Post-Next-Gen Disk?

[audiohominis]

Hi, guys. I’m a layman, so I may assume some silly things. Please be kind. :o

I hear there’s a new optical storage device in development, which will mark the terabyte era. One can speculate that with such tremendous capacity for video content at their disposal,
studios and developers will go in either of the following directions:

They will stick to the current HD resolution standard (1920x1080p), but they will see their way to delivering high-quality lossless video to the audience thus finally allowing them to enjoy a picture that is free of compression artifacts, segmented gradients, and generally the shortcomings that result from forced trading of quality for speed and compactness. This would be possible provided that the processing power of the players will grow along, which is more than likely.

Or... they will push the resolution even higher.


The way I see it direction B wouldn’t make much sense. Why? Because aside from fully CG animated films and such, (which can be rendered with any synthetic level of detail), given current cinema technology, there really is no motion picture camera in existence capable of capturing enough detail to necessitate even higher viewing definition. This includes 35mm film as well as high-end professional digital camcorders. Even a 70mm IMAX film is NOT too sharp for 1920x1080p to handle. There’s a limit to how much detail can be extracted from film when digitizing the frames. Once you reach that limit, no matter how high you scan all you get is bigger grains and nothing more. That’s why I think they should focus on making lossless video available to the consumer. What great time it would be then, ha? :)

Would anyone agree, or disagree on this? Maybe there is some new info on this somewhere on the web that I’m not aware of.

Cheerios!

[Doom9]

I've already reported about 4K resolutions.. so you know where things are going - but it's a long time off. The move to HD is going to keep us busy for at least another decade.
And at that point, we'll have to see where we're at bandwidth wise.. perhaps the next generation bypasses physical media altogether.

In try imagining the studio's paranoia about lossless content - BD+ will probably look like kindergarden compared to what the studios will cook up to protect lossless content.

And with today's disc formats, a good compression job often results in no visible artifacts.. if you can't see the loss, there's not much point in lossless anyway (highdefdigest just ran a lengthy editorial about the point or lack thereof of lossless). Lossloss mostly makes sense if you're looking at re-encoding and that's the part studios don't want us to do anyway...

[Shinigami-Sama]

by that time I expect consumer to get fed up with all DRM drek and form some sort of united front against it

then again I'm also expecting 3500 years of tyrannical rule...

[Soulhunter]

Uhm, iirc they already finalized the specs for UHDTV [7680×4320]. Japan plans to start airing it ~2025, BBC and EBU want to follow later. And I read Sony already has some TV prototypes in their labs that can do UHDTV @ 60Hz... Guess we will have enough demand for this "tremendous capacity" without going lossless!


Bye

[neuron2]

Uhm, iirc they already finalized the specs for UHDTV [7680×4320]. Japan plans to start airing it ~2025, BBC and EBU want to follow later. Where are they planning to get the transmission bandwidth for broadcasting this? Just hoping some great new compression scheme will be developed by then? Or will it just be for local disk drives?

[Shinigami-Sama]

Where are they planning to get the transmission bandwidth for broadcasting this? Just hoping some great new compression scheme will be developed by then? Or will it just be for local disk drives?

last I head Telus(Canadian ISP) is trying to bring 'fibre to the house', which really isn't to the house but to the neighbourhood, where the houses will hook up to it VIA 10gigE cat6A

I don't have the sauce on hand though so take it with some salt

[audiohominis]

Uhm, iirc they already finalized the specs for UHDTV [7680×4320]...
Holly cow! Is it for displaying shows shot with... alien cameras or something?

[salehin]

The other very important factor is the rates at which the cost of bandwidth and hidef hardware (, both computer and non-computer based,) go down... without their cost going down, and thus making it accessible to many end users (both computer and non-computer oriented users,) this venture will not be sustainable in the long run

[Doom9]

Is it for displaying shows shot with... alien cameras or something?Studios already make 4K transfers today.. don't think for one second that 1080p is the end of the line. 4K is still not mainstream but today's movies are transfered in 2K so that's still twice the horizontal resolution of today's HD stuff.

[Morte66]

I'm pretty happy with 1920x1080 at my customary viewing distances of 1.25 to 2x the diagonal. I wouldn't be that excited about a resolution increase.

What I would like to see is:

- An end to film grain at source, except as VFX (e.g. grainy flashback scenes). Putting the Matrix HD DVD rip through mvdegrain2 at 1fps is a pain in the arse.

- An increase in frame rate for smoother motion, less motion blur, and an end to the need to limit creative content to what 24p can handle.

- Greater bit depth. 8 bit video bands on a contrasty modern display with PC gamma, whichever way you slice it. Only adding noise (be it proper dither or film grain or basic noise) will help you. 8 bits was enough for consumer use when DVD was created but it isn't now.

- An end to colour subsampling and those 2x2 chroma blocks, which soften the picture and accentuate banding/blocking issues.

[Soulhunter]

Where are they planning to get the transmission bandwidth for broadcasting this? Just hoping some great new compression scheme will be developed by then? Or will it just be for local disk drives?
From the SMPTE Journal...
...an optical transmission system with dense wavelength division multiplexing (DWDM) technology for contribution, and a codec system based on MPEG-2 for program distribution. A live transmission of uncompressed material with a 260-km single-mode fiber was also performed. For program distribution, some experiments were conducted at bit rates of 200 to 640 Mbits/sec via experimental satellite transponders or IP networks. These transmission experiments demonstrate the feasibility of using UHDTV in future broadcasting...

Bye

[foxyshadis]

So I guess they just don't expect to broadcast terrestrially or via satellite? Unless they expect to find a way to get signals in the hundreds of GHz or THz out there. Wow.

[Doobie]

They can put about a TB on a disk, but they're not about to pull the rug out from under the current standards.

I don't see a need for higher resolution. There would be little return of image quality for the bandwidth cost. But, I'd like to see higher frame rates and 3D.

[Mug Funky]

UHDTV looks a bit overkill to me... what possible gains in the movie experience would there be? do we go to cinemas to see sharp edges or a good movie?

besides, the current state of the art in film scanning is around 6k for super 35, and generally you can't do a DI on 6k frames without breaking the software. 4k is currently the max for that, and 2k is enough for most cinemas (film can hold more detail, especially print-stocks which aren't limited by exposure time like camera neg is, but in the end the projector in the cinema is almost never focused properly).

also, just pushing 2k through cables is a challenge. the full 2k 2048x1556 has to be run at a slower framerate to be viewable at all, though 2048x1152 (16:9 cropped) can run at 24fps.

*sigh* i wish i had a spirit 4k to play with. those are good, though even they can't do 4k in realtime (yet... in theory they can be retrofitted).

[edit]

@ Morte66:

- film grain is very very low on slower film stocks, like 50D or 64D, or even fuji's new 160T. of course the lighting budget increases if a DoP chooses to shoot on these films. movies these days are very grainy because faster films allow more freedom to shoot in other ways (like handheld, or available light, or documentary-style, or for no disceernable reason like in spiderman 3). also, some movies are over-graded or underexposed in parts, and this can push the grain up a fair bit.

- any decent camera can be overcranked to quite high framerates (around 75-90fps), and special cameras can go all the way up to around 300fps. all you need is fast film and lots of light.

- DVDs have been (mainly) mastered in 10-bit since the format's inception. the banding tends to come in at the decoding end, or in the TV during YUV to RGB conversion (some screens use 7-bit!).

chroma subsampling is annoying, i agree. but if the movie is interesting enough you wont see it :)

[Morte66]

@ Morte66:

- film grain is very very low on slower film stocks, like 50D or 64D, or even fuji's new 160T.

"Very low" is relative, it's compared to older film. It's still bloody noisy.

Put it this way, on still photographic film I've only ever gotten clean 1920xnnnn out of KI50, which was far and away the cleanest chromogenic colour neg film on the market, and that was with twice the frame area "35mm" cinefilm uses.

Bring on 3 chip digital with twinned light/dark sensors per pixel for film-like dynamic range. :)

Any decent camera can be overcranked to quite high framerates (around 75-90fps), and special cameras can go all the way up to around 300fps.

They can, but they aren't. I'm asking for an industry to change its ways here, not a technical thing. [Apparently the technical people wanted to go 60p for the US HDTV launch, but for some reason ABC TV executives dug their heels in over 24p, and it has pretty much knocked on since then.]

[Mug Funky]

"Very low" is relative, it's compared to older film. It's still bloody noisy.

not sure what you mean there. vision2 50D is pretty recent, and vivid 160T is brand new (out a few months ago i think). i haven't really seen many older stocks, though i've seen enough to know they're generally grainier than current ones of the same speed.

anyhoo, the 50D is pretty much completely grainless off the scanner, and barely shows any after heavy "stretching" in DI. it's hard to tell whether it's just the dither down to 10 bits in this case. "very low" here means below the 8-bit noise floor in some cases - it's hard to observe more than that as the eyes and the monitoring equipment reach their limits here.

but this is a little OT i think.

Bring on 3 chip digital with twinned light/dark sensors per pixel for film-like dynamic range.

check out the arri D-20. comes pretty close, but not there yet. if you underexpose to save the highlights, you get weird streaky sensor noise in the blacks. you need a really good DoP and a lot of lighting to get the same result as on even 250D. say goodbye to outdoor shoots on anything but overcast days.

i'm not anti-digital, just saying it's not quite caught up yet. the next few years will be very interesting. i suspect quality will follow convenience here - once the data can be moved around quickly and securely enough, then people will be able to shoot digital cheaper than film, and the quality will follow. right now, though it's counter-intuitive, high-end digital is more expensive and lower quality.

[Soulhunter]

So I guess they just don't expect to broadcast terrestrially or via satellite? Unless they expect to find a way to get signals in the hundreds of GHz or THz out there. Wow.
Japan is assigned with the 21-GHz band for future satellite broadcasting. If UHDTV broadcasting were to begin, satellite broadcasting would be feasible due to availability of a wide bandwidth...

UHDTV looks a bit overkill to me... what possible gains in the movie experience would there be? do we go to cinemas to see sharp edges or a good movie?
Looks like overkill... But its a big step closer to the resolution a human eye works at... And if it technically possible, why no!? :]



- film grain is very very low on slower film stocks, like 50D or 64D, or even fuji's new 160T. of course the lighting budget increases if a DoP chooses to shoot on these films. movies these days are very grainy because faster films allow more freedom to shoot in other ways (like handheld, or available light, or documentary-style, or for no disceernable reason like in spiderman 3). also, some movies are over-graded or underexposed in parts, and this can push the grain up a fair bit.
True! Shooting night scenes for example is rather difficult when using fast film material... They have to use much more artificial light which could make the scene look unnatural, plus it costs more money! So they probably pick faster film which results in a higher amount of visible grain... And if they use it for low-light condition scenes, they probably use it for normal light condition scenes as well... Coz otherwise the "look" of the movie would change to much from scene to scene!


Bye

[audiohominis]

I've already reported about 4K resolutions.. so you know where things are going - but it's a long time off. The move to HD is going to keep us busy for at least another decade.
Yeah, I kinda know that digital intermediates are 4K, 6K scans. They’ve been doing this for years. That’s logical because you always want to make sure to have some extra resolution for post-production and editing (better more than not enough), not to mention that eventually it is all printed back to film, so you want to minimize the loss during that conversion process too. However, again, when comparing theater to HD, as far as sharpness is concerned, I really don’t see from film the level of visual information detail that a 1920 by 1080 image couldn’t carry. This even applies to the crisp, perfectly in-focus objects in the frame that were, like… in the very eye of camera’s focal center of interest when they were photographed. Again, even in 70mm IMAX camera footage the edges of the sharpest objects in the focus have somewhat thick and soft anti-aliasing.
If you care to get this notion out of my system, all you have to do is upload/link to a sample 4K frame scan from a movie, which I could reduce to 1920x1080 (with different down sampling algs) and see some effective loss of detail in certain areas. So I could say “Yeah, Doom9 was right; after re-size, I can no longer read the letters in that book on the desk” or”… I can no longer count the tiny squares on that actor’s checkered shirt” or something.:)

[Soulhunter]

If you care to get this notion out of my system, all you have to do is upload/link to a sample 4K frame scan from a movie, which I could reduce to 1920x1080 (with different down sampling algs) and see some effective loss of detail in certain areas...
Uhm... You have a screen capable to display 4k? Because otherwise a fair comparison between 4k and 1080p wouldnt be possible imo... Still, if you search around enough you should be able to find some 4k [or close to 4k] film scans via google!

However, again, when comparing theater to HD, as far as sharpness is concerned, I really don’t see from film the level of visual information detail that a 1920 by 1080 image couldn’t carry...
True in most cases imo... The problem is that theaters usually use sucky x-generation copys, made from negatives which were made from positives etc... Usually this will result in a virtual resolution much lower [could be even below 2k] than what the original film negatives contain [up to 4k, or even more...]. Lets hope UHD-DVDs [or UHD-HoloDiscs or whatever, lol] will use the masters as source, not some bad copy... Otherwise we wont gain anything! :\


Bye

[audiohominis]

Uhm... You have a screen capable to display 4k? Because otherwise a fair comparison between 4k and 1080p wouldnt be possible imo...
I don’t think it’s really necessary to have a monitor with enough dots to encompass the entire image at 1x1. To make an adequate decision on whether there’s actually more detail, all I’d need is to see whatever portion of the image is within my display’s range as long as I view it at 1x1 zooming, of course.
The problem is that theaters usually use sucky x-generation copys, made from negatives which were made from positives etc...
So, I am to understand that only the master is a thorough frame-by-frame laser print with a lot of love invested in it, but the reels distributed to theatres are merely quick and dirty copies of that master? On one hand, I’m thinking this wouldn’t make much sense because you wana deliver max quality to the viewer. On the other hand, I don’t know how time consuming and expensive it would be to laser-print every single reel, so it kinda does make sense. However, I’m pretty sure that DVDs and other movie disks are encoded transfers of the digital master, not a re-scan of the print of the digital master. In fact, there are specific titles that I’m sure this applies to.
if you search around enough you should be able to find some 4k [or close to 4k] film scans via google!
Trust me, I did, but I couldn’t find one. Sometimes you come across things by accident that you couldn’t find if you looked for them.

[um3k]

Hasn't anyone here ever heard of the Red One? http://www.red.com/

[linx05]

Um3k, I have. I drool over the thing. My friend who is in my production company has bought one. It will probably be shipped in a couple of years but hey, it still has his name on it.

[audiohominis]

Hasn't anyone here ever heard of the Red One? http://www.red.com/
Yeah, I knew about this one, but never saw footage shot with it. Downloading a sample now.

EDIT: The milk girls footage is superb. Also, it made my hardware scream for mercy because very little compression was applied. 62.68 mbits/sec is quite generous; especially considering that there's no audio whatsoever and that this is a relatively low-motion steady cam shot. 11 sec footage at 24fps is almost 90mb. The next question burning in my head probably belongs in more cinematography oriented message boards.
EDIT:
It combines low noise with superior charge capacity for dynamic range and color fidelity that allows you to finally have a no excuse digital imaging alternative to shooting 35mm film.I think that statement is still too early.

[zambelli]

- DVDs have been (mainly) mastered in 10-bit since the format's inception. the banding tends to come in at the decoding end, or in the TV during YUV to RGB conversion (some screens use 7-bit!).
Actually, the banding is typically introduced during preprocessing or encoding when converting the 10-bit 4:4:4 or 4:2:2 master to 8-bit 4:2:0 video that MPEG-2, H.264 and VC-1 use.

[linx05]

The Red One even does slow-mo!

EDIT:
I think that statement is still too early.
I don't think so. Some of the samples you see have been captured using pre-alpha versions.

[Morte66]

If you care to get this notion out of my system, all you have to do is upload/link to a sample 4K frame scan from a movie, which I could reduce to 1920x1080 (with different down sampling algs) and see some effective loss of detail in certain areas.


ftp://vqeg.its.bldrdoc.gov/HDTV/SVT_MultiFormat/2160p50_CgrLevels_Master_SVTdec05_/

3840x2160p50 frame scans from 65mm film, lossless, view with xnview. Not from a movie, but hey. There's 90GB of it if you want.

[audiohominis]

Oh, this does count as movie alright! A fine example it is (It ain’t Arriflex 765 by any chance, is it?).
As far as I can see, the frames in this particular shot support my point about the detail thing.

PS. It’s kind of tempting to save that entire sequence because it’s lossless and it features firm-breasted athletic women.

[Shinigami-Sama]

Oh, this does count as movie alright! A fine example it is (It ain’t Arriflex 765 by any chance, is it?).
As far as I can see, the frames in this particular shot support my point about the detail thing.

PS. It’s kind of tempting to save that entire sequence because it’s lossless and it features firm-breasted athletic women.

which means it'd be good for testing different quality profiles eh? ;)

[Morte66]

Oh, this does count as movie alright! A fine example it is (It ain’t Arriflex 765 by any chance, is it?). [/SIZE]

Browse up that tree and you'll find a PDF document somewhere with full details of the camera/film/scanner. There's a tool called sgi2yuv that will convert the frame scans to yuv 4:2:0 so you can see the effects of colour subsampling. You can also read the YUV into AviSynth with the RawSource() plugin to encode it with x264 etc.

[I downloaded all the 1080p versions some time back and converted them to 4:2:0 (about 22GB -> 7GB) then compressed that with lossless ffv1, so it now occupies 3.91GB on my hard drive. It never turned out to be as useful as I expected -- ultimately I mostly encode from DVD, and blocking + flicker + MPEG2 crap are a bigger problem than film grain. More recently I've denoised some HD DVD (after they went to all that trouble retaining the grain I hate), and it has rather different grain characteristics from whatever SVT were using. But for the purposes of this thread -- what post-nextgen could/should do -- it's interesting material.]

[audiohominis]

It combines low noise with superior charge capacity for dynamic range and color fidelity that allows you to finally have a no excuse digital imaging alternative to shooting 35mm film. I think that statement is still too early. I don't think so. Some of the samples you see have been captured using pre-alpha versions.
What I really meant to say is that “no excuse” is a pretty strong claim for any digital camera company to make, unless their product offers absolutely everything film does and more.