i've often wondered about the relationship between MP3 data compression artifacts and the source frequency - do you know if it's worse at higher frequencies (which was my assumption), or if it's independent of frequency?

i know that MP3 haters like to ballyhoo about the lack of low end, but i always assumed that the low frequencies would show the least artifacts.

"Regular" recordings (like audio tape, vinyl, .WAV, etc.) work by recording a sound wave directly. Analog recordings can 'sample' continuously, but often not accurately. Digital recordings sample accurately but not continuously. To approximate continous sampling, the data size will increase exponentially. It's impossible to get a perfect replication from a digital recording, but, if you let the data size get very large, then the approximation is better than most speakers can reproduce anyway.

MP3 does not record sound directly. Instead, it records which sine waves are, approximately, present in a sound during a small time period. (MP3's will quote that period as their 'sampling rate', but it's obviously a different meaning)

The math behind it says that any waveform, including a 5 minute waveform of any music track, can be perfectly replicated if you add up an infinite number of basic sine waves, each sine wave of a different frequency. Of course, it's impossible to list an infinite set of frequencies, but if you list just a few frequencies, and change that set often, then you can get a very good replication of a waveform without having to list very much. So, a decent sounding MP3 can closely approximate real music while using very little actual data (compared to direct sampling).

So, an MP3 can equally well replicate low frequencies as high frequencies, so to answer your question, yes MP3 quality is independent of frequency.

HOWEVER, higher frequencies tend to be more complex, that is, they might be speaking voices or instruments with timbre, etc. and this means that there are a great number of high frequencies added together to get the desired effect. It takes more data to list all those frequencies. A bass line has to be simpler by definition and so will always need less data to list accurately. Since the 'bitrate' of an MP3 is capped, there will need to be sacrifices, and proportionately those affect the high frequencies more, simply because more data is used there. To that end, high frequency (hi-hats, snare drums, michael jackson) sounds will, physiologically at least, seem more distorted than basslines (james earl jones).

-jM

A&D