5D Mark IV focus problem.

[Redcrown]

I did a quick/dirty micro focus adjustment for my 70-200 lens on my new 5D4. Got a -10, which was the same as on my old 5D3, so I felt confident. Went on a walk about and made about 50 shots at various settings. Most of the shots looked a little soft to me.

So I spent the last 3 hrs doing a serious MFA. Custom target, all distances measured twice. Camera as level and parallel as I can make it. Fired about 6 shots at each MFA setting of 0, -5, -10, +5, +10, plus live view. I set the lens to minimum focus distance or infinity before each auto focus. Used only the center focus point, which should be the most accurate. Even made a couple manual focus shots in Liveview at 10x with a loupe on the LCD. Used a heavy tripod, so no camera shake. All shots at 1/400 shutter speed, so no mirror slap vibration.

Found a problem, a big one I think. While accuracy could be achieved, repeatability was poor. I determined that -10 was my best MFA setting (for the max zoom of 200mm). So I made 20 identical shots at that setting, throwing the lens out of focus before each shot. Ten of those were thru the viewfinder, 10 with Liveview. In each batch, only 4 or 5 shots were OK. Close enough to the best reference shot to say they matched. But 5 or 6 of the bacth were not OK. Far enough off the reference to call it a bad focus.

I've had the 70-200 for over 10 years, used on 4 cameras. Used on a 5D3 for the past 4 years with no problems other than operator error.

So, has anybody else done any extensive focus testing? Any similar problems with repeatability? Are you getting better than a 50% success rate?

 

[Blake Cook]

How do your best MFA settings photos compare to shots with Live-View AF? They should be equally sharp. the the same DOF.

 

[Sage]

I felt that focus on my new IDXM2 was little soft and inconsistent, so I sent it to Canon and asked to have the focus re-calibrated. They found and fixed an issue and now all is well.

 

[Bruce sawle]

I MFA all of my lenses and will continue to obseve there behavior. So far they seem to all work great. Very little needed on each lens. The lowest was -1 the highest amount was +7. The numbers are very similar to my MKIII maybe a click or two difference.

 

[J A C S]

How far was the target?

 

[whumber]

Are you seeing this problem with liveview as well or just through the OVF?

 

[penums]

Have you test with a shorter Shutter Time, 1/800s oder 1/1200s.

30MPixel show more camera-shake

 

[Damoo]

I have been comfortable with my Zeiss 135mm f2 and 70-200mm f4L IS. I could not lurk around with the f2.8. But today I picked a 70-200mm f2.8L IS II today & also the 5D Mark IV.

Tried few shots & haven't been able to find fault with the AF.

I will step out this evening and try some low light shots (lantern festival decoration in HongKong)

 

[larry55279]

If you are getting very OOF shots in liveview, it's the lens, not the camera. Lenses wear and fall out of adjustment.

It's *possible* that the AF module in the camera has a dust glob on it. You can *very carefully* lift the main mirror up a bit, and blow air on the AF module to clean.

I had to do this on my 1Ds, a long time ago.

 

[Redcrown]

OP here - mini update. Another walkabout, some more testing.

My inconsistent focus is prevalent only with multi-point focus modes. The more focus points used, the more inconsistent the focus. The worst is "Large Zone" focusing, where the camera picks the closest object. In my tests, that mode lights up between 6 and 18 focus points. Often, some of those points are clearly not part of the object. One or more point may fall on the distant background.

Using "Large Zone" focus mode, my failure rate was over 50% on 30+ different shots. I had assumed that all the points that light up were in-focus. Now I'm doubtful. I wonder if the logic is averaging the results of all points used? I remember reading an ecomomics book once that said the average human has one breast and one testicle.

Using "AF focus point expansion" with 8 points gave me about a 30% failure rate. By failure, I mean the camera claimed focus lock, but produced an image that was too soft to accept as in focus.

Using single center point focus, my success rate was +95%. The only failure was a focus on the white t-shirt of a guy standing in the sun. No detail in the shirt, very bright. Camera reported focus lock, but the image was way off focus.

Using Liveview focus, my success rate was also +90%. I don't see any reason for the few failures. Being forced to use DPP I did notice something interesting, if we can believe the red boxes that DPP draws around focus points. On the 5D4, the Liveview focus box is 216 pixels square. That's about 46k pixels. On the 5D3, the liveview focus box is 574 X 766 pixels. That's about 429k pixels, almost 10 times as big as the 5D4. I wonder what difference that might make?

Another note: My 70-200 lens came out with a -10 for the MFA at 200mm, but a zero MFA at 70mm. I discovered that DPP shows the actual MFA used in the EXIF. So I look at several shots between the 70 and 200 range. Looks like it does a simple linear interpolation. Not exactly linear, but close.

 

[clager]

OP here - mini update. Another walkabout, some more testing.

My inconsistent focus is prevalent only with multi-point focus modes. The more focus points used, the more inconsistent the focus. The worst is "Large Zone" focusing, where the camera picks the closest object. In my tests, that mode lights up between 6 and 18 focus points. Often, some of those points are clearly not part of the object. One or more point may fall on the distant background.

Using "Large Zone" focus mode, my failure rate was over 50% on 30+ different shots. I had assumed that all the points that light up were in-focus. Now I'm doubtful. I wonder if the logic is averaging the results of all points used? I remember reading an ecomomics book once that said the average human has one breast and one testicle.

Using "AF focus point expansion" with 8 points gave me about a 30% failure rate. By failure, I mean the camera claimed focus lock, but produced an image that was too soft to accept as in focus.

Using single center point focus, my success rate was +95%. The only failure was a focus on the white t-shirt of a guy standing in the sun. No detail in the shirt, very bright. Camera reported focus lock, but the image was way off focus.

Using Liveview focus, my success rate was also +90%. I don't see any reason for the few failures. Being forced to use DPP I did notice something interesting, if we can believe the red boxes that DPP draws around focus points. On the 5D4, the Liveview focus box is 216 pixels square. That's about 46k pixels. On the 5D3, the liveview focus box is 574 X 766 pixels. That's about 429k pixels, almost 10 times as big as the 5D4. I wonder what difference that might make?

Another note: My 70-200 lens came out with a -10 for the MFA at 200mm, but a zero MFA at 70mm. I discovered that DPP shows the actual MFA used in the EXIF. So I look at several shots between the 70 and 200 range. Looks like it does a simple linear interpolation. Not exactly linear, but close.

Not sure I agree with faulty lenses or whatever. One of the big selling points of this upgrade is the AF, focusing system. I would go back to the shop where you bought it and have them check it out.

 

[paulcroft]

>> Using "Large Zone" focus mode, my failure rate was over 50% on 30+ different shots. I had assumed that all the points that light up were in-focus. Now I'm doubtful. <<

Think about it for a second and you will realise that, unless there is something exceptional in your viewfinder, there will only be one 'object' in your shot that is in perfect focus because there is only one distance which is perfectly in focus at any given setting. What the red lights are showing in the large zone focus mode are those parts of the shot which will be in 'acceptable' focus for the given aperture.

The various focus modes all have their specific uses and none is perfect for all situations which is why there are several. About the only time I use the large zone focus mode is when photographing groups of people when it's impossible for everyone perfect focus.

If you want to test focus accuracy use only the central square and make sure that there is only one thing in the square which the camera can possibly choose.

 

[Redcrown]

For those willing... here is a Dropbox link to a layered tif file. It's a small crop, so only 12.5mb.

I made 8 shots of this scene with the 70-200 lens at 200mm (after -10 MFA was set). With each shot I manually defocused, focused, and fired. The first 4 shots were with single point center focus. The next 4 were with Large Zone focus. All shot at 1/640, f/8, ISO 1600, from a distance of about 50 feet. DOF should be well over 5 feet.

The 4 shots with single point center focus were identical and good focus. Of the 4 shots with large zone focus, 2 were OK. Not a perfect match to the single point focus shots, but close enough. But 2 other large zone shots were soft.

This layered tif shows one of the single point focus shots and the worst of the large zone focus point shots. In theory, some part of that large zone shot should be in focus. At least one of the multi focus points should be in focus? I can't find one. All points look out-of-focus to me, compared to the "good" focus shot.

The layer of red boxes shows the large zone focus points that DPP displays. The red X shows the single point focus.

https://dl.dropboxusercontent.com/u/62166185/5D4-focus-compare.tif

 

[Thestructured]

I got mine in the first shipment from B&H. After less than 100 photos it was clear that there was a focus issue. The left side, bottom cross type focus points were way off, and I noticed some poor repeatability like you did. I returned it.

I have another one coming tomorrow, hopefully without that issue.

 

[Jeremy Park]

I just bought a 5div and ran it through Focal AFMA software which gave really odd results... all L series prime lenses and getting consistently +8 's upto +15 ! The new 35mm L series lens which is pin sharp on my 5DSR gave me a +6... which had me worried. I then took the camera and did some real world testing and found the focus completely inconsistent regardless of my micro-adjustements or defaulting back to no AFMA. I'll be taking mine back tomorrow. Not overly happy since I have 2 jobs tomorrow I was hoping to use it on.

 

[RedFox88]

Have you tried zero mfa and see how that did ? On real life photos not a piece of paper 5 feet away

 

[Thestructured]

This is similar to the results I'm getting when using sigma primes, which work perfectly fine on our 6d's.

I'm really close to the return window on the 5d iv.

 

[Redcrown]

I'm really close to the return window on the 5d iv.

My 2nd 5D4 body is definitely focusing more consistently that the 1st one I returned. And this is on all 3 of my Canon lenses (70-200, 25-105, 100 macro).

However, there are still some slightly out-of-focus shots that make me suspicious of a problem other than operator error. My return window is also fast approaching, but I think I'll keep it.

I'm really enjoying the benefits of better image quality at higher ISOs. Better noise control is one thing, but I'm noticing much better retention of detail at those higher ISOs. One example is the "eyelash" test on a close up portrait. With my 5D3, fine eyelashes would start to fade away at ISO 800. At ISO 1600 they were pretty bad. With the 5D4, I'm finding eyelashes still acceptable at ISO 3200!

I'm also seeing a big difference with the 5D4 when I raise shadows. With the 5D3, when I raise shadows in ACR/LR, noise comes on quickly. Adding noise reduction to counter that, and detail goes all muddy. With the 5D4, I'm getting much less noise and thus better detail retention.

 

[larry55279]

My 2nd 5D4 body is definitely focusing more consistently that the 1st one I returned. And this is on all 3 of my Canon lenses (70-200, 25-105, 100 macro).

However, there are still some slightly out-of-focus shots that make me suspicious of a problem other than operator error. My return window is also fast approaching, but I think I'll keep it.

Rationally, it's far more likely that it is partially user-error, and partially an unfounded expectation of perfection in AF. Prepare to be constantly unhappy in your digital photography endeavors.

 

[Redcrown]

Rationally, it's far more likely that it is partially user-error, and partially an unfounded expectation of perfection in AF. Prepare to be constantly unhappy in your digital photography endeavors.

Interesting point of view. I think my "expectation of perfection" is well founded. I found it in previous cameras and I found it in the majority of shots by the 5D4 that are "perfect". I just want a higher hit ratio. At least equal to older technology, hopefully better.

The possibility of user-error always haunts testing. I try to carefully analyze each shot to see if there was possible user error. I've found that too, and restructured and repeated tests to avoid it. Still get some mysterious off focus.

I'm prepared to be unhappy... but only some of the time and certainly not constantly. Doing something that makes you constantly unhappy would be a sign of mental illness, and my shrink says I'm OK.