The circuit allows you to eliminate the vocal portion of an audio signal, while leaving the instrumental portion. The circuit mixes two channels that must be 180 out of phase, so the signals that form the center-stereo image is canceled out.

Those signals usually appear in phase. Resistor R3 biases the noninverting input of U1 from a center tap formed by resistors R1 and R4, and capacitor C3. Resistor R4, capacitor C3, and potentiometer R6 form a negative-feedback circuit that establishes the closed-loop voltage gain of U1 at unity. The signal is inverted between the input and output.

Click on Picture for Full Size

Signals applied to the right input are coupled to the noninverting input of U1 through C4 and attenuating resistor R5. Resistors R3 and R5 make up a 6 dB attenuator, so once again, there is unity voltage gain between the input and the output. However, the right input signal is not inverted.

Therefore, a signal appearing at both inputs is phased out by the circuit and will not appear at the output. Even if the two input signals are at slightly different levels because of different source impedances, you can still adjust for full cancellation by carefully tweaking R6.