Ultra High Definition Video

mr.henry writes "Engineers at the Japanese Broadcasting Corporation (NHK) have developed a prototype ultra high definition video (UHDV) system. How good is it? When it was shown to the public, some viewers experienced nausea because of the ultra realistic visual effect of speed without the usual physical sensation of movement. 18 minutes of UHDV takes up 3.5 terabytes." 4,000 horizontal scanlines. Excellent.

Comments

Frame Rate

[augustz]

The question is what is the frame rate. At 60 fps (i) they may have experienced nausea from that. If it was 60 fps progressive that would be something very nice.

I'm starting to wish they would shoot movies at 60fps.

What compression did they use?

[JFMulder]

I mean, at 33 million pixels for a picture, let's say 25 (it's the number of PAL FRAMES, not fields, per seconds) times per second at 32 bits of depth and you get 3.3gigs per second.

Which makes you wonder if they used compression at all? Even if their system was doing 60 non-interlaced frames, you get roughly 8 gigs of uncompressed video per second. Compressed, it would have to be way less that 3.3GB/s.

And based on the numbers, you can see that they either didn't use audio, or it was included in the 3.3GB/S figure because 3.5TB / 18 minutes / 60 seconds = 3.3GB/S.

So, is there someone I forgot, or are these guys really using uncompressed video? And if they did, WHY? I know, uncompressed video will always be cleaner, but come on, this might be a little too much in this case.

Re:What compression did they use?

[MrResistor]

I repair high end digital video servers. Here are my thoughts on the subject:

First, they probably are using MPEG or MPEG2, or maybe MJPEG, but that's pretty unlikely these days. Just because they're using the codec doesn't mean they're going to use the full compression it allows them. Remember that there's a trade-off there; the more it's compressed the more quality they lose. The bottom line is, if they want to show off the technology they will be compressing it as little as possible, because the degradation will be visible to someone who's an expert.

One of the things that has hampered digital technology in filmmaking is the quality of the final image. At 2k lines digital becomes competitive, in fact with a 2k telecine (converts film to video) you can just start to make out the grain of the film. At 4k digital is better than film, and thats going to win over a lot of directors and producers who never would have considered it otherwise. For myself, movies in the theater look a little fuzzy to me. Not bad, but noticable. I would welcome the improved quality this tech will bring.

I strongly suspect that's the market this tech is aimed at, because nobody is going to be broadcasting uncompressed 4k video.

Additionally, the rule in the industry is that you never compress your source material. A lot of that is superstition at this point IMHO, but the fact remains that there is going to be a need for this. That stance is kind of ironic, considering that most broadcasters will compress what they're sending out as much as they can get away with.

I very much doubt that the problem is scaling processing power to do the compression. Any pro-level setup has dedicated hardware to do that, and if one encoder chip can't handle the bandwidth itself than they just use multiple chips. Moving the data around the rest of the system is a bigger design challenge than the compression is.

One more point (bordering on OT, but it is related), video is the most demanding application that hard drives are used in. I have to torture test every single drive we send out myself, using our own methods, because none of the standard drive testing tools/suites even compare to what we consider "normal" use of our product.

For most data applications one only needs to worry about capacity and bandwidth. Latency is rarely considered at all, it doesn't matter if the data arrives 500ms late. For video, that isn't the case, latency is a very big issue. That 500ms delay represents a very annoying glitch on the output.

Here's some numbers to chew on: What we consider high quality standard definition NTSC video is about 50Mbps (that's about 100,000 bits per field for MJPEG, double that for MPEG), TV broadcasts are typically in the 10-15Mbps range. A 5 drive RAID3 (4 data drives + 1 parity drive) array of 73GB Seagate Cheetah Vs (10kRPM) attached by fibrechannel can handle simultaneous record and playback of 2 50Mbps streams, with about 12 hours of record time total (less than that actually, since it's highly recommended that you leave about 10% of the drive free). That doesn't include audio, and I honestly don't know what accompanies the video on that stream, but I do know that there is some vertical synch info added to make editing MPEG less of a PITA.

I don't work on HD stuff right now, but I can tell you that we typically run it at 70Mbps and the RAID described above cannot handle 2 of those streams.

Finally, while I don't know all the details, my company does offer a 4k telecine, and IIRC it uses 16 1Gbps fibreoptic cables in parralell to move all the data around.

nice, but...

[Tumbleweed]

I'd rather see a move towards 1080p (not i, for criminey's sake!), with much higher framerate. Tests by the military showed that figher pilots can perceive framerates up to at least 200fps, and while a successful fighter pilot is almost certainly going to be hardwired to be able to process such information faster, certainly a framerate well over the current 24fps for movies and 30fps for TV (in the U.S.) is desirable. Certainly filmmakers would appreciate being able to pan side to side much quicker than they're able to, without having stop-motion effects all over the place. I think a nice compromise would be 120fps. This is evenly divisible by both 24 and 30 (making for easy downgrades to older formats).

Widescreen 1080p, 120fps. Now *that's* what I'd like to have. And interlaced formats should be banned from the face of the Earth. Suitable only for spammers to view. *bleh*