Welcome to the new Film-Tech Forums!
The forum you are looking at is entirely new software. Because there was no good way to import all of the old archived data from the last 20 years on the old software, everyone will need to register for a new account to participate.
To access the original forums from 1999-2019 which are now a "read only" status, click on the "FORUM ARCHIVE" link above.
Please remember registering with your first and last REAL name is mandatory. This forum is for professionals and fake names are not permitted. To get to the registration page click here.
Once the registration has been approved, you will be able to login via the link in the upper right corner of this page.
Also, please remember while it is highly encouraged to upload an avatar image to your profile, is not a requirement. If you choose to upload an avatar image, please remember that it IS a requirement that the image must be a clear photo of your face.
Thank you!
That's what I was thinking about and that's why I suggested looking at the video card. I did not understand that all the graphics processing was done in the CPU, not the graphics card.
There's an option for DCP-o-matic that will offload some of the processing (the colour space transform) to the GPU but the JPEG2000 decoder is CPU-based. It could (and probably should) make use of CUDA/OpenCL for doing that... maybe one day (nvidia released some libraries to make this easier, but supporting that is somewhere in the middle of a long to-do list...)
I've just discovered that this computer can play a 2k trailer using the dcpomatic player with no frames being dropped. But playing a 4k trailer causes it to become very jerky until I tell it to decode at half resolution. Then it's fine.
And my former computer couldn't do any of this at all.
This stuff obviously takes a lot of computer horsepower to accomplish.
Here's the question:
A cinema server has never struck me as being a real screaming machine. At least, the boot up sequence and the user interface aren't what I'd call super snappy.
So what's the fairy dust that gets sprinkled onto cinema servers so they can play dcp's more easily than a general purpose computer with (I assume) higher specifications?
Almost, there, but already here: Happy new year!
The bid difference between a DCI screen server and DCP-o-matic with a personal computer is that the screen server is not using open source software that relies on CPU. It uses proprietary decoding.
And the fairy dust is called application-specific integrated circuit or Field Programmable Gate Array.
An example (you might recognize the logo): https://www.intopix.com/2K-4K-cinema...ncoder-decoder
Last edited by Ioannis Syrogiannis; 12-31-2024, 06:52 PM.
Reason: Editing for spelling.
But playing a 4k trailer causes it to become very jerky until I tell it to decode at half resolution.
Don't forget that screen resolution vs. processing power is an inverse square relationship. Double the screen size and the amount of work needed to complete the render/display of one frame multiplies by four. Cut it in half and it only takes one quarter the work.
Don't cinema servers/projectors do all their processing in hardware, specifically tailored for the job?
I've just discovered that this computer can play a 2k trailer using the dcpomatic player with no frames being dropped. But playing a 4k trailer causes it to become very jerky until I tell it to decode at half resolution. Then it's fine.​
If you haven't already, you could try the "OpenGL (faster)" option for "Video display mode" in the player preferences. Then, at least, if your display is less than 4K the downscale would be done by the graphics card. The limit you're hitting could be the CPU J2K decode speed, in which case I have no good answer (yet!)
On speed, the usl cms series did video decryption. Jpeg decoding. And forensic marking in an fpga. All audio stuff was handled in The CPU (4 core arm,). Fpgas are pretty amazing. The one we were using has 1200 pins.
On speed, the usl cms series did video decryption. Jpeg decoding. And forensic marking in an fpga. All audio stuff was handled in The CPU (4 core arm,). Fpgas are pretty amazing. The one we were using has 1200 pins.
If you haven't already, you could try the "OpenGL (faster)" option for "Video display mode" in the player preferences. Then, at least, if your display is less than 4K the downscale would be done by the graphics card. The limit you're hitting could be the CPU J2K decode speed, in which case I have no good answer (yet!)
I just tried this setting and it doesn't seem to make much difference. The 4k playback is jerky and stalls out within a few seconds of starting. But the half-resolution setting plays back perfectly.
Ultimately it doesn't matter much as far as I'm concerned since I just have a 2k projector.
I just finished making my first dcp of a still image of my theatre logo. I'll be doing my test screening of Moana sometime after lunch when I get around to it so I'll tack it on there and see how it looks. I'm looking forward to seeing that more than the movie.
Then after that I guess I'll have to start reading up on sound. I don't want to rip the roof off with boing or a bang but I still want to be able to hear it. I'm wondering if there's a way to get a readout of the sound levels in a "real" trailer so I have some numbers to use for a pattern.
I can see where this dcp-o-matic is going to be a lot of fun to play with.
I just tried this setting and it doesn't seem to make much difference. The 4k playback is jerky and stalls out within a few seconds of starting. But the half-resolution setting plays back perfectly.
Ultimately it doesn't matter much as far as I'm concerned since I just have a 2k projector.
I just finished making my first dcp of a still image of my theatre logo. I'll be doing my test screening of Moana sometime after lunch when I get around to it so I'll tack it on there and see how it looks. I'm looking forward to seeing that more than the movie.
Then after that I guess I'll have to start reading up on sound. I don't want to rip the roof off with boing or a bang but I still want to be able to hear it. I'm wondering if there's a way to get a readout of the sound levels in a "real" trailer so I have some numbers to use for a pattern.
I can see where this dcp-o-matic is going to be a lot of fun to play with.
Yeah there is a bit of a learning curve to adjusting for sound/levels when making DCPs. Often stereo video files will come out very hot if you don't adjust. The analysis and graphs are your friend, as well if you have time to make it twice and preview it... you can just play it in the auditorium, adjust the fader, and then go back into DCP-o-matic and use the features that allow you to specify what fader you had to play it at, and it will adjust to give you a 7.0 output equal to that SPL. Then there are the features to generate a smart 3.0 mix from a 2.0 mix etc... best for heavy dialog clips and non-feature content since it's a guess at a matrix mix.
DCP-o-matic offers a very useful LUFS and LEQ(m) analysis to hit the right ballpark. Easy to memorise numbers are -20LUFS or about 80dB LEQ(m). Depending on your room and leveling habits, this may need some fine adjustment, but these numbers never failed for me. They may be off for full length features, as there is no single number that will express full length feature loudness so that humans will comply with it. But for trailers, shorts, music, etc., these numbers will bring you into the right ballpark.
If not enabled, you need to activate 'Find integrated loudness, true peak and loudness range when analysing audio' in prefs. But it should be enabled by default in a fresh install. You then need to select the content/audio in the content window and click on 'Audio' - >'Show graph of audio levels' - you will see LUFS and LEQ(m) numbers at the bottom of the audio graph window that opens.
You may load DCP trailers and compare them - they will usually be much louder. If have seen some trailers hitting -14LUFS and 85dB LEQ(m) (which meets the TASA specs precisely, which is probably what one should expect from a high profile action movie trailer). But it depends a bit on your preshow or main feature leveling habits and wether the DCP you are trying to create should be heard with an impact or rather softly. Keep in mind, LUFS are negative numbers, so, -14 LUFS is 6dB LOUDER than -20 LUFS, and 85dB LEQ(m) is 5 dB louder than 80dB LEQ(m).
Note that both LUFS and LEQ(m) are not target levels, but complaint levels. You should not always aim to reach them with every content, but once you go near them, you are sure to get some complaints by your patrons.
Last edited by Carsten Kurz; 01-01-2025, 05:04 PM.
Tweet
Comment