Film-Tech Forum ARCHIVE: DCP size variations...

This is topic DCP size variations... in forum Digital Cinema Forum at Film-Tech Forum ARCHIVE.

To visit this topic, use this URL:
https://ft-forum.com/ft/cgi-bin/ubb/ultimatebb.cgi?ubb=get_topic;f=16;t=000500

Posted by Carsten Kurz (Member # 5396) on 02-02-2010, 06:54 AM:

continued from: 'Avatar's Aspect Ratio Craziness' in the Film handler's Forum:

German DCP sizes:

Bloody Valentine 3D: 154 GB
Ice Age 3D: 147 GB
Final Destination 3D: 140 GB
G-Force 3D: 117 GB
Christmas Carol 3D: 107 GB
Avatar 3D: 154 GB
Meatballs 3D: 148 GB

Anyone with other international versions? Hard to believe that 162min of AVATAR compresses to the same size as IA3 oder Meatballs - and only in germany. Finnland, UK and US report AVATAR with 247 oder 276GBytes. COCOM restrictions? Scenes with high-profile MILgear being replaced by Pandora nature shots?

- Carsten

Posted by Olivier Lemaire (Member # 5579) on 02-04-2010, 03:15 PM:

> continued from: 'Avatar's Aspect Ratio Craziness' in the Film handler's Forum:
(...)
> Anyone with other international versions?
French DCP sizes (for some I could find quickly) :
Ice Age 3D: 148 GB
Christmas Carol 3D: 103 GB
Avatar 2D: 186 GB
Avatar 3D: 155 GB
MeatBalls 2D: 106 GB
Final Destination 2D: 131 GB
G-Force 3D: 120 GB

Hope this'll help.

> Finnland, UK and US report AVATAR with 247 oder 276GBytes.
??? 276 GB for only one CPL ???

Posted by Pietro Clarici (Member # 4937) on 02-04-2010, 03:44 PM:

I can confirm the Italian DCP for AVATAR 3D is 154GB as well.

Still cannot figure out why, I thought the DCDM was the same for every version. Looks like Deluxe UK used a different compression ratio, I'm guessing it's a VBR encoding versus a CBR one. I remember Harry Potter 4 being more or less 70GB VBR.

Maybe they were told to save time (and money) when duplicating Euro DCPs... same reason why our 35mm prints are consistently worse than those released in the US.

Posted by Carsten Kurz (Member # 5396) on 02-04-2010, 04:35 PM:

>??? 276 GB for only one CPL ???

Yes, but even that would be still below the DCI max datarate spec of about 30MByte/s

DCI says: 'For a frame rate of 48 FPS, a 2K distribution shall have a maximum of 651,041 bytes per frame (aggregate of all three color components including headers). Additionally, it shall have a maximum of 520,833 bytes per color component per frame including all relevant tile-part headers.'

That's about 30Mbyte/s max, same as for 2D/24fps. AVATARs 162min would come out as roughly 290GBytes only for the image part at that rate. The mentioned 276GBytes for the US/UK version equal roughly 28MByte/s - just perfect for such a high-profile release.

The german/european 156GB would give only about 16Mbyte/s - just half of the allowed rate. Strange...

I have multiple independent quotes of 156Gbytes.

JPEG2000 itself does not implement some sort of VariableBitrate directly, but of course the encoding could do on a frame-by-frame basis. That's the only explanation I have - the encoding for the US version maxed out every frame to spec, while the encoding for the 'european' version used a more or less constant or conservative sub-optimal compression ratio in order not to peak over at some very few very detailed sequences. But why did they make a different encoding at all?

Any US/UK quotes for AVATAR 2D?

- Carsten

Posted by Julio Roberto (Member # 4976) on 02-04-2010, 04:58 PM:

DCI specs allows for just about any video data rate as long as it never peaks above 250Mbs. If you wanted the maximun quality (minimun compression), you would try to get as close to a constant 250Mbs rate as possible. If you want to save space, you could compress variably with some parts at say 150Mbs and others at 250Mbs, or all of the film at 150Mbs. But you can't gain more quality than keeping a (fairly) constant 250Mbps.

This rate is for the image part of the DCP. The audio/subtitles/etc has its own separate alloted data rate.

Notice that the video rate is INDEPENDENT on whether the film is 2D, 3D (=2K-at-48-fps) or 4K. So 4K and 3D are MORE compressed (i.e. worse quality/more compression artifacts) than plain 24fps 2K.

Why doesn't Hollywood regularly spit out DCP's that are "top quality" (i.e. aprox. 180GB+sound for a 2 hours movie)?

Well, because the data rate is over what's needed to carry the information necessary compressed so that a human eye can't distinguish any more of it.

DCI, i.e., used a 12bits code scheme when the whole xyz color spectrum alloted can be addressed with 11 bits per channel with a few colors to spare. 12 bits was totally unnecessary, not really used, and was only included because 12bits is more "digital friendly" than 11bits and to have some leaway just in case some people in the future can address more colors through genetic engineering [Razz]

.

Even more so, the film scanned to use as Digital Intermediate are usually scanned at 10 bits per channel precision to begin with. It's logarithmic, but obviously only 10 bits of information are saved, so even if you expand to 12bits-linear, which DCI xyz is not as it's gamma 2.6, but even if you did, once you compress it, it's back to nearly "10 bits" of information for a good efficient compression codec.

Think about it. A REALLY good codec can fit a very good 1920x1080x8bits movie, including multi-channel sound/subtitles etc, in about 35GB of data in blu-ray for a 2 hours film. And a lot of the coding is noise (grain etc), which a CGI film like Avatar doesn't need to have as much of. With a DCP, which is IN PRACTICAL terms a very similar 1998x1080x10bits (it's 12bits, but usually only 10 carry real information), ideally it would be 180GB+sound for the same film. That's like 6 times more. Sure the JPEG intra-only codec is innefficient as hell compared with AVC, but six times, is a lot. So usually around 130GB for a 2 hours movie DCP, including the inneficient uncompressed 6 channels discrete sound, is plenty good if you are not striving for perfection.

Since when does Hollywood strive for image perfection when it abandoned film for HD/DCI? [Wink]

Posted by Carsten Kurz (Member # 5396) on 02-04-2010, 05:16 PM:

It still doesn't explain why the US/UK version has 276GByte and the 'european' only 156GBytes. Given the DCI spec it is obvious that 3D features have roughly the same size as their 2D counterparts. If there are signal components that result in 28Mbytes/s for the US version - where are these components in the european version? Compared to the other mentionened 3D DCPs, the 156Gbytes seem more likely to be 'normal' for a 162min feature than 276GByte. Could it be that the latter contains actually two versions in different aspect ratios? Someone in this forum would have mentioned it, wouldn't he, given the discussions about AVATARs different AR versions and the delivery problems this caused for some. They would have had the choice of two different CPLs as well when programming the show.

- Carsten

Posted by Pietro Clarici (Member # 4937) on 02-04-2010, 05:16 PM:

quote: Julio Roberto
(...) the data rate is over what's needed to carry the information necessary compressed so that a human eye can't distinguish any more of it (...)

That makes sense, and I didn't account for the missing grain. But, as Carsten said:

quote: Carsten Kurz
(...) why did they make a different encoding at all?

Why decide to go all the way for the ENG version, only to change their minds and realize that they could release a *much* smaller DCP for every other market? Cost? Or they just want to help us Europeans ingest content faster? [Wink]

Posted by Julio Roberto (Member # 4976) on 02-04-2010, 05:20 PM:

Well, it comes to show that it doesn't matter if it's 2.35 cropped, 1.77 (original Avatar shot), 1.85 cropped, 2D 14fl, 3D 4fl, 30MBps compressed, 16MBps compressed, Dolby color corrected, RealD ghostbusted, IMAX film, digital Imax, 2D 35mm film or 2D digital DCI or whatever, it can still be the #1 box office movie and get plenty of Oscars and Golden Globes.

People just can't even tell the difference. We won't even mention Sony/DLP 4K ...

[Confused]

Anyway. If the version is Scope vs Flat (2D or 3D, doesn't matter), then there are fewer pixels to encode (i.e. flat=1998x1080=2.15 megapixel, Scope=2048x857=1.75megapixel per frame). At least in Spain in 2D 35mm film, only Scope was distributed. It seems in the USA there was a choice between flat and scope depending on which yielded the larger image in a particular theater's screen.

Posted by Carsten Kurz (Member # 5396) on 02-04-2010, 05:30 PM:

Jussi Siponen in the 'Avatar's Aspect Ratio Craziness' thread wrote:

'Our copy of Avatar (scope, with Finnish subtitles) is in the 250GB ballpark in size, which is what I'd expect given the run time.'

As subtitles for 3D have to be rendered/enclosed as bitmaps, not as standard DCI titling, I could imagine that a separate encoding has to take place. The titles themself wouldn't add much, but of course when you do another encoding, you're free to use whatever is allowed in the specs.

AVATAR in germany played only in CS - 3D and 2D/35mm as far as I know.

Hmm. Has anyone seen a 3D feature with added subtitling (I don't mean titles/typo as part of the main feature)?

I think the PNG text/subpicture DCI feature in principle could be used with 3D movies as well. However, you need free screen estate to place subtitles, don't you? You can not simply place them into the bottom part of the image. No matter where you'd put them, they could interfere badly with the stereoscopic arrangement, wouldn't they?

Has anyone seen AVATAR with subtitling? What aspect ratio, and where have the subtitles been placed?

- Carsten

Posted by Julio Roberto (Member # 4976) on 02-04-2010, 05:37 PM:

No doubts that if you decide to use the highest possible quality, ANY dcp film that long would take over 250GB no matter if it was 3D, 2D, flat or scope (or theoretical 4K for that matter).

So, obviously, it was compressed more in several markets, including most of Europe, it seems.

As to why aren't most DCP movies maxing out the available data rate to guarantee the maximun possible quality? Who knows. Obviously using smaller hard drives and replication times is a small bonus, but hardly worth the money. To avoid hickups during playback? Sure, but you could stay just below your max rate by about say 10% instead of a consistent 40% used today.

I don't have an answer other than "they can't really put more information on the frame, so why waste the space?"

Posted by Carsten Kurz (Member # 5396) on 02-04-2010, 07:35 PM:

Leaving out the question why they recompressed the european versions at all: It could indeed have been a completely different datarate control scheme:

The european version may have been compressed with constant compression parameters that make sure that at no point in the movie the maximum datarate of 30Mbyte/s is exceeded. Which means they waste bandwith budget/quality in most other parts of the movie - resulting in an average datarate of 16Mbyte/s - half of the max allowed.

The US version with it's 276Gbyte computes down to 28MBytes/s. That's average. And since no more than 30MByte/s is allowed, it means that no significant part of the movie can have either much less or much more than that. And that means, they must have optimized more or less each and every frame to the allowed maximum of about 651,041 Bytes per frame. That means monitoring and adjusting of compression parameters for every frame. That's advanced and means some added effort and, depending on encoding software, a lot more compression time. But for a high-profile feature film like AVATAR, I would expect nothing else than that.

If the US DCP really contains only a single main feature with MXF tracks of the same length as the european version, that's the only explanation I could think of.

- Carsten

Posted by Elise Brandt (Member # 5552) on 02-05-2010, 08:37 AM:

We're running a Finnish subtitled version, the subtitles are located on the bottom of the picture, on top of the picture. They don't interfere much with the 3D because they're coded exactly like the english subtitles used in the film.

I have seen an example of a 3D-film that used subtitles the cheap way -just splat it on as a "normal" added package as you do in some 2D-films. Ouch. The result is like trying to focus your eyes back and forth between a wall thirty metres away and the palm of your hand placed four inches from your face. Headache and nausea waiting to happen...

Just checked; our Avatar is only 149GB, which is odd if the version Jussi mentioned is 250? Same projector, as far as I know.

Posted by Jussi Siponen (Member # 5572) on 02-05-2010, 10:02 AM:

I admit I gave the size by memory -- Elise on the other hand has an exact number: consider 149GB to be the correct size for the Finnish subtitled version of Avatar. I should know better than to trust my memory by now... mea culpa.

Re: subtitle location, I think Elise is trying to say the subtitles are at the bottom-center of the picture (x/y axis) and in front of the picture (on z axis).

Subtiles are usually approximately at the screen plane, but they do move back and forth depending on what else is happening in the movie. The trick seems to be to have the subtitles at same depth or in front of at least the primary subject of the picture. Sometimes the subtitles are temporarily relocated from their default bottom-center position if something is "crawling into your lap" at the bottom.

The projector built-in subtitle system can only render on the screen plane, so subtitles can end up "drilling a hole" into what ever happens to be behind them in the picture. Like Elise said, this is extremely unpleasant to say the least and can be avoided by having properly placed (on z-axis) subtitles rendered in the content itself.

Posted by Elise Brandt (Member # 5552) on 02-05-2010, 10:25 AM:

Thank you Jussi. For the translation Elise=>English [Big Grin]

not nearly enough coffee today it seems.

Posted by Carsten Kurz (Member # 5396) on 02-05-2010, 11:14 AM:

Yep, I thought that added subtitling is a lot more complicated with 3D features. Their placement would really have to follow the main subject focus plane. I can not even imagine how much work that means to figure out on a scene-by-scene basis.

And I don't think the production will give something like a 'depth-cue-sheet' to prepare for the titling. Although they are certainly wise to think of something like that in the future.

Thanks for the updated DCP size. Makes more sense.

- Carsten