Research Projects

Thursday, December 20, 2007

Audio Lesson

I've taught this lesson before, but to the uninitiate the terminology is meaningless. The popularity of low fidelity audio formats like MP3 virtually proves that either the general public cannot discern audio quality or does not care. But still, for those of you who are podcasting for the first time, or merely wondering why the mic sounds terrible when you pot up all the way.. allow me to explain. Dont' worry I'll use audio visual aids. Todays lesson is all about distortion.



In this example above I take a tiny clip of the song "Christmas Cha Cha" by the Merry Macs. Note the position of the peaks and valleys in the wave form. If anyone cares, I'm using an open source audio editor named Audacity. I recommend it.



Here (above) I've amplified the audio. It's louder than the original clip but no data was lost. I've not compressed or limited the audio nor process it. in fact when you play the sample you'll find the difference to be barely discernible.



But here (above) I've over-amplified the audio. The peak data has been lost. As one technician I know once described it "I've colored outside the lines." In reality it's much like that. I've increased the amplitude beyond the capacity the bit rate and bandwidth will allow. It's called Amplitude distortion. The technical definition is "the output amplitude is not a linear function of the input amplitude. " Some operators use the terminology "hot," as in "the levels are too hot."



Here I've corrected the volume. It's less loud, but the audio still sounds terrible. That's because the peak data is lost forever. The waveform is full of "flat tops." Lowering the amplitude after you've had clipping is useless. that distortion has already happened. The audio cannot be "undistorted." No amount of post processing can fix it. My personal recommendation is to create the highest quality source audio BEFORE editing or broadcasting your recorded content.

Now don't be downloading my audio and checking one me. Your own audio editor will try to "rebuild" the peak data in it's own GUI. New peaks will appear, with fewer and lesser flat tops, but the distortion will remain.

No comments:

Post a Comment