motion artifacts

[directore]

fresh from testing and results ain't that good.

Set up: HV 20 mounted (relatively rigidly) on the side window of my car (RAV 4, somewhat stiff suspension). Plenty of daylight, shutter priority 1/500, infinity focus, 24p. Took it for a spin in the neighborhood, normal surface quality, bumps here and there but nothing major, 10mph max.

Resulting footage quite sharp, no motion induced blurring, but some frames show strong intra-frame swimming like distortion, clearly motion induced.

This was unexpected - with the exposure of 2ms there should almost be nothing motion related in the frame.

One hypothesis, optical stabilization in this cc has an extremely low time constant, say <1ms, hard to believe but in principle possible which would mean I'm seeing desperate but unsuccessful attempt to compensate for motion. Wrong, disabling optical stabilization (which should lock up their system physically) made that intraframe swimming distortion much much worse.

So either the system has a problem (too busy implementing that 60i crap in real time?) moving the charges from the chip to the A/D synchronously or there is some timing issue somewhere or I am missing something else. Which one is it ?

Canon?

-----
(here is the footage: http://rapidshare.com/files/44450441/testmotion2.mov.html ~8M, just closing the driver side door, not moving yet, camera on the other side).


(one more hypo is this: the chip exposes and collects charges sequentially line by line, which to me at least would be very unusual, but I admit I'm not up to speed on how chips operate these days.)

[Ian-T]

Sounds like the infamous wobble issue that has been discussed here before. Why don't you give us a sample of what you are seeing.

If this is the case then I believe it has a lot to do with the OIS (as you seem to suggest) but I believe most people think it has to do with rolling shutter in CMOS chips coupled with 24P. Next time maybe try shooting in 60i because 24p tends to exxagerate the rolling shutter artifacts.

[directore]

search button, wasn't it?
---
yeah that's it, my intuition was right, "rolling shutter" it clearly is http://www.fieldrobotics.org/~cgeyer/OMNIVIS05/final/Geyer.pdf, kind of a bummer I must say, how come I haven't seen this discussed in any published review? CamcorderInfo?
So no more true progressive footage to he had anywhere short of going back to the old celluloid?

Really?
---

P.S. sorry for cluttering the board rehashing old issues.

[Ian-T]

search button, wasn't it????

Unfotunately the clip only has audio. Tried it on 3 different players and no picture.

[directore]

("search button, wasn't it?" was a reminder to myself)

The clip is in AIC (Apple Intermediate Codec). Will repost in something more pedestrian.

And here it is (H.264 codec, mp4 enclosure) on a different hosting site though:

http://www.megaupload.com/?d=IEO1G52V

[Worley]

Yes, the information is read out line by line.

Old television cameras used tubes, and the output from them was sent line by line. So, if it we were still using tube-based cameras, would they suffer from the rolling shutter effect?

[directore]

Not really, depends on how they exposed those lines to light, if all at the same time, no rolling shutter artifacts. Which is, btw, how things work in the competing technology, CCD devices, there you expose every line at the same time, and read them out line by line afterwards. The original CMOS could not do that, they had to expose and read out lines individually, thus those horrible "rolling shutter" artifacts in the HV20, fortunately they came up with a solution since then, it's called "global shutter". Time for Canon to implement it!