The guitar-oriented singer/songwriter is as old a paradigm as the guitar itself, but there’s an inherent conflict: There’s quite a bit of frequency response overlap between voice and guitar. The usual solution is to play more softly when singing, but then the frequencies that don’t overlap are softer, too. Another option when recording is to feed the guitar signal into a compressor or noise gate with sidechaining capabilities, then use the mic signal to drive the sidechain input. This will compress or reduce the guitar level (when using a noise gate, you need to set it for relatively little level reduction, not a full gating effect), but it also lowers all frequencies—not just those in common with your voice.
However, there’s yet another approach I stumbled on when designing an amp sim optimized for acoustic guitars with piezo pickups (the sim was later included with Cakewalk Sonar as the “Acoustic Piezo Amp”). This technique feeds a guitar into a multiband compressor. However, only a midrange band is set to compress—all the other bands are bypassed by setting their thresholds as high as possible.
Set the midrange band to cover a range similar to your voice (around 200Hz to 1kHz) with a fairly low threshold (Fig. 1). If you’re not playing too loudly, then the guitar level will be below the threshold, and the midrange level won’t be reduced. As the guitar level gets louder, the guitar’s midrange will exceed the midrange band’s threshold and compress in your vocal range, while leaving the higher and lower frequencies intact. Note that I often raise the gain for the high and low bands to add a little more spice to the guitar. After all, if a multiband compressor isn’t compressing, then it’s essentially a graphic equalizer.
The main caution is that many multiband compressors have a “look ahead” function so they can prepare in advance to catch peaks. The look ahead time delays the audio, so if you have a choice, choose a multiband compressor with minimal delay.
With software-based recording systems, advanced plug-in jockeys can take this effect further to where it happens only when you’re singing. Split the guitar into three tracks, with an EQ in each band. You want shelving response for the high and low bands, and parametric response for the midrange. Insert a compressor with sidechaining into the mid frequency band, and feed the signal from your vocal mic into the compressor’s sidechain input. Now, the guitar’s midrange level will correlate only to your vocals. Pretty cool, eh? It’s almost like having a robot sound engineer to dip the guitar in just the right places when you’re singing.
Craig Anderton has played on or produced more than 20 major label releases, mastered hundreds of tracks, and written dozens of books. Check out some of his latest music at youtube.com/thecraiganderton.