AF issues using adapted EF lenses / AF Shootout vs 6D2

I was fooling around with the camera and found that it would focus amazingly well sometimes and not at all others.

I remembered reading that all 5000+ focus points are horizontal. Since there are not vertical or cross type AF points, the camera will not focus on a perfectly horizontal line. In good light with good contrast, a slight angle will do for focusing on a vertical line. In low light with low contrast, you may need a 45 degree angle of more to get the camera to focus on a horizontal line. On a vertical line, the camera focuses instantly and perfectly even in low light with low contrast.

Shades of the old EOS 620.
Interesting although sounds a little goofy to start framing shots with an angle to nail focus!
Nah... you don’t need 45 degrees. But you do need that line to cross the two sub/pixels at your focus point. You can easily test it. If you are having problems focusing on horizontal lines then rotate the camera 90 degrees and try again.
Yes you do need 45 degrees, or close to it, in low light with low contrast. My test was on the line where the wall meets the ceiling in a dark corner. On a brighter part of the room, or on a black table against a white wall, a slight angle is perfectly fine. For a vertical line, I was unable to find anything low contrast or dark enough to give the camera any trouble at all. It provides instant and perfect focus on vertical lines of any contrast in any light.

--
Ed Rizk
 
Last edited:
Ah ya... adapted tamron lenses are not OEM EF lenses. So that is not on canon.

And long glass... meh, not surprising.

This whole thread so far was not about those super long lenses. Why someone would want to take the R for wild life is beyond me. In burst shooting you get a stop motion type view rather than live view update, so it was already subpar for wild life in the first place.
 
I was fooling around with the camera and found that it would focus amazingly well sometimes and not at all others.

I remembered reading that all 5000+ focus points are horizontal. Since there are not vertical or cross type AF points, the camera will not focus on a perfectly horizontal line. In good light with good contrast, a slight angle will do for focusing on a vertical line. In low light with low contrast, you may need a 45 degree angle of more to get the camera to focus on a horizontal line. On a vertical line, the camera focuses instantly and perfectly even in low light with low contrast.

Shades of the old EOS 620.
Interesting although sounds a little goofy to start framing shots with an angle to nail focus!
Nah... you don’t need 45 degrees. But you do need that line to cross the two sub/pixels at your focus point. You can easily test it. If you are having problems focusing on horizontal lines then rotate the camera 90 degrees and try again.
Yes you do, or close to it, in low light with low contrast. My test was on the line where the wall meets the ceiling in a dark corner. On a brighter part of the room, or on a black table against a white wall, a slight angle is perfectly fine. For a vertical line, I was unable to find anything low contrast or dark enough to give the camera any trouble at all. It provides instant and perfect focus on vertical lines of any contrast in any light.
What I meant simply was that one need to have a feature such as a verticals line to cross both parts of the phase detection subpixels to give the camera something to compare in subpixel A to subpixel B.

The canon representative pretty much said just a small tilt is typically enough. In a low contrast situation it is about having enough “feature” for the sensor to do its work.

ALL mirrorless have this issue, since NONE of them are cross type or double cross type AF points.

Someone else bragged that his Leica Q is better and that they have it sorted while canon does not... which is another false assumption and not understanding how it works. Give a CDAF system a target with no contrast than it took will fail (it is contrast detection after all).

Focus on an actual subject. Your cat, dog, person, tree, etc etc and try to focus on the actual think you want to pull attention too (99% of the time it has contrast, like aomeone’s eye instead of their forehead) and it should grab focus np. So first of all... how about real world scenerios and realistic usage.

I mean have a look at the video posed in OP. The guy is testing how that camera in full auto AF mode. Would you trust your DSLR to focus on the correct subject in auto AF mode in a scene like that? No camera on earth will. And since the guy changes the framing or at least he doesn’t keep the situation even close to the same it isn’t even a valid adapt at a valid comparison.
 
Last edited:
lawny13 wrote:All the reports I have read so far say that AF is stellar with adapted EF lenses. I wonder if you are an outlier or rather if your particular unit is an outlier.
Just reporting my findings, there is a video linked in my OP that seems to corroborate this as well. really doubt my unit is a lemon..
Ya... watched the video. And it would hardly call it conclusive. We all know AF systems tend to lock onto the nearest subject. So the framing for the “worse performing” lens was such that I am not surprised that it grabbed onto the fence at times. While for the best one which is the sigma the framing is such that the white dog is not only bigger but also closer to the camera than anything else. I would be more curious if he used expanded AF pint mode which would limit the AF to one area rather than such a wide area to let it roam around.

Heck if I were to shoot wide mode with the A7III I also wouldn’t trust the system to magically select the subject I intend since as I said before it should pick up what was nearest. It is more effective to place an AF point on what I want it to focus on while in AF-C and once it grabs on it sticks.

Generally the issue you are mentioning is definitely a global mirrorless AF problem. They are either horizontal or virticle type PDAF points. Many DSLRs have all cross type and some double cross type AF points which makes them perform better on low contrast targets. BUT, that is only up to a point. They need some feature to focus on.
cross points vs only vertical PDAF may be the main culprit here for all i know..

that doesn't negate the fact this is a real issue.

I do agree the guy shouldn't have used wide mode and use single point instead.

FWIW my own personal findings were done using single point, which is how i focus 90% of the time.
 
lawny13 wrote:All the reports I have read so far say that AF is stellar with adapted EF lenses. I wonder if you are an outlier or rather if your particular unit is an outlier.
Just reporting my findings, there is a video linked in my OP that seems to corroborate this as well. really doubt my unit is a lemon..
Ya... watched the video. And it would hardly call it conclusive. We all know AF systems tend to lock onto the nearest subject. So the framing for the “worse performing” lens was such that I am not surprised that it grabbed onto the fence at times. While for the best one which is the sigma the framing is such that the white dog is not only bigger but also closer to the camera than anything else. I would be more curious if he used expanded AF pint mode which would limit the AF to one area rather than such a wide area to let it roam around.

Heck if I were to shoot wide mode with the A7III I also wouldn’t trust the system to magically select the subject I intend since as I said before it should pick up what was nearest. It is more effective to place an AF point on what I want it to focus on while in AF-C and once it grabs on it sticks.

Generally the issue you are mentioning is definitely a global mirrorless AF problem. They are either horizontal or virticle type PDAF points. Many DSLRs have all cross type and some double cross type AF points which makes them perform better on low contrast targets. BUT, that is only up to a point. They need some feature to focus on.
cross points vs only vertical PDAF may be the main culprit here for all i know..

that doesn't negate the fact this is a real issue.

I do agree the guy shouldn't have used wide mode and use single point instead.

FWIW my own personal findings were done using single point, which is how i focus 90% of the time.
Well... all mirrorless have this issue.

Mirrorless has its advantages but it also has its disadvantages. And it isn’t only single line va cross type that is the issue. A separate AF module in DSLRs also have A. Far large AF pixels. How else do you think they dealt with having less light reflected down to it? B. Some AF points are double cross type, so it has even more contrasting features to use. C. It doesn’t have an IR filter on it. So it also has that light available to it. Not sure how much the latter contributes.

The mirrorless systems don’t have micro alignment adjustments to deal with, and the nature of the thing allows for near full frame AF coverage. And the “issue” as you call it is a YMMV type thing. From the A7II to the A7III I have never one across it. Then again I never tried to take a picture of the corner line of a wall meeting g the ceiling (was both walls the same color?)
 
So far I have used my EOS R with the following:

24-70 2.8 II - Instant acquisition of subject with sharp focus wide open. On par with 5D4.

85 1.2 - Same as on 5D4, slower to focus with about the same sharpness wide open.

Can't wait to see what the RF lenses do.

Joe
 
  • diness wrote:
  1. DeeNikon wrote:
Omg.. and reviewers are all super happy about AF? Worlds best autofocus is dual pixel? Rly?
And don’t get me started on Studio photography.
Modeling light is not enough?
Depends. If the strobe is close to the person, yes, it helps quite a bit.

But if it’s a group and the strobe is farther away. And I’m at f8, the AF will struggle.

That’s a quirk of the Sony I don’t like.

I’m happy the Canon focuses wide open.
I’ve heard that Sony has changed this with the new firmware. Have you updated yet?
I read in the Sony thread about the new firmware update. Haven’t had a chance to do it yet. I hope you’re right about addressing the stop down focusing.

Thanks for the reminder.
Don’t!! Never ever update a song FE immediately upon release. Always always research and read reviews. I came across someone who said it bricked his camera and other that said his 3rd party batteries all stopped working.

i am also curious if they deal with their focus shift issue
 
lawny13 wrote:All the reports I have read so far say that AF is stellar with adapted EF lenses. I wonder if you are an outlier or rather if your particular unit is an outlier.
Just reporting my findings, there is a video linked in my OP that seems to corroborate this as well. really doubt my unit is a lemon..
Ya... watched the video. And it would hardly call it conclusive. We all know AF systems tend to lock onto the nearest subject. So the framing for the “worse performing” lens was such that I am not surprised that it grabbed onto the fence at times. While for the best one which is the sigma the framing is such that the white dog is not only bigger but also closer to the camera than anything else. I would be more curious if he used expanded AF pint mode which would limit the AF to one area rather than such a wide area to let it roam around.

Heck if I were to shoot wide mode with the A7III I also wouldn’t trust the system to magically select the subject I intend since as I said before it should pick up what was nearest. It is more effective to place an AF point on what I want it to focus on while in AF-C and once it grabs on it sticks.

Generally the issue you are mentioning is definitely a global mirrorless AF problem. They are either horizontal or virticle type PDAF points. Many DSLRs have all cross type and some double cross type AF points which makes them perform better on low contrast targets. BUT, that is only up to a point. They need some feature to focus on.
cross points vs only vertical PDAF may be the main culprit here for all i know..

that doesn't negate the fact this is a real issue.

I do agree the guy shouldn't have used wide mode and use single point instead.

FWIW my own personal findings were done using single point, which is how i focus 90% of the time.
Well... all mirrorless have this issue.

Mirrorless has its advantages but it also has its disadvantages. And it isn’t only single line va cross type that is the issue. A separate AF module in DSLRs also have A. Far large AF pixels. How else do you think they dealt with having less light reflected down to it? B. Some AF points are double cross type, so it has even more contrasting features to use. C. It doesn’t have an IR filter on it. So it also has that light available to it. Not sure how much the latter contributes.

The mirrorless systems don’t have micro alignment adjustments to deal with, and the nature of the thing allows for near full frame AF coverage. And the “issue” as you call it is a YMMV type thing. From the A7II to the A7III I have never one across it. Then again I never tried to take a picture of the corner line of a wall meeting g the ceiling (was both walls the same color?)
Thanks I'm well aware of the benefits of off-sensor AF modules inside DSLRs & of the limits of MILC's.

I've used all the Sony full frames including A73, own an XT2. Have demo'd the Z7.

I'm also not shooting walls or ceilings but 3D objects.

It shouldn't hunt or take 6 seconds to go to green box imo.
 
I was fooling around with the camera and found that it would focus amazingly well sometimes and not at all others.

I remembered reading that all 5000+ focus points are horizontal. Since there are not vertical or cross type AF points, the camera will not focus on a perfectly horizontal line. In good light with good contrast, a slight angle will do for focusing on a vertical line. In low light with low contrast, you may need a 45 degree angle of more to get the camera to focus on a horizontal line. On a vertical line, the camera focuses instantly and perfectly even in low light with low contrast.

Shades of the old EOS 620.
 
I was fooling around with the camera and found that it would focus amazingly well sometimes and not at all others.

I remembered reading that all 5000+ focus points are horizontal. Since there are not vertical or cross type AF points, the camera will not focus on a perfectly horizontal line. In good light with good contrast, a slight angle will do for focusing on a vertical line. In low light with low contrast, you may need a 45 degree angle of more to get the camera to focus on a horizontal line. On a vertical line, the camera focuses instantly and perfectly even in low light with low contrast.

Shades of the old EOS 620.
I know nothing about sensor manufacturing, but I wonder if it possible for Canon to rotate select pixels 90 degrees so they act as vertical AF sensor points.

The combined sub pixels would still offer the same light gathering capability. But the pixel would be divided into sub pixels horizontally instead of vertically.

Like I said, I don’t have the technical knowledge to know if that is 1.) possible, and 2.) if the DPAF would still work that way.
I don't see why not, and think it's a good idea. However, focus would not be uniform throughout the frame. You would have to know where to focus how, like you do with a DSLR.

That wouldn't bother me, but it would make the AF less "idiot proof". Heck, the existing system doesn't bother me. After two days, I can tell exactly how to focus on what.
 
It shouldn't hunt or take 6 seconds to go to green box imo.
Which lens took 6 seconds to give you a green box?
35mm 1.4, 50mm 1.8... sometimes it will just hunt forever, i'm talking pretty stressfull situations (low light, low contrast subjects) but nothing the 5D3 can't normally handle.
 
I was fooling around with the camera and found that it would focus amazingly well sometimes and not at all others.

I remembered reading that all 5000+ focus points are horizontal. Since there are not vertical or cross type AF points, the camera will not focus on a perfectly horizontal line. In good light with good contrast, a slight angle will do for focusing on a vertical line. In low light with low contrast, you may need a 45 degree angle of more to get the camera to focus on a horizontal line. On a vertical line, the camera focuses instantly and perfectly even in low light with low contrast.

Shades of the old EOS 620.
Interesting although sounds a little goofy to start framing shots with an angle to nail focus!
Nah... you don’t need 45 degrees. But you do need that line to cross the two sub/pixels at your focus point. You can easily test it. If you are having problems focusing on horizontal lines then rotate the camera 90 degrees and try again.
Yes you do, or close to it, in low light with low contrast. My test was on the line where the wall meets the ceiling in a dark corner. On a brighter part of the room, or on a black table against a white wall, a slight angle is perfectly fine. For a vertical line, I was unable to find anything low contrast or dark enough to give the camera any trouble at all. It provides instant and perfect focus on vertical lines of any contrast in any light.
What I meant simply was that one need to have a feature such as a verticals line to cross both parts of the phase detection subpixels to give the camera something to compare in subpixel A to subpixel B.

The canon representative pretty much said just a small tilt is typically enough. In a low contrast situation it is about having enough “feature” for the sensor to do its work.
On my R, anyway, that depends on how low the contrast AND light are. The line between my wall and ceiling in a dark corner at night requires a lot of tilt, just short of 45 degrees. I guess I could turn on the level and film it with my phone tonight if nobody believes me. Then again, maybe I'll just go out for a drink.
ALL mirrorless have this issue, since NONE of them are cross type or double cross type AF points.
And it's no problem, as long as you know how it works. It's almost irrelevant in bright light.
Someone else bragged that his Leica Q is better and that they have it sorted while canon does not... which is another false assumption and not understanding how it works. Give a CDAF system a target with no contrast than it took will fail (it is contrast detection after all).

Focus on an actual subject. Your cat, dog, person, tree, etc etc and try to focus on the actual think you want to pull attention too (99% of the time it has contrast, like aomeone’s eye instead of their forehead) and it should grab focus np. So first of all... how about real world scenerios and realistic usage.
Exactly. The eye, the face, almost any object has vertical lines or points of contrast anywhere you want focus 90+% of the time. When they don't, you just have to tilt the camera with the half press then tilt back and shoot. No big deal.
I mean have a look at the video posed in OP. The guy is testing how that camera in full auto AF mode. Would you trust your DSLR to focus on the correct subject in auto AF mode in a scene like that? No camera on earth will. And since the guy changes the framing or at least he doesn’t keep the situation even close to the same it isn’t even a valid adapt at a valid comparison.
If you let the camera pick the focus point, there is no telling what you will get.
 
It shouldn't hunt or take 6 seconds to go to green box imo.
Which lens took 6 seconds to give you a green box?
35mm 1.4, 50mm 1.8... sometimes it will just hunt forever, i'm talking pretty stressfull situations (low light, low contrast subjects) but nothing the 5D3 can't normally handle.
How do you have your AF-assist beam firing set in the AF 4 menu?

I tried the 35mm 1.4 II with the assist beam off and it performed about the same as the 5D4.

With the assist beam on, it focuses in a dark hallway if given just a little contrast or an edge or corner to focus on. Of course, you would never use it in that condition since you would have to set ISO to H2 which is too noisy to be useful.
 
Somehow Leica (Panasonic?) figured out low-light focus acquisition on mirrorless, because my Leica Q will focus without hunting in near-darkness, with minimal "false positives." I love it for astrophotography, because it can quickly easily focus on individual stars.
and are you focusing on a low contrast target? Cause that is what OP is complaining about. Both CDAF and PDAF methods requires some contrast. Take that Leica Q and try to focus on a smooth blank piece of paper with no shadows or highlights.
Having said that, I never work with continuous af on the Leica. I don't really have a need for it, and I assume it wouldn't really work well anyway, as it has no phase detection.

I own cameras by all the major brands, including three full-frame Canon dslr's (but I have not tried the Canon EOS R yet), but I keep coming back to the Leica Q whenever possible, because it's just so damn reliable. Having said that, it only has a 28mm fixed lens and one card slot.

I say this not to criticize Canon--I adore Canon color science, lenses, and their professional service system--but to say there is hope. It sounds like they just need to sort out whatever Leica was able to sort out. Maybe Canon's 2019 "pro mirrorless" camera will be better.

All the same, I look forward to testing the Canon EOS R for myself as soon as I can get my hands on one.
I am focusing on the faces of people. Often in the dark, like dimly lit concert venues and people sitting outside at night.


I am not as one-person wrote, focusing on white sheets of paper in complete darkness. I am concerned with what a camera can do under practical circumstances.


The headline here is that if low-light autofocus can work on even one mirrorless camera model, then it's not a problem necessitative to mirrorless cameras, as some here have claimed. If one manufacturer can figure it out, then it is possible for others, as well. I have faith that Canon will be able to sort it out.
 
Last edited:

Keyboard shortcuts

Back
Top