A practical example

Everyone should know what Karaoke* is, you get drunk and sing a familiar song in front of all your drinking buddies. In order to sing you need music, but you don't want the original voice in the music, that would be cheating. What to do...

In stereo music, generally the voice is balanced between the left and right tracks and the instruments are split between the two tracks to give the music depth. If we perform a subtraction, left - right then any differences between the two channels will remain but anything that is in both channels should be cancelled out. That is x-y = x-y but x-x=0!

Let's give it a try with Year of the Cat (sorry Al, I need a snippet to make a point)

Hit the play button on the controller to hear it.

Year of the Cat normal

Year of the Cat after the difference is taken.

Pretty simple huh? This is an easy trick that can be done without DSP with a fairly simple circuit. The great thing about DSP is that when we use it we don't have to build the circuit, we don't add noise and we can change the way we do it without having to build another circuit.

To do the demo here I used a few programs, I used SoundApp to convert Year of the Cat from the CD to a Quicktime movie, then I used MoviePlayer to cut the snippet out, then I used SoundEffects to do the actual DSP difference operation. I also downsampled the snippet from 44.1KHz to 20.05KHz. Then I used MoviePlayer again to convert the sound to a Quicktime file so you can hear it (if you have Quicktime 3 or 4).

*Japanese for "making a fool of oneself in public"

Back