PDA

View Full Version : AF problem in Live View - Canon 7D Mk II



nardes
02-01-2018, 3:59pm
The problem
A whole series of frog frames with the AF Region centred on the eye have the eye out-of-focus.

Background
I noticed an AF problem (on a night-time shoot) of a series of shots of the small Eastern Dwarf Tree Frog (Eastern Sedge Frog (Litoria fallax)). The AF was consistently missing the focus on the eye of the creature despite the AF Point being placed on the eye of these static frogs. The frogs are typically 20-25 mm long.

I was in Live View Mode with the AF Region placed on the frog’s eye and every shot (40-50) had the eye noticeably OOF, with nil shots of the eye in focus. Other parts of the frog are in sharp focus, so this is unlikely to be a “subject movement” issue, especially since I was using flash as the main light source.

Testing
To check the AF, I later took a couple of frames of the Canon User Manual on my PC Screen, at roughly the same subject to lens distance as the frogs and the focus was sharp. The image without the Red AF Rectangle is a Raw file crop, whereas the image displaying the Red AF Rectangle is from a screen capture in Canon’s DPP V4.

Settings
For the frog and test images, the Live View AF Method was “FlexiZone – Single” with “AF-Continuous” and “AF Mode – One Shot AF”

As you can see, other body parts in the same plane as the near-side eye are nicely in focus – does anyone have an explanation for all frames showing an OOF eye?

Possible cause?
These frogs inflate/deflate their throat sac generating a strikingly loud noise. I suspect the eye moves during this inflation/deflation process but was surprised that not one shot had the eye in focus.

Thanks

Dennis

Full res crop in Canon's DPP
134091

Shot of screen in Live View
134092

Canon's DPP showing AF Region
134093

ameerat42
02-01-2018, 4:43pm
Strange, for sure!

Q1: That larger, red square corresponds in size to the smaller white square in the next picture?
Q2: Could you make that zone smaller, for more precise spot focusing?
Q3: There is another, unexplained red square in the 3rd picture, along with the smaller white one.

More tests: Try it again on an inanimate object of similar size and depth.
Wondering: In the flat screen test, could some other AF point/zone be determining the focus?

nardes
02-01-2018, 5:59pm
Strange, for sure!

Q1: That larger, red square corresponds in size to the smaller white square in the next picture?
Q2: Could you make that zone smaller, for more precise spot focusing?
Q3: There is another, unexplained red square in the 3rd picture, along with the smaller white one.

More tests: Try it again on an inanimate object of similar size and depth.
Wondering: In the flat screen test, could some other AF point/zone be determining the focus?

Hi Am

Thanks for your reply.

Q1-Answer: Yes, in #1 the Red Rectangle is how Canon's DPP V4 represents the Live View AF rectangle on the full size image. The frog image is a screen capture of part of the full size (100%) image.
Q2-Answer: Not in Live View - you only get the choice to spot focus & small spot focus in normal TTL mode when looking through the Viewfinder.
Q3-Answer: The small White Rectangle is the illustration in the Canon User Manual of where the AF Rectangle is. The Red Rectangle in #3 is where Canon's DPP V4 is telling me the Live View AF rectangle was on image of my computer screen which was displaying the pdf of the Canon User Manual.

Cheers

Dennis

arthurking83
03-01-2018, 8:40am
Can you do another test?

Using the same settings/gear as before, can you setup something more three dimensional as the test target for the test reference shot.
Something like a ball, or stuffed toy or whatever, but basically the frog subject has more depth as a focus target, whereas the manual on the PC screen is a flat plane target.

In Lv mode, there is little to no chance that backfocus is the issue, and looking at the images of the manual on screen, this appears to be confirmation that your sensor alignment isn't the issue?
In Lv mode there is 'no such thing' as back or front focusing, unless the sensor is misaligned somehow.
back/front focusing can be an issue with the traditional DLSR method of relayed PDAF system.

The reason for the request of a new test shot to create a reference point is:

If you understand how the AF system actually works, what it is, is that the red square is only a guide as to where the focus system actually is.
What the focus system actually is, is not a square(red or otherwise) but a series of 'lines', on a 7D MkII(as I understand it) most if not all at cross type lines.
(even lines isn't the most accurate way to describe it, but for the purpose of this reply 'lines' will do).

My thinking above tho relates to these lines, and how they're aligned within the red squares.
The lines can be of three major variations of types. The most common is the standard vertical line, which could be described as |.
Then the cross point types, that just looks like a +.
And the best type currently used by some manufacturers are the double cross lines that look something like #
What is supposed to be the situation is, that all those lines should be contained within the red square area. The red squares themselves are just a laser projection on the viewfinder screen(or LCD) and don't really exist in the focusing system.
What does exist is the phase detection micro prism system, which has an 'alignment' .. that alignment is what we call the 'lines' referred too above.

Most of these high end DSLRs, have the double cross phase detect(PD) system at the central focus point.
Some cameras, eg. your 7DII probably has 60 af points of which the vast majority are of the single cross point type.
And the very outer periphery af areas usually contain those vertical PD types.

Briefly, the way those AF lines work is just like the good ol days when SLRs had split prisms and microprism focusing aids. The typical split prism tho was horizontally orientated .. like this - . So that if you were to try to focus that split prism on a subject that also had a horizontal orientation, it was nigh on impossible to focus it accurately as there was no difference in alignment on the subject. So the best subject for a horizontal split prism, was a vertically aligned subject .. ie. |
Same is true for the vertical AF point types.. if the subject was vertically aligned, the vertical PD system can't differentiate any alignment issue and hence can't really focus correctly.

The cross points are the bees knees, and the double cross points are the bees nu! ... well you know the other more prized parts of the bees! :p
Those two cross types don't care about subject orientation, as there is sure to be something to differentiate in the subject to allow it to align the subject.

With all that boring stuff as background info, the underlying question of which focus point you used shouldn't come into the issue.
Why?
you have a round subject(the frog eye) which has more than enough contrast area on both the upper and lower arc of the circle of the eye to allow the vertical PD point to focus.
That is(if you can try to imagine) even with the focus point setup looking like this [|] .... where the squared edges are the red square af indicator, and the vertical line is the actual PD system, as long as the round eye is in the red square, that vertical PD point will see the upper and lower arc of the eye and align those contrasting points to not be broken up(as in the old split prism days). That's how PD works .. like a range finder in a manner of speaking. The contrast along the edge is assessed, and if the edge is broken, it's adjusted to not be broken. simples!

But with all that, we assume that the PD points(both single and double cross, and the vertical types) are aligned within the red AF square.
The actual chances of them not being aligned perfectly is actually quite high.
That is, my earlier little visual of the vertical line ... [|] may not actually be the vertical point within the square, but the vertical point could be anywhere near the red square .. could even look like this .. []|
The problem can be that no one really knows, and other than a careful rebuild by Canon .. remembering that the camera comes off a production lines! .. it's almost impossible to tell.
The issue is that the distances/clearances involved in the lining up of the system are measured in microns .. to keep the AF system compact.

So as an example of how this could affect the images the way it did, lets say you used a slightly mid AF point that does indeed use the + cross point, so we assume that the red af indicator would be aligned like this [+].
But if the actual physical alignment of the + PD point was slightly to the right .. ie. []+, then this would explain why the actual focus was on the frogs back.

That explanation is one of the most unlikely, but still possible reasons, and all the image would need to be focused in exactly the same way ... ie, focus appearing to be on the frogs back in all images AND definitely zero movement of the frog on whatever thing it's resting on.
That is, frog looks like it's on a frond or something. was it 100% certain that the frond didn't have a very low frequency fore/aft movement to bring the frog into and out of focus?

The more 3d subject AF test I referred too above, where the subject is 101% static on whatever solid surface you can use and tripod and all that usual solid immovable jazz as the base setup for the camera/lens .. will help to determine if your AF point to red indicator is misaligned in any way.
No point using a flat field subject as you won't see the depth of any focus errors. so the subject needs to have a small size, some depth and small details for you to 'focus on'. .. as a rough example I reckon a golf ball could work.
The focus subject won't be the ball, or the actual shape of whatever you use, it'll be some very fine small detail on the overall subject.

And the tests would be with that very fine detail in the red AF square, and just slightly out of the red AF square.
From that frog image, I'd say if you try the test, then you'd want to concentrate on how it focuses with the red square just to the left of the fine AF test subject.

ps. I've written up an reply in another thread about the same issues, with actual made up diagrams on the af point-indicator alignment system. If my description is a bit vague, I'll try to locate that thread and link to it.

pps. The most common cause of AF alignment issues is due to knocks to the camera body.

nardes
03-01-2018, 10:37am
Can you do another test?

Using the same settings/gear as before, can you setup something more three dimensional as the test target for the test reference shot.
Something like a ball, or stuffed toy or whatever, but basically the frog subject has more depth as a focus target, whereas the manual on the PC screen is a flat plane target.

In Lv mode, there is little to no chance that backfocus is the issue, and looking at the images of the manual on screen, this appears to be confirmation that your sensor alignment isn't the issue?
In Lv mode there is 'no such thing' as back or front focusing, unless the sensor is misaligned somehow.
back/front focusing can be an issue with the traditional DLSR method of relayed PDAF system.

>snip


Hi Arthur

Thank you for your comprehensive analysis and suggestions for further 3D testing - something that I had also planned to do from Am's earlier response.:):th3:

This is a quick reply to let you know the reason I used the Live View Mode is that my initial Viewfinder TTL shots (using Spot AF and Small Spot AF) with the AF Region placed on the eye of the frog (the AF Area just overlapped the periphery of the eye) also produced OOF eyes, so on the fly, I judged this to be Back Focusing hence the reason for my switching to Live View.

Will respond with more details after I have digested your full reply and will provide further test results when I have found a similar sized 3D object to practice on.:)

Cheers

Dennis

- - - Updated - - -

Based on Arthur's reply, I'll add my findings to this post as I come across them rather than in one "big bang".

So, here is a description of the Canon EOS 7D Mk II AF System from the Canon User Manual.

It looks like all 65 AF Points are the "Cross Type" for F5.6 lenses, with the Centre AF also Diagonal for F2.8 lenses.

The 100-400 Mk I is shown as a "Group B" lens so all 65 "Cross Type" AF Points can be used and I guess it will be the same for the 100-400 Mk II.

Cheers

Dennis


EOS 7D Mk II AF Points
134103

Group B should all work with 100-400 lens.
134104

nardes
04-01-2018, 9:49pm
UPDATE: 4/01/2018

Went out last night and had another go using various Focusing Modes (AF via TTL, AF in Live View and Manual Focusing in LV) and AF Areas (Centre Point, Spot and then Zone in LV). IS off in all cases.


With auto-focus set to AF One Point or AF Spot, and the AF region placed on the eye of the frog (as shown in Canon’s DPPV4) the eye was consistently out of focus. I used AI Servo (Canon’s “follow focus” Mode) to allow for the frog’s eye moving.
Next, using Live View (One Shot AF, Zone Focusing, Continuous Focus) the eye was consistently out of focus.
Finally, using Live View with the lens set to Manual Focus, I focused on the mouth of the frog at x10 on the rear LCD and each shot was in focus at the mouth, provided the frog did not move.


So, when using AF through the Viewfinder (TTL) with the AF Point placed over the eye, it looks like the AF is selecting a focus feature other than the frog’s eye.

Why this also occurs with AF in Live View Mode is puzzling, as I can successfully manually focus in Live View Mode.

The set up this time was the 100-400mm Mk II (at 400mm) with Extender EF x1.4 Mk III and a 12mm Extension Tube to allow for closer focusing.

I’ll run some further tests at home over the next day or so.:confused013

Cheers

Dennis

arthurking83
05-01-2018, 10:32pm
....

So, when using AF through the Viewfinder (TTL) with the AF Point placed over the eye, it looks like the AF is selecting a focus feature other than the frog’s eye.

Why this also occurs with AF in Live View Mode is puzzling, as I can successfully manually focus in Live View Mode.

....

About the only possible explanation is that the camera lens are miscommunicating somehow and the lens is being told to focus incorrectly.
You'd think that in Lv mode this couldn't happen tho as the lens would stop at a point where contrast at the focus point is best.

:confused013

I'd test another lens to be sure.

nardes
06-01-2018, 8:31am
About the only possible explanation is that the camera lens are miscommunicating somehow and the lens is being told to focus incorrectly.
You'd think that in Lv mode this couldn't happen tho as the lens would stop at a point where contrast at the focus point is best.

:confused013

I'd test another lens to be sure.

Thanks again Arthur, I really appreciate your insights and help in trying to understand what is going on.:th3::)

I think you may be right with it being "lens based" as the same issues arose on a different body - my Canon 5D Mk IV.:confused013

I also fitted a Sigma 50mm F1.4 on the 5D Mk IV and the AF was perfect (daylight) on the eye of an Eastern Water Dragon (see attached image).

I will do some still life tests at home to see if anything else pops out of the data.

I am left wondering if it may be a close-focus issue on the 100-400, as the AF is perfect at the 400mm end where I do 99% of my shots, or simply that the AF process is being fooled by the amount of low contrast green tones in the frog image and so finds something else of higher contrast, albeit not in the same plane as the nearest part of the anatomy that DPP V4 indicates in the AF region.:confused013

Cheers

Dennis

Mark L
10-01-2018, 9:55pm
Haven't read all of the above and without getting to technical maybe the frog moved it's head a fraction at precisely the wrong time?
Nah.

nardes
10-01-2018, 10:02pm
Haven't read all of the above and without getting to technical maybe the frog moved it's head a fraction at precisely the wrong time?
Nah.

Not in all the 40-50 shots...I'd expect at least one lucky in-focus shot if that were the case.:confused013

nardes
16-01-2018, 6:52pm
Continuing with the investigation, I took some photos of an Eastern Water Dragon’s eye with the 100-400 Mk II and with the 100-400 Mk II PLUS Canon Extender EF x1.4 Mk III. Based on the 100% resolution crops below, I was pleasantly surprised to find that with the AF point on the eye of the EWD, I could not see any evidence of front or back focusing.:)


#1: Eye of an Eastern Water Dragon taken with the 100-400 Mk II at the 400mm end with no extender attached.
#2: Same EWD taken with 100-400 Mk II and the 400mm end PLUS Canon Extender EF x1.4 Mk III fitted.


This came as a bit of a relief as I didn’t really want to get into manually adjusting the AF to compensate for any front/back focusing.:)

The EWD was some 1.5 to 2 metres away and did not move its position whilst I fitted the Extender x1.4 Mk III and recomposed. The main environmental differences were that the original OOF scenario was at night time, with pre-flash illuminating the frog and a green frog on green leaf vs. daytime EWD with on-camera Speedlite with plenty of contrasty texture to focus on.

My next step is to re-visit the frog pond at night and try again.

Cheers

Dennis

#1: 400mm
134282

#2: 560mm
134283

PS - ignore EXIF of 560mm on 400mm file as it was a Layer in the 560mm file.

swifty
16-01-2018, 7:40pm
This might be a stupid question from me but is there a discrepancy between how close you can focus in LV AF and MF?
Is it possible that you can focus closer using MF (the opposite to how you can focus past infinity) and you were already at your closest AF focusing distance when you tried it with LV AF?

nardes
16-01-2018, 8:38pm
This might be a stupid question from me but is there a discrepancy between how close you can focus in LV AF and MF?
Is it possible that you can focus closer using MF (the opposite to how you can focus past infinity) and you were already at your closest AF focusing distance when you tried it with LV AF?

Hi Swifty

Thanks for your reply, I appreciate it.:)

When I tried all the 3 AF methods, (Viewfinder AF, Live View AF and MF) I made sure that I was at a greater distance from the subject than the lens' minimum focus distance. I did this by physically moving myself (closer) to the point where I could not achieve focus via any of the methods and then I stepped back so there was room for the focus to breathe as it were. I think that closest focus is around the 1 metre mark for this lens and generally, I was hovering around 1.5+ metres. To test this even further, I fitted a Canon 12mm Extension Tube which allows the lens to focus even closer (and prevents infinity focus as a result) but kept approx. 1.5 metres from the subject, so I think that the close focus point is unlikely to be the candidate here.

Thanks again!:th3:

Cheers

Dennis