PDA

View Full Version : OGG Codecs and constant bitrate



flips
10-23-2006, 03:59 PM
Hi!
I'm having good experience with the aoTuV Ogg Codec (SSE2), it's very quick, but I have discovered that VBR doesn't work very well with classical music (in the quiet parts, fragile violin etc. gets distorted/assumed to be noise, at least in my q=8 VBR OGG's.). I'm therefore experimenting with CBR OGG vs. LAME MP3. (If none are good enough, I would have to store my classical CD's in FLAC format.) The only downside then, besides taking a lot of space, is that my Dreambox doesn't understand FLAC, at least not yet. (WMA, AAC, mp4 or mpc is not an option either, and plain WAV's doesn't have ID tags ...)
Mpeg-2 audio I don't know much about ...

I can't seem to find any way to make the optimized OGG encoder make CBR OGG's, so I would have to use the much slower OGG encoder, right? (Or could I somehow hack the optimized codec to make CBR?) :)

--
flips

LtData
10-23-2006, 06:01 PM
Ogg doesn't really have a CBR mode, IIRC. It just limits the bitrate range to what you specify and tries to stick to it. It is still basically VBR though.

Deano
10-24-2006, 04:52 AM
I'm not sure where you get the idea that CBR would be better than VBR for your classical music. In my personal experience, VBR will always produce better results than a CBR file (unless, of course you are doing a 320CBR mp3, but I find that wasteful).

You can't make CBR OGG, it just doesn't happen as the codec is simply made to be a VBR codec.

If the results you are getting aren't good enough (although I find the idea that a Q8 OGG isn't enough to be very unlikely), keep turning it up. To be honest though, if you are sure that the lossy files aren't good enough then lossless is probably the only way to deal with it.

I would recommend ABX testing to see if you can really hear the errors and it's not just placebo, but a lot of people look at me as if I'm talking a foreign language.

flips
10-24-2006, 12:00 PM
I thought OGG q=8 ought to be sufficient, and for pop, rock, electronica, trance etc. it has always been, I haven't been able to tell the difference from the CD I ripped the track from. But one day here, on the train, I was listening to some classical music (Arvo Pärt) on my DMC HDD OGG/MP3 player, and this music really has lots of very faint/quiet/low volume details and such, and I found that it scrambled the violin and partly voices in these quiet parts of the track -- it sounded to me like some algorithm mistook the music for being just white noise or something.

Until some months ago I had most of my music ripped to FLAC, as I really don't like the idea of something being lossy compressed, but at a certain point I just realized I don't want to buy more HDD right now, so I searched for a nice program to start mass-converting to OGG, and I bought dbpowerAmp Music Converter. :) Another good motive for using OGG or MP3, is that both my Dreambox and my portable audio player will be able to play the files directly, so I'm consider ripping just about all my records to 320 or similar OGG or MP3.

When I first encountered this problem with my classical music being (just slightly, but still, corrupted), I decided to do some testing, and that's just what I'm doing. My plan is to rip some tracks in lots of different resolution MP3 and OGG's and compare to see if I can hear the difference. My theory was/is that maybe the VBR made the OGG encoder a little to aggressive in these quiet parts of the track. So I also wanted to compare VBR/CBR, bot MP3 and OGG.

(You mention ABX testing, but I don't know what ABX means, and even Google couldn't define it.)

If I choose the unoptimized/plain OGG codec, I can choose constant bitrate, and the dbpowerAmp info popup in Windows explorer reports it so be constant. But I was puzzled that Winamp seemed to think it was still variable.
I guess I will try to compare all combinations. I don't consider 320 AltInsane LAME MP3 to be to much loss of space, either (even when it takes even more space that "Constant/350 kbps" OGG). I used to have it all in FLAC, remember ... :D
(Read something, somewhere, some while ago -- someone claiming OGG/Vorbis really was unsuitable for classical music. But I don't buy that unless I can confirm it with my own hearing.)

--
flips :cool:

flips
10-24-2006, 12:09 PM
Found the meaning for ABX, though:
"ABX test, a double-blind methodology commonly used in codec listening tests and in high fidelity."

When I first discovered the problem, I'm sure it wasn't placebo, though, I was enjoying some music on the train, with closed eyes, not even thinking about what kind of codec or bitrate or anything. But the distortions was making the violin and partly the voices sound like some weird synth or a dolby filtering error ...