CBR vs. VBR - Does it matter in this case?

[echo1434]

Hi,

I'm doing a laserdisc conversion and have a DV AVI file of approximately 1 hour. I'm using Cinema Craft SP to convert it to MPEG-2. At first I thought I would just save time and use CBR since I can already keep the bitrate very high because it is a relatively short video. I decided on a CBR bitrate of 8.1 Mbps and left the audio as uncompressed PCM.

I've already got my DVD authored up and ready to burn, but then I read some other stuff that said VBR mode is always better to use regardless, since it does more analysis and such. I'm already aware of this, but does it really do anything more other than try to make the use of bits more efficient?

I know that I'm already on the upper end of the bitrate spectrum here, so I really don't care if a few parts of the video are going to get 9 Mbps from VBR instead of the 8 I'm getting from a constant rate.

Is there any real advantage to going ahead and doing this over at VBR? Thanks in advance.

[lenti_75]

at this bitrate not really.
it's better to use VBR for lower bitrates, but anything higher than 6000 it;s hard to find any differencies, personally, don't think there is any, at least to see it.

some pleople will say there is, some that there isn't...from my experience, there is no difference if the bitrate is higher than 6000, there is no dif not even at 5000...

as you said it tries to use the bits more efficiently....

[whitejremiah]

i personally ALWAYS use vbr......but at that type of bitrate, i doubt you will be able to tell the differance really................and 1 hour of video onto a dvdr at 8.1mb/s.......i dont think you will be able to tell it apart from the source video.......you may want to recapture it using mjpg or huffyuv if space allows though, cuz DV isnt the best thing to be capturing to from a capture card....

[greymalkin]

yep..re-capture using huffyuv or raw avi then encode video @ 9500bps and audio @ 224bps AC3. This will give best results.

[OmegaSupreme]

http://www.mediachance.com/dvdlab/tutorial/bitrate.html

[jman98]

You might indeed care if your video bit rate goes up to 9 Mbps because with LPCM audio your DVD will be out of spec (combined bit rate too high) and some authoring programs might refuse to author it. I have done similar laserdisc captures and I can tell you that you won't be able to tell the difference between 8.1 Mbps and 9 Mbps. You might as well just use 8.1 and LPCM audio to have better audio. I often do that.

VBR is "better" than CBR only if your maximum bit rate is quite a bit higher than the CBR bit rate. At bit rates above 7 Mpbs it's arguable that there's not really any appreciable difference in going higher, so you might as well just use CBR. I usually capture my laserdiscs at a bit rate of 9.1 and use CCE to re-encode them to a lower bit rate to fit on a DVD disc. I've used CBR and you can use VBR with settings of a low of 7.6 and a high of 8.4 with an average of 8.1 if you want. I think that's within spec but if 8.4 is too high for LPCM audio, I know 8.3 will be OK.

[SLK001]

CBR is always better than VBR (but at a significant size price). LPCM will consume about 1.5Mbps, so using 8.1Mbps for your video is okay.

The real advantage for you will be encoding times. If your videos are short, what you are doing is fine. Go for it...

[FulciLives]

I would not use a video bitrate higher than 7500kbps when using LPCM WAV audio.

Why?

If you use 8000kbps with LPCM WAV audio you are really close to the MAX limit of the DVD Video format and it seems that home made recordable formats (like DVD-R and DVD+R) cannot handle bitrates that are too close to the MAX allowed.

Also with CBR the bitrate will go up and down a bit so if you set the video bitrate to 8000kbps it may "spike" up as high as 8300kbps depending on the encoder used and that would be no good with LPCM WAV audio since you risk the bitrate limit.

Anyways I highly doubt the human eye can see a difference between a video CBR of 7500kbps vs 8000kbps.

That's my 2 cents worth.

- John "FulciLives" Coleman

[Nelson37]

The improved motion search prediction resulting from more than one pass is the often overlooked secondary advantage of (multipass) VBR.

Whether the improvement is worth the extra time is a judgement call, as with so many other decisions.

[echo1434]

The improved motion search prediction resulting from more than one pass is the often overlooked secondary advantage of (multipass) VBR.

Thanks for all the replies. This is what I was concerned about. I thought we pretty much had it summed up to "It does not matter in this case," but this comment has me wondering again...

Does the VBR actually help the converting program do something bitrate-independent to improve the look of the video (motion scenes, etc.)? Or is it simply helping it by using the extra bits during these portions? Again, I know I'm on the upper limit so I don't think the bitrate alone is going to affect anything. But could there really be some sort of extra "magic" provided by VBR mode?

[gadgetguy]

Since you already have the conversion done at CBR, review it and see if there are any visible flaws. If there aren't, then it would be a waste of time to re-encode it. If there are flaws that could be fixed by improved motion search prediction, then by all means, encode it again.

[SLK001]

The improved motion search prediction resulting from more than one pass is the often overlooked secondary advantage of (multipass) VBR.

The IMSR only increased the bitrates of a VBR stream. With a high CBR encode, it has no advantage.

[edDV]

Again, I know I'm on the upper limit so I don't think the bitrate alone is going to affect anything. But could there really be some sort of extra "magic" provided by VBR mode?

The "extra magic" in two pass VBR is to fit average bitrate exactly. This is important if you are working to the full capacity of the DVDR and want insurance to avoid a few MB of capacity overshoot.

The cost of this insurance is a second pass.

[greymalkin]

although the motion search precision is definitely a factor..remember you are converting from a laserdisc source which I assume is a professional video to begin with and will not have the motion nightmares of shaky, noisy home videos

[lenti_75]

for your sake convert it as VBR and convince yourself if there is any difference.

[mikiem]

Nothing that hasn't been posted here already, but in case it helps...

In general CBR is nice because it doesn't cut the bitrate for what the encoder judges are less important scenes, and it can make mistakes. VBR is nice because when it does cut the bitrate for a scene, any bitrate left over can be used somewhere else.

Is VBR worth it? Retail DVDs, digital cable/sat, all sometimes have high motion scenes where the encoding is poor, kind of fell appart in a mess of blocks. Chances are more bit rate would have cured the problem, so setting their vbr properly would have been a very good thing.

At the same time we've probably all seen video where dimmly lit, low motion scenes are terrible. To me it looks like you're watching a movie through a beat up piece of plexiglass. Improperly set vbr likely cut the bit rate too much. IMO the only problem with vbr is actually the settings used.

Now recording already encoded material at DVD rates probably can't do an awful lot to hurt or help, but it might happen if the content is encoder challenging. For final encoding original footage I personally can't see taking a chance on CBR, but also think you should be preparred to encode more then once (or join encoded scenes when possible) to adjust the vbr settings.

[edDV]

...
For final encoding original footage I personally can't see taking a chance on CBR, but also think you should be preparred to encode more then once (or join encoded scenes when possible) to adjust the vbr settings.

? High bitrate CBR avoids all the problems you mentioned. High bitrate CBR is the gold standard. VBR is used to increase compression or to fit a target file size.

[ZippyP.]

High bitrate CBR avoids all the problems you mentioned.

I agree. How can VBR improve upon a high bitrate CBR?

[echo1434]

For those who are interested... Since I had already authored up my DVD from the 8100 kbps CBR MPEG-2, I went ahead and burned a test copy to watch on my standalone player. I thought it looked great, so I'm going to call it good here.

For future reference, I'm still interested in this whole CBR vs. VBR debate. It was easy to use a high CBR here because the video was only an hour. But I also have things to transfer that will be longer. 1 1/2 hours, 2 hours, sometimes more. So what exactly is the definition of "high bitrate"? 6000? 7000? More? For these longer videos my bitrate will be more important because I'll have to choose between using a single DVD or splitting between two discs (or using dual layer).

VBR can certainly save space, but this talk about errors has me a little spooked. Is the Cinema Craft SP really that prone to this kind of thing? It is a rather high-end encoder.

[edDV]

All encoders face the same physics. DVD bitrate maximizes 8000-9500 Kb/s depending on how you do audio. To go longer than an hour you need to apply more compression.

CBR is efficient for high bitrates, VBR attempts to adjust scene bitrate to a motion/noise analysis and applies more compression to low motion scenes and more bitrate (bounded by specified max rate) for high motion areas*.

Two pass analyzes motion in a first pass and applies results to the second pass. The main advantage to two pass is exact fit to specified average bitrate (file size). Quality may improve slightly.


* Pro DVD authors will touch up frames in high motion areas to prevent bitrate spikes or motion artifacts and will not rely on VBR analysis only.

[guns1inger]

commercial transfers use diffent encode setting, often tweaked by hand, for different sections of the movie. This is done for a number of reasons, but mostly to encsure correct bitrate control. Mpeg has problems with certain scenes, and not just the expected high-motion stuff. For example, fades to black and cross fades are problematic and required a higher bitrate to prevent blocking. The encoder will be told by the operator to apply more bitrate to deal with this. This manual eintrevention is also used to ensure that I-frames are set at chapter stops so the chapter starts on the correct frame, not a near-enough-is-good-enough frame like many consumer (and pro-sumer) authoring apps.

Most quality encoders are pretty good at getting the bitrate allocated correctly. Of course, nothing is perfect, but as a general rule you are not likely to see many mistakes made by the encoder. Mostly it will be down to opertaor error. If you read up on the latest version of CCE SP, for instance, it has many optimisations built-in for just these types of issues.

[ZippyP.]

So what exactly is the definition of "high bitrate"? 6000? 7000? More?

A better question might be, "What is a high enought bitrate?"
It's a judgement call that depends on the source and how fussy you are. Ask 10 people here and you will get 10 different answers. You need to make up your own mind about that.

[FulciLives]

I do a CBR if the bitrate is 7000kbps or higher as I see no point for a VBR at that point.

I've even done some TV captures (where 100% quality is not needed) at CBR as low as 6000kbps and they look fine to me except on some very fast motion like when the camera in CSI pans and zooms quickly into a clue but even then these moments are so short that I don't care.

For things where quality is a big concern though I still do a VBR if under 7000kbps.

Personally I think it also helps to set a high MIN when doing a VBR encode. I usually set my MIN at 2500kbps and if the AVG is 6000kbps or more then I'll even set the MIN to 3000kbps. With longer stuff when I have to use an AVG of less than 5000kbps but above 4000kbps then I'll use 2000kbps as the MIN.

- John "FulciLives" Coleman

[mikiem]

"High bitrate CBR avoids all the problems you mentioned. High bitrate CBR is the gold standard. VBR is used to increase compression or to fit a target file size."
"I agree. How can VBR improve upon a high bitrate CBR?"

For Final encoding, when by nature you have a fixed size (otherwise why encode? ) vbr allows the max bitrate headroom for those scenes that need it. cbr lowers the ceiling. Cbr will help make sure low lite scenes don't get insufficient bit rate, but so will correctly setting vbr minimum and average bitrates in any codec that uses them. At any rate, not improving on a cbr hi-bitrate, but when/if re-encoding for final distribution.

"So what exactly is the definition of "high bitrate"? 6000? 7000? More?"

Probably shouldn't use the term I guess because it is vague -- high in relation to what? You're right. I consider 20 or so high, but for DVDs the max is much lower then that, usually vbr with 8 max & 6 average generically for about 1.5 - 1.75 hour's worth I think.

"this talk about errors has me a little spooked. Is the Cinema Craft SP really that prone to this kind of thing? It is a rather high-end encoder."

Any time you ask an encoder to make your decisions for you, which scenes get how much bandwidth, you can have problems. You need enough compression to stay within size limits, and try to balance everything with settings. Maybe not as big a deal with already encoded material as if it's going to get screwed up, probably already is -- but laserdisc might be an exception as perhaps automatic encoding wasn't as common then. It's why they used to always employ encoding folks -- that was their (only?) job.

"Personally I think it also helps to set a high MIN when doing a VBR encode"

Probably the single most important factor when doing vbr of any kind! A very low min is only good for tiny files of talking heads IMO.

[edDV]

"High bitrate CBR avoids all the problems you mentioned. High bitrate CBR is the gold standard. VBR is used to increase compression or to fit a target file size."
"I agree. How can VBR improve upon a high bitrate CBR?"

For Final encoding, when by nature you have a fixed size (otherwise why encode? ) vbr allows the max bitrate headroom for those scenes that need it. cbr lowers the ceiling. Cbr will help make sure low lite scenes don't get insufficient bit rate, but so will correctly setting vbr minimum and average bitrates in any codec that uses them. At any rate, not improving on a cbr hi-bitrate, but when/if re-encoding for final distribution.

This was in the context of a quality DV handheld source using high CBR 8000-9500 Kb/s bitrates. If you want two hour, then use two layers to maintain max quality.

VBR is appropriate when higher compression is desired for sources that are less shaky.

"So what exactly is the definition of "high bitrate"? 6000? 7000? More?"

Probably shouldn't use the term I guess because it is vague -- high in relation to what? You're right. I consider 20 or so high, but for DVDs the max is much lower then that, usually vbr with 8 max & 6 average generically for about 1.5 - 1.75 hour's worth I think.

For this context DVD high bitrate means near maximum for a 62 min DV tape or ~8200 Kb/s CBR for 2 ch PCM audio or ~9500 for 224Kb/s MP2 audio. One might back off from maximum if DVDR media tests show read buffer errors. 8000 and 9400 Kb/s adds some safety.

"this talk about errors has me a little spooked. Is the Cinema Craft SP really that prone to this kind of thing? It is a rather high-end encoder."

Any time you ask an encoder to make your decisions for you, which scenes get how much bandwidth, you can have problems. You need enough compression to stay within size limits, and try to balance everything with settings. Maybe not as big a deal with already encoded material as if it's going to get screwed up, probably already is -- but laserdisc might be an exception as perhaps automatic encoding wasn't as common then. It's why they used to always employ encoding folks -- that was their (only?) job.

Even high rate CBR encoding can reveal errors because 25Mb/s DV (entact fields) is being compressed ~3x to 8.2 Mb/s using mostly interframe motion compression. If the video is shaky to the point the encoder can't achieve sufficient motion compression, it will turn to increased intraframe (spatial detail) compression and if it can't manage that, it will repeat frames.

This is a major problem for realtime MPeg encoders but software encoders also bail out on difficult scenes. That is why the pros turn to hand video optimization in scenes where encoders fail. VBR modes have the same problems when the "Max" rate can't cope.

"Personally I think it also helps to set a high MIN when doing a VBR encode"

Probably the single most important factor when doing vbr of any kind! A very low min is only good for tiny files of talking heads IMO.

Talking heads, Ken Burns films and Powerpoint presentations.

[FulciLives]

I don't know if this is the "correct" thread to bring this up but something has bothered me for a long time and I've never brought it up nor have I really ever seen anyone else bring it up.

Dark scenes in a movie.

I've always read and heard things that dark scenes need less bitrate than bright scenes and that a really "dark" movie (like one that takes place mostly at night ... for instance BLADE or something like that) needs less bitrate.

Huh ???

It seems to me that "dark" scenes always seem to suffer compression aritfacts more so than "bright" scenes. Haven't we all scene the black shadows in a dark scene "dance" around ... at least on some moves?

This is something I even noticed prior to DVD. Look at a movie on VHS and the "bright" scenes always seem to look better than the "dark" scenes. I also noticed this back in the day when I would use a stand alone digital converter box to convert PAL VHS to NTSC ... the bright scenes would look good whereas the dark scenes would have strange artifacts due to the conversion. The PAL to NTSC conversion artifacts were either not there in bright scenes or not very noticeable.

So how come people are always saying that "dark" scenes need less bitrate? I think that's B.S.

- John "FulciLives" Coleman

[edDV]

Some thoughts come to mind.

TV gamma puts more bit precision in the grays when done correctly. The correct gamma curve for digital video is ITU-R BT.709 reproduced below


Ref http://www.poynton.com/PDFs/TIDV/Gamma.pdf

First not all computer RGB video filters are following this curve correctly so bit precision may be reduced in the low grays when the correct gamma curve is applied after linear filtering.

Second issue is signal to noise from analog captures. Black has ~3x lower signal to noise than white by definition because black occurs at 0.3Vp-p and white at 1Vp-p while noise is a constant for both. Therefore I would think black encoding has more quantizing and motion detection error due to the influence of noise.

If the source is a consumer camcorder CCD then even more relative noise exits in the dark grays.

Just some first thoughts.

[mikiem]

""dark" scenes always seem to suffer compression aritfacts more so than "bright" scenes. Haven't we all scene the black shadows in a dark scene "dance" around ... at least on some moves?"

Few years back doing low bitrate wmv, real, that sort of thing this always drove me nuts... Sometimes got away with adding noise (actually just encoding those scenes at lower quality setting mjpeg). Not sure it would even apply, or how much today, the encoder saw the noise as motion and recorded it. The encode was noisier true, but I felt the noise was less objectionable then the alternative.

[swiego]

A little while ago, I did one encode using CCE at 9.0Mb CBR (I chose 3 passes) and another at I think 6.0Mb average VBR (2.0 min, 9.8 max) 3-passes. I then wrote an avisynth script that loaded both and merged them (first video = top half, second video = bottom half) and looked at them frame by frame in virtualdub to see if I could see any difference in the fast motion pans or more jerky moments. (This was all hand-held camcorder footage.)

I couldn't see any difference at all.

Then I tried comparing the clips using avisynth scripts just to see if anything showed up, for example,

A=DirectShowSource("c:\cbr.mpv")
B=DirectShowSource("c:\vbr.mpv")
#StackHorizontal(A,B)
coloryuv(Overlay(A, B, mode="difference"), analyze=true)

And there were some really minor differences based on the numbers coloryuv threw out, but they too were smaaaaaall.

IMO, it doesn't make much of a difference at all.