6D autofocus capabilites: let's bust some myths about it

[Sporgon]

Consider that we are imaging three-dimensional space, and compare the XY dimension (FoV) to the Z dimension (DoF). When magnification is low, the DoF is relatively large compared to the FoV. For example, with a 50mm lens and 8 m distance, Z/X is ~0.4. At higher mag, e.g., a close up of the dog with the 50mm lens at <2 m, Z/X is <0.1.

Basically, misfocus is less obvious with a lower magnification.

Interesting, this explains why with longer lenses AF accuracy can be problematic even though there is a relatively large dof.

All I have tried to show is that the outer points on the 6D give infinitely more accuracy than an f5.6 dof - because there has been confusion in this thread about the meaning of an 'f5.6 precision' AF point, and that the outer points are perfectly useable in many circumstances whether near or far.

 

[bholliman]

Yeah, the 6D doesn't handle subjects moving towards or away well at all, especially towards, probably more noticeable due to the shallower DOF. Side to side is OK.

Did you try all af options? "Release priority" is the certain way to doom, and in my experience "Tracking sensitivity" responsive is better for side movement while "locked on" does better for forwards/away movement.

Thanks, this helped. I had my AF AI Servo priority set to release priority. I changed it to first priority "focus" and AI Servo accuracy is much better.

 

[bholliman]

I have tried to show is that the outer points on the 6D give infinitely more accuracy than an f5.6 dof - because there has been confusion in this thread about the meaning of an 'f5.6 precision' AF point, and that the outer points are perfectly useable in many circumstances whether near or far.

+1 I have a 50 1.4, 85 1.8 and 135 2.0 that I use all the time on my 6D at f/2 and below with very good focus accuracy. I primarily use the center AF point, but have good success with the outer points as well in decent light. Thanks for posting the field trials.

 

[Loswr]

Consider that we are imaging three-dimensional space, and compare the XY dimension (FoV) to the Z dimension (DoF). When magnification is low, the DoF is relatively large compared to the FoV. For example, with a 50mm lens and 8 m distance, Z/X is ~0.4. At higher mag, e.g., a close up of the dog with the 50mm lens at <2 m, Z/X is <0.1.

Basically, misfocus is less obvious with a lower magnification.

Interesting, this explains why with longer lenses AF accuracy can be problematic even though there is a relatively large dof.

All I have tried to show is that the outer points on the 6D give infinitely more accuracy than an f5.6 dof - because there has been confusion in this thread about the meaning of an 'f5.6 precision' AF point, and that the outer points are perfectly useable in many circumstances whether near or far.

Agreed, it's confusing. It sure doesn't help that Canon's literature on the subject treats precision and accuracy as interchangeable terms when they're not...and the difference is relevant here. For the average of a given set of tests, accuracy is 'closeness to true' and precision is 'closeness to each other'. A single shot says nothing about AF performance. A hundred shots says something.

An f/2.8 point has a wider baseline, therefore it's more accurate than an f/5.6 point. Canon apparently does not specify accuracy values for their AF systems. AFMA adjusts the global accuracy of the system.

The center point of many bodies is a 'high precision' point, that can be either f/2.8 or f/5.6 in terms of baseline. Canon does specify precision, albeit incompletely. It's specified as depth of focus, which is the sensor-side equivalent of DoF, but distances are in microns. High precision means 'within 1/3 the depth of focus', standard precision is 'within one depth of focus'. In both cases, that's the depth of focus for the attached lens' max aperture. So, with the 50/1.4 attached, an f/5.6 point is precise to within one depth of focus at f/1.4. We still don't know exactly what that means – will 95% of shots fall within that zone, or 99%, or 99.99% (that's what I mean about Canon's incomplete specification of precision).

Personally, I found the 5DII's outer points to be ok in good light, but with close subjects and fast lenses they'd miss critical focus with noticeable frequency. If I took 2-3 shots of the same subject, if be reasonably sure if getting a keeper.

 

[Marsu42]

I've made them fit the page but have included a magnification in each one where you can see that the very narrow dof has been achieved over the focus point.

Are the corner pips really 100% crops? If so, the dof is indeed to wide to test the outer pt. performance. I don't doubt the outer af points focus *somewhere*, but they either tend to hunt (mostly on my 17-40L for some reason or another) or miss critical focus (indeed as speculated 70-300L @200-300mm, focus on the eyes).

You have to look for the problem a bit which might explain the "I just got my 6d and the af is fine!!!" posts. But if taking non-repeatable shots in a hurry and the focus hunts or you have narrow dof and they eyes are just a bit out of focus the consequence is better to use focus & recompose, at least that's what I ended up with.

Personally, I found the 5DII's outer points to be ok in good light, but with close subjects and fast lenses they'd miss critical focus with noticeable frequency. If I took 2-3 shots of the same subject, if be reasonably sure if getting a keeper.

+1, that's about it, and one problem of this method is that I burn through my shutter in no time as for potential keepers it's about 3 shots (at least) for peace of mind.

 

[Sporgon]

I've made them fit the page but have included a magnification in each one where you can see that the very narrow dof has been achieved over the focus point.

Are the corner pips really 100% crops? If so, the dof is indeed to wide to test the outer pt. performance. I don't doubt the outer af points focus *somewhere*, but they either tend to hunt (mostly on my 17-40L for some reason or another) or miss critical focus (indeed as speculated 70-300L @200-300mm, focus on the eyes).

What I was trying to show was that the AF accuracy at f1.6 was way more than f5.6. To this end the 2m dof was 'very narrow' compared with the 12m dof that f5.6 would have given.

The examples were in reply to the comments you made about this, and could be demonstrated by attaching the very small, but fast loading files.

The page has had around 1000 views overnight ( UK time) so hopefully some people have found it interesting, despite the ridicule of Mr Skirball.

The smallest dof that I can achieve with my gear is 6mm. This is with the 50mm at minimum focusing distance at f1.4. Please note: the 50mm f1.4 at minimum focus distance and f1.4 and off centre is crap. I took three consecutive frames using the 2 o'clock outer point using the bottom of the eye as a target in low light; EV 3 to be precise. Each time, as before, I turned away focus manually so the camera had to re focus each time.

You will see that in each consecutive frame the camera has produced focus within the 6mm depth of field. There is perhaps a 1 or 2 mm variation, but the end result is the same: it's in focus. ( In fact at this distance this lens/body combo is back focusing a touch).

Now I would argue that for the vast majority of 6D owners ( and 5DII for that matter) this performance is absolutely acceptable.

The last frame is me attempting to manually focus using live view, but as you can see I have not succeeded; the manual focus on the 50 '1.4 just isn't good enough. I also tried live view contrast focusing but the focus zone is too big.

I am not saying the 6D/5D AF is anything like the 1 series and 5DIII, but I am saying that people who state the outer points are 'useless', 'pointless' etc are talking out of their collective behinds.

The crops are 100% of full resolution. As I said, don't expect it to be razor with that lens at that aperture at that part of the frame.

There is one other thing to consider though: dust. I find that the single orientation points are highly susceptible to contamination by dust / pollen etc, and if you are struggling with your outer points on these cameras try using a rocket blower or even blower brush to get the dust off your AF module.

 

[bholliman]

What I was trying to show was that the AF accuracy at f1.6 was way more than f5.6. To this end the 2m dof was 'very narrow' compared with the 12m dof that f5.6 would have given.

The examples were in reply to the comments you made about this, and could be demonstrated by attaching the very small, but fast loading files.

Thanks again for the sample shots and all your work on this topic. Your samples match my experience with the 6D AF. Its very precise and accurate for non-action photography and works well with wide apertures.

 

[BPLOL]

What I was trying to show was that the AF accuracy at f1.6 was way more than f5.6. To this end the 2m dof was 'very narrow' compared with the 12m dof that f5.6 would have given.

The examples were in reply to the comments you made about this, and could be demonstrated by attaching the very small, but fast loading files.

Thanks again for the sample shots and all your work on this topic. Your samples match my experience with the 6D AF. Its very precise and accurate for non-action photography and works well with wide apertures.

Been reading this discussion for days and days. I also find my 6D outer points very, very accurate for still objects and portrait stuff. It is reliable, and way better than my 400D's focus system. I know the XTi is a 2006 camera, but it too had only one cross-type point and 8 non cross points, however the 6D's outer points are in another league. Just because the outer points are not cross type, doesn't make it equal to the Rebel. At least in my experience.

I cannot, however, compare it with a newer Rebel with 9 cross-type points.

Peace!

 

[3kramd5]

To further add, my 1Ds3 and 1Dx never miss at f/1.4 on the outter points. Never. Always exact. No variability. Nails it everytime and that's even without AFMA.

it is physical impossibility because the AF measurement angle accuracy in Canons AF is F-2,8 ( eg.3,4)

That's not correct.

The specified precision is within the depth of focus at the max aperture of the lens for a standard precision AF point, and within 0.33 depth of focus (0.5 for some models) at the max aperture of the lens for a high precision AF point.

I discussed this issue with Chuck Westfall (Canon USA's technical mouthpiece), and this is part of his response:

"The fact that the AF points are functional with apertures as small as f/5.6, f/4 or f/2.8 respectively depending on the camera model and AF point under discussion does not imply that their measuring precision is limited to the depth of focus at those apertures. The AF detection system has the capability of calculating depth of focus based on the maximum aperture of the lens, whatever it happens to be."

I've been doing some reading and came across a pertinent point. Your post above relates specifically to high-precision AF points. The outer points are not high-precision, unless I misunderstand him here: "...one of the consequences of the TTL-SIR AF system is that except for the center point, AF precision is not proportional to the maximum aperture of the lens in use."

Of course, that's a 70D-centric dialog. Maybe some of the outer 1Dx/5D3 cross points are included.

 

[3kramd5]

"Point"less ;D unscientific data

5D3, EF50mm f/1.4, ISO100, AF point shown in screenshot. Always focused on the 20. Shown are 100% crops of the top right cross, bottom left cross, and center AF point, as well as a manual focus in each setup.

 

[3kramd5]

"Point"less ;D unscientific data

5D3, EF50mm f/1.4, ISO100, AF point shown in screenshot. Always focused on the 20. Shown are 100% crops of the top right cross, bottom left cross, and center AF point, as well as a manual focus in each setup.

 

[Marsu42]

The examples were in reply to the comments you made about this, and could be demonstrated by attaching the very small, but fast loading files. The page has had around 1000 views overnight ( UK time) so hopefully some people have found it interesting, despite the ridicule of Mr Skirball.

Not to be misunderstood: it's great you took the sample shots, so I know for average shooting with a fast lens the 6d outer points should be ok. I admit that very thin dof I'm still skeptical though

There is one other thing to consider though: dust. I find that the single orientation points are highly susceptible to contamination by dust / pollen etc, and if you are struggling with your outer points on these cameras try using a rocket blower or even blower brush to get the dust off your AF module.

Thanks, good point, maybe that's really the reason why my 6d has such a mediocre performance with the outer points! I never use them a lot, so if they have collected dust I haven't much of a comparison and my 60d is all cross points.... general spec'ed precision/accuracy of the 6d is not worth anything if doesn't lock on first try and hunts and this is the horse picture my 6d's outer point thinks is in focus:

 

[mackguyver]

Marsu, that's some nice bokeh . Also, sort of off topic, but the biggest issue I've had with AF on any camera is when I've bumped my diopter adjustment. There's nothing worse than locking focus and seeing what looks like your photo in the viewfinder

I would look into the dust issue as I've had Canon reps tell me that dust and pollen can throw a lot of things off in the camera.

 

[bholliman]

Thanks, good point, maybe that's really the reason why my 6d has such a mediocre performance with the outer points! I never use them a lot, so if they have collected dust I haven't much of a comparison and my 60d is all cross points.... general spec'ed precision/accuracy of the 6d is not worth anything if doesn't lock on first try and hunts and this is the horse picture my 6d's outer point thinks is in focus:

If its not dust, there may be a problem with the AF system on your 6D, I'd recommend sending it in to Canon if its still under warranty.

My "hit" rate with outer point focus is roughly 95%+ in decent light, even with shallow DOF shots with fast primes (using One Shot and slowly moving or stationary targets). I took nearly 300 shots of the kids at a theme park last weekend with my 6D and only 4 or 5 were out of focus. Roughly 1/3 of the shots were focused using the outer points.

 

[Marsu42]

I took nearly 300 shots of the kids at a theme park last weekend with my 6D and only 4 or 5 were out of focus. Roughly 1/3 of the shots were focused using the outer points.

Ok, that's a reference, I'll try to clean it and then get it serviced (I've got the free European CPS) if it still struggles. Btw the shot above was of course absolutely worst case, but often the outer points hunt a bit and then lock - by then the scene is gone or the focus isn't spot on @100% crop.

I also have to experiment with focus priority when doing tracking with the topmost point in portrait mode, no way around it really when tracking a horse running towards you... my problem might be a combination of dirty af array, non-cross point and bad 6d tracking behavior. Good to hear from some people without super powers, now I've much more to go on than before

 

[Sporgon]

@ Marsu42; thanks for your comments. Incidentally don't think that you are alone in struggling with focus on horses. Animal fur is notoriously difficult to focus on, and with horses in particular I have always had trouble focusing with every camera I have used.

 

[bholliman]

I also have to experiment with focus priority when doing tracking with the topmost point in portrait mode, no way around it really when tracking a horse running towards you... my problem might be a combination of dirty af array, non-cross point and bad 6d tracking behavior. Good to hear from some people without super powers, now I've much more to go on than before

I wasn't aware the horse was running toward you.... I don't have much success with those types of shots either! The 6D AF can handle side-to-side movement pretty well, but it definitely struggles with targets moving at anything over slow speed toward the camera. I'd say my hit rate with those types of shots is pretty low, maybe under 60% depending on the DOF. :-[ I could nail focus with any of the AF points on the kiddie rides at f/2.8 with my 70-200 using One Shot with no problem, but shots of my older son on some of the faster rides moving toward me was iffy with either AI Servo or One Shot. I really didn't even try these shots much since I knew they wouldn't come out.

 

[Sporgon]

I also have to experiment with focus priority when doing tracking with the topmost point in portrait mode, no way around it really when tracking a horse running towards you... my problem might be a combination of dirty af array, non-cross point and bad 6d tracking behavior. Good to hear from some people without super powers, now I've much more to go on than before

I wasn't aware the horse was running toward you.... I don't have much success with those types of shots either! The 6D AF can handle side-to-side movement pretty well, but it definitely struggles with targets moving at anything over slow speed toward the camera. I'd say my hit rate with those types of shots is pretty low, maybe under 60% depending on the DOF. :-[

+1, if you want to track subjects coming towards you at speed successfully I suggest a trip down to your local professional camera shop with a thick wad of cash

 

[bholliman]

+1, if you want to track subjects coming towards you at speed successfully I suggest a trip down to your local professional camera shop with a thick wad of cash

Agreed! Time for a 1Dx!

 

[Marsu42]

@ Marsu42; thanks for your comments. Incidentally don't think that you are alone in struggling with focus on horses. Animal fur is notoriously difficult to focus on, and with horses in particular I have always had trouble focusing with every camera I have used.

Ah, right, that might be another reason why I find it that hard to work with the 6d af system while a lot of people seem to find everything's peachy for their holiday shots :-o

+1, if you want to track subjects coming towards you at speed successfully I suggest a trip down to your local professional camera shop with a thick wad of cash

Unfortunately way over my budget, and currently I find it much more rewarding to invest into lighting gear (tipods, softboxes, reflectors, flashes, ...) than into a camera that will be outdated again in no time.

Actually I also really like my good ol' 60d, did some great frog macros the other day @iso800, the 6d is really for thin dof and shooting in cloudy weather, the autumn and evenings. I admit I still have problems that a €1600 camera isn't "allround" capable but has (imho) deliberate design problems, other than horses running towards me it's mostly ok... but it worries me that I cannot trust the af system in critical circumstances.

Edit: If you really want to laugh from pain or sarcasm, try the 6d's top point on a *dog* speeding towards you - zero shots in focus