I started working with feedback in 2006 and particularly this kind of non-linear mixer feedback in 2007. At that time, I was enthralled with David Tudor’s untitled and the raw, visceral quality of no-input mixer seemed to be very much of that world. Later on, thanks to the amazing research of You Nakai and Michael Johnson, I learned that the instrument that was untitled consisted primarily of a series of amplifiers and filters in feedback loops.

My instrument initially consisted solely of a Mackie 1604vlz Pro (a beast in and of itself). It’s a great mixer for feedback because it has 8 direct outs that are post fader, allowing for use of the filter section within the feedback loop. Soon, I began building very simple circuits to alter and control the behavior of the system. Several designs (and the occasional guitar pedal) came and went with varying levels of success, but eventually I landed on a 6×6 matrix, 6 VCA’s and 6 momentary and toggle switches wired in parallel. Two ring modulators were added for processing on the aux sends.

The switches, along with the mixer faders (which remember are inside the feedback loop) provided a really responsive performance interface. Certainly a major part of my attraction to the system is it’s autonomous behavior. At the same time I come from a jazz background and having that kind of gestural…I won’t say control, but influence over the system allowed a degree of fluidity in performance.

The diagram above shows a single channel feedback loop. Ch 1 direct out –> Switching Matrix–>VCA–>Switches–>Ch 1 input. This forms a pretty straightforward oscillator. More complex behavior can be achieved by routing subsequent channels in series. Ch 1 direct out –> Switching Matrix–>VCA–>Switches–>Ch 2 input–>Ch 2 direct out –> Switching Matrix–>VCA–>Switches–>Ch 3 input–>Ch 3 direct out –> Switching Matrix–>VCA–>Switches–>Ch 1 input.

I found that three channels in series provided a nice balance of instability and control, so I landed on using six total channels for feedback. I tapped each channel from it’s insert out and ran each signal into another mixer channel, for, well mixing. But it doesn’t stop there!

Each channel is tapped from that point and sent to the audio interface where it’s analyzed for pitch, loudness, brightness or noisiness. That data can then be scaled and sent back out the interface to the CV inputs of the VCAs. Of course any external source (for instance saxophone or voice) can be analyzed and used for control.

There’s also the option to drive the CV of the VCA’s with oscillators in Max/MSP. I control the oscillator frequency either with the pitch analysis or via midi sequencing/keyboards. Each oscillator also has index and offset control.

Overall the system can be thought of in three parts. The feedback section, mixing/processing section and the software/control section (via the audio interface) for something like this: