battle of the CBR/VBR/CQ modes

[musicnyman]

hi guys,
i've been reading and searching the forum for days before i posted this, and i've gotten so many different oponions on these different types of modes. i've read the article from tangentsoft and they recommend VBR over CQ and it makes sense. most of the posts were for VCD or DIVX or DV conversions. I'm sure a lot of people are backing up their dvds straight to SVCD's, so for you guys, what has worked out best? i have done this for almost 6 months now so i've experimented a lot but wanted to know if i have missed anything from all the posts on this board and my experimentation. so here are the questions:

1) as far as i know, it's not worth using VBR if you going to use the max bitrate in VBR to 2520 for a SVCD anyway right? many times i have a 3 cd SVCD backup that I set to 2520kbs since 2 cds would cause a very low bitrate. in cases where the max bitrate is the avg bitrate in VBR, use CBR?

2) is CQ even worth experimenting with in SVCD's cds? does anyone use this for DVD->SVCD conversion? i personally don't like it since target file size is very unpredictable and 2 pass VBR looks the same.

basically just want to make sure i'm not "ruining" my backups and should be using something else instead before removing the VOB files. let me know if 2 pass VBR is the way to go for DVD -> SVCD conersions or if someone else has had better results.

Thanks,

[adam]

1) You are correct. A vbr encoded file with a max bitrate of 2.5mbits should look no better than one encoded at 2.5 cbr, though it can achieve the same level of quality in a smaller filesize. In this situation cbr is better to use simply because its faster.

2) CQ encodes per a quality setting only. It keeps quantization set to a constant level and adjusts bitrate around that, not only giving unpredictable filesizes but also wasting bitrate in many cases because it is overly generous with it, which simply results in a larger file. Multipass vbr encodes according to a quality and a compression setting.

At the same avg bitrate CQ and 1-pass vbr should look identical. The only difference between the two modes is how the amount of bitrate to use is selected...so at the same bitrate they effectively result in the same encode. Similarly, At the same bitrate setting multi-pass vbr is no different from 1-pass vbr or CQ either, except that it analyzes the file before enoding which allows it to allocate bitrate more effectively.

CQ is useful if compression is not an issue or if you want to speed up your encodes, of course this is assuming that you hit the target size on the first try which rarely happens. If you are making svds than I really don't see how compression can't be an issue. Nobody wants a 5-6 cd movie and if they did, they could just use cbr. The only practical purpose for CQ that I can see for authoring svcds it to speed up the encode.

Simply put, multipass vbr is the most effective form of automatic bitrate allocation. The only way to be beat it would be to do it by hand. Trust me, your encodes are not suffering by using 2-pass vbr.

[banjazzer]

CQ will actually give you better quality than x-pass VBR.

http://tangentsoft.net/video/mpeg/enc-modes.html

[musicnyman]

awesome, thanks guys, that's what i thought. since i need a predictable file size, since 2 pass VBR gets me an 800 meg file size every single time, nothing over, nothing under, it would be preferable for SVCD's. i'll keep using VBR for anything below 2400kbs and CBR for SVCD's that are going to stay 2520 anyway.

Very helpful! Thanks!

[SatStorm]

It is very easy to understand, see:
If you go straight for quality, at the less possible time and you want to stay in standards, then CBR is the way to go.

For quality freaks, there is a possibility to use 2 Pass VBR and get better quality than CBR at the same bitrate and in standards, but it will take the double time and IT IS NOT WORTH IT!

If you want to try what I am talking about, then set 2 pass VBR, highest quality, with minimum, average and maximum bitrate @2520kb/s. That way, you have (in theory) better allocation of data per frame. It is like 2 Pass CBR or something. Totally theory, is praxis ain't worth even to try...

I am finding the same thing CQ in Tmpgenc is giving be better results than x-pass VBR.

[musicnyman]

i wish i could use CQ, but i'd have to encode the file twice anyway to find the right source file size since i like to fit all my svcds on an 800 meg file to squeeze every lit bit of resource on the cd i have.

what do most people use for the vbr min bitrate? the default in tmpgenc is 300. seems kinda low, do most people use 500 or 1000? i don't want to go too high because then it defeats the purpose of VBR and minimizing the bitrate at blank or very low motion scenes.

[offline]

vbr min bitrate for SVCD as required by various film trading
groups for widescreen letterbox is 1600... but 300 works well enough
for home stuff.

[musicnyman]

so for backing up my dvds, and being a quality freak that i am, i should keep min bitrate to 1600 and max bitrate to 2520 and play with avg bitrate to fit this on 1-2 cds. will this also help get rid of those annoying macroblocks? i've tried multiple things like increasing bitrate to the max using CBR and VBR, highest motion precision search, dc bits to 10. i don't like using the soften block noise since it makes the movie blurry. any other suggestions? if you do use soften block noise, what's your setting?

[banjazzer]

If you want to get the best quality in the shortest time, encode at CQ using a setting of 80+. Forget the filesize, and cut what you get into 800MB pieces, using the number of CDs that it takes. This is very quick, excellent quality, and cheap, since the price of CDs is so low. 8)

[portow]

My experiance has been that CQ in TMPGEnc provides a smoother, less blocky file than 1 or 2 pass.

I'm mainly an archivist and as such, for various captures, I prefer quality over file size (within reason, of course! )

[adam]

CQ will actually give you better quality than x-pass VBR.

encode at CQ using a setting of 80+. Forget the filesize, and cut what you get into 800MB pieces.

My experiance has been that CQ in TMPGEnc provides a smoother, less blocky file than 1 or 2 pass.

I prefer quality over file size

Typical. Every single person who says CQ yields superior quality to x-pass vbr always says the same thing. Did anybody actually read my post? CQ looks better because you are CHOOSING to throw compression out the window and only set a constant QUALITY level. You could do the exact same thing with x-pass vbr by simply CHOOSING to use a higher bitrate except that it would analyze the file before encoding, ensuring better bitrate allocation.

If you want to compare CQ and x-pass vbr you must first encode in CQ then run the resulting mpg through a bitrate viewer to determine the avg bitrate used. Then in x-pass vbr you must use that avg and the same min and max settings as the CQ encoded file and only then will you have an accurate comparison between x-pass vbr and CQ. If you are comparing CQ and x-pass vbr and you didnt use the EXACT same bitrate levels then your comparison is flawed and is most likely placing a significant handicap on x-pass vbr...no wonder CQ wins.

Sorry if I sound irritated but I am. The same misinformation perpetuates on internet forums because people don't realize how inaccurate their own tests are. If you research mpg encoding you will see that CQ is nothing more than 1-pass vbr. At the same settings CQ and x-pass vbr are literally doing the exact same thing, except x-pass vbr does it in more passes. The real question here is not whether CQ is better than x-pass vbr, its whether 1-pass vbr is better than (x+1)pass vbr, which it isn't. TMPGenc is an incredible encoder and for the price, its really outstanding but are people really gullible enough to believe that it has somehow stumbled onto some sort of holy grail of mpeg encoding? Why does no other mpg encoder in existence have a CQ mode? The answer is because you don't need one! You can achieve the same result with any other form of 1-pass vbr.

banjazzer that link is interesting and I have seen it before but its hardly evidence of anything. Its simply that guy's subjective opinion and no more credible than say, this forum post. It seems his tests are as flawed as every other CQ supporter. Sorry I'm not trying to pick on you, this post is directed towards anyone who claims CQ is superior without really testing first. If this doesn't apply to you than simply disregard this rant.

musicnyman, all I can tell you is to try each mode in several situations and judge for yourself, but do not make the same mistake that so many other people make. Ensure that your x-pass vbr encoded file is actually the same size as the CQ encoded one before concluding that it is inferior.

[Laddydaddy]

For what my two cents are worth I've used the CQ setting in tmpg v2.54 with the maximum bitrate set at 1600. By using virtual dub as a frameserver and the following v-dub filters, I was able to get an excellent quality rip with a small file size. The filters are;
Smart resize
Hue/Intensity/Color
Sharpen
The file size was actually smaller than if I encoded it at a CBR of 1150, the standard for vcd. I've tried different combinations in tmpg and this gave me the best quality with the smallest file size. A 50 minute file showed up as a 320MB file with a time of 32 minutes when burned using Nero.

[banjazzer]

banjazzer that link is interesting and I have seen it before but its hardly evidence of anything. Its simply that guy's subjective opinion and no more credible than say, this forum post. It seems his tests are as flawed as every other CQ supporter. Sorry I'm not trying to pick on you, this post is directed towards anyone who claims CQ is superior without really testing first. If this doesn't apply to you than simply disregard this rant.

Well, the link I mention seems quite convincing to me, but I'm not trying to make a religion out of it. Personally, I am happy with the quality of 2-pass VBR and CQ. I like the predictability of VBR and the speed of CQ. But, unless I have missed something, with x-pass VBR the size-predictability comes because the average bitrate is fixed, ie no more bits are available whatever may be required. The bits can be assigned where needed admittedly, but the total of bits used is constrained by the average bitrate. With CQ no average bitrate is specified, hence the unpredictability of size. If more bits are needed they are used, within the limits of the max bitrate. The total of bits used is constrained by the maximum bitrate. It would also seem to me that occasionally with difficult subjects, VBR is more likely to get caught out than CQ precisely because the overall bitrate is fixed beforehand. I have done no tests on this. I am only going by what others have written, and by what would appear to me to be logical.

[smokingweed3000]

so Adam, r u saying that CQ and VBR at lets say 2 or 3 passes r exactly the same except for the file size?

Im gonna have to try that out and to compare those two yr way.

[offline]

i don't like using the soften block noise since it makes the movie blurry. any other suggestions? if you do use soften block noise, what's your setting?

I don't (at least for SVCD's) I use temporal smoother using
Avisynth frame server (at default settings) for most projects.
You are 100% correct about the blurry effect of soften block
noise!

If there is one great filter internal to tmpeg, it has to be the de-noise
filter. Fantastic for clearing up noisy VHS/TV captures but
not much use when dealing with DVD quality sources. Other
than that I tend to use external filters.

[adam]

banjazzer: Like I said, CQ and x-pass vbr are basically the same thing, they are both vbr. For both, the total amount of bitrate that can be applied to any given sec of video is set by your max. In x-pass vbr the total amount of bits to be used is set by your avg and for CQ it is set by your quality setting, except of course you have no way of knowing how much compression will actually be used. They are both doing the EXACT same thing except that x-pass vbr does it better because it has the benefit of analyzing your video first.

unless I have missed something, with x-pass VBR the size-predictability comes because the average bitrate is fixed, ie no more bits are available whatever may be required. The bits can be assigned where needed admittedly, but the total of bits used is constrained by the average bitrate. With CQ no average bitrate is specified, hence the unpredictability of size. If more bits are needed they are used, within the limits of the max bitrate.

The total number of bits is always constrained by your avg, thats what your avg is. It doesn't matter whether you manually set an avg or your avg is determined as a result of the quantization level you set, either way both encoding modes have the same amount of bitrate to work with. The important part is where your peaks and valleys are, THATS what makes one encoding mode better than another and thats why x-pass vbr is superior. Your argument for why CQ has an advantage over x-pass vbr is simply that its bitrate allocation determines the avg automatically rather than allowing you to set it manually. Well setting your avg manually is no more arbitrary than setting your CQ level! In fact its less.

I really think I summed it up pretty well in my original post...if compression is of no concern than CQ is a viable method of encoding to use. If compression is of any concern than CQ can be used to speed up the encoding process at the expense of quality. The original poster wanted to know what the highest quality encoding method is and the answer is x-pass vbr. CQ will NOT produce higher quality video than x-pass vbr.

so Adam, r u saying that CQ and VBR at lets say 2 or 3 passes r exactly the same except for the file size?

Im gonna have to try that out and to compare those two yr way.

Don't forget that CQ is still vbr. I think you misunderstood what I was saying. At the same bitrate levels yes CQ and x-pass vbr are doing the same thing except multipass vbr will always result in better quality because will have better bitrate allocation. If the bitrate levels are the same then the size will be also so at the same filesize x-pass vbr will ALWAYS be higher quality than CQ. Since x-pass vbr allocates bitrate more effectively you could use less bitrate and still maintain the same relative quality as a CQ encoded file at a higher bitrate. This difference in bitrate would depend on an incredible multitude of things such as the source, additional encoding settings, and most importantly your subjective opinion. So the answer to your question is yes, it is possible to create a file using x-pass vbr which is basically the same as a CQ encoded one except that it is smaller in size. However to do this your going to have to adjust the bitrate settings of one encoding mode or the other and it will basically just be a matter of trial and error. Simply put, at the same bitrate levels x-pass vbr will always be higher quality than CQ. At higher bitrate levels CQ MAY be higher quality than x-pass vbr but of course this means a larger filesize.

[offline]

At higher bitrate levels CQ MAY be higher quality than x-pass vbr but of course this means a larger filesize.

I agree 100%. x-pass VBR can give the best results under many
conditions with a restricted filesize, while CQ is a great timesaver
if a certain quality level is required without strict constraints
on the final filesize (it may be more or less ..mostly more.)

It is difficult to tell which is better as picture quality is a
subjective beast and conditions vary wildly between video
sources and encoder qualities. For me VBR seems better
for SVHS and CQ for vcd (I say "Seems"!)

[smokingweed3000]

ok, thanks a lot adam, i got it now and im gonna try it out on a couple of movies.

[SatStorm]

About the speed of CQ_VBR.. (and if we talking about encoding with tmpgenc)

2 Pass VBR at motion etimate search (fast) mode takes almost the same time to encode as CQ_VBR @ Highest quality mode.

2 Pass is always better if you know what you are doing.

If you have a CPU with sse support the speed of CQ_VBR ain't an issue any more! For older cpu's maybe...

Just did an encode of Planet of the apes Using CQ mode in Tmpgenc, (3 CDs) Min 300 max 3400 Q=92. (1hr54mins long)

All I can say is the quality is awsome, and much faster than the 2passVBR.

Just another coment,
I made over 20 samples from Ghosts of Mars,Using various settings(res& bitrate) for both 2pass VBR & CQ, played them over & over on my standalone. CQ came out on top and the filesize was not a hell of a lot bigger.

[adam]

Well at those bitrate you are not going to see much difference between CQ and 2-pass vbr because your quality topped off a long time ago, unless of course your using the dvd's source resolution of 720x480/576.

I'm not sure if you knew this or not but a quality setting of %100 in CQ is literally CBR. It will simply take your max setting and use that for the whole movie. So at a quality setting of 92 you are practically using cbr. So at 480x480 ~3200 CBR you are getting at least as many bits per pixel as the source. The source quality is the limiting factor obviously so like I said, your quality topped off a while ago. I think you would be suprised to see just how much smaller a 2-pass vbr encoded file could get and still look just as good as that CQ encode. And even if you are encoding at the source's resolution, you should still be able to decrease your filesize fairly significantly using 2-pass vbr without losing any relative quality.

[banjazzer]

The total number of bits is always constrained by your avg, thats what your avg is. It doesn't matter whether you manually set an avg or your avg is determined as a result of the quantization level you set, either way both encoding modes have the same amount of bitrate to work with. The important part is where your peaks and valleys are, THATS what makes one encoding mode better than another and thats why x-pass vbr is superior.

Whilst I don't think we disagree by very much, actually, I would say there is a lot of difference between choosing your average as a bitrate rather than as a quality setting. The former is constrained to a finite number of bits, whereas the CQ one isn't, or else every CQ encode would be a fixed size if the max and min settings were constrained equally. They are not - it very much depends on the video material. Both encoding modes do not have the same amount of bits to play with. Try a 2-pass VBR and a CQ encode on 2 samples of video. Of the 4 encodes, the 2 2-pass VBR will be the same size, the 2 CQ ones will almost certainly not. An x-pass VBR will be a fixed size no matter what the source material, given a certain average setting. Remember also that one reason why people choose 2-pass VBR is that they want to constrain the finished product to a fixed filesize, which bears no relationship to the type of source material. With CQ you hope for a certain filesize, but this will depend on the compressability of the source.

Don't forget that CQ is still vbr. I think you misunderstood what I was saying. At the same bitrate levels yes CQ and x-pass vbr are doing the same thing except multipass vbr will always result in better quality because will have better bitrate allocation.

I know that both methods are VBR. However the x-pass VBR has a fixed amount of bits it can use, so therefore, in order to best use these, it needs a preliminary pass. Since the CQ method is constrained by the maximum bitrate and the quality setting there is no need for a preliminary pass. It is not constrained to a total amount of bits for the encode.

I think both methods have their uses, but I would disagree that x-pass VBR is intrinsically better than CQ, except in being able to predict filesize.

[Helpy Fairy]

inny minny minny moe
catch a tiger by its toe
if it hollars let him go
my mama sez
"you"
"are"
"going"
"to"
"be"
"the"
"one"

yey!















2-pass vbr is the blessed one!

I think both methods have their uses, but I would disagree that x-pass VBR is intrinsically better than CQ, except in being able to predict filesize.

Banjazzar I think you summed things up the best in your last comment, I think maybe some of the rest of you should maybe make a few tests. From my personal experiance CQ is giving better Quality, is much faster and the file size 5-15% bigger. If someone has an FTP I`d be willing to post some samples& the template I´m using.

[adam]

banjazzer you are still not grasping what I am saying. How can you possibly make an encode at an avg bitrate of 2.5mbits and compare it to an encode at an avg of 2mbits and draw any conclusion about the encoding mode? If given the same amount of bitrate to work with x-pass vbr will always look better, as I have already explained. Well what is the limiting factor when determining your avg bitrate? Of course its the size of your media. So when compression is an issue, basically anytime you plan on burning your resulting mpg to a disk, you know what the avg must be. Either you can set this avg in x-pass vbr or you can hope to achieve it with CQ but either way the x-pass vbr encoded file will always be higher quality. Like I keep saying over and over again, if compression is of any concern x-pass vbr is always better and will achieve a higher quality. And lets be honest, is compression ever not an issue? What's the point in transcoding a perfect source except to make it smaller?

If compression isnt of any concern than yes you can simply use CQ and let the encoder throw as much bitrate as it wants at it to maintain your quantization level, and like I have said, this is one instance where CQ is viable. But whatever avg bitrate CQ ends up using you could in turn use that in 2-pass vbr and the result would always be higher quality. If you don't believe me here then I give up because this fact should be painfully obvious by now.

Given the same movie, what differentiates a quality level of 80 versus a quality level of 75? Is it something tangible that you can see? How do you determine the quality setting to use? Isn't it often different depending on the nature of your source? Don't you just get a feel for what works and what doesnt based on past experiences? And isnt it all relative to your max bitrate setting anyway? Your quality level in CQ is arbitrarily picked. If I were of the mindset that compression was out the window then I could also arbitrarily pick an avg in 2-pass vbr and, given your CQ settings and my knowledge from past experiences, I could probably come pretty close in either quality or filesize. Both encoding modes will use roughly the same amount of total bits if you make logical and comparable settings, and they will use exactly the same amount of bits if set it up to do so(encode CQ first and determine avg bitrate.) How can you make a comparison between the two modes if you don't do this? Is CQ a higher quality encoding method simply because you set it to use more bitrate? If thats the case then I can prove that an RM file is higher quality than a DVD.

"I would disagree that x-pass VBR is intrinsically better than CQ, except in being able to predict filesize."

I did not say it is intrinsically better than CQ. Obviously CQ has its uses which may make it more appealing to some, (faster encoding at the expense of quality and predictability.) What I did say is that x-pass vbr is intrinsically higher quality than CQ and this is a fact. You think the only point of doing multiple passes is to guarantee a set avg and a set filesize? No wonder your not understanding the point of my post.

The original poster stated that he wanted to use the best encoding mode, obviously something which is subjective, but he also stated that he wanted to ensure that he didn't "ruin" his backups. Well I don't think longer encoding time can possibly ruin a backup so I think it was obvious that quality was his primary concern. I suggested that x-pass vbr was the highest quality encoding method and you specifically said, "CQ will actually give you better quality than x-pass VBR." I'm sorry but you could not be more incorrect. Again I don't mean to pick on you but this is a widespread misconception and I really think alot of people are mislead because of it. CQ is not higher quality than x-pass vbr and there is simply no way to justify that statement.

CQ is giving better Quality... and the file size 5-15% bigger.

No offense D_Head but you obviously didn't understand a word of this thread.

[energy80s]

Why not just use CBR mode and have done with it! At normal SVCD bitrates you will have 40 minutes of video per disc and the encoding time will be much less than using any type of VBR. Picture quality is fine (even better if you use CVD standards of 352x576(480) rather than the SVCD size adn willl be more compatible too.

[smokingweed3000]

just read what adam says... it makes sense. u cant just put the same bitrate in for both modes and compare the quality and size, cause CQ is gonna use the max bitrate to achieve the quality that u set it to and x-pass vbr is gonna use that bitrate for the file size, therefore the VBR encoded file is gonna be smaller and better quality then the CQ file.

if u wanna compare the CQ mode to the VBR mode u have to have exactly the same bitrate and other settings, therefore encoding the CQ file first and opening that file in a bitrate viewer and then use the same min, max, and avg bitrates in the X-pass VBR mode. Only after u have done that u could compare the two files.

u cant just take two files one at 1000 and the other 2000 and compare them.

[smokingweed3000]

cause CBR is a waste of time and disc space, the only thing u r right there is that CBR is more compatible

[banjazzer]

adam

Just as I appear not to be understanding you, I don't think you are understanding me. Given any specific source material to encode you decide on a starting point, either with a quality setting (CQ) or with a filesize in mind using the average bitrate setting (2-pass VBR). The 2-pass VBR will produce the given filesize no matter what. With the CQ, you do a sample with an arbitrary quality setting, say 75. If the source material is not particularly compressible you will get a larger filesize than you anticipated. You then decide whether to go for more discs or to drop the quality setting. You do not get this choice with 2-pass VBR. The decision is made at the outset regardless of the compressibility of the source material. On one level, what you say is correct - 2-pass VBR may give you the best use of your bits. But it gives no indication of how many bits to use. In practice, how many people do you suppose make a 2-pass VBR encode and then decide to do it again because the quality is not as good as they hoped? With a CQ setting you can take account of the nature of the source material, and you get adequate warning of this by a simple proportion calculation after doing one or more samples. So, although you may not agree, I would say x-pass VBR may be theoretically best if you know how many bits you are going to use, but CQ is practically best because you don't actually know how many bits are going to be needed. x-pass VBR takes only account of the amount of bits you decide to give it, which is decided at the outset. Since it does 2 passes it makes best use of them. CQ mode takes account only of the quality setting, so if more bits are needed they will get used. IMO this does not make x-pass VBR a superior method, regardless of taking twice the time to encode. On the contrary, I would say that CQ is the superior method, if quality rather than size is the overriding aim. x-pass VBR may give absolute best use of the bits you give it, but gives no indication what that number of bits should be.

Although you seem to think the reference I gave http://tangentsoft.net/video/mpeg/enc-modes.html
not worthy of consideration, I am not quite sure why.
I certainly agree with the conclusion.

"VBR modes are worth using, when you're not forced to use CBR. And when you can get away with it, pure CQ modes are the best VBR types of all. You have to work harder to get good video, but you can achieve absolutely stunning results with a properly-tuned CQ encoder."