Mal: Merci beaucoup for the 60i test. That seems to leave out the processing part of the chain.
Question: does the HV20 optical stabilizer, on and off, influence your results in any way ? It seems strange that a stabilizer that works so well, at least in my tests, would be the cause of the wobblies, but it could if the optical low frequency damping is off or bad.
On the other hand (another possible learning point for me like the previous one) I cannot see how the CMOS sensors cause the wobblies.
(So much to learn...so little time...)
I've just two points:
1) Is this a PAL issue? Or is the NTSC version less noticeable on this wobbly hassle?
2) Are there defective PAL units being distributed and delivered?
Last edited by Emanuel; 2007 May 1st at 23:45.
Yeah, it's certainly possible that we are discussing two different things in this thread anyway; the Norbert-wobble might NOT be what I am pointing out (with the crooked line). In the end, I guess CANON will have to weigh in on this, as others have mentioned.
But anyway, WELCOME Emanuel!
Nope, it doesn't seem to have any effect on the wobbles I'm getting, just the regular shakyness.
Originally Posted by Melachrino
After watching the samples, I've been changing my mind. Here's my post @dvxuser.com on this [LINK]:
Originally Posted by Mal
«I had a different opinion before. After downloading all the files (or almost), my guts are:
There's some with these PAL version units -- too much sensitive in terms of roller shutter comparing with other units. Austin Meyers' sample, for instance:
Is there better example?
The question is:
Is there or not a noticeable difference between NTSC vs. PAL versions regarding the roller shutter issue?
Taken these examples I'd say: YES.
To notice that there is a similar problem with the NTSC version units, it's to put a mask on it. 'cause of course the CMOS is there too, the roller shutter as well. But taking both examples (Norbert vs. Meyers) there's a great difference.
What's up? Defective units? Or "just" a defective PAL manufacturing?
Unless, other units may prove this conclusion as a wrong conclusion.»
Thanks for your kind words!
i contacted canon uk on the "wobblys" i spoke to a person from support
who said that there was no issues with this camcorder(personally i think this
person was not very technical minded as she was confused many times what i was saying ?) so i have contacted them by e-mail to try and get a knowledgeable techi who may be able to give a better reply i await there response.
That's good. We need a techy answer from Canon. I was thinking of contacting Canon Sweden about this but I had a feeling I would just end up talking to the support team and they are usually not the techheads.
By the way, a guy over at dvinfo had the exact same problem I have, and he has an NTSC HV20 from when it was recently released. Here's the thread: http://www.dvinfo.net/conf/showthrea...ghlight=stable
Download the footage he posted in the first post. It shows the same kind of distortion of the image that I'm getting.
Last edited by Norbert; 2007 May 3rd at 10:00.
Check the two Austin Meyers' samples:
On the 1st it is the same thing than yours, Ron/Norbert. But on the 2nd? There's no "the" same outcome. Why?
It seems this problem affects the PAL version more than the NTSC version.
There's a new thread here (it seems something similar):
PAL unit again.
Shall it be fair to call it more noticeable with PAL than NTSC?
Good point, Emanuel. I have a feeling that if I put my HV20 on that stabilizer and ran around with it like that it would still be wobbly as ****.
Numbox, would you be able to post a small clip walking with the camera handheld at full wide-angle with IMG STAB turned on?
ok, will do it when i get the apropriate shot. what am i looking for - moving objects or non moving ones?
Originally Posted by Norbert
also, can you tell me how to turn OIS on/off and is there a way to change the forum color from default black/green to something like gray or whit/black?
EDIT: Aaaaaaaaaaah, damn it, i've only seen one clip today at work, the one at the swimming pool or something and i completely missinterpreted what you ment by wobbling. The way video was encoded, it seemed to me there was wobbling in the moving objects. Don't know how to explain it really, but i was completely missreading the whole thread, what you meant by wobbling i always thought of as shaking. Now that i've downloaded your clip, it's clear to me what you meant by wobbling. And unfortunately, it's present in my recordings too, it's just that i've always contributed that to my own shaky hands. Strangely enough, i have some random shots made in a driving car, just shooting the road, and those were much steadyer, at least to me.
Last edited by Numbox; 2007 May 4th at 11:45.
It seems that while standing or sitting still with the camera in your hand, it doesn't wobble like that... much. At least not as bad as when you take it for a little walk.
You just confirmed my theory that all HV20s may have this problem and that very few owners see it as a problem. If this had been my first video camera I surely would have just blamed my shaky hands too, but I have owned other video cameras, both consumer and prosumer, and never experienced anything like this before.
This is however my first personal encounter with HDV. Does the image behave like this on all HDV cameras? If that is the case then I'm no longer so sure that HDV is something for me.
Last edited by Norbert; 2007 May 4th at 14:18.
Im no CMOS/CCD engineer, but the example of the straight line going sideways at fast pans (on page 2) could be explained as to how the sensor performs its image aquisition.
1 ) CMOS sensor captures entire frame at once and stores entire frame as a single picture then the image is transferred to the DSP/Digic processor. The same way that an image is aquired on film as one complete snapshot.
2 ) CMOS sensor is given a 'signal' to capture the image, this signal is not sent to all the pixels at once but is rather 'passed' along from pixel to pixel and/or the captured image is streamed to the DSP in a top to bottom, left to right nature.
If it is scanning the sensor from top to bottom, left to right, then perhaps as the CMOS sensor is doing its 'capture' by the time the image is captured as a single frame (remember we are scanning over 2 million elements) the image being captured has moved as the camera is moving. Given this it would explain the apparent distortion..
Im not sure if higher end cameras/imaging elements have additional circuitry so that the cmos element captures and stores an entire frame in a single 'click' and then streams the data to the DSP, which would mean each cmos element would require memory on chip in so as to allow the entire frame to be captured in one go... Given the price of the HV20 I'm inclined to think that the cmos is read one element at a time in a progressive nature at a very high speed (but not fast enough to counter the sheering effect at high speed pans). This effect would therefore also be visible on fast moving objects that are close.
Apparently the leaning vertical line example is caused by the rolling shutter.
Thanks for that... I guess my hypothesis was the long winded way of saying 'rolling shutter' <grin>
Page 83 on the PAL HV20 manual actually briefly talks about the phenomena..
I did run into a good page that demonstrates this effect
I also ran into this piece of 'ooh ahh' hardware
While I know its a different beast, it shows what CMOS technology can do.
Yup. Nevertheless, I'm also asking myself if the problem would be or not a PAL vs. NTSC issue on rolling shutter.
Ron, wouldn't it be also a shutter speed hassle at progressive mode?
Here is a 25p@1/25 example you probably already know, smooth enough challenging or even defeating the PAL theory:
Interesting update here:
«This issue is definitely rolling shutter. The reason the PAL version seems more affected than the NTSC version is because of the higher clock speed to get to the 60i/60P required for NTSC, rather than the 50i/50P for PAL. Higher clock speeds equate to faster read-out times (faster time for the read-out from the top to the bottom of the sensor), which means less rolling shutter.» ~ Jason Rodriguez, Silicon Imaging Team
Here's the clip. It's a very uneven surface and i wasn't being careful how i was walking and also i didn't even try to hold my hand as still as possible, just normal walking and shooting. The cam was set at 25p, TV@25.
Originally Posted by Norbert
Maybe slower shutter speed means less rolling shutter artifacts?
Numbox's footage = Norbert's footage
(from what I can tell....)
So, is that something that could be corrected with a firmware upgrade or is it here to stay? Does mounting something heavy on the camera really help to steady the image?