Mark III AI-Servo test on static subject

[joseph papa]

My Mark III is due back from Canon on Tues or Wed. It went in for submirror and one shot issues. Rest assured I will try this test and let you all know if I have good luck with an affected body that has been "fixed"

 

[Scott Larson]

I wish you had put all the required info in your first message too

Sorry for making you read the whole thread.

So you had your affected 1dmk3 replaced by another 1dmk3 also without
blue dot and when ?

No, I have a Mark III without the submirror fix. I sent it into Irvine because it wouldn't focus consistently in AI-Servo or One Shot. Irvine said that it would not need the submirror fix immediately because that only affects the camera in light and/or hot conditions. They were able to fix the One Shot problem completely and improve the AI-Focus somewhat but as you can see, it's still not quite fixed yet so it looks like Irvine was mistaken.

I'm still waiting for Canon to call me to schedule my submirror fix. They said maybe next week.

 

[Christopher Glaeser]

Have you reviewed the RG images with the very slight pixel blur? Any other ideas what could cause this? Note the rather high shutter speed.

Best,
Christopher

 

[Honor]

Hi Scott,

My message was not intended to you but DavidSwensson. You have been very clear in all your posts, there was no need to type that again. People missing your point simply did not read the whole thread. I have read it to get as many owners feedback as possible.
BTW that is a great test, simple and revealing.

Honore

 

[pikeface]

has given me second thoughts to if it puts it to bed.

The word constantly suggests all the time. Well I don't know about every body else but mine does not constantly move around, only every so oftern. And it always seems to come back to the same in focus point.

"Even if AI Servo AF is used to focus on a still subject, stable control is enabled so that the lens drive isn’t constantly fine-tuning. If the subject starts to move, the focusing responds immediately to detect the subject constantly".

--
Have a look and leave a message
http://www.pbase.com/jeremy_broome_smith

 

[Victor Engel]

This may seem like comparing apples to oranges, but I had the same issue with my Canon 10D and 50mm f/1.8 lens. The problem was resolved after I sent it in for calibration twice (after the first time, it performed worse than before I sent it). Another symptom was that if the camera was tripod mounted in AI Servo mode and I pointed it at a static subject, the lens would rapidly shift focus while the shutter button was half depressed.

I think the problem happens because of the lack of feedback in Canon's focusing algorithm. Simplistically, I think it goes like this:

• Focus button is pressed.
• Camera calculates whether image is in focus. If not, an algorithm determines how far the image is out of focus.
• Camera sends command to lens to move by X.
• Lens moves by Y (presumably, X = Y).

Now if everything is calibrated, the whole process results in an image that is in focus. But if there's a miscalibration, then whatever was calibrated for X may be incorrect, but different from before the focus operation started.

If this is what's going on, then the motion of the lens during this operation will be chaotic, in the mathematical sense.
--
http://www.pbase.com/victorengel/

 

[Honor]

You are on the right track. I will expand that saying that after each new iteration the image should be more in focus until the focus is below the focusing detecting threshold of the camera (all of that being extremely fast). IF anything is wrong in this dynamic feedback loop it will never converge meaning the OOF evaluation will not decrease below the focusing detecting threshold but keep jump around and the loop will keep active indefinitely. This unstability can be due to a faulty element (the submirror assembly for example). Here is one analogy: You've seen those old B&W movies where a car driver is constantly correcting the steering wheel to get the car going straight on the road whereas nowadays it is not necessary. The difference is a more precise and rigid steering.

Another point is if the in focus threshold is set lower on the latest camera there is going to be a greater rigidity requirement for all elements in servo dynamic loop to be able to maintain the focus and rest the servo motor.

This may seem like comparing apples to oranges, but I had the same
issue with my Canon 10D and 50mm f/1.8 lens. The problem was resolved
after I sent it in for calibration twice (after the first time, it
performed worse than before I sent it). Another symptom was that if
the camera was tripod mounted in AI Servo mode and I pointed it at a
static subject, the lens would rapidly shift focus while the shutter
button was half depressed.

I think the problem happens because of the lack of feedback in
Canon's focusing algorithm. Simplistically, I think it goes like this:

• Focus button is pressed.
• Camera calculates whether image is in focus. If not, an algorithm

determines how far the image is out of focus.

• Camera sends command to lens to move by X.
• Lens moves by Y (presumably, X = Y).

Now if everything is calibrated, the whole process results in an
image that is in focus. But if there's a miscalibration, then
whatever was calibrated for X may be incorrect, but different from
before the focus operation started.

If this is what's going on, then the motion of the lens during this
operation will be chaotic, in the mathematical sense.
--
http://www.pbase.com/victorengel/

 

[Victor Engel]

These pictures aren't exactly high contrast. The background appears to be grainy brown, so I'm wondering if you're using a high ISO and the weakest possible color. Do you get the same results with black and white? Different lighting?
--
http://www.pbase.com/victorengel/

 

[JKovack]

i repeated a test with my 50 1.8 and it DOES shift round......

none of my USM lenses shift.

This may seem like comparing apples to oranges, but I had the same
issue with my Canon 10D and 50mm f/1.8 lens. The problem was resolved
after I sent it in for calibration twice (after the first time, it
performed worse than before I sent it). Another symptom was that if
the camera was tripod mounted in AI Servo mode and I pointed it at a
static subject, the lens would rapidly shift focus while the shutter
button was half depressed.

 

[J A K]

"NT" means no text.

Joe Kurkjian, Pbase Supporter



SEARCHING FOR A BETTER SELF PORTRAIT

 

[Curator]

i repeated a test with my 50 1.8 and it DOES shift round......

none of my USM lenses shift.

Are you testing this on a fixed sub-mirror Mark III? I just tried this test with both a 100 2.8 Macro and 70-200 2.8 IS lens, and the focus in servo mode is jittery. I tried this test on a tripod and used wired shutter release cable, and the focus is very jittery for about 1 to 2 seconds and then settles down, but still shifts occasionally as I continue to hold the focus button. It is easiest to see by just watching the distance scale on the lens rather then looking through the viewfinder. I tried the same setup with my 20D, and it doesn't have this problem.

By the way, this is with a Mark III that has not had the sub-mirror fix yet. I also tried different micro-adjustment settings, and it didn't seem to improve, or worsen, the jittery behavior.

 

[JKovack]

blue dot, serial 545XXX, firmware 1.1.3

so yes the sub mirror was apparently fixed before i purchased it (late november, production date is october)

 

[Curator]

blue dot, serial 545XXX, firmware 1.1.3
so yes the sub mirror was apparently fixed before i purchased it
(late november, production date is october)

That's good...now I won't feel bad sending my camera in for the repair. I had not seen any focus issues, but I hadn't done much to stress the focus in servo mode. Perhaps this is the easiest test to determine if the focus problem exists.

 

[blackhawk13]

--That's normal behavior from all I've been told. One shot is a better choice many times anyway.

-nothing beats a fast lense, except a fast girl-

 

[Curator]

--That's normal behavior from all I've been told. One shot is a
better choice many times anyway.

I used to believe that also, but I noticed the following statement in the Mark III White Paper:

"Even if AI Servo AF is used to focus on a still subject, stable control is enabled so that the lens drive isn’t constantly fine-tuning. If the subject starts to move, the focusing responds immediately to detect the subject constantly".

I was able to reproduce his results with my Mark III that has not had the sub-mirror replaced yet. I could also see that when in servo mode, the focus is jittery for 1 to 2 seconds and then settles down, but would still shift focus very slightly while shooting. This can be seen by just watching any movement of the distance scale on the lens while shooting or even while just pressing the focus button on my shutter release cable. My 20D, once locked on focus, does not shift focus while holding down the focus button in servo mode, so I'm starting to believe that this is an easy way to demonstrate the focus issue of the Mark III.

 

[Victor Engel]

The submirror issue is a mechanical problem. The issue described in this thread is not a mechanical one related to the mirror. If so, it would manifest only when the mirror is being activated. I believe it to be a calibration issue or possibly an issue with loose/dirty contacts.
--
http://www.pbase.com/victorengel/

 

[blackhawk13]

--That's normal behavior from all I've been told. One shot is a
better choice many times anyway.

I used to believe that also, but I noticed the following statement in
the Mark III White Paper:

"Even if AI Servo AF is used to focus on a still subject, stable
control is enabled so that the lens drive isn’t constantly
fine-tuning. If the subject starts to move, the focusing responds
immediately to detect the subject constantly".

I was able to reproduce his results with my Mark III that has not had
the sub-mirror replaced yet. I could also see that when in servo
mode, the focus is jittery for 1 to 2 seconds and then settles down,
but would still shift focus very slightly while shooting. This can be
seen by just watching any movement of the distance scale on the lens
while shooting or even while just pressing the focus button on my
shutter release cable. My 20D, once locked on focus, does not shift
focus while holding down the focus button in servo mode, so I'm
starting to believe that this is an easy way to demonstrate the focus
issue of the Mark III.

--Interesting point. I haven't played with mine much since it's fix, but it doesn't seem as twitchy as it use to be in AI-servo with static subjects.
All good news.

Again there are multiple presettings to consider as well, so a lot of time, and experimentation is needed to fully understand this camera's unique personality.

The 20D isn't much of a comparison though. The MK-3 is more light sensitive and much faster, and using different hardware as well as software.

Personally I really like this body, much more than my 5D or 20D.

-nothing beats a fast lense, except a fast girl-

 

[Bartek]

Are you testing this on a fixed sub-mirror Mark III? I just tried
this test with both a 100 2.8 Macro and 70-200 2.8 IS lens, and the
focus in servo mode is jittery. I tried this test on a tripod and
used wired shutter release cable, and the focus is very jittery for
about 1 to 2 seconds and then settles down, but still shifts
occasionally as I continue to hold the focus button. It is easiest to
see by just watching the distance scale on the lens rather then
looking through the viewfinder. I tried the same setup with my 20D,
and it doesn't have this problem.

Hi Curator,

in my case of blue dot 1D3 with 1.1.3 and the 100 2.8 macro there is also a jittery sound coming from the lens every 1-2 seconds when holding the button, however the sequence comes perfectly spot on throughout the entire batch. I was watching the fast scroll of frame by frame sequence and there was not a slightest shift in the focus. When I hold the button and the jittery sound comes from the lens the shift in focus point is unnoticable in the VF and also the barrel movement is actually unnoticable as well.

I was thinking about the jittery movement as my target (line of black, well lit 12p text) was scewed at 45deg to the sensor plane so in fact not flat. The upper part of the text line is 1-2mm further that the bottom line and both are withing the focusing point frame so iit may confuse the camera.

With the 70-200 2.8 non IS I don't have this jittery sound however I'm further away from the same target of text line so this time it may be considered by camera's focusing system as "single & flat" enough.

I both cases my distance was determined by the relative height of the same single text line to equal the height of a single (center) focusing point.

my the test conditions (heavy studio column stand, cable etc are described here):

http://forums.dpreview.com/forums/read.asp?forum=1032&message=26422159

The 100 2.8 macro is brand new as my previous 3 year old copy of the same lens was acting wired in terms of focusing with my 1D3's left side focusing points (while being fine with my 5D) and the new copy of the lens is perfect on both cameras and all focusing points. I also made me wondering why two copies of the same lens act differently on the same body (maybe there was some minor revision change in 100 lens in the meantime).

--
Bartek

 

[Leon Wittwer]

See "I did that test" message above.
--
Leon
http://homepage.mac.com/leonwittwer/landscapes.htm

 

[Andy Wai]

The submirror issue is a mechanical problem. The issue described in
this thread is not a mechanical one related to the mirror. If so, it
would manifest only when the mirror is being activated. I believe it
to be a calibration issue or possibly an issue with loose/dirty
contacts.

Based on the number of data points on this thread, I doubt very much that this is a contact issue.

Try testing with very high contrast target, like the AF micro-adjustment here:

http://www.openphotographyforums.com/forums/showthread.php?t=4708

The target is made up of black and white dots. Put it on an LCD screen and something like a 135/2 can resolve it to the dot level from 10' away. AI Servo latches on immediately and stay put. No jittering at all. Now put some continuous tone picture on the same screen and AI Servo starts hunting.

In fact, it isn't just AI Servo. With single shot, the micro-adjustment target is dead on every time. With something less sharp, you can see slight variation in focus.

This is on a Blue Dot that went back to Canon before being sent back to dealer. And it's also one of those listed on the exclusion list. The AI Servo tracking speed custom function has no effect on the jitteriness and firmware 1.1.0 and 1.1.3 makes no difference either.

I'd say it's either calibration issue or problem in the way the AF algorithm deals with low contrast target.

Andy