December 19, 2014, 03:26:40 PM

Author Topic: New MFA method  (Read 23338 times)

East Wind Photography

  • 1D Mark IV
  • ******
  • Posts: 878
  • EWP
    • View Profile
Re: New MFA method
« Reply #90 on: February 20, 2013, 02:11:22 PM »
To give credit where credit is due (because I have been seriously harsh about Dot Tune)  I applaud horshak for thinking out of the box to come up with an alternative solution to AFMA calibration.  We need more people that think like that to come up with new technologies to improve the tools we use.

I just feel that AF detect was never intended do such and therefore is not accurate enough at this time to make it reliable.  Now that Dot tune is out there, lets hope the vendors take him seriously and start to refine the AF detect system and then adopt it as a a standard firmware option built in to cameras to do this automatically.

It is simply groundbreaking.  Time will tell if it gets adopted by Canon, Nikon or any other maker and I hope it does.


canon rumors FORUM

Re: New MFA method
« Reply #90 on: February 20, 2013, 02:11:22 PM »

horshack

  • PowerShot G1 X II
  • ***
  • Posts: 60
    • View Profile
Re: New MFA method
« Reply #91 on: February 20, 2013, 03:28:22 PM »
To give credit where credit is due (because I have been seriously harsh about Dot Tune)  I applaud horshak for thinking out of the box to come up with an alternative solution to AFMA calibration.  We need more people that think like that to come up with new technologies to improve the tools we use.

I just feel that AF detect was never intended do such and therefore is not accurate enough at this time to make it reliable.  Now that Dot tune is out there, lets hope the vendors take him seriously and start to refine the AF detect system and then adopt it as a a standard firmware option built in to cameras to do this automatically.

It is simply groundbreaking.  Time will tell if it gets adopted by Canon, Nikon or any other maker and I hope it does.

I appreciate that, thanks. I welcome all feedback about DotTune, both positive and negative. I put DotTune out there because I think it has the potential to help a lot of DSLR owners succeed at a process that has traditionally been error prone and frustrating, so any feedback which improves and evolves the method is highly encouraged.

Regarding the AF detect and the VF confirmation dot, I have a hunch that the "slop" in the dot might actually be by design. Since the VF dot was designed primarily as a manual-focusing aid, I'm thinking both Canon and Nikon thought making the dot too precise would mean manual focusing would be too slow and require too much dexterity of the focus ring by the shooter. For static shooters with lots of time it would be ok, but for anyone shooting dynamic stuff like street-shooting I think a very-tight range of dot-confirmation would be a negative in practical terms. This is partly why I think using the midpoint of the dot-confirmation range for AF tuning works so well.

Kent

  • SX60 HS
  • **
  • Posts: 4
    • View Profile
Re: New MFA method
« Reply #92 on: February 20, 2013, 03:48:45 PM »
Automation investigation request started at Magic Lantern.
http://www.magiclantern.fm/forum/index.php?topic=4648.0

PhotoCat

  • Rebel T5i
  • ****
  • Posts: 117
    • View Profile
Re: New MFA method
« Reply #93 on: February 21, 2013, 10:57:31 AM »
Just successfully tuned my 50/1.8 & 85/1.8 with this dot-tune method with my 5d2.
There is no question in my mind that horshack's new MFA method works and works consistently.
Thanks for sharing this brilliant idea horshack!

To get a consistent result in my 50/1.8, I had to gaffer tape the focus ring of the 50/1.8  ;)
so it doesn't move right after manual focusing (x10) in LV.

Getting the AFMA value for 85/1.8 was easy as it was in range: -18 to -1, meaning a AFMA value of -10.

However, it was a bit more tricky for the 50/1.8 as the dot remains red at -20 and the upper range was -4.
-4 should be credible but -20 is questionable and it could be -21, -22, -23 etc but the scale doesn't show it.

So I was guessing horshack's "detuning method" and detuned the focus a bit.
I did so by first setting AFMA to +5 (guessing) and turn the focus ring manual until I get a beep (red dot).
So now I know +5 is within range.  Based on this detuned focus, using the dot-tune method gives me a range
of -9 to +9.   So I know the half range is 9.     Based on the credible upper value of -4 and subtracting
9 (half range) from it gave me a final value of -13.

Using the obtained AFMA values gave me much sharper images at f1.8 at the distance I performed the calibration. (distance suitable for a full-length shot)

So in my mind, the dot-tune method is solid. What is questionable to me is the
implementation of AFMA, as it seems subject distance dependent.

For example, my 85/1.8 with AFMA of -10 works great with full length portraits but I will get soft images at a distance of 10 feet and closer.  The 50/1.8 seemed to work well with AFMA of -13 at all distance with the calibration target and lighting. However, it doesn't work very well with somce other subjects and lighting. So I just have to turn off AFMA in case of doubt.

I feel this kind of AFMA is a bandage solution anyway and I am hoping Canon will incorporate their factory lens
calibration routine into EOS Utility and I will be laughing!  No more "bad copies"  of lens... sigh...
Note the competition already has "USB lens dock" for factory strength lens calibration LOL!

Mehmetski

  • Power Shot G7X
  • **
  • Posts: 20
    • View Profile
Re: New MFA method
« Reply #94 on: February 21, 2013, 11:17:46 AM »
Repeat these steps after restarting your camera. It will give you different values.

East Wind Photography

  • 1D Mark IV
  • ******
  • Posts: 878
  • EWP
    • View Profile
Re: New MFA method
« Reply #95 on: February 21, 2013, 11:19:34 AM »
So you say it works consistently but you also put in caveats that it doesnt work if the lighting or distance to subject changes? 

So it seems to me that it's not consistent in your case and the results are questionable.  Did you verify that the selected AFMA was indeed correct using a 3D type of focus tester?  You can set up one easy enough using a yard or meter stick slanted at a sharp angle and fixing a flat high contrast target next to it to focus on.

I've been seriously trying to explain that unless you have a way of verifying the AFMA setting is correct most people should just disable it.  As you noticed that it can get you into some trouble.

Some lenses just dont focus well on a linear scale.  Some need different AFMA settings focused far than focused close.  To me that's a manufacturing flaw or the lens is just out of spec and needs adjusting.  In those cases you either need to set the AFMA for the distance you use it the most, or split the difference in the near and far AFMA tests and use aperture to increase the DOF and mask the error.

Either way you need an accurate method of verifying the settings as AF detect is not very accurate in most of the current models.

Just successfully tuned my 50/1.8 & 85/1.8 with this dot-tune method with my 5d2.
There is no question in my mind that horshack's new MFA method works and works consistently.
Thanks for sharing this brilliant idea horshack!

To get a consistent result in my 50/1.8, I had to gaffer tape the focus ring of the 50/1.8  ;)
so it doesn't move right after manual focusing (x10) in LV.

Getting the AFMA value for 85/1.8 was easy as it was in range: -18 to -1, meaning a AFMA value of -10.

However, it was a bit more tricky for the 50/1.8 as the dot remains red at -20 and the upper range was -4.
-4 should be credible but -20 is questionable and it could be -21, -22, -23 etc but the scale doesn't show it.

So I was guessing horshack's "detuning method" and detuned the focus a bit.
I did so by first setting AFMA to +5 (guessing) and turn the focus ring manual until I get a beep (red dot).
So now I know +5 is within range.  Based on this detuned focus, using the dot-tune method gives me a range
of -9 to +9.   So I know the half range is 9.     Based on the credible upper value of -4 and subtracting
9 (half range) from it gave me a final value of -13.

Using the obtained AFMA values gave me much sharper images at f1.8 at the distance I performed the calibration. (distance suitable for a full-length shot)

So in my mind, the dot-tune method is solid. What is questionable to me is the
implementation of AFMA, as it seems subject distance dependent.

For example, my 85/1.8 with AFMA of -10 works great with full length portraits but I will get soft images at a distance of 10 feet and closer.  The 50/1.8 seemed to work well with AFMA of -13 at all distance with the calibration target and lighting. However, it doesn't work very well with somce other subjects and lighting. So I just have to turn off AFMA in case of doubt.

I feel this kind of AFMA is a bandage solution anyway and I am hoping Canon will incorporate their factory lens
calibration routine into EOS Utility and I will be laughing!  No more "bad copies"  of lens... sigh...
Note the competition already has "USB lens dock" for factory strength lens calibration LOL!

PhotoCat

  • Rebel T5i
  • ****
  • Posts: 117
    • View Profile
Re: New MFA method
« Reply #96 on: February 21, 2013, 11:52:00 AM »
Sorry for the confusion if any.
By "consistent" result, I meant dot-tune always gives me the same AFMA value with the same
target and lighting.  No, I did not mean consistent focusing under whatever shooting condition and
distance after AFMA.

I will try the 3D target as u had suggested.  No, both of my 85/1.8 and 50/1.8 did not magically
become perfect after AFMA. It just improved focusing accuracy at a specific distance I chose while it suffered
at other subject distance.  It is a tradeoff.
As I had said, AFMA is a bandage solution at best but that has nothing to do
with the dot-tune procedure which is giving consistent value for me for a fixed target, fixed lighting and
distance.
« Last Edit: February 21, 2013, 11:54:04 AM by PhotoCat »

canon rumors FORUM

Re: New MFA method
« Reply #96 on: February 21, 2013, 11:52:00 AM »

PhotoCat

  • Rebel T5i
  • ****
  • Posts: 117
    • View Profile
Re: New MFA method
« Reply #97 on: February 21, 2013, 02:00:40 PM »

Some lenses just dont focus well on a linear scale.  Some need different AFMA settings focused far than focused close.  To me that's a manufacturing flaw or the lens is just out of spec and needs adjusting.  In those cases you either need to set the AFMA for the distance you use it the most, or split the difference in the near and far AFMA tests and use aperture to increase the DOF and mask the error.



Yup, that has been my feeling too and I am glad I can finally find someone to agree with me  :)

For all the lenses I have, I have never been able to AFMA it such that it works for any subject distance.
The ones that worked for any subject distance had AFMA turned off LOL!

Yet I do recall someone on this forum had great success with it... It is puzzling.
Perhaps some particular lens flaw happens to be compensated by AFMA perfectly.

East Wind Photography

  • 1D Mark IV
  • ******
  • Posts: 878
  • EWP
    • View Profile
Re: New MFA method
« Reply #98 on: February 21, 2013, 03:08:36 PM »
Generally speaking the best AFMA is the one that on average gives you the best image across the entire range subject distance range....unless the lens is really out of kilter.

That has been a long standing issue with zooms.  You not only have the subject distance issue to worry about but different AFMA settings across the entire zoom range.  Canon at least now with the later bodies allows for a separate AFMA for each end of the zoom range but the middle range cannot always be extrapolated linearly between the two settings.

Anyway some of the error can be made up by stopping down the lens.  The best thing to do is to send the lens back to canon and have it adjusted so that the focal plain relative to AFMA is consistent from minimum to infinity focus distance when AF is activated.


Some lenses just dont focus well on a linear scale.  Some need different AFMA settings focused far than focused close.  To me that's a manufacturing flaw or the lens is just out of spec and needs adjusting.  In those cases you either need to set the AFMA for the distance you use it the most, or split the difference in the near and far AFMA tests and use aperture to increase the DOF and mask the error.



Yup, that has been my feeling too and I am glad I can finally find someone to agree with me  :)

For all the lenses I have, I have never been able to AFMA it such that it works for any subject distance.
The ones that worked for any subject distance had AFMA turned off LOL!

Yet I do recall someone on this forum had great success with it... It is puzzling.
Perhaps some particular lens flaw happens to be compensated by AFMA perfectly.

digital paradise

  • Canon 70D
  • ****
  • Posts: 290
    • View Profile
    • Zenon Char Photography
Re: New MFA method
« Reply #99 on: February 21, 2013, 06:40:44 PM »
Repeat these steps after restarting your camera. It will give you different values.

This should apply to all MFA methods then. 

East Wind Photography

  • 1D Mark IV
  • ******
  • Posts: 878
  • EWP
    • View Profile
Re: New MFA method
« Reply #100 on: February 21, 2013, 09:35:40 PM »
No because Dot Tune uses AF detect, others do not.

Repeat these steps after restarting your camera. It will give you different values.

This should apply to all MFA methods then.

digital paradise

  • Canon 70D
  • ****
  • Posts: 290
    • View Profile
    • Zenon Char Photography
Re: New MFA method
« Reply #101 on: February 21, 2013, 10:48:34 PM »
No because Dot Tune uses AF detect, others do not.

Repeat these steps after restarting your camera. It will give you different values.

This should apply to all MFA methods then.

So then what difference would it make using AF detect compared to other methods. Does the power only shut of to AF detect and not to other parts of my camera?

East Wind Photography

  • 1D Mark IV
  • ******
  • Posts: 878
  • EWP
    • View Profile
Re: New MFA method
« Reply #102 on: February 22, 2013, 10:49:16 AM »
Other methods such as FoCal actually analyze the image using a special target to determine sharpness or resolvability at various AFMA settings.  It actually takes pictures of the target to do it's analysis.  It does not use AF detect at all.

They do have a beta version out which appears to do something similar as Dot Tune but I have not tried it yet.

I also have found that FoCal does not work very well with some camera lens combinations.  I've used it and then done checks using a lensalign 3D target and found it to be quite off in its AFMA selection with some lenses.  One of the things that it is biased towards is AF repeatability and consistency.  Not necessarily the highest sharpness or resolvability.  I agree that consistency is important but the goal is to provide the highest resolvability by calibrating for an in focus image.  So even with FoCal you STILL have to do manual checks and compare shots to make sure it's recommended AFMA is sane.  So I just do it manually now since the manual verification has to be done anyway. Manual meaning the sloped ruler and high contrast target process.

No because Dot Tune uses AF detect, others do not.

Repeat these steps after restarting your camera. It will give you different values.

This should apply to all MFA methods then.

So then what difference would it make using AF detect compared to other methods. Does the power only shut of to AF detect and not to other parts of my camera?

canon rumors FORUM

Re: New MFA method
« Reply #102 on: February 22, 2013, 10:49:16 AM »

digital paradise

  • Canon 70D
  • ****
  • Posts: 290
    • View Profile
    • Zenon Char Photography
Re: New MFA method
« Reply #103 on: February 24, 2013, 03:02:18 AM »

Harry Muff

  • Canon 6D
  • *****
  • Posts: 415
    • View Profile
    • My Flickr:
Re: New MFA method
« Reply #104 on: February 24, 2013, 04:54:54 AM »
I can't see any mention of this on the Magic Lantern site yet.


Will the 5D2 be included?
Some cameras… With Canon written on them. Oh, and some lenses… Also with Canon written on them. Oh, and a shiny camera with Fuji written on it too...

Feel free to have a wander round my flickr

canon rumors FORUM

Re: New MFA method
« Reply #104 on: February 24, 2013, 04:54:54 AM »