HowTo? Make appropriate target for AFMA auto-dot-tune?

Started by l_d_allan, January 11, 2014, 05:51:39 PM

Previous topic - Next topic

0 Members and 1 Guest are viewing this topic.

l_d_allan

With my 6d and 5d2, I rarely, if ever, get a successful Dot-Tune AFMA.

There is almost always a warning:
"Double-check the Focus Target"

90%+ of the time, there is a value for AF micro-focus that seems valid ... mid-way between the left and right side of the focus confirmation squares ... such as
__xxXXXXXXXXXXXXXx_________
results in something like -3, for example, which seems valid.

This happens with both cameras, and all Canon EF lenses that I've tested, including 35mm f2 IS, 50mm f1.8ii, 24-105mm f4, etc. I get the focus set with LiveView and 10x magnification, then turn AF off and LiveView off.

I've tried several targets, like a red brick wall with almost white mortar, 19"x22" laser printer patterns for high contrast, outside in direct sun, inside in room with quite a bit of light from south facing window, etc.

Am I doing something wrong? Some setting I'm over-looking?

I did come across a post for the 6d that mentioned:
'In "INFO button display options" -> "Display shooting functions" has to be enabled'
but I'm unclear what that means.


a1ex

Back then when I tried it, I simply focused on some fine print. Since I don't use AF lenses, I have no idea how well (and if) it's working these days (but from lack of bug reports, I thought it was working).

I took a look at the code, and I think I found a bug introduced here: https://bitbucket.org/hudson/magic-lantern/commits/13de0bb601dc0d26c22135616a13f719cd33a2ac

There is a message saying "Double-check the Focus Target" in the "OK" code path, in the automatic scanning routine. The manual scanning routine seems OK, let me know if it helps.

ayshih

I just tested the module, and the automatic scanning still works fine.

Quote from: l_d_allan on January 11, 2014, 05:51:39 PM
With my 6d and 5d2, I rarely, if ever, get a successful Dot-Tune AFMA.

There is almost always a warning:
"Double-check the Focus Target"

90%+ of the time, there is a value for AF micro-focus that seems valid ... mid-way between the left and right side of the focus confirmation squares ... such as
__xxXXXXXXXXXXXXXx_________
results in something like -3, for example, which seems valid.

It sounds like you are, in fact, getting a successful AFMA determination, and you just need to check whether your Canon setting has been properly updated to that midpoint value.

Quote from: a1ex on January 11, 2014, 06:12:07 PM
There is a message saying "Double-check the Focus Target" in the "OK" code path, in the automatic scanning routine. The manual scanning routine seems OK, let me know if it helps.

That message can probably just be removed.  I imagine it was put there to encourage the user to make sure that the camera was not bumped during the process, but it's prone to misinterpretation, and it overwrites the AFMA message that's printed out just a few lines above.
Canon EOS 50D | 17–40mm f/4L & 70–300mm f/4.5–5.6 DO IS | Lexar 1066x

a1ex

That message appears twice, so I think it was just forgotten here.

Somebody merged that pull request before I had a chance to try it, and after that I assumed he understood the code and tried it. Then, these guys got angry at me when I've asked them to actually try things before commiting or merging patches...

l_d_allan

Quote from: a1ex on January 11, 2014, 06:12:07 PM
There is a message saying "Double-check the Focus Target" in the "OK" code path, in the automatic scanning routine.

That would explain the issue. IIRC, it did work fine on the 5d2 some months ago ... which really had me wondering when the 6d seemed glitchy, and then the 5d2 had the same issue.

Thanks for checking.

QuoteThe manual scanning routine seems OK, let me know if it helps.

So it will be in the next Nightly-Build, at least for the 5d2? And eventually in the TL build?

At this point, I haven't attempted to compile my own builds.

a1ex

Hopefully yes, if anyone confirms the bug and commits the fix.

l_d_allan

Quote from: a1ex on January 11, 2014, 06:12:07 PM
Since I don't use AF lenses

Out of curiosity, I'm wondering if you have "focus confirmation chipped lenses"?

In addition to several Canon EF and EF-S lenses, I have the Samyang 14mm and 8mm fisheye lenses, which are completely manual with no "focus confirmation chips". My impression has been that DotTune AFMA doesn't apply to those lenses. Or not?



ayshih

Quote from: a1ex on January 11, 2014, 06:50:34 PM
Hopefully yes, if anyone confirms the bug and commits the fix.

I removed the message and double-checked again that the automatic scanning is working properly.  Pull request submitted.
Canon EOS 50D | 17–40mm f/4L & 70–300mm f/4.5–5.6 DO IS | Lexar 1066x

a1ex

I have 3 focus confirmation chips, and DotTune works just fine with them.

The problem I found with these chips is they tend to interact in strange ways with the camera. One of them (no-name) works just fine on 5D2, but on 5D3 the camera switches between modes and menus without me pressing anything (also with plain Canon firmware). The other two are Optix v6, advertised to work on 5D3, they seem to work better, but they get into program mode just from taking pictures. When that happens, you usually have to remove the lens and put it back.

So, I'm no longer using these chips. Zoom on half-shutter and sharpness tricks work best for me.

I also have a focusing screen like this: http://www.magiclantern.fm/forum/index.php?topic=9916 but I couldn't get used to it.

l_d_allan

Quote from: a1ex on January 11, 2014, 06:50:34 PM
if anyone confirms the bug

IIRC, when I first used the 5d2 nightly build for DotTune AFMA, it seemed to "just work" ...

When I got the 6d and attempted to do DotTune AFMA, the builds for both the 6d and also the 5d2 seem less stable than before. I'll get really unexpected patterns like:


  • Capital X stands for "confirmed focus"
  • lower case x stands for "marginal focus"
  • _ stands for "not focused"
  • XXXXXXXXXXXXXXXXXXXXXXXXXXXXXXXXXXXXXXXXXXX
  • everything seems to have focus confirmation ... invalid
  • ___xx____XXXX____XXXX___xxx______
  • seems unexpected to have several "not focused" between a number of  "confirmed"

Has anyone else experienced this? I've had this happen outside in bright sun, on tripod, with a secured high contrast target, with little or no wind. It doesn't seem like the camera or target is moving, but the pattern above suggests that something is moving.

Or operator error?


a1ex

Quite possible.

You can try some very old builds to check if there's indeed a problem with the new algorithm. The ones before 2013-07-29 are using my implementation, the later ones are using the algorithm from YMP.
http://nanomad.magiclantern.fm/nightly/archive/

If the problems appeared right after that date, we know where to look. If not, it would be very helpful if you can point out the date when the problems appeared.

l_d_allan

Quote from: a1ex on January 11, 2014, 07:13:43 PM
So, I'm no longer using these chips.

With the u.w.a. Samyang 14mm and fisheye, the DOF tends to be so great, that I more or less do "zone focusing". The 14mm distance scale seems reasonably accurate, but the distances on the 8mm fisheye seem flawed. To focus at infinity, it seems to work best to be "past" infinity. I've been tempted to get a "focus confirmation chip" for the Samyang 8mm fisheye.

QuoteZoom on half-shutter and sharpness tricks work best for me.

I'll give those a try. I've seen them in the menu, but I was fuzzy on how they worked.

QuoteI also have a focusing screen like this: http://www.magiclantern.fm/forum/index.php?topic=9916 but I couldn't get used to it.

I suppose those would give a "split range-finder" effect? That's all we had back in the early 70's with my TLb and then A-1.

I did have the interchangeable screen on my 5d2 pop out, which was hard to work around. IIRC, LiveView sort of worked. Fortunately, the camera was still within warranty. Whew.

To me, the interchangeable focus screens are just something to go wrong.  For me, 10x magnification in LiveView is great, but it sure slows things down. YMMV.

I'm ignorant about "focus peaking". I suppose if I did more "sports/action photography", the split-range-finder would make more sense.

l_d_allan

Quote from: a1ex on January 11, 2014, 07:22:33 PM
If the problems appeared right after that date, we know where to look. If not, it would be very helpful if you can point out the date when the problems appeared.

I'll put that on my "to do list", but it may be a while before I have anything to report back. Sorry.

Interesting ... and also frustrating to have working code get less stable over time. It happens. Something of an example of "bit rot"?



a1ex

Yes. That's why good bug reports are very important; if no developer uses some feature for a few months and nobody reports the problem, we have no idea about the presence of the bug.

Note: when I find this kind of bugs reported after half an year, I usually remove the affected feature. But this one seems to be too useful to do that.

Audionut

Just tried the latest build here on a 5D3.  Worked like a charm.  The were the odd focus confirms outside the main bunch, but this is to be expected IMO.

The only messages I received were telling me to make sure I have perfect focus, and leave the camera still.

I use a focus confirmation target downloaded from focal.  I even had it printed on some high quality paper when I was doing wedding pics for a customer.  The light was EV 6 and target and camera were not affected by movement.

With lenses that have a minimum aperture of f/4.0 or so, I would run the test with the camera at minimum focus distance to the target.  This will ensure you have the narrowest DOF and should help the routine by only using a narrow range of AFMA values.  From there, you can move the camera further from the target to whatever range you deem to be your average and run the routine again.  Don't be surprised to learn that the AFMA value will be different at different focus distances.

I would also use the manual dot tune method to check for issues.  If you have the same problems with the manual setup, this suggests a failure with the camera/lens/user.

r12

is there anywhere the module with w and t for the 5D3. At the beginning both values could be set separately

l_d_allan

Quote from: l_d_allan on January 11, 2014, 05:51:39 PM
With my 6d and 5d2, I rarely, if ever, get a successful Dot-Tune AFMA.

There is almost always a warning:
"Double-check the Focus Target"

Seems fine now ... I'm not longer getting that warning. Thanks.