posts: 202
QuoteQuote:
I have read these but still cannot see why lens manufacturing variations should result in poor focus, because the AF system uses feedback to maximise contrast.
The AF system uses phase detection feedback from line sensor arrays to calculate defocus errors...the AF sensors are not in the image sensor, but in a seperate CCD. (bottom of the mirror box - light passes through the main mirror and is reflected down into the AF module by a secondary mirror). So what if the exact distance between the AF module that contains the AF sensor is ever so slightly off from the image sensor? Master lenses are used to calibrate the AF module vs the camera body position using mechanical screws, and then via software if needed (the infamous "firmware debug mode" for us end users, but provided for service techs via special software) to microns to minimize this error.

The defocus amount determined by the AF sensors is used to determine how far the focusing unit in the lens has to be moved. This amount is calculated on half-press, and then used by the CPU to calculate the number of pulses an encoder connected to the AF drive motor must feed back (count down) when the motor moves. What if the exact amount of lens focus unit movement based on the AF motor/encoder isn't exactly the same from lens to lens? Then there is a "focusing width" value (a constant, or calculated based on information from the lens re:diaphragm/focus depth/focal length) that the AF-determined defocus error is compared to to decide if things are "in focus enough". What if there are slight errors in this info due to tolerances in the lens build?

Hopefully we can see how error can be inherent in all these different aspects. Not to mention the number of multi-sensors in auto select mode - all supplying possibly slightly different info depending on what they 'see'.



EDIT: I would suggest checking out US Patent 7,496,290 B2 - Pentax's patent for it's multi-point AF system... Pretty cool!

Last edited by jmg257; 02-20-2012 at 08:24 AM.
02-20-2012, 08:04 AM   #36
Inactive Account




Join Date: Jan 2012
Posts: 202
QuoteQuote:
In my understanding ALL is the "camera offset", ONE is the individual lens adjust. They should be additive, in the menu both values show in the same time. Otherwise one or the other should grey out, isn't it?
The manual memo on page 127 gives a hint - it states that an [ALL] selection OK'd will overide a [ONE] selection. But it isn't very clear - it does operate as I noted...the last setting [ALL] or [ONE] OK'd will override the other.

BTW - although I tested powering off and on in between and the OKd value sticks, I did NOT test what happens when you switch a lens with 1 [ONE] setting from another lens that has a different [ONE] value.

Last edited by jmg257; 02-20-2012 at 08:18 AM.
02-20-2012, 08:45 AM   #37
Veteran Member




Join Date: Jan 2009
Location: London
Posts: 466
I wonder why the camera cannot do AF using the main sensor.

Presumably it is performance limited. They cannot get the data off the main sensor fast enough, and they have to read the entire main sensor to get the data from the AF locations. I am suprised though because they can evidently do 50fps HD video...

Plus they must be using the main sensor for AF when in the live view mode. Somebody here posted that all focus problems disappear when you use LV.
02-20-2012, 08:51 AM   #38
Inactive Account




Join Date: Jan 2012
Posts: 202
When AF is on the image sensor it is Contrast Detection type - which hasn't been as fast as phase detection, and has its own limitations. It is getting better though and is the type of AF on mirrorless systems.

QuoteQuote:
"Phase Detection Autofocus
This AF method is most prevalent in digital SLR cameras. It requires dedicated hardware (line and cross-type autofocus sensors) which can increase the overall cost of a camera. Phase detection AF still uses contrast to assess where to focus, but contrast just isn't used as iteratively or throughout the image...only at the autofocus sensors. A disadvantage with the phase detection method is that it requires pre-calibration, and can therefore always misfocus if incorrectly calibrated.


Contrast Detection Autofocus
This AF method is most prevalent in compact cameras and newer SLR cameras with Live View. This is because these cameras can be made less expensive by using the sensor itself to perform autofocus, with the (relatively) cheaper processing power used to determine maximal contrast. A disadvantage to contrast AF is that it cannot gauge whether the camera lens is front or back focused-- just that it is out of focus. This is primarily why it has to be used iteratively; contrast measurements at at least two lens positions are needed to assess which direction the camera should be focusing (ie, in front or in back of where it was previously focusing). The contrast detection is generally more fail proof (if given lots of light and enough time to focus) since it not reliant on pre-calibration."
Phase Detection vs Contrast Detection Autofocus (AF)

http://www.cambridgeincolour.com/tutorials/camera-autofocus.htm

02-20-2012, 08:58 AM   #39
fuf
Forum Member




Join Date: Jan 2012
Posts: 52
QuoteOriginally posted by peterh337 Quote
I wonder why the camera cannot do AF using the main sensor.

Presumably it is performance limited. They cannot get the data off the main sensor fast enough, and they have to read the entire main sensor to get the data from the AF locations. I am suprised though because they can evidently do 50fps HD video...

Plus they must be using the main sensor for AF when in the live view mode. Somebody here posted that all focus problems disappear when you use LV.
The main sensor is hidden behind the mirror which reflects the light to the viewfinder and lets through part of the light to the phase detection AF sensors. If it was used for focusing you wouldn't be able to use the viewfinder when the camera focused.

Sony uses a fixed-position translucent mirror in their select models which can combine phase detection and contract detection. However the sensor then receives less light (== higher ISO or longer exposure time is needed) and in some images with high contract in them one can notice ghosting. DPreview wrote about the issue with example images to show the issue.

Contract detection AF has problems of its own as well -- sometimes it simply cannot focus when there is not enough contract in the scene. Oftentimes the whole focus range is gone through to find the highest contract and the power consumption is way higher.
02-20-2012, 11:45 AM   #40
Veteran Member
octavmandru's Avatar

Join Date: Jan 2009
Location: somewhere around
Posts: 615
Original Poster
QuoteOriginally posted by jmg257 Quote
The manual memo on page 127 gives a hint - it states that an [ALL] selection OK'd will overide a [ONE] selection. But it isn't very clear - it does operate as I noted...the last setting [ALL] or [ONE] OK'd will override the other.

BTW - although I tested powering off and on in between and the OKd value sticks, I did NOT test what happens when you switch a lens with 1 [ONE] setting from another lens that has a different [ONE] value.
Oh brother... Now I'm really confused about how they imagine people should adapt to the machine.
02-20-2012, 12:02 PM   #41
Pentaxian
philbaum's Avatar

Join Date: Oct 2007
Location: Port Townsend, Washington State, USA
Posts: 3,659
QuoteOriginally posted by octavmandru Quote
That is something new!

In my understanding ALL is the "camera offset", ONE is the individual lens adjust. They should be additive, in the menu both values show in the same time. Otherwise one or the other should grey out, isn't it?
I had the same impression that you did, i.e. that the 2 could be cumulative. I wonder if the K20 wasn't programmed differently for this. They could have figured that if the camera microfocus was out by more than 10, then the customer should be sending it in for calibration anyway.

I just hope that the "new" Pentax communicates better. The fact that manuals have to be issued in a variety of languages may have caused companies to reduce the amount of explanation in their manuals.
02-20-2012, 06:44 PM   #42
Veteran Member
Smeggypants's Avatar

Join Date: Sep 2010
Posts: 1,536
QuoteOriginally posted by adi3000 Quote
I repost this, maybe you missed and is good to try it.

Quantitative Method for Micro Focus Adjustment for DSLRs - Enzo's Home

I tried and it works very well, unexpected. I have a DA* 16-50 and I been unhappy with it even if I adjusted the focus before. But using this method give me another setting better for every situation, even in low light now is perfect.

- in my situation the compiled class was not working. I use only the .java file and use "compile and run" (you need jdk for this).
I first started adjusting the AF on my lenses using that method, but I'ev since found you don't need to go into such scientific detail to calibrate them. Now I just take a few pictures handheld narrowing the setting down until it's correct.
02-20-2012, 06:49 PM   #43
Veteran Member
Smeggypants's Avatar

Join Date: Sep 2010
Posts: 1,536
QuoteOriginally posted by peterh337 Quote
I wonder why the camera cannot do AF using the main sensor.

Presumably it is performance limited. They cannot get the data off the main sensor fast enough, and they have to read the entire main sensor to get the data from the AF locations. I am suprised though because they can evidently do 50fps HD video...

Plus they must be using the main sensor for AF when in the live view mode. Somebody here posted that all focus problems disappear when you use LV.
The K-5 can do AF using the main sensor in LV mode. It's the contrast method and thus is a lot slower than the phase method used by the main AF system which has the separate AF sensor located in the bottom of the camera
02-20-2012, 09:40 PM   #44
Veteran Member
octavmandru's Avatar

Join Date: Jan 2009
Location: somewhere around
Posts: 615
Original Poster
QuoteOriginally posted by peterh337 Quote
I wonder why the camera cannot do AF using the main sensor.

Presumably it is performance limited. They cannot get the data off the main sensor fast enough, and they have to read the entire main sensor to get the data from the AF locations. I am suprised though because they can evidently do 50fps HD video...

Plus they must be using the main sensor for AF when in the live view mode. Somebody here posted that all focus problems disappear when you use LV.
QuoteOriginally posted by Smeggypants Quote
The K-5 can do AF using the main sensor in LV mode. It's the contrast method and thus is a lot slower than the phase method used by the main AF system which has the separate AF sensor located in the bottom of the camera
Wonder how fast is the CDAF on the K-01...
Come to think of it, on mirrorless cameras the accuracy of the AF should be better since the image sensor itself is the AF sensor.

I suppose the limitation on current CDAF system is not knowing the amount of focus correction needed for perfect focus. Since it is "contrast detection", AF works in a feedback loop, that making it slow. Just give it a couple of years.
This is most likely the future of professional cameras. Lightning speed CDAF.
02-21-2012, 03:51 AM   #45
Veteran Member




Join Date: Jan 2009
Location: London
Posts: 466
You have to do "feedback" no matter how you do it.

Otherwise, you must be somehow measuring the distance to the target (laser/ultrasonics) and then somehow telling the lens to "focus for 10.5m" and there is no way to do that accurately, given manufacturing tolerances, etc.

Cameras which did that in the past used narrow apertures on which focus didn't matter much.

The current AF system which uses separate sensors is still using negative feedback and has to "hunt" around the best focal point.

I am pretty sure it is physically impossible to drive a lens directly to the optimal focal point, because the image looks exactly the same when it is slightly out of focus in one direction or the other direction.

So it has to be a sensor performance issue which causes Pentax etc to use the separate focus sensors.

One day they will be able to use the main CCD rapidly and that will do away with all these lens manufacturing issues. I bet you that a lot of R&D is going into that right now, because software is a lot cheaper than that separate sensor array etc etc.
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, da*50-135, dslr, k-5, k-5 ii, k-5 iis, k20, k5, lenses, life, light, pentax k-5
Thread Tools Search this Thread
Search this Thread:

Advanced Search



All times are GMT -7. The time now is 01:10 AM. | See also: NikonForums.com, CanonForums.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