Followup: 25mm is FRONT FOCUSING on G7 (examples)

Big D in SP wrote:

The 14-42II kit zoom, 14-140, and 60mm Sigma all focus perfectly. That would seem to indicate a lens problem, but it puzzles me. Contrast Detect focusing (theoretically) should eliminate this kind of focus inaccuracy. Since it's happening wide-open as well as stopped down it shouldn't be attributed to focus shift upon stopping down, should it?

Warren, I used to think the same thing, but a few months ago I got the Olympus 14-42mm EZ pancake lens and discovered a similar problem.  In my case all focal lengths were fine except for at 42mm.  At 42mm the lens was horrible, but at other focal lengths it was just fine.  The focus at 42mm seemed to be way off.  I sent it to Olympus Japan and they sent it back fixed (took 2 tries).  They didn't give details of what they did, but I suspect there is calibration data in the lens memory and that got updated by them.  I can imagine that probably at the service centers they have a device that they can mount an m4/3 lens to and it can run though some automatic tests of the lens to do calibration.  No doubt there is a mode to do a manual override to allow the tech to do more testing.  I had to send mine in twice.  The first time they returned it they made it much, much worse.  It was horrible at all focal lengths!  I sent it back and this time when it came back it was good at all focal lengths.  I suspect that the second time a more experienced tech took a look at it and fine tuned the calibration.  Maybe the first time automatic calibration really screwed things up.

Again, I don't know that my suspicion is correct, but it is a reasonable suspicion, I think.

Henry Richardson wrote:

Big D in SP wrote:

The 14-42II kit zoom, 14-140, and 60mm Sigma all focus perfectly. That would seem to indicate a lens problem, but it puzzles me. Contrast Detect focusing (theoretically) should eliminate this kind of focus inaccuracy. Since it's happening wide-open as well as stopped down it shouldn't be attributed to focus shift upon stopping down, should it?

Warren, I used to think the same thing, but a few months ago I got the Olympus 14-42mm EZ pancake lens and discovered a similar problem. In my case all focal lengths were fine except for at 42mm. At 42mm the lens was horrible, but at other focal lengths it was just fine. The focus at 42mm seemed to be way off. I sent it to Olympus Japan and they sent it back fixed (took 2 tries). They didn't give details of what they did, but I suspect there is calibration data in the lens memory and that got updated by them. I can imagine that probably at the service centers they have a device that they can mount an m4/3 lens to and it can run though some automatic tests of the lens to do calibration. No doubt there is a mode to do a manual override to allow the tech to do more testing. I had to send mine in twice. The first time they returned it they made it much, much worse. It was horrible at all focal lengths! I sent it back and this time when it came back it was good at all focal lengths. I suspect that the second time a more experienced tech took a look at it and fine tuned the calibration. Maybe the first time automatic calibration really screwed things up.

Again, I don't know that my suspicion is correct, but it is a reasonable suspicion, I think.

Henry,

Thank you for the info.  I will certainly pursue this.

Warren

Hi Warren, Testing for focus accuracy can present some  technical difficulties. Appropriate technique is required for meaningful results. Here are some thoughts, observations and suggestions

* You do not say what AF area was used. With the setup which you presented you should use Pinpoint. Anything larger just causes the lens to focus on the closest part of the target under the edge of the AF box, which appears to have happened here.

* The f1.7 lens has less depth of focus than the f2.8 lens. Other than that I see no significant difference between them.

* For a proper AF test you should set up a series of non overlapping vertical targets like headstones at a cemetery or packs of cards on edge each with the flat face towards the camera each at a different distance and make sure the AF box is definitely smaller than each target.

For a quick cheap and cheerful test you can lay a page of newspaper on the floor. Draw a line across the middle. Focus using pinpoint on the line with the lens axis at about 45 degrees to the plane of the floor.

* Panasonic M43 cameras are reknowned for single AF accuracy and hardly ever misfocus given a decent target on which the AF system can lock.  Your setup does not provide such a decent target.

Good luck

Andrew

Astrotripper wrote:

Bhima78 wrote:

Its the lens. I had the same problem with 3 different copies of the PL 15mm on my E-M10 about a year ago. I have since gotten the PL15 again for my work camera (GX8) and have also put it through its paces on my E-M10. My new PL15 does not exhibit this issue at all. I even did 45-degree A/B testing like you did. What I think is going on:

Our contrast detect lenses open to the widest aperture to find focus, then stop down to the desired aperture once focus is achieved and then the shot is taken. Something with the electronics is telling the camera that yes, the lens has found focus (prematurely), so stop down (if you aren't shooting wide open), and fire.

Contrast detect SHOULD prevent this from happening.  However, all the camera system is doing is deciding whether something in the depth of field focus zone is in focus or not.  Practically, that means that the point might be in the very front of the zone, the very back of the zone, or the middle of the zone - all of those locations would read to the camera AF system as "in focus."  Focus peaking has the same issue - I've got images from a recent concert where focus peaking told me the eyes were in focus, and they were... but were at the very front of the focus zone, so the images were un-useably soft across the performers' faces.

This is why, when shooting situations where precision control of DOF is important (think portraits, close in, with an f1.2 or f1.4 lens), you really should either manual focus, or autofocus and then manually adjust.

I've seen this many, many times over the years with all kinds of cameras.  Phase detect focus systems have the same issue - they detect if a target zone is in focus, NOT whether it's in the middle of the DOF focus zone.  That's why you have the setting that allows you to AF and then manually tweak.  It's also why many of my professional photographer friends are switching away from DSLRs from Canon and Nikon, to mirrorless systems.  (Four of them in the last three months.)  The focus screens on DSLRs are NOT optimized for manual focusing - what they show you is effectively f2.8 (at best) to f4 (most frequently), so trying to manual focus an f1.4 lens is a crap shoot.  I know people will tell you that they can but if they do, they're not shooting enough in critical situations to know that it's impossible to do accurately.  You can't beat design parameters.

Used to be you could change out focus screens for ones that showed with more precision - the tradeoff was usually that they went dark and unusable after a certain focal length and max aperture, but the last high pixel count Canon that could do that was made several years ago.  The last Nikon that could do that was the D2X.  This is what's now causing a number of pros who love wide aperture shooting to switch, usually to Sony (full frame so you get the max benefit of a wide aperture lens, high pixel count since they print large).

However... if a lens is de-centered, it WILL produce results like people are talking about here.  That's something that sending it in for repair should fix.  I've been told many times by Panasonic, Olympus, and Fuji reps at sessions they've held, that there's no focus tuning adjustment capability on contrast detect cameras and lenses.  They're not the engineers so maybe they're wrong, but multiple people, multiple brands, multiple times, makes me believe them until I find out otherwise.

Stopping down might be an issue only when the lens suffers from serious spherical aberration and thus closing down the aperture causes focus plane to shift. I doubt any native MFT lens exhibits a problem like that (but I might be wrong).

Yes, many of the MFT lenses do have spherical aberration issues. I can see focus shift in the 42.5 f1.2 - although the depth of field increases usually cover that up reasonably well.  But that's at the cost of precision of placement of the DOF.

One thing people miss is that m4/3 cameras (maybe others, but I know it for certain on m4/3 because I've looked through the lens from the front) do stop down when focusing.  I don't know why, but the aperture does jitter down and back with AF engaged.  You get people posting in forums about "my lens is making noise when I'm on constant focus" all the time.  I suspect it's for dealing with focus shift.

As for contrast detect systems, you can get an idea how it works if you have a slow focusing lens. Take a note of how focus changes back and forth, but pay special attention to the final stage. You will notice that the final change into focused position is immediate. There is no further confirmation, there's no fine tuning at the last moment. It goes from well out of focus straight to focused in a single quick move and it's done. If autofocus motor is faulty in some way, or if another kind of issue prevents it from working properly and consistently, you will get focus errors because that final move does not lead to a result that the camera expects.

Remember that the final move is to being in focus - NOT to centering the DOF zone around the point being focused on, because it's impossible to determine that algorithmically and still be fast.  Hence, it's very, very common to get images that look front focused or back focused.  I'm in a long email conversation with someone who's a well known DSLR expert (sells books, popular web site, frequent conference speaker) on this.  He maintains that contrast detect focus isn't a panacea for flawless AF on wide aperture lenses even when manually focusing.  I used to disagree until my recent trip where I shot a couple thousand images with the 42.5 f1.2 and 75mm f1.8.  The number of images where the DOF zone is just slightly off is massively annoying, auto focused, or manually focused.  I need to learn to work with and tolerate magnified focus assist windows popping up.

Contrast detect is not a magic bullet that never misses. I mean, you could make it work in a way that it would never ever miss, even with faulty lenses. But we all want AF to be fast, so we have to live with the compromises that make it possible.

Placing thin DOF zones has always been an issue, even in manual focus days.  The increased reliance on AF makes it seem worse, because now it's "happening to us",  but I remember going through image review in classes 10 years ago and taking a beating from the instructor for not managing placement of the in-focus zone.

It does it in two ways.

1) Camera has anti-flicker enabled and aperture is used to control light all the time

2) using a automatic Exposure mode, no matter what, and aperture is used to adjust Exposure for focus and live view.

Use manual mode (M) and aperture will not be closed down in any focusing mode no matter how bright subject.

The aperture doesn't get changed even when using wide open or closed down ratios. Only when the image is taken the aperture closes down.

Big D you should do the exact same test with your ruler coming from right to left, and also flat on the floor (= bottom to top) to check whether your lens is actually decentered which could be giving this fault from left to right but a different fault in other cases.

Also do the discrete object tests as has been mentioned, 'gravestones' type.

axlotl wrote:

Hi Warren, Testing for focus accuracy can present some technical difficulties. Appropriate technique is required for meaningful results. Here are some thoughts, observations and suggestions

* You do not say what AF area was used. With the setup which you presented you should use Pinpoint. Anything larger just causes the lens to focus on the closest part of the target under the edge of the AF box, which appears to have happened here.

Smallest focus box.  Shot approximately 100 shots with 5 lenses.  The 25mm front focused, the others nailed focus.  I only posted a few shots to show the difference.

* The f1.7 lens has less depth of focus than the f2.8 lens. Other than that I see no significant difference between them.

* For a proper AF test you should set up a series of non overlapping vertical targets like headstones at a cemetery or packs of cards on edge each with the flat face towards the camera each at a different distance and make sure the AF box is definitely smaller than each target.

For a quick cheap and cheerful test you can lay a page of newspaper on the floor. Draw a line across the middle. Focus using pinpoint on the line with the lens axis at about 45 degrees to the plane of the floor.

* Panasonic M43 cameras are reknowned for single AF accuracy and hardly ever misfocus given a decent target on which the AF system can lock. Your setup does not provide such a decent target.

Good luck

Andrew

Today I did a different test with two cameras: The new G7 and one of my G6's.  I set my camera on a tripod at 48" from the target with the 25mm lens.  I focused on the target, locked focus and then carefully advanced the tripod a tiny amount forward, taking about 15 shots in succession.

Then, I examined them closely at 100% to determine which were the sharpest.  Those which were taken approximately 3-4 inches forward of the first shot at f1.7 were the sharpest.  Those at f3.2 were sharpest at approximately 5-6 inches forward of the first shot.  Other lenses tested were sharp from the original distance.  The results were the same with both cameras, so, it appears that the lens is definitely the problem.  I'll be contacting Panasonic for warranty because with this test I now know how sharp the lens can be when proper focus is achieved!

A few cropped shots follow to illustrate the difference.

Thanks to all for your comments and suggestions.

G7 25mm @ f1.7 original distance

G7 25mm f1.7 tripod moved forward @ 3-4 inches

G7 25mm f3.2 Original distance

G7 25mm f3.2 Tripod moved forward @ 5-6 inches

G6 25mm f3.2 Tripod moved forward after focus locked on first shot @ 5.6 inches

Regards to all,

Warren

I might add this comment:  The differences are considerably more apparent with original files than these cropped, compressed examples.  But, looking closely one will see the differences.

Warren

JConrad wrote:

My 25/1.7 also front focuses, most notably at infinity. Closer up it seems fine (though it has more problems on the GX8 than on my E-M10 II). Near infinity, it almost always front focuses a fair bit, causing softness even stopped down.

Landscape shots were one of the first to suggest to me that the softness of indoor shots was something more than poor resolution and why I tested the focusing accuracy.

It's interesting that the lens seemingly focuses better on the EM-10 II than the GX8.  Why, I wonder, would that be?

Warren

robjan245 wrote:

I had the same issue with my G7 with the 14-140 lens. I ended up returning the G7 for the Sony A6000.

So you prefer poor lowlight AF and 1080p video? Ok.