Max bitrate question

[brikin]

I am using .50a and am getting a max bitrate of 7200 for progressive sources. Is this correct? In the change log for this build it talks about the max bitrate for interlaced sources.

Thanks

[The_Flash]

I'm a little confused by this too. I've recently processed several titles which are progressive and I've noticed that the max bitrate being used is 7200. If anyone can shed some more light on this topic I'd be interested as well.

[jdobbs]

Originally posted by brikin
I am using .50a and am getting a max bitrate of 7200 for progressive sources. Is this correct? In the change log for this build it talks about the max bitrate for interlaced sources.

Thanks I'll look at it. Are you sure you haven't modified the max bitrate variable in the INI?

[jdobbs]

DAMN!!! I'm amazed no one noticed this before. I had the variable backwards... it was reducing the bitrate for progressive and not for interlaced (which is where it is needed). That would seem to indicate that the bitrate stutter was never fixed... but amazingly enough it has been reported to have gone away.

[The_Flash]

One more point for the placebo effect.

Thanks for taking care of this jd

[krispy]

I've noticed it.. been using RB Keeper to set max bitrate to 9000

Any chance that DVD-RB can detect max bitrate for each cell & set it to the max bitrate for CCE on a per cell basis?

[jdobbs]

Originally posted by krispy
I've noticed it.. been using RB Keeper to set max bitrate to 9000

Any chance that DVD-RB can detect max bitrate for each cell & set it to the max bitrate for CCE on a per cell basis? I may be a little confused. The maximum bitrate would be the same for the entire DVD.

If you're asking that the adjustment (for NTSC interlaced sources) to maximum bitrate be assigned at the cell level -- it is being done that way now (in v0.51 -- which should be posted tomorrow).

[krispy]

RB assign the peak to each cell & set the value from the orginal source.

eg


| Min. | Avg. | Peak^
--------+-------+-------+-------
Cell 1: | 0 | 5446 | 7341
Cell 2: | 0 | 5980 | 7680
Cell 3: | 0 | 5705 | 7603
...
...


^ Peak value is from bitrate viewer off orginal. If RB could do this it be great.


I've just encoded the first 2 cells with a peak set @ 9000 and compared ran it through bitrate viewer & compared it to the original peak



| Orig. | RB
| Peak | Peak
--------+-------+--------
Cell 1 | 7341 | 7554
Cell 2 | 7680 | 8821



I don't know if there is any benifit in keeping the orginal peak value when re-encoding. But I've always set my peak to be the same as the original..

[Amenophis]

Originally posted by jdobbs
I may be a little confused. The maximum bitrate would be the same for the entire DVD.


yes, the complete stream has a maximum bitrate, but not every cell or title has the same amount of audio streams. so the maximum bitrate for video only could change.

[jdobbs]

@Krispy

What you are discussing is not "maximum bitrate" but instead "peak bitrate" -- sounds alike but it isn't. "Maximum bitrate" as we are discussing it here is the setting that gets passed to CCE limiting the top bitrate that should ever be present in the stream (along with minumum bitrate and average bitrate). The encoder then uses those limits when it generates the MPEG stream. Since one of the goals of VBR is to allocate only the bits needed and distribute it evenly based on demand across the stream -- you will normally not see a number that high anywhere when you review the generated MPEG (using bitrate viewer for example).

The "peak bitrate" is what you might see when reviewing the file with bitrate viewer. You might very often see a case in which the maximum bitrate is 9000 and the peak never exceeds 7000 -- because it wasn't calculated as needed.

DVD-RB doesn't assign "peak bitrates" -- they are a result of the VBR encoding process. They will naturally follow (generally) that of the original since it was also VBR, but doesn't necessary match it. CCE is a very good encoder -- probably the world's best -- so I don't think I would question the way in which it allocates bits.

There really is no benefit to setting the maximum bitrate to match the peak in the original. When the source is reencoded it is best to let CCE decide how much and where to allocate bits. Changing the value to a lower one will probably hurt the picture and not affect the size at all ("average bitrate" affects the size)

@Amenophis

There is a DVD standard for maximum video bitrate and maximum overall bitrate. If the video stays within it's limits the overall will as well. The audio/subpicture streams of the original had to live within the same standards.

The maximum bitrate shouldn't change between each cell. If you do that you can limit the quality of that cell. The "peak" will change based upon demand. Maximum bitrate has a lot less impact than you might imagine. The "average bitrate" is the far-and-away most affective driver, and the "minimum bitrate" is the second most affective.

[jdobbs]

Here's an experiment to emphasize the point I'm making:

Create an AVS with this content:

Trim(ColorBars(720,480),0,10000))
ConvertToYUY2()

Then do a multipass VBR encode of it using CCE with:

Minimum bitrate=0, Average bitrate=4000, Maximum bitrate=9000

You will see that although the maximum is set to 9000, the peak never goes above 1000...

[Noah]

Originally posted by jdobbs
DAMN!!! I'm amazed no one noticed this before. I had the variable backwards... it was reducing the bitrate for progressive and not for interlaced (which is where it is needed). That would seem to indicate that the bitrate stutter was never fixed... but amazingly enough it has been reported to have gone away.

I noticed back here (http://forum.doom9.org/showthread.php?s=&postid=499823#post499823) but I hadn't done anything interlaced since then, and I thought the bitrate stutter was an issue on progressive sources as well? :confused:

I've seen bitrate peaks/set top DVD player stutters with progressive sources...

[jdobbs]

Originally posted by Noah
I noticed back here (http://forum.doom9.org/showthread.php?s=&postid=499823#post499823) but I hadn't done anything interlaced since then, and I thought the bitrate stutter was an issue on progressive sources as well? :confused:

I've seen bitrate peaks/set top DVD player stutters with progressive sources... It would be if the bitrate went above the maximum allowable by DVD. But that shouldn't happen when it is set to 9000. Notice I say shouldn't not doesn't. I've never seen it on any I've done -- but there is more on earth than my encodes.

What I know could happen is this: I encode all sources at a constant framerate (either 25fps or 23.976fps). I then reinstitute the original framerate upon rebuild. If a 29.97fps source gets encoded at 23.976 and there was a point in the film where a bitrate of 9000 was realized (rare but possible) -- it would become 10800 when converted/played at 29.97fps. That's what had to be corrected.

You can still manually set the maximum bitrate for progressive. But it really shouldn't be necessary (again "shouldn't" doesn't necessarily mean "isn't").

[Amenophis]

Originally posted by jdobbs
It would be if the bitrate went above the maximum allowable by DVD. But that shouldn't happen when it is set to 9000. Notice I say shouldn't not doesn't. I've never seen it on any I've done -- but there is more on earth than my encodes.


for most of the dvds 9000 kbit/s for the video should be ok. but the total bitrate has to be below 9800 kbit/s, too. so if you have 2 dd5 and 1 dts-stream (=1664 kbit/s) everything above ~8100 kbit/s could result in stutter.

[Noah]

Originally posted by jdobbs
It would be if the bitrate went above the maximum allowable by DVD. But that shouldn't happen when it is set to 9000. Notice I say shouldn't not doesn't. I've never seen it on any I've done -- but there is more on earth than my encodes.

What I know could happen is this: I encode all sources at a constant framerate (either 25fps or 23.976fps). I then reinstitute the original framerate upon rebuild. If a 29.97fps source gets encoded at 23.976 and there was a point in the film where a bitrate of 9000 was realized (rare but possible) -- it would become 10800 when converted/played at 29.97fps. That's what had to be corrected.

You can still manually set the maximum bitrate for progressive. But it really shouldn't be necessary (again "shouldn't" doesn't necessarily mean "isn't").
Well, "shouldn't" isn't "isn't" for me. ;)

I did Daredevil and Blade 2 with v0.49 and vbr_brate_max=9000. Everything I've got tells me they're progressive frame source at 23.976, yet both came out with peaks ~11000.

Wasn't it the conclusion of the "Still getting stutter" thread that CCE could peak above vbr_brate_max? Or was that dependent on the framerate conversion?

Anyway, I think I'll be keeping vbr_brate_max at 8000 or lower. All my tests indicate that CCE will exceed that at low % reduction and default vbr_bias settings, regardless of framerate or source type.

[jdobbs]

Originally posted by Amenophis
for most of the dvds 9000 kbit/s for the video should be ok. but the total bitrate has to be below 9800 kbit/s, too. so if you have 2 dd5 and 1 dts-stream (=1664 kbit/s) everything above ~8100 kbit/s could result in stutter. 9800 is the maximum bitrate for video, the overall bitrate (video, audio, and subpictures) is 10.08Mbs