Friday, November 20, 2009

Can you hear the difference when audio is compressed?

We have seen that images can often be compressed significantly without noticeable degradation.

A recent review asks whether one can distinguish between a file that is ripped from a CD with no loss and a compressed MP3 version of the same file.

There were only seven subjects in this small test. They were given two recordings of two rock songs and asked to tell which sounded best. One was a lossless recording ripped from a CD, the other the same file compressed. Some compared the lossless files to 192 kbps MP3s, others compared them to 320 kbps MP3s. (Apple iTunes music is 128 kbps).

Only one person could accurately pinpoint which tracks were MP3 in every case. Of the rest, three subjects picked the lossless track twice and the other three only made one correct choice.

This is obviously a limited test -- there were few subjects and few controls. The results might be affected by things like

  • the quality of the playback equipment
  • the hearing acuity of the subjects
  • the familiarity of the music to the subjects
  • the nature of the particular song or other audio
Still, a test like this makes one wonder whether bandwidth is wasted by a lossless file.

You could easily create a similar test and try it on your friends (and yourself).

It would also be interesting to do a test of recorded voice. Podcasts are typically delivered as 64 or 128 kbps MP3 files -- can listeners tell the difference? Would a higher data rate matter?

As storage cost falls and bandwidth increases, how is the audio compression tradeoff of file size versus quality affected?