PDA

View Full Version : HCenc - VBR Bias working ?


Movie_Maker
3rd August 2008, 19:59
Hi there, first post here. I have been playing around with HCenc 0.23 for some time now when I started to convert my personal MiniDV tapes to DVD. The camcorder is a SONY TRV30.

My question is about the VBR Bias setting. My understanding is that the more I set it towards 100 the less would be the bitrate deviation from the average bitrate(like CBR). I tried values of 0, 50, 100 for the same input file(avi) but I could not see any difference in the Avg and Min bitrates. I was also expecting to see a difference(reduction) in the Average Quantizer value also but that stays the same as well. The files size remains same too.
Maybe I am not understanding or setting something correctly, but does this sound normal ?
I am using an Avisynth script as input that has nothing but Avisource and ConverttoYV12 in it.

Thanks!

Boulder
5th August 2008, 04:59
Have you tried comparing the different bias encodes in Bitrate Viewer?

blutach
5th August 2008, 13:12
Please see jdobbs' answer in the DVD-RB forum. In a way, this is a cross post - please see rule 11.

Regards

Movie_Maker
5th August 2008, 15:38
I tried Bitrate Viewer, and the charts for two files - one with 0 bias and one with a bias of 90 - and they are as good as identical! There is as good as no difference between the two in their BR and QL throughout the file.
This doesn't sound right, unless I am not understanding something correctly ? :confused: Here's the hc.ini if that helps -

*INFILE c:\test.avs
*OUTFILE C:\test.m2v
*LOGFILE C:\test.log
*LLPATH C:\HCE\HC023
*BITRATE 8000
*MAXBITRATE 9900
*FRAMES 0 1268
*NOSMP
*PROFILE best
*ASPECT 4:3
*GOP 5 0
*DC_PREC 10
*DVSOURCE
*CLOSEDGOPS
*BIAS 90
*LASTIFRAME
*MATRIX mpeg
*PRIORITY normal

Boulder
5th August 2008, 16:50
Have you tried with a lower average bitrate?

Movie_Maker
6th August 2008, 03:11
hmm... I tried with a bitrate of 6000 and then I could see the VBR<->CBR behaviour of the bias setting. With 0 bias I could see noticeable swings in bitrate and with 90 bias everything was centered around the average bitrate. So why doesn't it do the same at higher bitrates ?
It seems like it would be impossible to force a high CBR-type bitrate i.e. if I want the bitrate to always stay above 8000 no matter what and also maximize all the available bitrate(9800) for the tough areas in the video.

Another thing I noticed is that even if I specify a very high avg and max bitrate, it won't use all that bitrate. For example, even if I specify avg and max as 9800 it will still stay below 9000 even in the hardest hitting areas of the video.

Boulder
6th August 2008, 04:40
I think both of those issues are due to VBV buffer underflows. When you set a high average bitrate, there is a good chance that HC needs to re-encode many GOPs using higher quantizers to avoid underflows.

Movie_Maker
6th August 2008, 05:15
I switched the 'View Buffer' option ON and watched it through the process, it never went under ~75% i.e. the progress bar always held at 75% or higher.
I do see some areas in Bitrate Viewer where the Q.Level jumps quite a bit even though the bitrate doesn't change much.

Boulder
6th August 2008, 07:03
I'm not sure if HC023 shows all the intermediate steps, you might be seeing the buffer of the last encode. In HC022 you could see the buffer being underflown and thus GOPs being re-encoded.

Movie_Maker
6th August 2008, 19:25
assuming thats the case, is it a limitation of the encoder ? Is there anyway to increase the buffer so that it doesn't run out ?
or that's just the way mpeg2 for dvd compression works ?

Boulder
6th August 2008, 19:28
That's just the way DVD specs state things should be. If you get VBV buffer underflows, you can expect stuttering during playback. You can test it by disabling the buffer check, with DV material, I'm quite sure the result is stuttering video and skipping audio when playing the result on a standalone DVD player.

czerro
7th August 2008, 08:31
Is there a noticeable differance between the quality of the CBR and VBR images? 6000 KBps is quite high and I would not expect a VBR encode to look much different than a CBR encode at anything less than a 1000-some pixels x 1000-some pixels. It seems you have enough "data storage space" to "existing data" to achieve a transparent encode. If you desire a transparent encode then CBR is the way to go. It is much less time consuming.

I was unaware of a HCenc 0.23 and I'm unsure of the setting you are referring to. I would suggest that you perform a target size calculation. Encode in CBR and VBR and compare the qualities to the quality of the encode at the "VBR bias" mode. The idea makes no sense to me, and I believe it is likely a test for speed functionality in VBR encoding.

Once you hit 6000kbps you are at essentially oversaturating the video with bits/sec.

Again I know nothing about DV, but I imagine the resolution cannot be that much higher than DVD.

Boulder
7th August 2008, 09:28
6000kbps might be enough for some material but for some it might not be. DV is usually very bitrate-hungry.

Movie_Maker
8th August 2008, 06:26
The output m2v file reports a vbv buffer size of 112. What should it be for a mpeg file targeted for dvd ? And is that a fixed limit ? I mean does it have to exactly that, or can it be something else too ?

Boulder
8th August 2008, 07:03
That's what it should be for a DVD.

Movie_Maker
8th August 2008, 15:13
somewhere I read that a general 'rule of the thumb' is that vbv=bitrate*0.04. If that is true, or even close to true, then that yields a bitrate of less than 6000. Now I am confused why they say the max bitrate of dvd is 9800 ??