Film-Tech Forum ARCHIVE: Race to witch mountain hard drive capacity label

This is topic Race to witch mountain hard drive capacity label 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=000329

Posted by Robert LaValley (Member # 4319) on 03-13-2009, 04:46 AM:

this has happened to me once before with a technicolor drive but I got a drive for race to witch mountain that was labeled 111GB for the feature but upon ingesting it was only 99GB. Has this ever happened to you? You would think the size of a file they could get right on the label.
The other thing I found wierd was 99GB for an hour and 30 minute action movie seemed kinda tiny... what do you think?

Posted by Julio Roberto (Member # 4976) on 03-13-2009, 11:37 AM:

What I think is that that's potentially far worse quality due to compression than a (theoretical) 50GB blu-ray disc which normally uses far, far, far, far more efficient codecs for video (h264 vs m-jpeg2k) and audio, and has to do so in a smaller colorspace (4:2:2 8bits).

In that sense, a well done blu-disc could easily have some 30 times less loss of detail DUE TO COMPRESSION than this DCI. It will have more loss of detail due to the format differences between DCI and blu-ray, but the compression loss itself would be orders of magnitude better in the domestic format than in the "professional" format.

A 90m movie "done right" in DCI should take at least 150GB. If it takes much less than that, somebody is throwing away detail unnecessarily.

Even if the detail wasn't there to begin with (i.e. the movie was shot in 16mm or something), it would still not make sense to compress it much more than what's limited by the bandwidth.

It only has an advantage in the amount of time it takes to replicate the hard drives and to ingest the content in the servers, of course.

It would be not possible to notice the quality difference for us mere mortals, so perhaps they figured they could just save the replication time. A replication farm could put out a batch of hard drives with a verified 100GB "print" in about half and hour or less.

But it makes you wonder about how much "superior" a movie in theater at say 1998x1080 4:4:4 12bits at 99GB is compared to some home systems capable of up to 1920x1080 4:2:2 8bits at 50GB, considering the home system codecs of superior efficiency. That's if the aspect ratios were the best possible 1.85 at the theater and the best possible 1.78 at home. If we compare a Scope DCI movie with a HD blu-ray program at home, the difference could be even more in favor of blu-ray.

Posted by Todd McCracken (Member # 4718) on 03-13-2009, 01:13 PM:

Robert,
The 111gb listed was the full DCP including the trailers.

Julio,
The distribution end has no say whatsoever on package size, that is entirely up to the encoding house.

Posted by Lyle Romer (Member # 1266) on 03-13-2009, 01:23 PM:

There are a bunch of "Constant Quality" M-JPEG2000 variable bit rate encoders out there. In theory they are saving bits by encoding frames that don't need them with less bits. In theory the quality will be equivalent to the full 250 MBPS versions. In practice who knows.......I also don't know if they are capable of encoding complex frames with more bits since they are available.

Posted by Todd McCracken (Member # 4718) on 03-13-2009, 01:42 PM:

I seem to remember American Gangster was a whopping 60gigs for a 2 1/2hr feature.
It looked great.

Posted by Julio Roberto (Member # 4976) on 03-13-2009, 02:56 PM:

My motto is: if the bandwidth is available, use it. If you target constant Q and it gives it to you at under your bitrate ... just increase the Q.

But why would you want to do constant Q instead of automatic bit rate? Bitrate is your limit and not Q for this application. Analyze 24 frames, obtain max Q under 250Mbs and under max peak bitrate (audio etc taken in acccount) and spit it out.

For something like Dcinema, where the bandwidth and the storage is available, it would make sense. Slightly Q variations in consequtive frames at such high quality levels shouldn't be apparent to the naked eye anyway, so no need for constant Q.

Even if you decide to do constant Q, if you see your bitrate 50% less than the limit and no frame hit anywhere near the peak limit, just ran another freaking pass with a little higher Q, for pits sake.

American Gangter etc was probably MPEG2 encoded instead of M-JPEG2K, I would guess.

With more efficient compression codecs, like just about any interframe one, you can get smaller file sizes at same quality levels as M-JPEG2K (if they are developed for the same bit depth and colorspace).

But for Dcinema, file size is not such an important parameter. It should be quality of image, as the storage space and the bitrates are available by definition and the only difference is, say, 30 minutes more or less to ingest a movie and a few minutes more or less to replicate the package.

This is just your usual lazyness and saving $1, like when they put out blu-ray transfers that are only (say) 15GB (including extras or whatever), when they have at least 25GB available, i.e.

Of course nobody is going to notice the difference in quality, but since it basically doesn't cost you anything, why compress more than necessary? More compression always equals throwing away more detail. The detail may just be noise (grain) or not visually important, but it was there in the master and you are throwing it away for no good reason.

I can see them heavily DNR'ing the DCI video stream next to save even more space in file size after compression [Wink]

I don't mind that they don't hit the maximun bitrate, but a reduction of more than 30% like we are talking about here for no apparent reason, seems a bit worrying to me.

If we get into conspiracy mode, perhaps some studios are giving Imax a break by lowering the quality of regular DCI and allowing Imax to max out bitrates etc in their own masters to claim better quality than mom-and-pap multiplexes. [evil]