Re: A few bits about EX1 shutter, blurs and transfer

Mark,

[Mark Palmos] "2/ maybe its my imagination, but shooting at 100th of a second 1080 50i HQ even a very gradual pan shows significant softening of the edges once the pan starts. When it stops again, everything is sharp. At the speed I was panning I wouldnt have expected the softening to be as extreme at 100th of a second. This happens regardless of Interlaced or Progressive."

I think that I can answer this question adequately, and no, it's not your imagination. Several things are at work here. First of all a shutter speed of 100th of a second is not that fast. It oughta be a bit better at a significantly faster shutter speed.

Secondly, and probably the bigger culprit, you are seeing the limitations of the long GOP Mpeg2 codec. The Sony 35mbps codec is better than it has any right to be, IMO, but it's still compression defined over multiple frames of video. Particularly on a pan of a detailed subject where everything in the frame is changing with each frame, however slowly, it's impossible for the codec to keep up with that amount of change, so the algorythm automatically cuts the detail to avoid blocking artifacts. You hear folks say the codec is nearly impossible to break, but they are essentially talking about blocking and other highly visible artifacts. Some situations, usually involving Pans, the codec has to break down the detail to keep up and avoid the more visible garbage. You'll see much less problem if you have a subject that takes up far less than the whole frame, moving through a stationary shot. There's considerably less to change from frame to frame and the long GOP algorythm is more successful at preserving the detail.

This speaks directly to the Sony/Pany controversy over the relative benefits of interframe (long GOP), v.s., intraframe compression. All other things being equal, although in most instances the opposite is true, the XDCam 35 mbps interframe codec is not going to look as good while panning a detailed subject or in subject movement that fills much of the frame as the DVCPro100HD codec because the Pany codec compresses each frame separately.

Matt Devino's moving shootout footage, illustrates what I'm saying pretty nicely even though we are seeing full res footage that has all been compressed long GOP again for web delivery. In his case, all other things are not equal in that the EX1 has far better native resolution on the camera and in the codec than the HVX 200, yet when you watch the comparison when the respective cameras pan with the skateboarder, there's very little difference in apparent comparative resolution, the HVX holds it's own because of the codec differences. Yet on the wide locked frames of the skate through, the background looks sharper with the EX1, while it's hard to see a lot of difference in the apparent sharpness of the skater himself.

It'd be great to see Matt's tests natively from each camera. Perhaps he could comment on what he sees prior to compression for web delivery.

[Mark Palmos] "what on earth is so nice about progressive. I mean, ok, when you freeze it looks nicer, ie not interlaced, but when you pan it jerks and looks simply awful... almost as bad as having fields reversed by mistake. If our little EX1 did 1080 50p that would probably be the best of both worlds, but alas. I personally prefer a smoother looking 50i to the jerky 25p. I also know progressive is better for compositing as well as watching on computers, but for watching on a TV monitor, i far prefer interlaced."

I gotta say that like you I don't understand the 24/25p or die phenomenon, however things are a little different for us in NTSC land. When 24p goes to TV here pull down or pull up or whatever the heck it is, is applied here to deliver a 1080 60i signal, and that tends to mask some of the judder that you see in 25p but also lowers detail. IMO the most logical way to shoot relatively low action things here in 1080 is 30p where you get some of the best from both worlds.

Ron Shook