New MFA method

[East Wind Photography]

It's of no use to me when the AFMA setting is off by 8, a full DOF.  Maybe Nikon has better AF confirmation.

Nope. As mentioned previously I had the same kinds of error.  So in essence the method is flawed.  Relies on an accurate AF confirm which we all have known for some time is pretty darned useless.

It's a very nice concept and one that someday may be refined enough to actually be built into firmware.  Right now it's even worse than FoCal.

I tried this today on the 5d3 with my 50L 1.2 and 16-35II's long end and it gave me very inconsistent results. I did this on a controlled light setup, marked my positions at max distance for both lenses, wide open aperture and aligned the target (from lens-align) parallel to the sensor. I tried this technique 3 times on the same positions with both lenses. I shut down the camera and restarted the test after each try. These are the results:
50mm 1.2
first try -5 to +12
second -9 to +9
third -8 to +16

16-35II on 35mm
first try -9 to 10
second -5 to +14
third -6 to +12

Did I do something wrong?

I'm actually surprised at how few people have reported DotTune isn't working for them. Not because I don't have confidence in my method, but because I expected the teething period to last longer as people sorted out the nuances of the viewfinder focus confirmation system, not to mention the percentage of users who report MFA issues using even existing methods. In my original dpreview Nikon thread the success rate was close to 100%, whereas on the Canon FM thread there were quite a few more people with issues. But nearly everyone who had problems later reported back and said the method worked for them once they followed the additional guidance I posted in my follow-up posts on that FM thread.

As for the accuracy and precision of the viewfinder confirmation system, I've actually found it to be extremely accurate on both the Nikon and Canon bodies I tried. But it takes patience, diligence, and some practice work through the nuances of its feedback.

[Meh]

As for the accuracy and precision of the viewfinder confirmation system, I've actually found it to be extremely accurate on both the Nikon and Canon bodies I tried. But it takes patience, diligence, and some practice work through the nuances of its feedback.

What "nuances of it's feedback"?  The light either comes on, or it doesn't.

[digital paradise]

Not sure if it was here or elsewhere that the author and testers determined that when you get to the point where the AF green dot flickers or delays then you went too far. It can do that for 3 or more values - negative or positive.

The last value where the green dot shows AF lock instantly is the stopping point. If the next value flickers or hesitates even for a bit use the previous value. Not sure if the members that are having mixed results are following this. It happened to me when I did not stop and kept going until there was no AF confirmation at all. Negative side was consistent but the positive side was +16 one time and +20 the next. 

I have yet to test this but Fred Miranda at FM forums stated it appeared more accurate.               

[nicku]

Interesting.... I have a Sigma 50mm f/1.4, that need AFMA. i will try it

[J.R.]

Not sure if it was here or elsewhere that the author and testers determined that when you get to the point where the AF green dot flickers or delays then you went too far. It can do that for 3 or more values - negative or positive.

The last value where the green dot shows AF lock instantly is the stopping point. If the next value flickers or hesitates even for a bit use the previous value. Not sure if the members that are having mixed results are following this. It happened to me when I did not stop and kept going until there was no AF confirmation at all. Negative side was consistent but the positive side was +16 one time and +20 the next. 

I have yet to test this but Fred Miranda at FM forums stated it appeared more accurate.             

Thanks for the information. I've been testing this till the AF confirmation is no longer available and have been getting incorrect results.

I'll test again and see what happens.

[AlanF]

To those have made repeat measurements and have found different values - you are experiencing the life of a scientist.  All measurements are subject to random and systematic errors.  Just repeat the experiment several times and use the average value, unless some of the values are absurd. Errors can obviously arise from the initial focussing or during the dot confirm.  I would not consider a series of AFMAs of say for example 2, 6, 4, 3, 3, 0 to be wrong but merely a mean of 3 with a not unsurprising standard deviation. 

[cervantes]

snapsy aka horshack aka the guy that killed Reikan Technology Ltd   

I've never used AFMA but I'll give it a try tonight!

[East Wind Photography]

And then you will find that your focus is no better than if you just turned off AFMA.  The setting is not something that you can average based on a set of bad values.  You can be off by 1 but any more than that and you are off.  1 AFMA is 1/8 of the depth of field.  For most telephoto lenses where AFMA makes the biggest change being off by 1 wide open makes a difference wether the eye of you subject is in focus or not.  There can be no guessing and no averaging of test values.  The setting has to be exactly where it needs to be.  The dot method doesn't get you there.  You still need to run your photo tests and dial in the setting.

To those have made repeat measurements and have found different values - you are experiencing the life of a scientist.  All measurements are subject to random and systematic errors.  Just repeat the experiment several times and use the average value, unless some of the values are absurd. Errors can obviously arise from the initial focussing or during the dot confirm.  I would not consider a series of AFMAs of say for example 2, 6, 4, 3, 3, 0 to be wrong but merely a mean of 3 with a not unsurprising standard deviation.

[digital paradise]

I jokingly asked snapsy when he was coming out with software. I had in mind doing this test several times to include variation and get the best value. Obviously this was meant to be a quick method so software just slows it down. It is so quick to use that it would take no time to go through it several times putting the lens OOF between the tests. Next week I will give it a good run.

So East Wind. Are you saying that the same issues apply to FoCal and Focus Tune? I normally send my lenses to Canon because I prefer a trained technician with the correct tools to tune my lenses. It is a pain so I am always looking for new method.

I have not tried FoCal but did try Focus Tune with mixed results. Also I tried every other method out there in the last several years. I won't use any method where I have to visually make a decision. I want the method to tell me what value to use.       

[AlanF]

And then you will find that your focus is no better than if you just turned off AFMA.  The setting is not something that you can average based on a set of bad values.  You can be off by 1 but any more than that and you are off.  1 AFMA is 1/8 of the depth of field.  For most telephoto lenses where AFMA makes the biggest change being off by 1 wide open makes a difference wether the eye of you subject is in focus or not.  There can be no guessing and no averaging of test values.  The setting has to be exactly where it needs to be.  The dot method doesn't get you there.  You still need to run your photo tests and dial in the setting.

To those have made repeat measurements and have found different values - you are experiencing the life of a scientist.  All measurements are subject to random and systematic errors.  Just repeat the experiment several times and use the average value, unless some of the values are absurd. Errors can obviously arise from the initial focussing or during the dot confirm.  I would not consider a series of AFMAs of say for example 2, 6, 4, 3, 3, 0 to be wrong but merely a mean of 3 with a not unsurprising standard deviation.

The nature of making any measurement is that your results will fit a normal or skewed Gaussian distribution. The more measurements you make, the closer your result will be to the true one.  FoCal fits its measurements to a Gaussian curve.  Taking an average of a series of measurements is surely better than turning AFMA off. Canon does not claim to have an accuracy in any one shot to 1 unit.  Look at the spread of data in the lens rental analysis where they quote the equivalent of a standard deviation for repeat measurements on different bodies etc. 

[neuroanatomist]

1 AFMA is 1/8 of the depth of field.

Point of clarification...  1 unit of AFMA is 1/8 the depth of focus, not depth of field.  The AF system doesn't know subject distance (that's estimated by the lens after focus is achieved).  Depth of focus is measured at the AF sensor itself (micron distances).  The practical consequence is that a unit of AFMA has more impact on depth of field the closer the subject is to the camera.

[East Wind Photography]

I would have thought it would have more impact on subjects at farther distance and less effect at closer distances.  Otherise we would be AFMA testing at closer distances.  either way it must be coincidence then that 8 units appear to equal the depth of field on my 600 F4.

1 AFMA is 1/8 of the depth of field.

Point of clarification...  1 unit of AFMA is 1/8 the depth of focus, not depth of field.  The AF system doesn't know subject distance (that's estimated by the lens after focus is achieved).  Depth of focus is measured at the AF sensor itself (micron distances).  The practical consequence is that a unit of AFMA has more impact on depth of field the closer the subject is to the camera.

[David Hull]

Nope. As mentioned previously I had the same kinds of error.  So in essence the method is flawed.  Relies on an accurate AF confirm which we all have known for some time is pretty darned useless.

It's a very nice concept and one that someday may be refined enough to actually be built into firmware.  Right now it's even worse than FoCal.

I tried this today on the 5d3 with my 50L 1.2 and 16-35II's long end and it gave me very inconsistent results. I did this on a controlled light setup, marked my positions at max distance for both lenses, wide open aperture and aligned the target (from lens-align) parallel to the sensor. I tried this technique 3 times on the same positions with both lenses. I shut down the camera and restarted the test after each try. These are the results:
50mm 1.2
first try -5 to +12
second -9 to +9
third -8 to +16

16-35II on 35mm
first try -9 to 10
second -5 to +14
third -6 to +12

Did I do something wrong?

I'm actually surprised at how few people have reported DotTune isn't working for them. Not because I don't have confidence in my method, but because I expected the teething period to last longer as people sorted out the nuances of the viewfinder focus confirmation system, not to mention the percentage of users who report MFA issues using even existing methods. In my original dpreview Nikon thread the success rate was close to 100%, whereas on the Canon FM thread there were quite a few more people with issues. But nearly everyone who had problems later reported back and said the method worked for them once they followed the additional guidance I posted in my follow-up posts on that FM thread.

As for the accuracy and precision of the viewfinder confirmation system, I've actually found it to be extremely accurate on both the Nikon and Canon bodies I tried. But it takes patience, diligence, and some practice work through the nuances of its feedback.

It seems simple enough for a peerson to run your proposed methodology and compare it to what he done using some previous method and see if the results are similar.  I did it quickly on several lenses last weekend and got roughly the results I expected (i.e. it appeared to work -- not surprising since it seems technicaly sound).  I find that all of my glass (on the 5DIII anyway) is within +/- a couple tics of 0 and that didn’t change from what I did with my previous methodology.  The 50 1.4 wanted -7 which was a little more than previous for some reason (maybe I screwed the pooch somewhere on that one).

[AlanF]

Disappointingly, it didn't work for me. My 300mm f/2.8 is +8 on FoCal and sloping ruler tests, but +2 on the Dot. With a 2xTC, it is +6 on FoCal and slope and -8 on the dot. For the latter, the actual image at +6 was the same quality as live view and -8 out of focus.

The method did work spot on for the 100-400mm L.  The series II telephotos do have sophisticated focussing algorithms and perhaps they don't work so well with the manual focussing dot.  It is surprising that the green dot lights up over a range of 20 units. 

[AlanF]

I checked the method again today with the 300 f/2.8 series II plus extenders using a high contrast black cross on white in good light. In all cases, the dot method was way out. Recalibration using the same target and sloping ruler was within 1 unit of previous.