MP3 "lossy" compression algorithm doubt

[paulscotti5]

Hi everyone,

I just had a doubt about the .mp3 compression method.

I know that when you compress a raw sound file, like a .wav, with the mp3 algorithm, some chunks of data are discarded.
(It may sound kinda stupid), i.e when you convert an mp3 file to a wav file, it will not have the same quality as before the compression, right? and in theory, if you keep compressing infinitely a .wav file with the .mp3 algoritm, it will end up weighting 1 byte?(Thus making the sound file useless)

Thanks.

 

[rgd1101]

no you can't keep compressing it until it 1 byte. But yes it lossy so it going to loss some high/low end.

 

[Pinhedd]

Hi everyone,

I just had a doubt about the .mp3 compression method.

I know that when you compress a raw sound file, like a .wav, with the mp3 algorithm, some chunks of data are discarded.
(It may sound kinda stupid), i.e when you convert an mp3 file to a wav file, it will not have the same quality as before the compression, right? and in theory, if you keep compressing infinitely a .wav file with the .mp3 algoritm, it will end up weighting 1 byte?(Thus making the sound file useless)

Thanks.

Nope, won't work that way.

.wav files are raw, uncompressed, unencoded waveform representations. It's nothing more than a sequence of 8/16/24 bit values sampled at a fixed rate all nicely packed in below a header.

MPEG Audio Layer 3 encoding works by first breaking the time domain samples into chunks and then converting each chunk from time domain to frequency domain. This process is lossless (using a modified discrete cosine transform). The frequency domain samples are then analyzed using Psychoacoustics to eliminate frequency components which would be inaudible or barely audible. The resulting blocks are then compressed using a standard lossless block compression algorithm. Removing the high frequency information greatly improves the efficiency of the standard compression.

MP3 decoding simply reverses the process. The blocks are decompressed and reconstructed into frequency domain chunks, transformed into time domain chunks, buffered, and played.

Repeating this process will continuously degrade the quality of the audio track until it converges on a quality level which is the parameters of the phychoacoustic process result in no further data removal.