battle of the CBR/VBR/CQ modes

[vidkid®]

I've been following this thread with great interest because I have been testing CQ vs. 2-pass VBR recently on my own trying to get the best quality out of some so-so captures from the tv tuner.

IMO (<- note), given my source and my tests, I found CQ to be better. I just assumed I was doing something wrong with 2-pass because I know that 1. better quality with CQ didn't seem right and 2. what all the experts here and elsewhere say about multiple pass being better. At least this is what I thought until I saw the link mentioned above (http://tangentsoft.net/video/mpeg/enc-modes.html)

So, when I saw Adam's post, I decided to put it to the test.

adam wrote:

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.

Anyway, my tests yielded an interesting finding that I'm hoping someone who knows a lot more about this than me can explain.

When I set TMPGEnc to MPEG-2, 480x480, CQ 80 with 500 min. (pad not to be lower than min.)/2496 max. (I got these settings from banjazzer.) the resulting file run through bitrate viewer says Peak=2253 KBit, Avg.=1289, the "Q." level Peak=6.46 and Avg.=3.89. The "nom. bitrate" is reported at 2496000 Bit/Sec.

When I set TMPGEnc to MPEG-2, 480x480, 2-pass VBR with 500 min. (pad not to be lower than min.)/1289 avg.(from above)/2496 max., and run the resulting file through bitrate viewer, it says Peak=1762, Avg.=1207, and the "Q." level Peak=10.18 and Avg.=5.45 The "nom. bitrate" is reported at 2496000 Bit/Sec.

The CQ file is ~7% bigger. Watching the result with my eyes, the CQ file looks slightly better. However, judging from bitrate viewer, there's some room for improvement from the 2-pass encoding.

I guess I'm wondering:
- What is the Q. level and why is it SO different between CQ and 2-Pass??
- Can anything be done to get TMPGEnc to more closely honor the Avg. (1289 here) I set for 2-pass AND bump up the peak?? without purchasing PLUS
- Where did Adam go on his vacation so I can get my answers today?? (that's a joke - I hope he's enjoying himself)

[banjazzer]

without purchasing PLUS

As an aside, what exactly is the difference between PLUS and the ordinary version. I'm using PLUS, but I can't see a lot of difference!

[vidkid®]

Well, according to this http://www.pegasys-inc.com/e_release02.html not much. I was (mistakenly apparently) under the impression that you could do more than 2-pass with Plus. The info. says Dec. 2001, so maybe something has changed.

[portow]

I ran three files over the last 24 hours, three of them at 2-PassVBR:

average bitrate of 2300,
max at 2450
min 300,
with noise reduction HQ turned on (10-1-10).
Bitrate Viewer reported an avg of 2173 (Q of 13.61)

I compaired the results to three files at

CQ 90,
2400 Max,
300 Min
Bitrate Viewer reported an avg of 2296 (12.64)

I thought the files encoded at CQ looked slightly better. Was it my imagination? The VBR was smaller (95,000KB vs 99,000 for a 5'18" minute file) but I could swear the CQ looked better.


So I had my wife look at the files without telling her what was what. "Which looks better... 1.... or 2"

I did it a few times and CQ won out each time. I understand in theory why 2-Pass would be better at the same setting, compressing the file more efficiently where needed, but my tests don't hold out. I need to print this thread and digest it further.

[Nelson37]

Just to throw in my 2 cents worth -

1. CQ makes it easier to get good results as less knowledge about bitrate is needed to make a good looking video. It is not impossibe to make a 2-pass that looks better for the same size, it is just more difficult.

2. As seen above, it is difficult to match filesizes exactly to get an accurate comparison. As file size nears equal, quality difference disappears.

3. I believe the tolerance factors, or adherence to avg or max, is somehow different with CQ. (see #4) The fact remains that IF YOU PICK THE RIGHT PARAMETERS X-pass is better

4. I use 4-pass VBR, max 3000, min 500, avg 1600-2400. My results blow away anything I got with CQ. I arrived at these numbers partly while attempting to get CQ's quality with multipass file size predictability. Note the significant difference in bitrate PEAK in the above example, hence my max 3000.

[kwag]

CQ is NOT equal to VBR. It's a complete different result. Both in quantization scale and bit rate. Completely different curves.
http://www.tecoltd.com/enctest/concepts.htm

kwag

[banjazzer]

4. I use 4-pass VBR

Which encoder do you use? I presume not TMPGEnc as I can only see the option for 2-pass.

[vidkid®]

CQ is NOT equal to VBR. It's a complete different result. Both in quantization scale and bit rate. Completely different curves.
http://www.tecoltd.com/enctest/concepts.htm

kwag

Thanks for the link kwag. Judging from this, I've been getting better Q. levels with my CQ encodes in about the same file space.

[Nelson37]

banjanzzer - I currently use CCE, Although I derived my min and max numbers while testing with TMPGenc 2-pass. My evaluations were based on HUNDREDS of test encodes with incremental changes. It appeared to me that CQ mode was somehow more closely keeping to the higher rate, to get equivalent quality (AND smaller filesize) with 2-pass raising the max bitrate was effective.

NOTE - Q factors and many mathematical tools can give some useful information, however, they will sometimes indicate answers which will conflict with perceived visual quality. The eyes have it.

NOTE #2 - we seem to have agreement that with roughly equal filesizes, quality is very close. Would you consider that a LARGER x-pass file would be of BETTER quality than CQ? Because, once I have decided to use the same number of discs as CQ, my X-pass file will ALWAYS be larger (higher avg bitrate) because I will use ALL of the disk capacity while CQ will leave some space unused.

[banjazzer]

NOTE #2 - we seem to have agreement that with roughly equal filesizes, quality is very close. Would you consider that a LARGER x-pass file would be of BETTER quality than CQ? Because, once I have decided to use the same number of discs as CQ, my X-pass file will ALWAYS be larger (higher avg bitrate) because I will use ALL of the disk capacity while CQ will leave some space unused.

I suspect we only have agreement because adam is on holiday. 8) However, I would assume provided max and min bitrates were set equivalently in both modes, given a high enough average bitrate (2-pass VBR) and a high enough quality setting(CQ) the results would be near as dammit indistinguishable, given roughly equal final filesizes (if that lot makes sense!) Or put another way, I don't know. 8) I have enough difficulty keeping up with my encoding without doing lots of tests!

At the end of the day, though, I think both methods can give exceptionally good results, provided you start out with a high enough av. bitrate, or a high enough quality setting. I do think, though, that there is a big temptation to decide from the outset how many discs you want to use with 2-pass VBR, and to do the encode to that. Depending on the compressibility of the source, the bits allocated may not be enough. CQ mode will take account of this, and you will just get a larger resultant filesize. Which is why, particularly if I am pushed for time, I use a CQ setting of 80 and use as many discs as it takes. OK, I may decide to tweak it a bit if the encode was 2CD+ say 20MB, but generally I would be tempted to use however many discs were required.

If you are asking about more than 2-pass encoding, we are outside the remit of TMPGEnc, but you would hope that there was some reason for doing the extra passes. Where I disagree with adam, though, is that if the av. bitrate is unsufficient for the material, it doesn't matter how many passes you do. 8) CQ is more likely to give good results in most situations than 2-pass VBR, which is not to say that CQ is better than 2-pass VBR.

[energy80s]

One thing that is very obvious from this thread is that there is VERY LITTLE DIFFERENCE between CQ and VBR encoded files except that the CQ ones encode MUCH QUICKER. And I also know of no-one who uses 4 or 5 passes on video files. Who the hell could be bothered with that! If you want that good quality just buy the bloody DVD and have done with it!

[banjazzer]

One thing that is very obvious from this thread is that there is VERY LITTLE DIFFERENCE between CQ and VBR encoded files except that the CQ ones encode MUCH QUICKER.

That is very true, but I would also add that the fundamental approach of the 2 methods is totally different. 2-pass VBR approaches it from a filesize/average bitrate, and CQ approaches it from a quality setting. The unpredictability of the ultimate filesize is CQ's biggest plus - if the file needs to be bigger to satisfy the quality setting and taking into account the source material, it will be.

[dmonty]

Ok but my dad can kick all your dad's asses.

[banjazzer]

My dad only has one ass. 8)

[dmonty]

doh!

[baker]

THIS POST IS SO FUKING STUPIT!


I AM GOING TO SET YOU GUYS BLOODY RIGHT NOW.

USING TMPEG AS AN EXAMPLE.

C.Q. Mode-

A mode where you set the quailty of the image or the ammount of distortion you will allow. This can be set from 1-100 100 being perfect quailty and 1 being a lot of distortion.

2 PASS VBR MODE-

A mode whre INSTED of you setting the q level you set an average bitrate for the encoder to follow. This means you can quite accuratly guestimate the size of your file. Two pass will take longer then q pass BUT will have a smaller file size AND look better at an equal bitrate.

Because q sets the avg bitrate at whatever it likes the q file may appear to look better as it will be able to do what it likes.

IF ANYBODY HAS ANYMORE QUESTIONS ASK ME.

THIS STUPIT POST HAS GONE ON LONG ENOUGH.

[banjazzer]

Two pass will take longer then q pass BUT will have a smaller file size AND look better at an equal bitrate.

Because q sets the avg bitrate at whatever it likes the q file may appear to look better as it will be able to do what it likes.

Someone once said something to the effect "Better to say nothing, and have people suspect you are a fool, than open your mouth and remove all doubt."

Just re-read what you wrote again, and then have a long think. Really, if you can't cope with this thread, then don't bother reading it. For some of us it is interesting - not least because it questions quite a few preconceptions.

Now, if you want people to go to your site, you first have to say something sensible.

[baker]

FOR **** SAKE I DO NOT HAVE A WEB SITE!!!!! THE AMMOUNT OF E-MAIL I GET THANKING ME FOR HAVING SUCH A GREAT SITE IS UNREAL!!!! BUT I DO NOT HAVE A WEBSITE!!!

Baker

[baker]

whats the mistake?

Two pass will take longer then q pass BUT will have a smaller file size AND look better at an equal bitrate.(that sentence is correct)

Because q sets the avg bitrate at whatever it likes the q file may appear to look better as it will be able to do what it likes.(thats true too)

Forgot to add in that last sentence,up the bitrate insted of able to do what it likes.

Baker

[banjazzer]

FOR **** SAKE I DO NOT HAVE A WEB SITE!!!!!

OK. Chill! If you want people to read your guide, stop having a tantrum. Just because you don't like what you see doesn't mean you can suppress it. If you can't stand the heat, get out of the kitchen!

[banjazzer]

Two pass will take longer then q pass BUT will have a smaller file size AND look better at an equal bitrate.(that sentence is correct)

So, you have evidence that with 2 files coded from the same material and with same overall filesize, the 2-pass VBR is better quality than the CQ one? 8) My whole point, which you appear to have missed, is that choosing the filesize at the outset is a bad thing, unless you know beforehand what the compressibility of the material will be. With CQ you don't have that problem. Look, I don't care if you believe what you want to believe, but don't try to suppress the thread because you don't agree!

[baker]

whats my moistake anyhow?

Baker

[baker]

So, you have evidence that with 2 files coded from the same material and with same overall filesize, the 2-pass VBR is better quality than the CQ one? My whole point, which you appear to have missed, is that choosing the filesize at the outset is a bad thing, unless you know beforehand what the compressibility of the material will be. With CQ you don't have that problem. Look, I don't care if you believe what you want to believe, but don't try to suppress the thread because you don't agree!

So, you have evidence that with 2 files coded from the same material and with same overall filesize, the 2-pass VBR is better quality than the CQ one?(thats not what I said at all. of course say a 5000 avg bitrate 2 pass would look the best) My whole point, which you appear to have missed, is that choosing the filesize at the outset is a bad thing, unless you know beforehand what the compressibility of the material will be.(its very easy to work that out using dvd. A 352x576 avg bitrate should be around 1600) With CQ you don't have that problem. Look, I don't care if you believe what you want to believe, but don't try to suppress the thread because you don't agree!(?)

Baker

[banjazzer]

Have you been drinking? 8)

[supercrew]

Ok, now we know CQ is for maximum quality and Speed, and x-pass vbr is for known filesize.

I don't know about anyone else, but as my collection of movie's has grown I have leaned more towards trying to get my movies on 1 disc if at all possible. I know this means giving up a little quality, but even cd-r are dirt cheap they all weight the same therefore when my movies are on a single disc I'm saving 1/2 the weight in my collection.

I'm on the road at times and it can be hard lugging around a bunch of movies if each is on 2 or 3 disks each.

Just my thoughts...

Supercrew

[baker]

yes, a bit. but that dosn't matter. I still can't c my mistake. plz point it out.

Baker

[banjazzer]

Here are my conclusions, seen through my eyes, because I can't see through anyone else's. 8)

If you are concerned foremost about the number of discs you use, and you have time on your hands, go for 2-pass VBR.

If you are concerned foremost about quality, go for CQ, at a reasonable quality setting, and accept the number of discs it takes.

If you are not concerned about the number of discs, both CQ and 2-pass VBR will both give excellent results, although the 2-pass VBR will take a lot longer. If you never want to be caught out quality-wise, go for CQ.

[baker]

I agree apart from

If you are concerned foremost about quality, go for CQ, at a reasonable quality setting, and accept the number of discs it takes.

Baker

[baker]

I agree apart from

If you are concerned foremost about quality, go for CQ, at a reasonable quality setting, and accept the number of discs it takes.

Baker

[baker]

I agree apart from

If you are concerned foremost about quality, go for CQ, at a reasonable quality setting, and accept the number of discs it takes.

Baker