The need for audio compression has been apparent for some time. Even with a high-speed connection, the 10 MB/sec of redbook audio is hardly acceptable for streaming over the Internet. For some years, MP3 has been the compression scheme of choice even though it is lossy; lossless compression can rarely save more than 30% of the size of the file, where 90% reduction is common with MP3.
A bit of history first. The Motion Picture Experts Group (aha! that's where MPEG comes from!!) covers audio, video and still-image formats. The audio formats are labelled MPEG Level 1, Level 2, etc. Each is a superset of the lower ones so that MP3 includes MP2 which includes MP1. Each also uses "perceptual coding" in which less significant components of the audio are sacrificed for smaller file size. Note that as more compression is used, maintaining constant quality means more computing.
With the Fraunhofer Institute's development of L3ENC, high quality and substantial compression were realized. A serious listener with excellent source material can hear the degradation of the Fraunhofer engine even at high rates, but the penalty is not great and storage requirements are reduced in many cases by an order of magnitude - a payoff well worth the price to most users. For similar compression at low rates, other engines produce more artifacts and poorer frequency response; again, they may be acceptable to many users. In all that follows your judgement on matters of sound quality, cost and ethics should be applied; I am trying only to report the facts as I have observed them.
The Fraunhofer Institute focussed development on the MP3 encoding algorithm and owns the rights to the one in general use. (WARNING: I will not enter debate on the moral or legal implications of algorithm patents or of copying commercial material.) For the rates of interest to MPEG, everything up to Level 3 is handled by the high-quality Fraunhofer codec at least as well as by anything that has shown up since. When one moves to the higher sample rates and higher bitrates of Level 4 (and above?), the rules are different.
Originally, the Fraunhofer codec was licensed at a substantial fee for a higher-speed, lower-quality version and a very steep one for the full package. An individual license for general-purpose use on your computer cost of the order of $300 either directly from them or from companies which licensed it to wrap in their own shell. I signed on then, buying the AudioActive Studio Pro at a painful price. That installed the Fraunhofer codec into Windows and let me run programs such as ECDC which used it to encode audio. Along came a number of other codecs with their own algorithms. They were priced low - or free. Some were extremely fast, though not very good. Still, they satisfied many users' needs and they killed the market for the Fraunhofer.
So Fraunhofer offered a different way to use the high-quality encoder: contain it within a program. That brings the license down well under $100 and products such as CoolEdit can afford to offer an MP3 plugin and AudioActive Studio Pro can be priced competitively. Unfortunately, that means that their codec is not accessible to other programs - it is not installed into Windows. I suppose one could still license the codec directly from Fraunhofer, but it is probably still priced out of reach. So lower-quality codecs are available which may well install into Windows for you. I had some cluttering my system before my recent move to Win2K. For now I'm prepared to run CoolEdit Pro with the plugin and AudioActive Studio Pro and not compress audio in programs such as ECDC. When I get a good solution, I do not want to have to figure out how to remove garbage, even garbage I paid for in my more ignorant days.
High-quality codecs are available as freeware - both BLADE and LAME are generally regarded as superior to Fraunhofer above 128 Kbps - but I have not found an implementation which installs one into Windows. If someone has a method for Win2K, please let me know. I have both BLADE and LAME on my system for use in Exact Audio Copy, for example. Note that they demand substantial computing power; on most machines, they run far slower than real time.
E-mail me at cdrecording@mrichter.com
Return to Mike's home page