CCE vs 2-pass and CQ?

[davidian]

Hey, I've been around this forum for some time now, and I've been making VCD's and SVCD's since I found Vcdhelp. However, there is something that I find a little strange.

Everyone says that CCE is the encoder that gives best SVCD quality, and that tmpeg's 2-pass vbr is very good too (not as good as CCE but very good)..

So what's the problem!? When I encode with CCE I get terrible results with maximum macroblocks, and when I use tmpeg's 2-pass VBR I get them too, lots.

But then, when I use Tmpeg's CQ setting, the quality gets a lot better. Why is this? People have been telling me that I'm doing something wrong, but I can't find anything.

One more question. When using VBR, the encoder is supposed to allocate sufficient bitrate when needed, right? So if there's a lot of movement, according to my logic, the bitrate should be as high as possible.

So if I set the max bitrate to 2600 kbps, I think the encoder should put 2600 kbps at these scenes, but it obviously doesn't.

I tried this, taking a short clip with a lot of movement, and encoded it with 2-pass vbr and cbr, fixed to 2600 kbps.

This was all done in tmpeg. max: 2600, min 600 and average 1900.

The results was that the CBR was bigger, but had a lot better quality.. shouldn't the 2-pass vbr file had been about the same size, since I think it should have put more bits into it?

Ah just a question.

[Truman]

<TABLE BORDER=0 ALIGN=CENTER WIDTH=85%><TR><TD><font size=-1>Quote:</font><HR size=1 color=black></TD></TR><TR><TD><FONT SIZE=-1><BLOCKQUOTE>
I tried this, taking a short clip with a lot of movement, and encoded it with 2-pass vbr and cbr, fixed to 2600 kbps.

This was all done in tmpeg. max: 2600, min 600 and average 1900.

The results was that the CBR was bigger, but had a lot better quality.. shouldn't the 2-pass vbr file had been about the same size, since I think it should have put more bits into it?

Ah just a question.
</BLOCKQUOTE></FONT></TD></TR><TR><TD><HR size=1 color=black></TD></TR></TABLE>
A file encoded with 2pass VBR 1900kbps avg bitrate will have the same size as a file encoded with 1900kbps CBR.

What a source are you encoding?

[Kdiddy]

Ok man Ill try to give my best opinion or take on this

truman: "A file encoded with 2pass VBR 1900kbps avg bitrate will have the same size as a file encoded with 1900kbps CBR"
Man I know you know your stuff, but while in theory this is true, more often they are of different sizes in real world...because of the few # of passes...I'll explain below

D: with VBR, the both encoders try their best to stay at the specified avg bitrate, the max & min is what it is allowed to go up, WHEN IT DEEMS NEEDED.....now on 2 pass, it uses a mathimatical equation...basically, scanning the the clip first, finding the max & min values, and computes a avg value that is closest to the specified value.....often I have done like 1300 min, 1900 max, 1600 avg, 3 passes...and will get a file at like 1615, and sometimes 1580 depending how much action movement....which why more often than VBR & CBR have different file sizes...however the more passes you do, the closer it will be to specified average, but your quality increase becomes nill after 3-4 passes.

now as far as CCE vs TMPG, to my eyes, I cant see much difference tween the 2 (with mpeg 2 video)...I use a method that sorta combines the best of both that Truman in fact told me about...so I use CCE just for its sheer speed advantage.

[Truman]

Yes, I noticed that too. I use cce VBR 3-5pass, when I set for example 1600kbps avg bitrate, mostly I end with a 1590-1595 avg bitrate. Would like to know, whether the cce encoder is more accurate in CBR mode. I will check it out after finishing my current conversion.

[Kdiddy]

havent done a CBR encode in awhile..but I thnk I seem to remember it not being too accurate in CBR mode...not sure though

[davidian]

Ok, I dunno, so maybe this is a stupid question, but why doesn't the bitrate go up to like 2600 kbps (max) if there's a lot of movement, to make it look better, instead of trying to stay very close to the average and get worse quality?

[teamhawaii]

same here dude.. i've tried CCE using what I feel to be the same settings I use in TMPGEnc.. and guess what..

the TMPGENc encodes come out way better... i don't know why either..

[bbb]

Look.

The key is average bitrate.

Tmpgenc can only use 1900 throughout the encoding because your clip is ALL fast motion and you set the average bitrate to 1900. In other words, there is no normal motion scenes low bitrate to average out the fast motion scenes' high bitrate to 1900.

If you encoded an entire movie that has generally normal motion with a few fast motion scenes, then those fast scenes would get 2600 and the normal motion scenes would get less, with the entire encoded movie averaging out to 1900.

[Kdiddy]

yes the key word is AVERAGE:

"This was all done in tmpeg. max: 2600, min 600 and average 1900."

The prob with this is that your min value -1300 below your specified average and your max is +500 above that average..which leads the enocder to think you would prefer more "lower the average specified" bits to be used than above average...basically your bitrate will probably fall somewhere tween 1600-1900...depending on amount of movement..but still at that bitrate, you should NOT see a LOT of blocks, actually should much at all...so that leads me to believe your doing something wrong in your encode process...like maybe too high of resolution or bad source video..remember the GIGO rule.

[davidian]

Yeah, well that's what I thought all the time. So then I specify max 2600 and avg 1900, the encoder should boost it up to ~2600 when there is a lot of action. Yeah that makes sence.

But it still doesn't explain, why I got the strange results.

Source: DVD (that leaves the GIGO out)
Encoder: Tmpeg / CCE

So I compared, CBR 2600 and VBR, avg 1900, min 600, max 2600.

There was a lot of action in my clip, and I thought, locically, the VBR file should have about the same size as the CBR, and the same quality, since the encoder should boost the bitrate up to make the quality good, but that didn't happen.

I chose 4-pass VBR in CCE (2.5 SP) and 2-pass in Tmpeg. Both files were much smaller than the CBR file and had much worse quality. My conclusions are that the VBR files didn't boost the bitrate to a sufficient level.

Oh yeah, what I mean with "bad quality" is LOTS of macroblocks.

And, strangely enough, when I tried LSX encoder (v 2.0 I think) I got no macroblocks, very good quality.

I'm almost thinking about extracting some pictues of my mpeg-streams and posting them, to compare, and see what people think.. if anyone cares

It's just so strange, most people say they get superiour results with CCE, and NO macroblocks, when I get more macroblocks with CCE that with any other encoder.

[Kdiddy]

well that could be lot of variables in that, I still say one is that your min value is too low, you need to raise your min value to at least 1200. I dont think understand the concept still. Your are giving the encoder WAY more bits below your average than above, just cuz you say 2600 is max, doesnt automatically make the encoder encode that high during a fast scene so far ha ha's and to make me shut-up, raise your min value to 1200, also what settings do you use in CCE and how are you frame serving?? because at 1900 you should NOT get blocks at all with a DVD source...and again, your CBR & VBR will most likely NOT be the same file size, rarely does VBR actually encode at specified avg rate... "Both files were much smaller than the CBR file and had much worse quality", well then we know for sure its not at 1900 bps, use an app called "mpeg properties" and see what bitrate your video is actually being encoded....

Again, I get no blocks with CCE, and using it combine with TMPG I get excellent results...it might be best to post your FULL process from ripping to end to see where the problem may lie.

[davidian]

Ok, Kdiddy, thanks a lot for taking the time!

DVDripper: Claddvd 1.61
Then I load the VOB into dvd2avi and create a .wav file and a .d2v file.

Then I load the .d2v into tmpeg to resize, and save the project into a .tpr file, and then load it into vfapi converter to make a dummy avi. (When using CCE that is)

Then I just load the avi file into CCE. What I have not tried, is to use 1200 as min value, I shall try that and see what results I get.

/Johan

[ThaSmig]

Kdiddy: do you mind sharing your best-of-both-worlds guide, that truman gave you? would be nice to see what it is..
cheerz

[jonyutah]

I have had the same problems and when ever i think that i have VBR sroted out it goes and changes the formula on me. Differs from sorce to sorce I think. I have leaned away from VBR and use CQ, I find it to be more reliable even with the expense of size.
Wayne