Beni Cherniavksy
2001-Oct-25 00:39 UTC
[vorbis] Fwd: Re: Clarification on pshycho-acoustic in Vorbis (your non-MP3 guide)
After reading http://mp3.radified.com/mp3.htm I sent Rad an explanation of some things as I understand them. He liked it and posted it on his site (still unlinked, use the URL below). Can somebody with better understanding of psycho-acoustic terms and the vorbis model check it and comment on it? In particular I didn't know how vorbis handles quantization noise. If you reply with corrections, CC: <Rad@radified.com>. On Wed, 24 Oct 2001, Rad wrote:> Beni, > > Thanks for the note. I found it most helpful, so much so that I posted it on > my site: > > http://mp3.radified.com/ogg_psycho_models.htm > > I cleaned up the grammar a bit, and tried to maximize readability. > > Perhaps you could review it & ensure that I didn't change the actual meaning > of any parts. > > Is it okay that I included your email addy? I can remove it if you don't > want it there. > > Thx again, > Rad > Laguna Beach, Ca. > > ----- Original Message ----- > From: "Beni Cherniavksy" <cben@techunix.technion.ac.il> > To: <Rad@radified.com> > Sent: Wednesday, October 24, 2001 8:01 AM > Subject: Clarification on pshycho-acoustic in Vorbis (your non-MP3 guide) > > > > Your guide (ripping, Page 6 of 7) says: > > > > While MP3's version of VBR uses psycho-acoustic models, Ogg is > > based solely on noise, masking, coupling and ATH thresholds. I > > have no idea of what that means, but I read that's how I heard it > > works. =) > > > > These _are_ exactly aspects of a psycho-acoustic model. ("Most people eat > > soups but he eats boiled water with vegetables and other additions"). > > I'm not a specialist but I've got quite a good idea of them (I hope :). > > BTW, I like Vorbis very much, so you'll see a bias ;). > > > > [rest omitted, see http://mp3.radified.com/ogg_psycho_models.htm]-- Beni Cherniavsky <cben@tx.technion.ac.il> (also scben@t2 in Technion) --- >8 ---- List archives: http://www.xiph.org/archives/ Ogg project homepage: http://www.xiph.org/ogg/ To unsubscribe from this list, send a message to 'vorbis-request@xiph.org' containing only the word 'unsubscribe' in the body. No subject is needed. Unsubscribe messages sent to the list will be ignored/filtered.
Gian-Carlo Pascutto
2001-Oct-25 04:23 UTC
[vorbis] Fwd: Re: Clarification on pshycho-acoustic in Vorbis (your non-MP3 guide)
On Thu, 25 Oct 2001, Beni Cherniavksy wrote:> After reading http://mp3.radified.com/mp3.htm I sent Rad an > explanation of some things as I understand them. He liked it and > posted it on his site (still unlinked, use the URL below). Can > somebody with better understanding of psycho-acoustic terms and the > vorbis model check it and comment on it? In particular I didn't know > how vorbis handles quantization noise. If you reply with corrections, > CC: <Rad@radified.com>.About the 'noise': he probably originally heard about noise & tone masking. Tone masking is probably familiar to you, this is just when louder frequencies mask out adjacent quieter frequencies. Things aren't always that simple though. Think about what happens when you hit a cymbal. A cymbal doesn't really play a tone, it generates broadband noise. Noise masking basically determines how much extra noise the encoder can introduce in that noise without it being audible. This is generally more tricky than tone masking and one of the things that's improved a lot in RC3. (but RC2 did it still much better than LAME for example...LAME developers are now trying to imitate the Vorbis method ;) As for the original page, there are a few others things that raise my eyebrows.>While Lame offers what most agree to be the best MP3 encoding quality currently available, and the one being developed most aggressively, some claim that Liquid Audio, a proprietary encoding format, based on the AAC codec, is the *best* lossy encoder of them all, especially when bit-rates are taken into consideration.>a) Vorbis development is much more active than LAME development b) if you say AAC is the best, you at least need to give reasons, but you give none I find it funny MPC is critiqued for HF harshness, while it generally gives better quality than AAC (even at lower bitrates). For an objective test of MPC vs AAC vs OggRC2 @ 128kbps, take a look at http://www.ff123.net (also much info about MP3 encoders)>MPC's strength is the quality of the files it encodes, and the method (subband) it uses.>Subbanding is no guarantee at all for good quality. MP3 also uses subbanding for example.>Ogg Vorbis is a new lossy VBR encoder receiving much attention. The Vorbis.com site has been down a lot. I>Vorbis itself is not limited to VBR (although thats the only mode the encoder currently offers). The site was only down after it got swamped by the RC2 release. It's nearly never down now.>It has many cool features (such as bit-rate peeling and wavelets).>Vorbis 1.0 will have no wavelets. They _should_ be in later versions but I feel it too early to tout that as a cool feature. (Bitrate peeling is closer, and will work with Vorbis 1.0)>I've heard that some files encoded with Ogg exhibit harshness in the high frequencies, and others have problems with the bass.>The harshness in the high frequencies happens with the RC2 release at low bitrates. (not with previous versions, and also not with the tuned encoders and new experimental versions). Vorbis has _no_ known bass problems. Every report I have seen since RC2 has been bogus. If you still believe it has, send samples. We cannot fix a problem that does not exist.>Because Ogg is open source, it's possible for folks with coding know-how to modify (improve?) the program to their liking. That's what Garf did. I used to have a link to his page containing these files, but the link is no longer valid.>The link is: http://sjeng.org/ftp/vorbis If you want 160kbps VBR, pick gt1, else pick gt2. (gt2 also contains the 160kbps mode but I did not test it). You enable the 160kbps mode in gt1 by asking for a very high bitrate (-b 999).>It doesn't look like they ever plan to offer bit-rates exceeding 96kbps. I'm not sure why they would want to limit bit-rates, and not give control of this to the user.>Simply because there is no use to the technique at bitrates higher than 96kbps. From 128kbps on the encoder can just store the info which MP3Pro would (often wrongly) predict.>Advanced Audio Coding (AAC). Whereas MP3 is based on MPEG-1 (layer 3), AAC is based on MPEG-2. MPEG-2 is what DVDs use (both audio & video), so it is an advanced, high-quality encoding format. Their catchy motto is Store less, hear more.>Nearly all DVD movies I have had in my hands so far used Dolby AC3, not AAC. AAC also exists in an MPEG-4 version AFAIK. There exists MPEG-2 extensions of MP3 too. There is no need to mention all this as it's both wrong and confuses matters. Why isn't the Liquid Audio section in the AAC section? It _is_ AAC after all, and the split up makes them look like something different. Perhaps there are corrections possible to my corrections. In that case, I'm sure someone on the list will point them out :) -- GCP --- >8 ---- List archives: http://www.xiph.org/archives/ Ogg project homepage: http://www.xiph.org/ogg/ To unsubscribe from this list, send a message to 'vorbis-request@xiph.org' containing only the word 'unsubscribe' in the body. No subject is needed. Unsubscribe messages sent to the list will be ignored/filtered.