VBR-CBR Challenge

[macm]

Hi all,

I know all the theory about VBR-CBR and do my caps switching between them, but a question has arised: Why some keep saying that VBR is best?
I can't see a reason other than file size (I mean,better compression!!!). If I set a bitrate high enough (from 6000+) with CBR, I get better quality than just setting a minimum/average/maximum bitrate with VBR.

Any ideas?

[mats.hogberg]

reason other than file size

There you go! VBR gives you better quality with a given file size. The discussion is pointless if you dont take file size in consideration.

/Mats
[/b]

[Adder_78]

Yes but it's kind of stupid to use high bitrate for scenes that dont require it. If the file size is not important then do high bitrate CBR... 8)

[macm]

thanks all,

that's it! The file size doesn't matter at all, because almost all my VHS sources are within the 1h30min limit, and waste some bitrate at some frames is worth because then all movie go smoothly. Wanna a test? Just play a DVD made from CBR and other made from VBR, both with same bitrate range, and in the same scene (of course with a subject movement!!), press the pause: you will notice that with the VBR DVD, it appears with square or pixelated noise. With CBR those almost won't happens.

[Gazorgan]

If *cough cough* size doesn't matter then go with 8500 CBR and be done with it. Very simple. Especially if you like changing DVD's in the middle of the movie.

Where size does matter, it's painfully obvious. SVCD/xSVCD/CVD all make a huge difference.

[AmuroRay]

thanks all,

that's it! The file size doesn't matter at all, because almost all my VHS sources are within the 1h30min limit, and waste some bitrate at some frames is worth because then all movie go smoothly. Wanna a test? Just play a DVD made from CBR and other made from VBR, both with same bitrate range, and in the same scene (of course with a subject movement!!), press the pause: you will notice that with the VBR DVD, it appears with square or pixelated noise. With CBR those almost won't happens.

I totally agree with you. I just recently made the switch from VBR (1000, 5400, 7200) to CBR (6000) and the reduction of noise and pixelation is quite noticeable. I am primarily author anime DVDs from DivX files and after seeing the CBR results I'm not sure if I will ever switch back to VBR.

[SLK001]

I totally agree with you. I just recently made the switch from VBR (1000, 5400, 7200) to CBR (6000) and the reduction of noise and pixelation is quite noticeable. I am primarily author anime DVDs from DivX files and after seeing the CBR results I'm not sure if I will ever switch back to VBR.

There really isn't any good VBR capturing hardware that most of us could afford. For captures, CBR is the only way to go. But, for encoding, VBR can be just as good as CBR, if given enough time (1 pass through for a .VAF file, then three passes through the actual video, so this is 4X the time of CBR).

If your VBR videos are pixelating, then you are bit-starving your video. You can get this same effect by bit-starving a CBR stream.

[bugster]

Using a good quality encoder (TmpGenc, CCE, Others), multipass VBR with min 2000, Avg 6000, Max 8000 wil beat a CBR of 6000 on some material. The reason being high motion scenes that need it can go up to 8000 in the VBR case but are stuck with 6000 in the CBR case. Lots of fast motion or very noisy source material are ideal candidates for this treatment.

Can Encoders that work 'in real time' during the capture process equal this, I can't prove it but somehow I doubt it.

[GKar]

The biggest problem I have with capturing VBR via hardware is the inability to guess-ta-mate the resulting file size without multiple captures of the same movie.

[Gazorgan]

You never capture VBR, because you can't actually do it. VBR is multi-pass, which is impossible when capturing.

[SLK001]

The biggest problem I have with capturing VBR via hardware is the inability to guess-ta-mate the resulting file size without multiple captures of the same movie.

VBR encoding needs both pre and post position video information to be able to make the proper encoding rate adjustments. I would postulate that your encoding hardware doesn't have this capability. If you looked at your VBR captures, you will probably see no BR difference between it and a CBR cap of the same video.

[GKar]

You never capture VBR, because you can't actually do it. VBR is multi-pass, which is impossible when capturing.

My Dazzle DVC II can capture VBR, and quite well too. My most recent attempt was a VHS James Bond movie. CBR around 5000MB size just about filled the DVD, VBR 5000MB size was about 3GB, same movie, same 5000MB bit rate. Biggest problem I noticed in VBR mode was some darker scenes had problems.

[Cornucopia]

You never capture VBR, because you can't actually do it. VBR is multi-pass, which is impossible when capturing.

Au contrare, there is such a thing as single-pass VBR and buffered single-pass VBR. With 1-pass, it just makes a guess, so the bitrate can vary widely--including making the filesize LARGER than what you'd expected. This obviously isn't the best way to go. Buffering this input (w/ memory store that analyzes single or multiple GOPs from the past second) helps to keep the swings moderated and closer to the expected BR. This is what is used in a number of set-top recorders.

Scott

[rhegedus]

My Dazzle DVC II can capture VBR, and quite well too. My most recent attempt was a VHS James Bond movie. CBR around 5000MB size just about filled the DVD, VBR 5000MB size was about 3GB, same movie, same 5000MB bit rate. Biggest problem I noticed in VBR mode was some darker scenes had problems.

Are you sure - same average bitrate but over a gig difference in file size? Something is not quite right.

[GKar]

My Dazzle DVC II can capture VBR, and quite well too. My most recent attempt was a VHS James Bond movie. CBR around 5000MB size just about filled the DVD, VBR 5000MB size was about 3GB, same movie, same 5000MB bit rate. Biggest problem I noticed in VBR mode was some darker scenes had problems.

Are you sure - same average bitrate but over a gig difference in file size? Something is not quite right.

I never said same average bit rate, I can only set the high end bitrate, then it adjusts itself downward from there I guess.

[rhegedus]

I've been doing some testing of my own since I started capturing my VHS collection and straight from broadcast.

Here are some results from a DV capture of a car show on TV (lots of action combined with still shots) than were later encoded using either CBR or VBR at a bitrate of 3000.

CBR3000



VBR3000




I've got some still shots at which VBR2000 looks identical to CBR3000. At the same average bit-rate, VBR (2-pass) beats CBR - no question.

I've been meaning to write a guide on this............

[SLK001]

To be fair, you must compare the two by setting the MAX BR in VBR to the CBR rate. For example, if you cap in CBR at 8500kbps, then the MAX for your VBR should also be 8500.

[GKar]

Just to make sure I wasn't blowing smoke up anyone's ass, I captured the first 45 minutes of "Tomorrow Never Dies" (James Bond) on VHS (including the Music Video and Ode to Q), using my Dazzle DVC II hardware VBR real time encoding at VBR 8000MB bitrate. I am using Windows 98/SE, total file size for the MPEG2 was 1.41GB, and it looked pretty good considering the source was tape. The CBR rate for the entire tape would have been 4850MB based on BitRateCalc for a 2 hour movie whereas using VBR at 8000MB I could have put the entire tape and music video and Ode to Q on 1 DVD.

Mike

[bugster]

You never capture VBR, because you can't actually do it. VBR is multi-pass, which is impossible when capturing.

Sorry but thats wrong, For example, TmpGencs CQ mode is VBR, but only single pass. If TmpGenc can do this, so can an 'on the fly' capture encoder. It may not give fully predictable filesizes like a 2-pass (or more) solution, but it is still VBR.

[SLK001]

Sorry but thats wrong, For example, TmpGencs CQ mode is VBR, but only single pass. If TmpGenc can do this, so can an 'on the fly' capture encoder. It may not give fully predictable filesizes like a 2-pass (or more) solution, but it is still VBR.

It's much easier for a software encoder to "peak" ahead than it is for a hardware encoder to try to see what it hasn't encoded. My hardware encoder will encode in VBR, but the variation in bitrate throughout the video is less than 2-300kbps from the selected BR (in other words, it is basically CBR).

The true test of VBR action is a screen plot of Bitrate Viewer checking the video.

[rhegedus]

I never said same average bit rate, I can only set the high end bitrate, then it adjusts itself downward from there I guess.

This is something new to me, thanks for pointing it out. The whole 'one pass VBR' is a mystery to me - I wonder what sort of buffer size is needed or over what time period the VBR scanning takes place.

If anyone is interested, here is a past thread from which I have learned a lot:

Could VBR ever be better than CBR??

Have fun!

[macm]

Thanks all.

I think the best I can do is decide based on the source VHS. If it is a VHS-C home footage I'll do a high MPEG-2 capture, 9600K bitrate CBR so it will fit exactly 1 hour per DVD-R.
When the source is any VHS (movie, TV capture) I'll do a VBR maybe 8000k max, 6000 avr, 4000 minimum, so it will fit 2 hours in a DVD-R.

what counts for me is more quality, despite the file size. So a CBR should go OK most of the time.

Anyone has a clue on what the best setting for IBP GOP structure?
I normally use the ATI MMC default, 1 I, 4 P, 2 B.

THX.

[rhegedus]

I think the best I can do is decide based on the source VHS. If it is a VHS-C home footage I'll do a high MPEG-2 capture, 9600K bitrate CBR so it will fit exactly 1 hour per DVD-R.
When the source is any VHS (movie, TV capture) I'll do a VBR maybe 8000k max, 6000 avr, 4000 minimum, so it will fit 2 hours in a DVD-R.

Probably a bit excessive:

Hi8/VHS to DVD: which bitrate do you recommend?

[lordsmurf]

Anyone has a clue on what the best setting for IBP GOP structure? I normally use the ATI MMC default, 1 I, 4 P, 2 B.
THX.

I find the encodes looks cleaner with less P frames, opting instead for 1I 2P 2B. This is noticed most on VBR.

The VBR capture is a partial mystery to me too, but an educated guess is that while some info is buffered in RAM, decisions are quickly made. And the VBR is likely made on a GOP-by-GOP or frame-by-frame basis, not a full temporal of the whole MPEG stream.

VBR is not about file size. Not at all. VBR is about giving you the amount of bitrate needed for a scene. Does this affect file size? Sure. Proper use of VBR does not "harm" it as compared to CBR.

A CBR encode of 4000k will look worse than a VBR of 0-5000k and the VBR will likely be smaller filesize. The same max bitrate for each is not a fair test, as it does not address the reasoning behind VBR: better quality in the same space (not merely to make it smaller, though it may happen).

[vhelp]

.
.

My .02 cents worth,

I think that after some planning has ben completed ie, movie length etc.,
one has to decide something (I find interesting) is something like this perhaps..

* If you are planning to someday re-encode this source (for whatever the
...reason) then perhaps a CBR might (I said might) be a better alternative.
...Of course, this will depend on the source, it's quality, and others such
...issues that could result in leaning towards a CBR (not a newbie technique)

* if you are not planning on any such future re-encodes, then perhaps a VBR
...is in order here.


VBR could be either of two things (modes) ie,

* 2-pass (or multi-pass)
* or CQ (for constant quality - a variation of (rather good one w/ TMPG) of
...VBR's 2-pass)

I've said elsewhere's that I'm in another testing stage w/ CQ's tunes min/max
param settings. I'm still testing this out.

Also, after messing around w/ Standard VCD's CBR 1150, and uping it up
to 2300, I found that this yielded a very good xVCD - quality'wise. So, I too,
have ben tinkering w/ CBR vs. VBR better alternatives.

I won't do a 2-pass (or more) because it's a complete waist of time, among
other nonsense (IMO)
And, if you are doing DVD authoring from your capture projects, then you
deffinately don't need to do any multi-pass type encoding (user descresion)
but instead, use CQ.., because DVD specs are 9.8k vs. SVCD's +/- 2.5k
and the audio, as I've used it, I set to 192k for un noticable sound difference

I always find these CBR vs. VBR debates interesting. They seem to almost
always stir up new facts or ideas. And I treat them as such.

-vhelp 2040

[SHS]

lordsmurf it is very to unstand the diff from VBR vs CBR
CBR mode dose all full frames all the time just like AVI, DV, etc but it dosen't give you very efficient compression.
VBR mode only dose part of the frame as in the part that change from certain scenes before it which in turn is more efficient compression as much as 15% smaller file size sometime even more.

[bugster]

VBR is not about file size. Not at all. VBR is about giving you the amount of bitrate needed for a scene. Does this affect file size? Sure. Proper use of VBR does not "harm" it as compared to CBR.

Have to disagree here, at least partly. Multipass VBR encodes, with a none realtime encoder (i.e not encoode whilst capture) allows you to make best use of the space available on your target medium (CD or DVD) whilst still using a high bitrate where needed. If encoding when capturing, the reason behind using VBR is probably different, and the quality you can achieve probably different too. I can't claim any experience in this matter but would expect that under such circumstances, a CBR encode at 6Mps vs a VBR encode at MAX 6Mps would look very similiar quality wise but the VBR encode COULD be smaller as bitrate is not wasted on still and low motion scenes.

A CBR encode of 4000k will look worse than a VBR of 0-5000k and the VBR will likely be smaller filesize. The same max bitrate for each is not a fair test, as it does not address the reasoning behind VBR: better quality in the same space (not merely to make it smaller, though it may happen).

The same Max bitrate is not a fair test. You should compare the VBR Average bitrate against the CBR bitrate

CBR mode dose all full frames all the time just like AVI, DV, etc but it dosen't give you very efficient compression.
VBR mode only dose part of the frame as in the part that change from certain scenes before it which in turn is more efficient compression as much as 15% smaller file size sometime even more.

Sorry SHS but you are mistaken. CBR and VBR (mpeg) use exactly the same compression schemes, its just a matter of how the bitrate is spread around. In fact a CBR mpeg is not really CBR. the 'Constant' is actually spread across each GOP, not per frame. What you are describing is an I-Frame only encode, which is not common in mepgs and is indeed inneficient (and is how DV works, which is not mpeg). So in a CBR encode, it is each GOP that is the same number of bytes in size, not each frame.
In fact the latest version of CCE has a 2-pass CBR mode. This allows it to determine the best use of its available bitrate between the various I,B and P frames within each GOP. But every GOP is the same total size.

I won't do a 2-pass (or more) because it's a complete waist of time, among
other nonsense (IMO)
And, if you are doing DVD authoring from your capture projects, then you
deffinately don't need to do any multi-pass type encoding (user descresion)
but instead, use CQ.., because DVD specs are 9.8k vs. SVCD's +/- 2.5k
and the audio, as I've used it, I set to 192k for un noticable sound difference

To me the main advantage of 2-pass in TmpGenc, over its admittedly very good CQ mode is that I can max out the space used on the DVD, when I have good quality source. Sure it takes longer than CQ but I find nothing more annoying than encoding a movie only to find it only 3.6Gigs or 4.8Gigs meaning I have either wasted space and possibly lost some quality or I have to adjust the encode and go again (or author and use DVDshrink). I can happily put up with the long encodes in 2-pass by encoding overnight. If you want to encode a large number of captures frequently then I can see why CQ or other faster encoders are attractive.

[BJ_M]

You never capture VBR, because you can't actually do it. VBR is multi-pass, which is impossible when capturing.

almost all high end capture encoders do VBR and CBR encoding ... some even can do 2 pass VBR (tape goes through twice)

[SHS]

Sorry bugster but CBR is Constant Bit Rate just like VBR is Variable Bit Rate you should look it up on the Glossary.
(use exactly the same compression schemes) but you miss read what I read efficient compression like let say we use winzip at 50% compression and but let say that wan't good enough so bump it to 75% compression which in turn get more efficient compression.
Firsl I like read the Glossary
http://www.mpeg.org/MPEG/DVD/General/Glossary.html#VBR
and when done with that then read
http://www.mpeg.org/MPEG/DVD/General/Glossary.html#VBV

[SHS]

You never capture VBR, because you can't actually do it. VBR is multi-pass, which is impossible when capturing.

Gazorgan you can do VBR with REALtime Software/Hardware Encoder but what can not do is Multi-Pass VBR with REALtime Software/Hardware Encoder this can only be done by Off-Line Software enocding and tell you the truth I see no diff with one pass vs 2 pass encoding.