|[Home] [Credit Search] [Category Browser] [Staff Roll Call]||The LINUX.COM Article Archive|
|Originally Published: Friday, 29 June 2001||Author: The Linux.com Staff with Joshua Pruitt, Kenny Sabarese and Jack Moffitt|
|Published to: enhance_articles_multimedia/Multimedia News||Page: 1/1 - [Std View]|
Ogg Vorbis: The Future of Digital Music
Take a look at Vorbis the truly open and interoperable digital compression format for audio and music. Vorbis is a part of the ambitious Ogg framework for open source multimedia, and is cultivated by the non-profit Xiph.org to make life better for you, me and the neighbors, too.
In fact, they already did. In 1988 Fraunhofer sent a now famous "letter" claiming royalty payments on several MP3 encoders. This didn't stop the encoders from distributing, but it forced some of them to go source-code only (LAME and BladeEnc being noticeable examples). More recently they have stepped forward again claiming (their right) to royalties from streaming MP3s and royalties from musicians selling MP3s. The future looks bleak: the terms surrounding MP3Pro are far more expensive.
Ogg Vorbis is an attempt to build a high quality truly open source media layer for the Internet. It's an ambitious project, and unusual as the results will directly benefit not just developers (like, say game developers and multimedia programmers) but end-users as well: People who enjoy music will not have to worry about "stealing". Content creators, in turn, will benefit from an Ogg Vorbis world. Musicians will not have to pay royalties to distribute their music in compressed format, or be locked into proprietary schemes like those from Microsoft. Many more people will enjoy their music, since it will interoperate with everything. The OGG project could well become the most important application of open source software there has ever been.
Ogg Vorbis is developed and supported by the Xiph.org Foundation, a non-profit software development group that stresses interoperability. Xiph.org is also, of course, the home of the popular Icecast streaming media server as well as many other important open source projects. Xiph.org license the core libraries of Ogg Vorbis under a Free BSD-like license in the hope that this license will help developers adopt the technology for closed platforms like Sony PlayStation 2 and portable music players. Xiph.org only retains the right to set standards for the format and maintain compliance. Note that the applications developed by the Xiph.org foundation are licensed under the familiar GPL including: oggenc, ogg123, cdparanoia and Icecast.
According to the Xiph.org web site Ogg is the name for an even more ambitious multimedia framework that will eventually include other codecs, like Tarkin (for video) and Squish (for lossless audio). Vorbis, on the other hand, is the name for the single compression codec used for digital sound and music.
The 'sweet spot' for Vorbis is probably 8kHz - 48kHz, 16bit, 16kbps-200kbps. These aren't limitations, but this is where most people will use Vorbis, and what the format is "tuned" for to a large extent. This places Vorbis in the same class as other audio representations including MPEG-1 audio layer 3 (MP3), MP3Pro, MPEG-4 audio (AAC and TwinVQ), and PAC.
Linux.com What advantage does variable bit rate deliver?
First of all lets make clear that that MP3 is also capable of encoding at variable bit rate (VBR), although the implementation is quite different, including the phsycoachoustic models.
For instance, LAME doesn't trust its psychoacoustics. It uses them as suggestions. In Vorbis, the psychoacoustics analysis is what determines how many bits get used.
By using a constant bit rate (such as 128, 192, etc.), you are using the same "level of compression" throughout the entirety of the song. This may be OK in many new-ish pop or hard rock songs, as they are traditionally monotonous, repetitive, and a steady volume throughout the composition.
However, there is a lot of music, (classical music and high quality rock are good examples) that are more dynamic: Softer at some parts, louder at others, simpler at some parts (fewer instruments), and more complex at others (more sounds and harmonics). By using a constant bitrate on this type of music, you're really shortchanging yourself, because the soft parts can handle more compression, and therefore lend the complex parts room for less compression, in order to maintain the same overall quality throughout the song. This optimizes compression for size and quality.
CBR must waste bits sometimes. For instance, if you encoded digital silence, you'd still be encoding it with 128kbps. In Vorbis you could probably encode that at 5kbps!
If you've ever listened to an MP3 of a complex song with too low a bitrate, you've probably heard the artifacts: "swishing sounds" at higher pitches (such as cymbals), warbles, "muffled tones" at lower pitches, etc. This is because MP3 is (more or less) an audio version of JPEG - with all its visual artifacts and boxes visible at super high compression levels.
With variable bit rate, you are able to keep the song at a stable relative compressed quality level by adjusting the compression according to the complexity of the music. Ogg (and VBR mp3) does this by automatically choosing a compression level for each micro-slice of a song based upon the relative complexity of that slice.
Say, for example, a song starts with a soft acoustic introduction and ends with silence. The introduction would sound CD quality for most people at a bitrate of, say, 112 or 96, and the silence, of course, would sound the same as CD at a tiny 32! But during the complex parts in the middle of the song, the compression is "bumped down" to say, 160, 192, or even 256 at really hard parts, to compensate for the complicated slices - keeping the quality uniform, and the complex parts still near CD quality.
In the end, the average bitrate is comparable to a CBR file, but the quality is much, much higher.
Of course, the goal is to use VBR to keep the quality at a consistent level throughout the song: a quality that has no audible artifacts (CD quality) to the average listener using relatively good equipment, and still have a pretty good file size savings. This is the holy grail of audio compression, and something that is simply not possible with CBR. You'll often have artifacts that any untrained ear can hear with CBR at some point in almost any given song.
Linux.com OK, so does that explain why a .mp3 file and a .ogg file of about the same size sound so different? I mean, anybody with ears can tell that the .ogg file sounds far better than the MP3.
Absolutely right. However, Ogg Vorbis is technically only a little better at compression than MP3 (perhaps 5-10%). Its strength lies in that the Ogg encoder automatically enforces good encoding practices, by not even allowing for constant bitrate encoding at all.
However, although CBR is not currently implemented in Vorbis, there is nothing in the specification that disallows it. Future versions of the encoder will allow you to constrain the VBR within certain limits (although it does a good job of this now), and even do very CBR-like stuff when that might be appropriate.
And while encoding a Vorbis file at 128 avg. bitrate is only a bit smaller than an MP3, the quality is drastically different. Vorbis should deliver transparent quality at 1.0 at around 96kbps and is currently pretty close now at 128kbps.
Unlike MP3 Vorbis does not have channel coupling, and it is certainly spanking MP3 in terms of quality, even in VBR vs. VBR tests. With channel coupling most things will be transparent at 96kbps or so, maybe even lower. Compare that with most people claiming MP3 is transparent near the 256kbps range. It's a big difference.
Another advantage of Vorbis over VBR MP3 is that the VBR is more tight and specific. MP3 can jump around with 128, 160, 192, etc., but Ogg can use any bitrate as needed in the song, such as 111, 167, 119, etc.
The current Vorbis encoder is a little more bounded as well. The VBR produced from the current encoder is quite stream-able, since it doesn't vary too wildly above the average bitrate, although it will certainly encode silence in as little space as possible no matter how many bits you have available.
And of course the most important advantage of Ogg versus MP3 is that Ogg is a royalty-free open standard.
Who can say what the future will hold? Could anybody have predicted the rapid rise of Linux a few years ago? For us it depends how much you believe in free software. As a consumer our feeling is the question will be answered not by the free software and open source movements, but by the proprietary software industry that has pitted itself against its own customers. How long will people put up with expensive, lock-in, buggy crap being shoved down their throats, whether that is DOS-based "network" operating systems, or Brittany Spears? Every day more and more people are demanding high-quality, hand-crafted products and technologies at a fair and reasonable price: Linux, Ogg Vorbis, Ben and Jerries. So long as we, the developers and consumers, demand that kind of value in what we use and want, and take action, take that extra step or two to get it, the further the proprietary software and entertainment industries will be fighting themselves into a smaller and smaller greedy little box.
Linux.com wishes to thank Joshua Pruitt, Kenny Sabarese and Jack Moffitt, Executive Director of the Xiph.org Foundation.