View previous topic :: View next topic |
Author |
Message |
Rally1970
Joined: 01 Jun 2003 Posts: 3
|
Posted: Sun 1 Jun 2003, 22:16 Post subject: Bitrate peak despite CBR using TMPGEnc |
|
|
Hi!
I have created a short movie in Premiere 6.5 in which I use the noise video effect to generate fake "no signal TV nosie".
Using TMPGEnc (2.510.49.157) to encode a PAL DVD, I get a nasty bitrate peak of 14.1 Mbit/s during the TV nosie scene. Since this peak is way above the maximum bitrate of a DVD, the authoring programs I have tried complains and refuses to create a DVD.
I have tried both CBR and 2-pass VBR but with the same problem. It is easy to understand that the noise scene requires a very high bitrate in order to get good quality but I didn't expect TMPGEnc to go beyond the maximum bitrate setting but instead lowering the quality.
Does anyone have any ideas/suggestions?
These are my settings in TMPGEnc:
Video: MPEG-2, 720x576, 4:3, 25 fps
2-pass VBR with the following limits
min. 2750
avg. 6000
max. 8000
enable padding
P spoilage: 0
B spoilage: 20
VBV automatic
Source: Interlace, bottom field first, 4:3 625 line
GOP: 1-4-1, seq. header = 1, detect scene changes |
|
Back to top |
|
|
RMN Site Admin
Joined: 04 Feb 2003 Posts: 587 Location: Lisboa, Portugal
|
Posted: Sun 1 Jun 2003, 22:41 Post subject: |
|
|
Hm... can't say that has ever happened to me.
Separate the noise segment from the rest and run a few tests just on that part. Namely:
1. If the VBV buffer size is set to 0 (auto), set it to 224. TMPGEnc may be generating frames that have too much information.
2. Don't bother with B-pictures, they take longer to encode and won't work well with random noise. Use I=1, P=14.
RMN
~~~ |
|
Back to top |
|
|
Rally1970
Joined: 01 Jun 2003 Posts: 3
|
Posted: Sun 1 Jun 2003, 22:52 Post subject: |
|
|
Good idea to run some test on the noise scene alone!
Perhaps the "Detect scene change" setting should be turned off? I guess that TMPGEnc detects a scene change between every single frame in the noise scene since it is noise (and in a perfect world the noise is white = no correlation between any frames at all). |
|
Back to top |
|
|
RMN Site Admin
Joined: 04 Feb 2003 Posts: 587 Location: Lisboa, Portugal
|
Posted: Mon 2 Jun 2003, 1:31 Post subject: |
|
|
You can turn it off for this test (thought about mentioning that), but since you'll have to turn it on again when you encode the full movie (unless you feel like setting the scene starts manually, which is a pain), it's not where I'd start.
Changing the video buffer will hopefully force TMPGEnc to behave.
RMN
~~~ |
|
Back to top |
|
|
Rally1970
Joined: 01 Jun 2003 Posts: 3
|
Posted: Mon 2 Jun 2003, 5:44 Post subject: |
|
|
I changed VBV, GOP structure, turned "Detect scene change" on/off etc. The changes didn't really make any difference. Finally, I tried to change the settings for the effect itself in Premiere. By changing the amount of noise from 10% to 100%, the bitrate was lowered to to the CBR level.
Strange! I would have thought that increasing the noise would yield even higher peaks. |
|
Back to top |
|
|
RMN Site Admin
Joined: 04 Feb 2003 Posts: 587 Location: Lisboa, Portugal
|
Posted: Mon 2 Jun 2003, 19:00 Post subject: |
|
|
Either way, it looks like you've clearly found a bug in the way TMPGEnc calculates the average bitrate. I would suggest contacting Pegasys / Hori and letting them know about it.
RMN
~~~ |
|
Back to top |
|
|
|