[ Home | Staff & Contacts | HiFi Playground | Listening tests | DIY & Tweakings | Music & Books ]

MPEG3: lossy compression ?

[Italian version]

You may have read about the HEPC project of Metaxas Audio Systems in an article of some weeks ago here on TNT, where Kostas Metaxas himself explained his views and projects about the Future of the source for Hi-end audio.
To cut a long story short, the main idea behind this project is: design and build a Personal Computer aimed to reproduce Music, which reads it as files from a CD-rom, directly from the Net or from *external* peripherals like DAT, DVD or any other digital source, present or next-to-come.
This of course requires hi-end level audio boards, hi-end power supply circuits and a general rethinking of the whole concept of PC.
Now, one of the possibilities to store Music into files that are reasonably sized (in order to make them downloadable from the Web, for example), compression algorithms are a must.

Thanks to the kind collaboration of Metaxas Audio I've had the possibility to listen to a direct comparision between standard CD audio recording and MPEG3 compressed tracks.
In particular I've received a CD with four tracks *standard* (called *raw*) each one followed by the relative compressed counterparts, one with a compression ratio of 10:1 and the other with 5:1.
The tracks were recorded via a standard Pentium-based CD (no HEPC was involved during the process) and taken from CD's *Harry Belafonte at Carnegie Hall* and *The Cowboy Junkies* - well-known recordings among audiophiles.

Lossy compression or not ?

Let's admit it: the MPEG3 compression is CLEVER but the performance depends strongly (and obviously) on the choosen rate.
The 10:1 compression rate is clearly excessive: it cuts both low and high frequencies, flattens the virtual perspective and subtracts liveliness to the recording.
The sound becomes harsh and artificial, dynamics diminish and it seems you're listening to a bad copy on a bad Compact Cassette.
But please remeber we're talkin' of a 10:1 compression rate which is a HEAVY one.
Still MPEG3 succeeds giving you an idea of how the recording will sound like and I'm pretty sure that on cheap low-fi systems it will be considered more than sufficient by most customers.
The 5:1 compression rate is another story. Though you can always *tell* which track is playing, the compressed or the uncompressed one, here the differences become subtler.
The first thing one can notice is the reduced soundstage, depth, width and height. Especially depth.
Then the air amiong the instruments: it seems everything has been *compacted*, losing the breath of the otherwise very good recording.
Also it seems the pace changes a bit (I know this is a psychoacoustic effect), and the compressed tracks sound somewhat faster, as if the players were late for dinner :-)
Despite of this, the overall effect is incredible and, without comparing it to the uncompressed version, you wouldn't believe you're listening to a 5:1 compressed data.
Clearly, if you give people a compressed recording, few of them will argue if it is compressed or not. They would probably say: well, the soundstage isn't so realistic or there's not enough *air* among instruments but how many commercial recordings sound this way? Tons of 'em, sadly.

The bottom line

Compression is compression, you just can't avoid this. We're sadly going towards a compressed version of our universe, compressed pictures, movies and audio.
Since the digital revolution (involution) started it was clear that the uncompressed HUGE analog data were doomed to an end.
The new DVD audio standard seems now to be the 96 kHz/24 (or less) bits instead of what we have now (44.1/16) so it seems there's still hope for a more analogue :-) digital sound.
But once you can use clever compression algorithms and hence store more data in a given medium, well, our ears will be required to decide which kind of compression is better or worse.
MPEG3 works fine and clearly can be already used to store small excerpts of Music into Web servers, for faster downloads and quick evaluation of what you're going to buy, so that you can decide if the recording is worth the money or not.
I simply can't wait to test-listen to the first samples of the HEPCs.
For more details please refer to
Metaxas Audio Systems

Copyright © 1998 Lucio Cadeddu

[ Home | Staff & Contacts | HiFi Playground | Listening tests | DIY & Tweakings | Music & Books ]