battle of the CBR/VBR/CQ modes

[banjazzer]

OK, OK, OK.

You agree with me, except about what I said.

[baker]

I agree with everything except one part.

Baker

[Murrdawg045]

Oh my gosh, some people are just not smart.

CQ has an average bitrate! Everything does! Thats why they call it average!!!

The difference is that with CQ, you don't have to set the average, it decides for you by the amount of action in the movie.

If you put in a max bitrate at 2520 and a CQ of 90, you will get an average bitrate somewhere around 2250, give or take. When you do a high CQ you are gauranteed a certain quality, no matter the length or amount of action in the movie. You are not, however gauranteed a filesize. If you set CQ to 90, it will use as much space as needed so that the movie is at a certain quality.

If you don't care about the size, than just set 2-pass var bitrate to a higher average.

If you still don't believe me, find out for yourself. Make a movie segment with CQ and find the average bitrate. Then plug that average in for 2 pass var bitrate and see which is better.

[Nelson37]

Folks, CQ's only advantage is to pick the bitrate for you. If you develop the skill to correctly set bitrate for good quality, with VBR you will know the EXACT filesize based on movie length. You can then CHOOSE to vary that down, or up, to fill 1, 2, or 3 disks. Pick a movie, then instead of picking some arbitrary number pick a real-world number of disks. My VBR will smoke your CQ EVERY SINGLE TIME, as my file will have a higher bitrate and larger size.

Think of each CD as a bitrate bucket, VBR will ALWAYS fill each more completely, and therefore yield a HIGHER bitrate for a given number of CD's. You can only come close to this with CQ through dumb luck or multiple encodes.

My particular heresy is that I think most folks are missing the boat by setting Max bitrates at 2500, or same as AVG. I use Max 3000 or 3500, and my tests indicate that CQ will spike higher than VBR unless higher Max rates are used for VBR. I do believe that CQ uses a different algorithm to set bitrates, and is difficult to compare to VBR in terms of encoder settings.

If you guys are REALLY hung up on quality and speed, try CCE 4-pass, 500, 2200, 3000. You will never go back, except for MPG1.

[baker]

You see in CCE it asks you how mcuh distortion you would like. I chose 7 and usually get a avg bitrate of 1.6.

Baker

[vitualis]

I think that adam's understanding of vbr is generally correct. I shall attempt to prove it by using a logical step-by-step example.

As I have stated before, CQ only apparently gives better quality because you are not using the same measuring stick.

CQ is where you want (or are willing to accept) a certain amount of quality and the size doesn't matter. The usage of CQ falls between 1-pass VBR and CBR.

I will illustrate how CQ, 1-pass VBR and 2+pass VBR works.

----------------------------------------------------------
Concept of quality in MPEG encoding
For any given encoder, the subjective quality of the clip is a function of the fidelity.

The fidelity of a clip is a function of:
1. bitrate allocation efficiency
2. and overall average bitrate

i.e., fidelity = fn(bitrate allocation) * fn(avg bitrate)
and subjective quality = fn(fidelity)

For example:

1. CBR=1000 vs. CBR=2000
   The "bitrate allocation efficiency" of both clips are identical (both being CBR).
   The "overall average bitrate" of the CBR=2000 clip is obvious higher
   Thus, the fidelity (and hence quality) of the CBR=2000 clip is higher than CBR=1000
2. CQ=60 vs. CQ=80
   The "bitrate allocation efficiency" of both clips are probably identical (both being CQ)
   The "overall average bitrate" of the CQ=80 clip will be higher
   Thus, the fidelity (and hence quality) of the CQ=80 clip is higher than CQ=60

----------------------------------------------------------
CQ encoding
Consider there being a video clip where most of the beginning contains easy to encode video and most of the end contains difficult to encode video.



Now, say for example you use CQ encoding (which remember, is no more than 1-pass VBR where the computer dynamically adjusts the bitrate according to what it considers to be "complexity" without actually trying to encode to MPEG).

And say the results of the encoding is that:
min=300, avg=2000, max=2500

We can assume that then (if the "complexity" algorithms are well written) that the majority of the bitrate went to the later half of the movie.


1-pass VBR
Now say we enter the values that are identical to that of the CQ encoding so that we are using the same measuring stick.

The encoder would in this mode value adhere to the average bitrate more than the "quality" setting. What does this actually mean? It means that the bitrate is less variable but the quality is more variable, when compared to CQ.

Now, the encoder doesn't have a priori knowledge of the video data in the video clip by definition (otherwise it would be at least 2-pass). Thus, when encoding the same video clip (clip1):

• for the easy to encode part, the average bitrate would be higher compared to the CQ encoding
• for the hard to encode part, the average bitrate would be lower compared to the CQ encoding

Now what does this actually mean? For 1-pass VBR compared to CQ:

• for the easy to encode part, 1-pass VBR has higher quality than CQ
• for the hard to encode part, CQ has higher quality than 1-pass VBR

So does CQ look better than 1-pass VBR? Logically, it should be apparent that even using exactly the same bitrate settings, the answer is only "maybe".

If we use the concept of fidelity from before:
fidelity = fn(bitrate allocation) * fn(avg bitrate)
The average bitrate between CQ and 1-pass VBR is the same
The "efficiency of bitrate allocation" is unknown. It is unclear whether or not CQ encoding is "more efficient" than 1-pass VBR. My gut feeling tells me that it "maybe" in some circumstances.

And so, a similar answer as before: CQ encoding MAY have better quality than 1-pass VBR at the same bitrate parameters.


2-pass VBR
This is where I think some of the confusion has taken place. VBR encoding is hamstrung by only 1-pass because the encoder lacks vital information for the best allocation of bits -- predominately, the complexity of the entire video clip.

In multipass-VBR encoding, this is no longer the case.

Let us consider encoding the same clip1 with 2-pass VBR encoding with the same settings (e.g., min=300, avg=2000, max=2500).

On the first pass, the encoder essentially maps out the complexity of the entire clip.

On the second pass (where the encoder actually does the encoding), it can do what CQ can do and what 1-pass VBR can't do as well -- that is, allocate most of the bitrate to the second part of the video clip where it is difficult to encode.

That is, the average bitrate for the first part will be relatively lower.
The average bitrate for the second part will be relatively higher.

However, what 2-pass VBR encoding can do much better than CQ encoding is in the allocation of bits. Multipass VBR encoding knows the complexity of the entire video a priori. CQ tries to get around this by not caring.

What is the implication for this?

The only way CQ encoding can look as good as say 2-pass VBR is by using more bits as it is less efficient. However, when using the same number of bits, that is identical bitrate settings, 2-pass VBR will always look better than the equivalent CQ encoded clip.

If we use the concept of fidelity from before:
fidelity = fn(bitrate allocation) * fn(avg bitrate)
The average bitrate is the same between CQ and multipass VBR encoding.
The "bitrate allocation efficiency", however, is always HIGHER in multipass VBR encoding
Thus, the same answer as before: multipass VBR encoding will have better quality than CQ encoding at the same bitrate.

Each subsequent additional pass (e.g., 3-pass, 4-pass, x-pass) further refines the bitrate allocation. Thus, the spread of bits throughout the ENTIRE video clip is increasingly efficient. This means, that the overall quality of the clip is improved on each subsequent pass.

----------------------------------------------------------

How can multipass VBR provide much better quality than cQ practically?
When making SVCDs (where this whole question is pertinent) we live in a world of fixed media size. Any part of a disc that is not used is wasted bitrate potential for VBR encoding.

For example, if I put a VBR media clip onto a disc where it had 100 MB of free space left, that 100 MB could have been better used to boost/lift the average bitrate of the clip.

The problem with CQ encoding is that you cannot know the size of the final clip BEFORE it has encoded.

Say you encoded a clip at CQ=70 and the final size ended up being 700MB. On an 80min/700MB disc, you can fit up to approx. 800MB. That means that you could have potentially encoded your clip at perhaps CQ=80 with even better quality.

With multipass VBR (or even manual 1-pass) encoding, however, you can set your average bitrate so that the resultant file will be exactly 800 MB.

What does this mean?
Not only is multipass VBR more efficient (i.e., better quality!) at bitrate distribution than CQ encoding at the same bitrate, you can practically get higher average bitrates (again better quality) for ANY SVCD.

----------------------------------------------------------

Conclusion

1. Multi-pass VBR (i.e., 2+) encoding WILL ALWAYS give better quality than CQ encoding if you use the same measuring stick (i.e., ultimately the same average bitrate or the same final file size)
2. Multi-pass VBR (i.e., 2+) encoding WILL FURTHER ALWAYS give better quality than CQ encoding practically on a disc media as you can encode your file to fill the entire disc thus boosting the average bitrate.

----------------------------------------------------------
Discussion
What are the benefits of CQ encoding?

• CQ encoding is nothing more than 1-pass VBR encoding so it is relatively much faster than multipass encoding.
• CQ encoding MAY offer better video quality than manual 1-pass VBR encoding.
• It is still more space efficient than CBR encoding
• Potentially less fiddling with settings

What are the disadvantages of CQ encoding?

• Lack of control of final file size means that any possible quality improvements over 1-pass VBR may be lost when you practically try to create a SVCD (an experienced encoder will try to completely fill a disc)
• Although it is faster, the quality is inferior to multipass VBR (i.e., 2+) when using the same measuring stick. When considering you can also control final file size with multipass VBR (thus boosting average bitrate), this may be very significant.
• "Q" is not a tangible varible. There is no easy way to compare CQ encoding between different encoders or even different versions of the same encoder.

Other Issues -- Response to some comments in this thread

• It has been mentioned by that CQ encoding can vary the bitrate more to try to stick to the same "quality" level when compared to manual VBR encoding. In other works, manual VBR encoding is less "variable". This has somehow been equated to better quality for CQ.
  
  It should be noted that this MAY only true for 1-pass VBR encoding. Multipass-VBR encoding (i.e., 2+) can vary exactly as much as CQ and furthermore, the bitrate allocation is more efficient/refined --> better quality AT THE SAME bitrate settings.
  
  Even when considering this point for CQ vs. manual 1-pass VBR encoding, you have to also consider it from another point of view. CQ encoding by keeping the "quality" at a certain level also prevents the quality from IMPROVING beyond that level. For any one place where manual 1-pass VBR doesn't look as good as CQ, there may be another where 1-pass VBR actually looks better. Thus it doesn't necessarily logically follow that keeping the "quality level" constant provides better quality.
  
  To emphasise again, this ambiguity doesn't even exist at all for multipass VBR and it will have better quality than CQ at all times when using the same bitrate parameters.
  
• Don't take this post as me saying that "you should" be using multipass VBR or that "you shouldn't" use CQ. This IS NOT WHAT I AM SAYING AT ALL!! I am simply aiming to clarify the confusion that is apparent on this thread.
  
  The argument that CQ yields better quality than multipass VBR is simply not correct and is not logically sound.
  
  What you actually do to encode MPEG files for SVCDs though, is completely up to you. That is the beauty of VBR encoding. It is very flexible. If you want ABSOLUTE MAX quality, then go CBR at 2500 kbit/s. If you want quick and easy, then go CQ, 1-pass VBR, or perhaps even straight CBR. If you want MAX quality while optimising disc space for your movies, you can't beat multipass VBR encoding.

----------------------------------------------------------
Caveats used in the above post
I am aware of these issues but I don't think they affect the overall arguements:

1. When I say "quality", I actually mean "fidelity". However, for a given encoder and all other things being equal, improved "fidelity" can probably be equated with improved subjective "quality".
2. fidelity = fn(bitrate allocation) * fn(avg bitrate) is a useful concept but you shouldn't overstep the bounds. Subjective quality is only a function of fidelity but there are large differences between people -- moderated by different cognition. There shouldn't be too much of an issue when using the same encoder but there can be large differences of opinion when trying to compare different encoders.
3. VBR encoding should theoretically improve quality over CBR as the "bitrate allocation efficiency" has been increased. This is, however, dependent on an intellegent encoder. If, e.g., more bits are allocated to easy to encode scenes, then in effect, the "bitrate allocation efficiency" has decreased and a VBR encoding mode could conceivably have worse quality than CBR. In such as case, you could right say that the VBR mode of such an encoder "broken".
4. The assumption involved when we are saying that VBR can have better quality than CBR is that the VBR encoded clip has the same average bitrate as the CBR encoded clip (same measuring stick) and further where the the VBR-max bitrate is not the same as the VBR-avg bitrate.
5. Similarly, the assumption that takes place when saying that multipass VBR encoding will always look better than CQ VBR encoding at a given average bitrate is that multipass VBR in the encoder is not "broken".
6. Although each additional pass (in multipass VBR encoding) improves final video fidelity/quality, there are diminishing returns.
7. For a standard SVCD (that is, within 2x CD spin bitrate limitations), the absolute maximum video quality for any encoder is CBR at approx. 2500 kbit/s. Thus, if the video clip that you want to put on a SVCD is < 37min (for 74min/650MB media) or < 40min (for 80min/700MB media), there is no benefit in using multipass VBR over CQ. This is simply because there is no benefit in using any VBR mode over simply CBR at the max. bitrate.

Regards,
Michael Tam

While I generally agree with you Michael.....

This time there are a few points which I dont think are 100% correct eg

Now, say for example you use CQ encoding (which remember, is no more than 1-pass VBR where the computer dynamically adjusts the bitrate according to what it considers to be "complexity" without actually trying to encode to MPEG).

better explained for some might be. With a very complex scene VBR will comprimise Quality for bitrate, acording to min,average, and max.

With a very complex scene, this will be compressed more than a simple scene.

You can do a 7 pass VBR encode, the fact is you will just optimise the the bitrate better to match the source for the same given inputs..Min-AV_MAX
(with a good encoder you will decrease more distortions than you introduce with each pass)

With CQ mode the the compression will remain constant(relatively) whether you have a complex or simple scene. This time contrained by Min & Max bitrate(Using Tmpgenc) no average meaning file size can not be predicted accurately.

CQ mode does not comprimise Quality for bitrate, this is why you have no average to preset in your encoder.

I am not saying CQ is better than VBR xpass
Personally I use both.

Cheers

[banjazzer]

The problem with CQ encoding is that you cannot know the size of the final clip BEFORE it has encoded.

Say you encoded a clip at CQ=70 and the final size ended up being 700MB. On an 80min/700MB disc, you can fit up to approx. 800MB. That means that you could have potentially encoded your clip at perhaps CQ=80 with even better quality.

This is equivalent to saying, you can always add a bit extra quality, which is true up to a point. This is a weak argument for 2 reasons. Firstly, throwing in extra bits beyond a certain level will have no distinguishable effect. Secondly, just as you could fill the blank space on a CD, you could if you wanted actually add an extra CD. However, as I said, beyond a certain level, throwing extra bits at the encode will have no appreciable effect. My argument against 2-pass VBR is that by working to a bitrate you can get it wrong, depending on the source material. No matter how many passes you have, if the bits allocated are insufficient they will remain insufficient.

However, what 2-pass VBR encoding can do much better than CQ encoding is in the allocation of bits. Multipass VBR encoding knows the complexity of the entire video a priori. CQ tries to get around this by not caring.

It is not a qustion of not caring. If you don't limit the amount of bits in the first place you do not need the first pass! The first pass is only because you have a fixed number of bits to start with. If you limit the bits, then multipass is essential. If you don't, then it isn't.

Say you encoded a clip at CQ=70 and the final size ended up being 700MB. On an 80min/700MB disc, you can fit up to approx. 800MB. That means that you could have potentially encoded your clip at perhaps CQ=80 with even better quality.

With multipass VBR (or even manual 1-pass) encoding, however, you can set your average bitrate so that the resultant file will be exactly 800 MB.

What does this mean?
Not only is multipass VBR more efficient (i.e., better quality!) at bitrate distribution than CQ encoding at the same bitrate, you can practically get higher average bitrates (again better quality) for ANY SVCD.

I don't agree multipass (or specifically 2-pass) is specifically more efficient. It is more efficient at allocating bits if you choose to limit them in the first place. And because you start from a fixed bitrate rather than a fixed quality setting, under certain circumstances the 2-pass VBR may give worse results, because there is no way the program can intelligently allocate more bits than you have chosen.

The argument that CQ yields better quality than multipass VBR is simply not correct and is not logically sound.

I do not think I actually said that, so I won't try to defend that position. What I am saying is that CQ does not suffer from the limitations of 2-pass VBR - namely that you are working to an average bitrate/filesize rather than to a quality setting. I am disagreeing with the position, which I think adam took, that 2-pass VBR is superior to CQ. It most certainly is not.

adam chose to ignore the quote from the TMPGEnc helpfile which I quoted (and I am only talking about TMPGEnc here by the way, because I have no experience of other encoders). This basically states the position from which each form of encoding starts. I repeat it, since it basically encapsulates what I have been saying. It is the crux of the whole argument. My emphaasis once again.

2-pass Variable Bitrate (VBR)]

Every scene in a movie has a different optimal bitrate. In this mode, a scene is analyzed to determine the best bitrate before it is encoded. One can set the maximum, average, and minimum bitrates to use.

[Constant Quality (CQ)]

When one is more concerned about picture quality than file size, one should use this mode. Using the CQ mode, one can set all of the aspects of quality to use in the encoding. This mode will guarantee a high quality movie; however, movies with many scene with rapid motion can become rather large.

[smokingweed3000]

banjazzer u should be quiet cause michael shut u down and basically closed this post at his last post.

[smokingweed3000]

forgot one thing,
do u guys recommend using 4-pass in CCE when making VCDs?

[vitualis]

This is equivalent to saying, you can always add a bit extra quality, which is true up to a point. This is a weak argument for 2 reasons. Firstly, throwing in extra bits beyond a certain level will have no distinguishable effect. Secondly, just as you could fill the blank space on a CD, you could if you wanted actually add an extra CD. However, as I said, beyond a certain level, throwing extra bits at the encode will have no appreciable effect. My argument against 2-pass VBR is that by working to a bitrate you can get it wrong, depending on the source material. No matter how many passes you have, if the bits allocated are insufficient they will remain insufficient.

Actually, your argument here does not stand up to logic.

True, throwing in extra bits beyond a certain level will not have a distinguishable effect. However, is this relevant to SVCDs? Even 2500 kbit/s CBR doesn't NOT yield flawless reproduction.

Increasing the bitrate for VBR encoding (up to the max) will always yield improved quality for a SVCD.

Secondly, your "working to a bitrate" arguement also does not stand up to logic.

I have demonstrated how with multipass VBR encoding the average bitrate WILL ALWAYS be higher than the equivalent CQ encoded clip. If the bitrate is insufficient for multipass VBR encoding, it is even more so for CQ encoding.

It is not a qustion of not caring. If you don't limit the amount of bits in the first place you do not need the first pass! The first pass is only because you have a fixed number of bits to start with. If you limit the bits, then multipass is essential. If you don't, then it isn't.

You did not read the flip side of this. 1-pass VBR limits the bitrate more than CQ yes. But, CQ encoding limits the quality more. 2-pass VBR does not limit the bitrate any more than CQ encoding as for practical encoding, the average bitrate can always be made to be higher than an equivalent CQ encoding.

I don't agree multipass (or specifically 2-pass) is specifically more efficient. It is more efficient at allocating bits if you choose to limit them in the first place. And because you start from a fixed bitrate rather than a fixed quality setting, under certain circumstances the 2-pass VBR may give worse results, because there is no way the program can intelligently allocate more bits than you have chosen.

Again, you are view this from only one perspective. This view is ONLY correct if you don't have fixed media sizes or if the max bitrate is unlimited. Neither of these conditions are true for SVCD.

In practice (as I described in detail before), the average bitrate using multipass VBR can always be made higher than CQ (that is, you have MORE bits to play with).

Furthermore, unless the encoder is broken, multipass VBR encoding is always more efficient than CQ encoding -- read my previous post.

What I am saying is that CQ does not suffer from the limitations of 2-pass VBR - namely that you are working to an average bitrate/filesize rather than to a quality setting. I am disagreeing with the position, which I think adam took, that 2-pass VBR is superior to CQ. It most certainly is not.

adam chose to ignore the quote from the TMPGEnc helpfile which I quoted (and I am only talking about TMPGEnc here by the way, because I have no experience of other encoders). This basically states the position from which each form of encoding starts. I repeat it, since it basically encapsulates what I have been saying. It is the crux of the whole argument. My emphaasis once again.

2-pass Variable Bitrate (VBR)]

Every scene in a movie has a different optimal bitrate. In this mode, a scene is analyzed to determine the best bitrate before it is encoded. One can set the maximum, average, and minimum bitrates to use.

[Constant Quality (CQ)]

When one is more concerned about picture quality than file size, one should use this mode. Using the CQ mode, one can set all of the aspects of quality to use in the encoding. This mode will guarantee a high quality movie; however, movies with many scene with rapid motion can become rather large.

Actually, adam's position is completely in agreement with the TMPGEnc's help file. I cannot help but wonder if you are misinterpreting it.

With multipass VBR encoding, the bitrate is optimised for EVERY scene with regards to the preset bitrate settings.

For CQ encoding, the bitrate is optimised for EVERY scene with regards to a predetermined "quality" setting.

If the multipass VBR encoding and CQ encoded file are the same size, then multipass VBR encoding will have distributed those bits more efficiently than the CQ encoded clip.

In other words, the multipass VBR encoded file will in effect have a higher "Q" co-efficient than the CQ encoded clip for the SAME SIZE.

Regards.

[banjazzer]

banjazzer u should be quiet cause michael shut u down and basically closed this post at his last post.

Perhaps in your mind because you cannot comprehend what has actually been written. However many times you repeat the same basic fallacy, trot out the same mantra, it will not suddenly become truth. You think what you want, - I have no wish to undermine religious fervour. 8)

[vitualis]

With CQ mode the the compression will remain constant(relatively) whether you have a complex or simple scene. This time contrained by Min & Max bitrate(Using Tmpgenc) no average meaning file size can not be predicted accurately.

CQ mode does not comprimise Quality for bitrate, this is why you have no average to preset in your encoder.

This is what I've been trying to debunk. This is ONLY true if you either don't have fixed media sizes and/or there is no maximum bitrate.

Neither is true for SVCD.

CQ mode doesn't compromise quality for bitrate only in a limited view (that you are e.g., willing to accept Q=70 and that's it).

If you use multipass VBR encoding where the average bitrate is the same, you are not limited to Q=70. As the bitrate allocation is better, the equivalent Q is higher.

Also, what do you mean by "adding distortion" by each additional pass?

Regards.

[vitualis]

I hate dogma as well which is why I posted in length and detail to work this out logically.

Please read through it slowly and carefully and think it out. Draw a few diagrams if you want.

It should be obvious after a while that multipass VBR encoding always yields better quality (in any measure, Q or otherwise) than CQ encoding if the average bitrate is the same.

If you find flaw with the logic, then post your argument. Please, well thought arguments only!

Regards.

My last comment, I`m sure it will fall on many deaf ears!!

For those of you who look at other sites beside this one, especially this thread, on the subject of CQ . You will find if you look objectivly and not blinded by fitting 799.9mb on a 80min CD!!
CQ if you take the time to experiment will yield better encodes in general than any form of VBR.!! Whether you us 1passVBR or 100 pass VBR

One suggestion ,Baldrick if your watching this thread, if you have any spare bandwidth maybe start a sample library 20sec to 1 min samples comparing VBR,CBR,CQ and also on different encoders. To make it fair Post a few sources on the site so everyone has the same start point. Then people can argue as much as they want, the resulting encode for everyone to evaluate will be the decider.
Maybe a 1 min action scene source, 1 min slow scene etc etc.






Its FuXking hot outside 35°C gotta go for a beer. 8)

Cheers

[banjazzer]

You did not read the flip side of this. 1-pass VBR limits the bitrate more than CQ yes. But, CQ encoding limits the quality more. 2-pass VBR does not limit the bitrate any more than CQ encoding as for practical encoding, the average bitrate can always be made to be higher than an equivalent CQ encoding.

You limit one mode by restricting the average bitrate, you limit the other by restricting the quality setting. How you can think the former is in some way superior to the latter escapes me. You limit the quality setting by what you choose with CQ mode. You limit the quality in 2-pass VBR by setting the average bitrate, and if you get it wrong there is nothing you can do about it, except encode again with a higher average setting.

Furthermore, unless the encoder is broken, multipass VBR encoding is always more efficient than CQ encoding -- read my previous post.

I read your other post, but nothing you say there demonstrates anything other than that multipass VBR makes best use of the bits you allocate. That is not the same thing as saying it is superior to CQ mode.

Actually, adam's position is completely in agreement with the TMPGEnc's help file. I cannot help but wonder if you are misinterpreting it.

With multipass VBR encoding, the bitrate is optimised for EVERY scene with regards to the preset bitrate settings

Not from my reading it isn't. It says quite categorically, and there is no other way of interpreting this, if you want quality as the prime goal, choose the CQ mode. If you want filesize as the prime goal, choose 2-pass VBR. If we cannot agree what that simple passage means, then words have lost all meaning.

If the multipass VBR encoding and CQ encoded file are the same size, then multipass VBR encoding will have distributed those bits more efficiently than the CQ encoded clip.

I have not run tests to dispute this, but others, I seem to recall, have. It seems you are arguing from what you think should be the case, rather than what is actually the case. However, regardless of that moot point, in real life the 2 situations would not yield the same size file. Why? Because with 2-pass VBR you set out with a filesize in mind. With CQ you set out with a quality setting in mind. No matter how many here are impressed by massive posts, and the same points made over and over, there is no way of getting away from the basic difference by which these 2 encoding methods operate. In one you start with a filesize/average bitrate in mind. In the other bitrate is not limited by this choice. You start with a quality setting in mind.

[banjazzer]

It should be obvious after a while that multipass VBR encoding always yields better quality (in any measure, Q or otherwise) than CQ encoding if the average bitrate is the same.

If you find flaw with the logic, then post your argument. Please, well thought arguments only!

I didn't think you would make it quite so easy. 8) Assuming for now I accept that for equal filesize the 2-pass VBR will yield better quality, which is actually a moot point, the basic fallacy is that in every real-life situation the average bitrate (and filesizes) will not be the same, except by accident. You know the filesize beforehand with 2-pass VBR, you do not with CQ. Although those who love order and certainty might find this an inconvenience, it is actually the reason why CQ mode is good.

[injunpana]

As quoted by Banjazzer

didn't think you would make it quite so easy. Assuming for now I accept that for equal filesize the 2-pass VBR will yield better quality, which is actually a moot point, the basic fallacy is that in every real-life situation the average bitrate (and filesizes) will not be the same, except by accident. You know the filesize beforehand with 2-pass VBR, you do not with CQ. Although those who love order and certainty might find this an inconvenience, it is actually the reason why CQ mode is good

I disagree a little bit with Banjazzer here. Assuming an equal size of say 10 mb the 2-pass VBR will yield certain quality of 1 minute. With the CQ mode, that same file size may yield a low quality movie of 1 + x minutes long if the CQ is set too low. If set too high, that 10 mb might yield an excellent quality movie of 1 - x minutes long. So theoretically, if you want the best quality, up your minimum and maximum bitrate and set your CQ to a high setting. Of course, you may end up with a file size that you have to put in 3 or even 4 cds. If you want to put in them in 2 cds, then you have to compromise and set your bitrates and CQ setting lower. This is where practice comes in in order to know which are the best settings for a certain movie of a certain length, type ( the movie may be always of dark scenes, fast scenes, etc.) so that you can come up with a file size that would optimize 2 cds.

I think Banjazzer is explaining very simply and correctly what CQ is all about. The others are just stuck in a paradigm that they can't envision it at all. They still keep insisting that CQ works with an average bitrate. For God's sake people! CQ does not care about average bitrates! It only cares about maintaining that quality that you set it to do!

This thread is giving me a headache!

[Nelson37]

Banjanzzer, D_head - You are truly missing it, based on invalid comparisons. My HUNDREDS of test encodes tell me that for ANY given quality level, X-pass will give a better video for equivalent size. Proper knowledge of bitrate control allows me to set MY quality level exactly where I want it, for ANY movie. Do YOU know that your "Q" of 60, or 70, or 80 will give the desired quality for a given movie?

CQ will make a good looking video, fast and easy. But X-pass will make a BETTER video on the same number of disks, every time. And CCE will do 2 or 3 passes in the same time as TMPGenc CQ.

Vitualis - Caveat #7 - I thought standard DVD drives were 1x, which translates to 8x CD speed? This relates to my bitrate spike concept, which I think you alluded to in your excellent post and I believe is one area where CQ differs from x-pass. My totally visual tests agree with your analysis, in that CQ seemed to maintain better quality in difficult scenes than equivalent x-pass, with a slight loss in more static scenes.
X-pass with a higher max level (3000-3500, which works in my Apex) was better than CQ in ALL scenes. As regards these "spikes" and the SVCD spec, does not an AVG bitrate of 2500 inherently imply that some areas will fall below and some ABOVE that 2500 level? Is there an actual Never to Exceed bitrate? all I can find is the overall average limit. If the 2500 to 2800 is purely the current mechanical limit, are not faster drives being introduced so that this limit will be raised?

[energy80s]

I'm amazed that this thread is still going! All that the poster wanted to know was if there is really much difference between CBR, VBR and CQ coding of an MPEG file with TMPGenc. And the simple answer is .... NO! The main difference between VBR and the other methods is that it takes a lot longer to encode and you can determine the exact (well, almost exact) filesize in advance. At the end of the day, use the method that you feel most comfortable with. If it looks good to you, then there is nothing wrong with it!

[smokingweed3000]

would u guys recommend using 4 pass VBR with max 3000, min 1000, and avg 2500 for a VCD in CCE

[smokingweed3000]

would u guys recommend using 4 pass VBR with max 3000, min 1000, and avg 2500 for a VCD in CCE?, i could go for SVCD but only if u dont recommned VCD.

[banjazzer]

energy80s

Absolutely use whatever mode you want to. They can all give excellent results with the right settings. However, the thread has taken on a life of its own independent of the original post, and I think that is no bad thing. At the very least it has made us all examine (I hope) our own practice, and why we do things the way we do. Personally I have used both 2-pass VBR and CQ, and been very happy with the results from both. I must admit I don't like the time taken with 2-pass.

The thread has become very big, but that is no problem, either. No-one is forced to pay any attntion to it at all. Personally I have found it all fascinating, and was completely unaware mpg encoding could become such an emotive topic.

[shardison]

My two cents.

I captured an old VHS clip at 6000kbs.

I re-encoded the clip to 4300kbs average 2X pass with a 2500 to 6000 range to work with. (using TMPGEnc) I also resized from 640x480 to 720x480 for DVD. (perhaps a clue?)

I then encoded the same clip using CQ, at CQ=70 quality; and by pure luck it ended up 10% smaller than the multipass clip.

The difference between the clips was obvious.

CQ looked better - it had the visual depth, detail, and contrast of the original capture. The multipass was "flat" looking, had less detail, and lost contrast. Not by much, mind you, but I could see the difference.

That shouldn't be true, but that's how I judged it.

[energy80s]

would u guys recommend using 4 pass VBR with max 3000, min 1000, and avg 2500 for a VCD in CCE?, i could go for SVCD but only if u dont recommned VCD.

err ... NO!!

[injunpana]

Banjanzzer, D_head - You are truly missing it, based on invalid comparisons. My HUNDREDS of test encodes tell me that for [bold]ANY given quality level, X-pass will give a better video for equivalent size[/bold]. Proper knowledge of bitrate control allows me to set MY quality level exactly where I want it, for ANY movie. Do YOU know that your "Q" of 60, or 70, or 80 will give the desired quality for a given movie?

But how would you know that you set them, 2-pass VBR and CQ, at the same quality when they operate at different parameters, 2-pass VBR works on average bitrates while CQ operates on average quality. You simply wouldn't know exactly.

The best gauge we could is to compare a movie done in 2-pass VBR that would optimize 1 cd and to compare to that done by CQ optimizing exactly 1 cd too. Now that would be fair, wouldn't it? Of course, you would find out, CQ wins everytime.

[Kdiddy]

The problem is that so many on these forums make the mistake of doing is taking biasly written encoders and making generalize statements based upon them. In THEORY were CQ & x-pass VBR done an truly unbias encoder at equal parameters, then there video clips would be same. As stated b4, with CQ you can expect a faster encode with an unpredictable filesize and vice versa for x-pass VBR. Because you feel that CQ is better is a testament to how TMPG was written, NOT CQ's supposed prowess over x-pass VBR. I find so odd that all the proponents of CQ never mentioned it in conjunction with CCE, only TMPG. In trying CCE's CQ option, you see that it falls well short of its X-pass counterpart.

I have yet to see any TMPG CQ mpeg2 480x480 at equal paremeters that can compare to that of a CCE's 3-pass mpeg2 480x480. however, TMPG's mpeg1 option at VCD specs, outperforms that of CCE. Again, merely a fact of how the encoders are written, but not a reflection on mpeg1 vs. mpeg2 and/or CQ vs. x-pass VBR .

[kinneera]

Vitualis - Caveat #7 - I thought standard DVD drives were 1x, which translates to 8x CD speed?

This is the equivalent data transfer speed when reading from DVDs. It is not directly related to the mechanical spin speed, but rather the density of data possible on the media. Thus, since an SVCD is still on CD media, it is more constrained, on the order of 2-3x.

Might as well toss my little drop into the bucket here on this thread. You can run any CQ encode, even a theoretical pure CQ in which no maximum is specified, and when the final result is produced you can obtain an average bitrate from it. If you then turn around and put this average into an x-pass encode (equal measuring stick as previously mentioned), the resulting file will be exactly the same size but will look better with each additional pass.

[injunpana]

Might as well toss my little drop into the bucket here on this thread. You can run any CQ encode, even a theoretical pure CQ in which no maximum is specified, and when the final result is produced you can obtain an average bitrate from it. If you then turn around and put this average into an x-pass encode (equal measuring stick as previously mentioned), the resulting file will be exactly the same size but will look better with each additional pass.

Again, people still insists on tying down CQ to average bitrates. Of course you can get the average under any line or curve - that's mathematics, but that average won't mean a thing to CQ. Because it's not concerned with the bitrates, it's only concerned in maintaining the quality you set it to.

Your proposal is just like saying, "what if we turn it around and say that you run your 2-pass VBR, then we get the average quantisization and use that quantisization and run it by CQ?" I would like to see the comparison result of that. And your measuring stick is not that practical, because in real life you don't encode by setting your bitrates and other parameters completely disregarding whether it would fit or optimize your chosen number of cds you would like to put it to.

In real life, you first choose how many discs you would like to use, then try to optimize that chosen capacity. What use would a beautiful encode be if it is way over the capacities of two cds you prefer. Then you would have to reencode again, or otherwise be contented with 3 cds.

What I suggested is better. Let's put a given movie with complete credits in a single cd using CQ and VBR and let's see what looks better. It's too easy to put a short movie in one cd with acceptable quality so let's select a long movie, say THE ABYSS so that we can really see the difference between the two. What say you people?

[kinneera]

Again, people still insists on tying down CQ to average bitrates. Of course you can get the average under any line or curve - that's mathematics, but that average won't mean a thing to CQ. Because it's not concerned with the bitrates, it's only concerned in maintaining the quality you set it to.

I'm not going to waste too much breath on this because you are obviously totally missing the point, which has been explained by numerous people including two moderators. To say CQ isn't concerned with bitrates is pure idiocy - all encoding is concerned with bitrates. In fact, CQ's biggest concern is ensuring that there are enough bits to encode a given scene at a given quality. So if you extract the average bitrate that results from the CQ encode - which you concede is possible - the very least that the x-pass VBR can achieve is the exact same result. But since it has a priori knowledge of the complexity levels throughout the whole move, which CQ does not, it can do better, and more so with each given pass. It's just a matter of how much time you have.

Your proposal is just like saying, "what if we turn it around and say that you run your 2-pass VBR, then we get the average quantisization and use that quantisization and run it by CQ?"

Well of course this is nonsensical because you would be using 2-pass to achieve higher quality in a limited storage space, not quality at any cost. But to answer your question, the two pass would produce a more efficient bitrate allocation at the specified bitrate, and thus a better average quantization (funny how those extra passes are useful like that) and so the CQ would produce either a somewhat lower quality video or a somewhat larger file if you did that.

And your measuring stick is not that practical, because in real life you don't encode by setting your bitrates and other parameters completely disregarding whether it would fit or optimize your chosen number of cds you would like to put it to.

The argument I made is theoretical extension of logic. If CQ can be beat when no constraints are made, then obviously it's going to be beat in more constrained situations, particularly the ones where bitrate allocation efficiency is so paramount. Other than that, this seems to be the same argument that has been made against CQ - too unpredictable just to get a lower quality result.

What I suggested is better. Let's put a given movie with complete credits in a single cd using CQ and VBR and let's see what looks better. It's too easy to put a short movie in one cd with acceptable quality so let's select a long movie, say THE ABYSS so that we can really see the difference between the two. What say you people?

Surely you realize that this has been done many times before, including I'm sure by most everyone who has made the point about the advantage of x-pass encoding? Do it again if you so desire.

To tell the truth, I can't think of any better argument regarding the superiority of multi-pass encoding than the fact that CCE costs $2000 to get that feature, and greater than 2-passes is only possible in TMPGEnc with the most expensive version.