MP3 Encoding at 128kb/s

 

MP3HQ128k.gif (4316 bytes)
128kb/s HQ MP3 Frequency Response

MP3HQ128kS.gif (40941 bytes)
128kb/s HQ MP3 Spectral View

Here we have the results of testing at 128kb/s. You can see that the frequency response is faithfully reproduced up to nearly 16kHz but falls of rapidly above this. The spectral graph also shows no content above 17kHz. Take note that in the 10khz to 16kHz range you can clearly see how the lower intensity content has been pruned. This is where the codec is exploiting the masking effect.

It is this lack of low intensity high frequency content below 16kHz I believe that is the big shortcoming of encoding music at 128kb/s.

Listening tests also reveal the shortcomings of encoding at this bitrate. Some encoding artifacts heard as odd noises are evident and would be distracting for serious listening. The harpsichord notes are slightly muffled and the trumpet valve noises almost undetectable. Overall balance and robustness of the sound is good though.

There is certain content that 128kb/s MP3 does not handle well at all. On another test recording I have done from a live performance of Coming Home from Local Hero the encoding artifacts are so bad that the content sounds artificial. At the start of the recording there is a lot of applause from the audience and then a synthesizer starts playing a chord. At 128kb/s the applause sounds unnatural and there is bad phase distortion on the synthesizer. At higher encoding rates the problem disappears.

My opinion is that 128kb/s MP3 is great for mobile use and less critical material such as metal or heavy rock. If your tastes run to instrumental music, "unplugged" vocal music or classical music you are definitely going to want to use higher encoding rates. This format is good enough for music center type use but definitely not good enough to be considered Hi-Fi.

The low CPU overhead of about 22% on my machine and efficient 12:1 encoding ratio makes it a great format for playing music while working on the PC.

[Back] [Next]

© 1999 Peter Miller for MP3'Tech - www.mp3tech.org