Originally Posted by Asr
Lossy formats like MP3 probably do the same thing. MP3 manages a much higher compression ratio since it works on a much more complex algorithm, and it tosses out inaudible frequencies (>20 kHz, <20 Hz).
Not quite. One of the principles that MP3 uses is that we cannot hear everything even in the 20 - 20K range due to masking, this is an effect where one frequency will mask another nearby frequency if it is sufficiently more powerful. How this works is quite complex and highly frequency dependent, some frequencies can be masked by others that may be just 20db higher, masking is less effective as the two frequencies get further apart, it is fascinating really. Most MP3 encoders cut almost everything above 19K down drastically. Run a awav file through a real time spectrum analyzer (such as the one in cooledit) then do the same for an MP3 version, the difference at the higher frequencies is stark. That it (at high bit-rates) is transparent to 98% of folks is a real testament to ingenuity.
When Mp3 arrived I was a real hard core skeptic and the early encoders could be very flaky but they are really good these days.
Linear Mode
