Still getting duplicated frames on 5D mk II

[Bill Binder]

I don't think you have ever told us whether you have been able to reproduce this in FULL MANUAL (e.g., M mode, video mode not exposure simulation, no auto WB, no auto ISO, no auto anything)?

I can get your results if not in full manual, but never when in full manual, which is where you ought to be anyway, no?

[Julian Frost]

Yes, FULL MANUAL EVERYTHING (M mode, manual ISO, manual WB, manual shutter, 24P, 30P, correct live view settings etc). It is NOT a settings or CF card issue, It is a defective camera. I've spent 2 years researching this problem. It needs to be fixed at the hardware level.

Sorry if I sound a bit snippy, but I am rather annoyed. I just got an email from Canon asking if I'd tried other CF cards!! And this in response to my 2 page email that detailed all the tests performed, settings used, and hardware used. I took the time to do all these tests and document them, and provide original evidence (MOV files from the camera) and supporting evidence (screen shots of the duplicate frames) so that Canon had everything they need. They could at least READ the documentation they asked for!

[Andrew Clark]

Hey Julian -

Bummer that you still are having this problem and that you are considering selling the camera because of it.

You'll definitely find a buyer; but the buyer will be a still photographer and could careless about the video functionality. My buyer didn't care about the video capabilities of the cam, since he was a still photographer. He bought my cam because he couldn't afford his first choice, the Nikon D3x.

But nonetheless, he wasn't the only one to inquire about the cam as I had no less that fifty inquiries...and that was within 2 hours from when I posted it!!

On the other hand, I do hope that the Irvine facility can find out what the heck is causing the skipping frames.

Please keep us posted; much appreciate it.

[Khoi Pham]

Keep up the pressure and demand to talk to higher up in management, once it took 3 times before they get it right for me on a repair. BTW your technician sound just like the same idiot tech that works on mine.

[Julian Frost]

The person I spoke with at CES, was *the* top guy at the Canon Service Center. I'll be leaving him a voicemail message today.

[Julian Frost]

I got a call from Canon yesterday. They were setting up a conference call with the techs in New York to discuss my camera's problem, and I should hear some news later today. They were glad I'd included sample footage on a CF card. Now they can use the EXIF data to set up their own test bodies to see if they can replicate the problem. Based on my own, extensive, testing, I'm 100% certain it's a hardware issue, but I'm very glad they're giving my camera the full attention it needs, and not just randomly replacing parts and saying "it's done!"

[Andrew Clark]

Great to hear the progress so far Julian!! Keep the pressure on them, now that they admit there is some sort of issue going on that's not card related.

Let's hope they can isolate the problem, fix it and get you back to shooting without any frame skip issues.

Appreciate you keeping us posted; look forward to your next (hopefully) positive post!!

[Julian Frost]

The update from New York is that as 5D mk II owners, we should be aware of a page in the printed manual that says something to the effect of "when shooting video with high ISO setting, if the camera has to record a lot of information, it's possible that the video could freeze for a moment." It was explained to me that having a high ISO, moving the camera, and having lighting conditions change, could cause the camera to momentarily pause video recording.

While the manual does in fact say something to that effect, I personally don't believe that's what's happening here. The manual was written before manual control was introduced, and we've all seen what happens to footage recorded in an automatic mode when the camera changes aperture during recording... It pauses, for almost second. What I'm seeing is a duplication of 3 frames, which results in a skip in the footage. Another reason I don't think this is the reason for what I'm seeing, is that sudden changes in the amount of data being recorded by the camera are exactly what Vincent Laforet captured during his shooting of Reverie. His camera was attached to a moving vehicle, driven down dark streets, with car headlights suddenly illuminating the way. His shots from a helicopter have more motion in them than a simple pan, or a locked off shot in an interview setting, shot at 100 ISO where the light doesn't change and the subject hardly moves. I've had duplicated frames in all these situations. I've had duplicate frames occur when I've been shooting at ISO 100, with the camera on a tripod, and the subject sitting and talking at the camera.

Anyway, I'm told that the sensor, and associated electronics, have all been replaced and that in side-by-side tests with one of Canon's reference bodies, they could not replicate the problem on my camera. I will pick it up tomorrow and run it through its paces. Hopefully it will be perfect. If not, well, we will have to see what the next step is. I'm hoping this will be the end of it. Canon's techs have worked hard on this for me, and I fully believe this issue will soon be resolved.

I'll post an update in the next few days.

[Antony Newman]

Julian,

I am not 100% convinced it is a hardware issue.

====
I could speculate that when the focus of the lens is changes, that this causes the DryOs to invoke a 'property' that runs a short lived operation at a higher priority than either the Fwrite() task of the writing data to CF, and also higher than the task that fills the Hd vram segment that the sesor data is being copied into.

Rotate the lens too much, combined with the .264 DIGIC compression doing too much processing (ie interframe picture changing too much, or Higher ISO) -> and this causes a task freeze.

This could hypothetically be induced by created a task with a stupidly low priority (eg 1), that runs for an increasingly large amount of time.

====
But, it is possible that changing of the focus is actually causing the image to change 'the most' from the perspective of the .264 compression. The being the case - they should. 1) Give priority to the DIGIC to finish the .264 compression, reduce the priority of the other tasks, to guarantee that it is completing within the required frame time. Or 2) they should use the DIGIC clock to count how much time is left on each frame. If half way through the frame, it's not going to have enough time to encode .. they want to dynamically change the encoding 'quality' ... thereby avoiding dropped frames.

====
Whatever the actual cause - I would speculate that it also involves the memory bus, and or the ARM chip - as I guess the performance of ARM only code probably also freezes.

AJ

[Julian Frost]

Al,

But I get the duplicated frames even when the camera is on a tripod and the subject is sitting a fixed distance away... No change in focus, no zooming, no camera movement, bare minimal subject movement, and the camera is in full manual (everything) mode.

Anyway, I'm off to pick it up now. Will let you know. My fingers are crossed!

[Andrew Clark]

Julian -

I hope it is GENUINELY repaired to perfect operating conditions.

I personally believe that there was just a bad batch of 5d2's that unfortunately have this most irritating problem. Why others have never mentioned it or encountered it, is kind of bizarre.

Whatever the problem(s) that are causing all this, let's hope it is 1.) resolved and 2.) won't pop up in the next version of this camera.

Keep us posted please!!

[Wayne Mann]

Hi Julian,

Do you have an update for us yet? Very curious to know if the problem has been resolved.

Have a great day


Wayne Mann

[Julian Frost]

I got my camera back, and was able to recreate the problem again later that evening. I spoke with one of the senior technicians about it and we're currently working on a solution (technical or otherwise), which may involve me shipping it to Canon New York for the techs to look at.

I can reliably recreate the problem at ISO 3200, which, while not a fair test (since that's pushing the video limits of the camera's single processor quite a bit), allows me to give Canon some footage to play with... and with it, all the EXIF data which they need to replicate the problem.

I get duplicated frames at -all- ISO settings, but at ISO 100, it happens quite infrequently... usually on a take I can't reshoot! :-) When Canon used their 5D mk II test body with all the same settings as I used, even at ISO 3200, they were unable to get duplicate frames in their footage. To me, this means it's a problem with my specific camera.

While speaking to a technician on the phone, I loaded a clip into Canon's ZoomBrowser to give her the EXIF data, as she'd been told (incorrectly) that I only had this problem at ISO 3200. The clip, which contained one set of 3 duplicated frames, was shot in bright daylight, at ISO 100, full manual mode. After checking the EXIF data, she admitted that she has seen this problem at ISO settings greater than 1600, but she hasn't seen it at ISO settings below 1600.

I'll keep this thread updated.

[Julian Frost]

I shipped my camera off to Canon's New York technicians on Thursday. We'll see what happens. I only had the camera for a short time, prior to shipping it off, but during that time, I was unable to get any repeated frames at ISO 100 to send along with it. That's the trouble with intermittent problems! Grr!

[Julian Frost]

Sorry for the long delay...

I sent my camera off to Canon in New York, at the request of Canon's Customer Service department. They kept the camera for a several weeks, after sending me a loaner, but ultimately they said they were unable to duplicate the problem. Since the only way to replicate the problem is to use the camera to shoot hours and hours of video, and then watch it all very carefully, in the hope of catching a very intermittent problem, I highly doubt they put in a whole lot of effort in trying to replicate the duplicated frames!

They asked me to send them sample footage of duplicated frames at ISO 100-1600, but since I don't normally keep unusable footage, I didn't have any readily available. They asked me to look through (IE. "watch") all my clips to find examples of footage containing duplicate frames. That's several hundred gigs of 24P and 30P footage! That would have been tedious, and probably pointless, for the reason given earlier.

So, about 3 weeks ago, I got my camera back. I haven't had many opportunities to use it since I got it back. If I shoot video and get duplicated frames, I will definitely keep the footage and send it to Canon along with the camera.