Listener Response – On MP3s
I received this response from one of our listeners to our most recent podcast on MP3s. Thank you Harold from Austria for your insightful email.
As a loyal listener as well as professional audio engineer I feel the need to respond to your podcast about MP3 in general and the idea of “hearing the difference” between uncompressed and MP3 audio.
Short story: Although I have no doubt that the fellow studio guy did a great job of producing theses files, it does prove NOTHING but instead your conclusion of this is what mp3 is missing is highly misleading!
Slightly longer story:
In all lossy compression algorithms the question arises: What is left out? It is NOT frequency range and it is also almost never dynamic range if one goes beyond 128kbs at MP3 for example. So what is left out? In essence: What you cannot hear can be left out! This gets you the compression ratios of 1:7 instead of the 1:2 for purely mathematically (=lossless) compression.
Much as you don’t hear the TV news when your kids run around the house screaming, louder parts of the music and percussive parts (drums, cymbals) do MASK other parts of the signal. As soon as the kids run into the other room, the TV host can be heard and understood again.
So to stay in the above example: The news guy is still in the recorded signal but because of the kids screaming, you cannot hear him. MP3 detects such masked signals and more or less omits it! Believe it or not but during a drum solo you are almost deaf as your hearing is VERY limited about 20ms after every cymbal crash.
What sounds very dramatic in theory is in fact pretty unnoticeable.
If you are really curious: http://en.wikipedia.org/wiki/Auditory_masking and http://en.wikipedia.org/wiki/Cocktail_party_effect
Of course I simplified the above, because the lower the bitrates get, the more the dynamic as well as frequency range IS compromised as well as more and more artifacts come in. These are signals, which were not originally there but came in by the process itself. Go ask your mobile phone for a demo!
Since MP3 takes the hearing of the average human into account, there are people who detect compression better than others.
Sad thing is, that youngsters tend to prefer nowadays MP3s over real music. It is assumed that it is “easier” to listen to and our hearing gets “lazy” because we are only presented with the most needed parts of the music.
Conclusion: The provided “differential” audio file indeed lets you listen to the difference between MP3 and uncompressed BUT it does NOT show you what you “missed” hearing. You would have not heard it anyhow!!
Just think about the CSI guys recording in your house and afterwards trying to extract the voice of the TV host masked by all this kids screaming!
BTW: This is the very reason why all this intelligence guys do NEVER record in MP3 format no matter how big the bitrate is. The recorder would simply throw everything away, what you cannot hear now. This means the masked voice would not be on the recording and therefor NOBODY could filter it out again no matter how cool the gear is.
So, MP3 can only be the END product, but never a production format! The encoder needs to know the whole picture before it knows what to omit because of masking. As long as all this mixing and mastering is not yet finished, there is no “big picture” yet.
Hope this is of interest to you and your listeners.
Again, thank you Harold. You make some very valid points.