p.12 #1 · ['NEW Fix' UPDATE!] - MkIII AF still broken
jkurkjia wrote:
Off the top of my head I can’t envision what aspect of IS might contribute to what we see (actually, what we don’t see) from a mechanical standpoint. It’s almost like one side of the IS optics is pivoting around one point and either vibrating on the other side OR just plane (no pun intended) cocked at a bad angle. I wonder if there is some Canon prepared document that describes IS from both an optical and mechanical standpoint?
Lenses are designed to poduce a flat image. i.e. the image focuses onto a flat plane rather than a curved shape. Turning that flat plane away from the correct alignment - if you could do it- keeps it flat but at the wrong angle so that part of it moves behind the image sensor and part of it move in front and part remains on the sensor. This is basically what the IS does.
The IS operates by moving a lens group off-axis far enough and fast enough to redirect the light that is coming from a now-off-axis viewpoint back to where it was on the sensor. That way we keep the old view on the sensor even though the lens is actually pointing at a slightly different view. This assumes that the lens motion is mostly angular rather than purely "sideways". i.e. the camera and lens face a new direction and the sensor initially sees a different scene from the new angle of view. The IS corrects that by restoring the original scene, within the mechanical limitations of the IS group to respond.
If the IS group has incorrectly moved to turn that scene when there has been no corresponding turning of the camera then in effect you have a sort of tilt-lens effect that can shift the plane of focus so that it is no longer parallel with as well as coincident with the sensor plane. It will now be at an angle to but probably cutting through the sensor plane. The IS group has not changed the focus distance, but just turned the focused image plane off-axis.
In the early days of IS it was a plastic bag full of fluid held between two flat pieces of glass - one fixed and the other twisting without actually moving off-axis - so that it was in effect a variable prism lens ("vari-prism"). Nowadays it is more like a double-concave lens that moves sideways off-axis without twisting. The middle bit of the lens has less optical turning effect than the outer bit because it is more like a flat piece of glass. The farther it moves the more significant the optical turning effect becomes due to the shape of the lens.
p.12 #5 · ['NEW Fix' UPDATE!] - MkIII AF still broken
Alan321 wrote:
Lenses are designed to poduce a flat image. i.e. the image focuses onto a flat plane rather than a curved shape. Turning that flat plane away from the correct alignment - if you could do it- keeps it flat but at the wrong angle so that part of it moves behind the image sensor and part of it move in front and part remains on the sensor. This is basically what the IS does.
The IS operates by moving a lens group off-axis far enough and fast enough to redirect the light that is coming from a now-off-axis viewpoint back to where it was on the sensor. That way we keep the old view on the sensor even though the lens is actually pointing at a slightly different view. This assumes that the lens motion is mostly angular rather than purely "sideways". i.e. the camera and lens face a new direction and the sensor initially sees a different scene from the new angle of view. The IS corrects that by restoring the original scene, within the mechanical limitations of the IS group to respond.
If the IS group has incorrectly moved to turn that scene when there has been no corresponding turning of the camera then in effect you have a sort of tilt-lens effect that can shift the plane of focus so that it is no longer parallel with as well as coincident with the sensor plane. It will now be at an angle to but probably cutting through the sensor plane. The IS group has not changed the focus distance, but just turned the focused image plane off-axis.
In the early days of IS it was a plastic bag full of fluid held between two flat pieces of glass - one fixed and the other twisting without actually moving off-axis - so that it was in effect a variable prism lens ("vari-prism"). Nowadays it is more like a double-concave lens that moves sideways off-axis without twisting. The middle bit of the lens has less optical turning effect than the outer bit because it is more like a flat piece of glass. The farther it moves the more significant the optical turning effect becomes due to the shape of the lens.
Alan, thanks for that lucid description. I "finally" successfully downloaded the patent and hopefully will be able to review it tomorrow. Again, thanks much and I appreciate your taking the time to pull together a comprehensive reply.
p.12 #6 · ['NEW Fix' UPDATE!] - MkIII AF still broken
Mark Powell wrote:
Jeff, thanks for hanging in there with maturity. I look forward to reading this thread to its conclusion.
I will 2nd that, with maturity indeed! It is a waste that someone with his experience 'lost it' but hey, self control is indeed a virtue.
Jeff, I am going to take some more test shots over the next day or 2 since you are camera 'less'. You are obviously far more of a Photoshop guru than me - no idea how to get multiple loupes on the same image!
p.12 #7 · ['NEW Fix' UPDATE!] - MkIII AF still broken
It's actually a post-processing gyration. Take extra screengrabs as you move the loupe around, then just copy/paste them onto the main image (re-positioning them accurately can be tricky), flatten and re-save. Voila!
p.12 #8 · ['NEW Fix' UPDATE!] - MkIII AF still broken
Jeff wrote:
The focus was on the tree in the center, and both sides of the rock were roughly equidistant to my position. Even if what you suggest were true, you should be able to move 'backward' on the right hand pillar of rock until you find the focal plane, but you cannot. There is absolutely nothing in focus in the right half of the frame, but there should be. At that distance to subject, the left side of the frame exhibits the depth of field one would expect with this lens and f/stop combination; the right is inexplicable for normally-operating equipment, in my opinion. ...Show more →
This has been noticed many times on the 16-35/2.8
One side of the image is just plain blurry and looks out of focus, while the other side is fine.
Sounds like a lens alignment issue.
That doesn't explain why it would be fine on a different camera, though.
p.12 #10 · ['NEW Fix' UPDATE!] - MkIII AF still broken
Alistair Watson wrote:
Or perhaps a sensor alignment issue, in which case this behaviour would be noticed on all lenses.
Disclaimer: I don't own a MKIII, but am interested in the cause of these focus issues.
As I originally stated, it could possibly be an issue with a loose anti-dust assembly (or AA filter if they share mechanism) that causes it to shift or tilt just enough to cause mis-alignment or cause slight defocus type symptoms.
Maybe the reason it only shows up with longer lens is due to vibration harmonics based on weight and overall camera balance when mated to different sized lenses. I don't believe it's related to a focus problem with the camera or lens continuing to focus during image capture. If that was the case the whole image would show the same amount of de-focusing (?) problem.
Hey, no one else has come up with a valid and verifiable explanation, so who's to say this is any less relevant?
p.12 #11 · ['NEW Fix' UPDATE!] - MkIII AF still broken
Alan321 wrote:
Lenses are designed to poduce a flat image. i.e. the image focuses onto a flat plane rather than a curved shape. Turning that flat plane away from the correct alignment - if you could do it- keeps it flat but at the wrong angle so that part of it moves behind the image sensor and part of it move in front and part remains on the sensor. This is basically what the IS does.
The IS operates by moving a lens group off-axis far enough and fast enough to redirect the light that is coming from a now-off-axis viewpoint back to where it was on the sensor. That way we keep the old view on the sensor even though the lens is actually pointing at a slightly different view. This assumes that the lens motion is mostly angular rather than purely "sideways". i.e. the camera and lens face a new direction and the sensor initially sees a different scene from the new angle of view. The IS corrects that by restoring the original scene, within the mechanical limitations of the IS group to respond.
If the IS group has incorrectly moved to turn that scene when there has been no corresponding turning of the camera then in effect you have a sort of tilt-lens effect that can shift the plane of focus so that it is no longer parallel with as well as coincident with the sensor plane. It will now be at an angle to but probably cutting through the sensor plane. The IS group has not changed the focus distance, but just turned the focused image plane off-axis.
In the early days of IS it was a plastic bag full of fluid held between two flat pieces of glass - one fixed and the other twisting without actually moving off-axis - so that it was in effect a variable prism lens ("vari-prism"). Nowadays it is more like a double-concave lens that moves sideways off-axis without twisting. The middle bit of the lens has less optical turning effect than the outer bit because it is more like a flat piece of glass. The farther it moves the more significant the optical turning effect becomes due to the shape of the lens.
So, what are your thoughts on #4, the shot that appears to be in focus on the left side of the picture only? Do you think this might be a case of a bad front focus situation (i.e. lens/camera system front focus problem independent of IS) that is somewhat mitigated on the left side (but further compromised on the right side) by a rotation about the vertical axis as commanded by the IS system? I'm sorry to say my thought here sure seems to be a stretch if the AF system was trying to acquire the tree; you know, that would mean Jeff's setup has a REALLY bad front focus issue.
I guess we might need additional input from Jeff regarding whether or not his 1DmkIII and this lens have a "tendency" to front focus or what ... Jeff?
p.12 #12 · ['NEW Fix' UPDATE!] - MkIII AF still broken
jkurkjia wrote:
So, what are your thoughts on #4, the shot that appears to be in focus on the left side of the picture only? Do you think this might be a case of a bad front focus situation (i.e. lens/camera system front focus problem independent of IS) that is somewhat mitigated on the left side (but further compromised on the right side) by a rotation about the vertical axis as commanded by the IS system? I'm sorry to say my thought here sure seems to be a stretch if the AF system was trying to acquire the tree; you know, that would mean Jeff's setup has a REALLY bad front focus issue.
I guess we might need additional input from Jeff regarding whether or not his 1DmkIII and this lens have a "tendency" to front focus or what ... Jeff?
I don't think image #4 can be put down to front focussing since front focussing tends to be visible throughout the focal plane, besides if it was front focussing that badly Jeff would have sent it in to Canon long ago. Collimation is where one side of the image can be out of focus in comparison to the other and this is due to the optical elements being misalinged on one side.
Equally, if Jeff's 70-200 performs perfectly on his 1Ds without this characteristic this torpedos the theory that there is an optical lens problem, rather the problem is more likely to be some kind of interface issue with the 1D Mark 3 body.
Just my 2p worth.
Also, FWIW we are discussing 2 things here. The thread was started regarding the ghosting visible in RG's test images and now we are discussing a focus alignment problem on Jeff's camera. With all due respect, this is a 1D3 body that is within the effected serial number range and needs the sub-mirror fix so therefore may not exhibit this behaviour once returned from Canon VA.
The image ghosting in RG's images is very interesting though I have been going over the rugby shots from last Saturday and so far don't really see this in my own shots. Very curious indeed though......
p.12 #13 · ['NEW Fix' UPDATE!] - MkIII AF still broken
DavidP wrote:
This has been noticed many times on the 16-35/2.8
One side of the image is just plain blurry and looks out of focus, while the other side is fine.
Sounds like a lens alignment issue.
That doesn't explain why it would be fine on a different camera, though.
It does not happen on a different camera during the same time. Jeff noticed the problem AFTER he switched to the 1DMK3 and he no longer has the 1Ds to do the same test. May be he was not looking so closely to his 1Ds images as of his 1DMk3 images . It is pretty obvious lens alignment problem and has nothing to do the body or the IS. If in dobut, shoot a few frames with IS and shoot a few similar frames without IS. It is very simple to rule out the IS is the problem. Of course, it has to wait for Jeff to get his camera back first. I wonder if there is problem with the lens, how much does it affect the testing results that we were assuming it was the camera problem ? Without the same problem with another lens rules out it is a camera problem. I just notices that Image #4 was shot with a 1.4X, which adds another variable in the equation. I assume that there is still problem without the 1.4X Tc?
p.12 #14 · ['NEW Fix' UPDATE!] - MkIII AF still broken
jkurkjia wrote:
I guess we might need additional input from Jeff regarding whether or not his 1DmkIII and this lens have a "tendency" to front focus or what ... Jeff?
No significant front- nor backfocusing. Plus, there's just no way that a 'simple' explanation like this could cause what is exhibited in the image (if you could see it at 100%, you'd understand; I can send a JPEG to whoever wants; e-mail me).
I went over my 1Ds shots in detail last night, and all images were sharp across the field. I will review recent 30D + 70-200 shots tonight, and try to shoot some flat-field tests in the morning. Hopefully I'll have the MkIII back in hand tomorrow, and will be able to compare with the real-time 30D shots.
p.12 #15 · ['NEW Fix' UPDATE!] - MkIII AF still broken
slau wrote:
It does not happen on a different camera during the same time. Jeff noticed the problem AFTER he switched to the 1DMK3 and he no longer has the 1Ds to do the same test.
Lenses both can and DO change with time, though. So the only way to really know would be to use the same lens on two different cameras on the same day, same lighting, and same subject.
Also, if it's IS (as some suspect), then one could simply take the same shot both with and without IS to see if it's having the effect.
p.12 #17 · ['NEW Fix' UPDATE!] - MkIII AF still broken
Jeff wrote:
You probably just missed my post above; CW confirmed that RG used IS for all ssequences.
If the lens were actively focusing during shutter release, wouldn't you expect a variability in focus from top to bottom, due to the vertically-travelling shutter curtains? We talked about this early in this thread in relation to the 'ghosting' issue (I think).
Yes, that's what I'd expect. I missed some stuff earlier; thanks for clarifiying.
p.12 #20 · ['NEW Fix' UPDATE!] - MkIII AF still broken
Jeff wrote:
Just ran across this 'ghosting' example; 24-70/2.8L:
Could that one be motion blur? It looks as if the boy and the rope were moving down and to our left, and the brightest areas are what show the blur. Okay, I'm just fishing for an explanation that's not as mysterious as the MkIII problems. Motion blur sometimes shows up as ghosting in highlights, while darker details can seem sharp.
POOF!
. It is pretty obvious lens alignment problem and has nothing to do the body or the IS. If in dobut, shoot a few frames with IS and shoot a few similar frames without IS. It is very simple to rule out the IS is the problem. Of course, it has to wait for Jeff to get his camera back first. I wonder if there is problem with the lens, how much does it affect the testing results that we were assuming it was the camera problem ? Without the same problem with another lens rules out it is a camera problem. I just notices that Image #4 was shot with a 1.4X, which adds another variable in the equation. I assume that there is still problem without the 1.4X Tc?
