Neurohedron: A Nonlinear Sequencer Interface

The Neurohedron

Read the full NIME Conference Paper!

Traditional music sequencers are designed fundamentally around predictability and repetition, and these are powerful elements that make them so ubiquitous.  More modern approaches to algorithmic composition heavily involve unpredictability and randomness that is then (sometimes) tamed and manipulated by the composer, resulting in a nonlinear compositional and performative process.

The Neurohedron is a novel music instrument and modal software controller that I conceived of as a nonlinear sequencer.  The simplest traditional sequencers may employ eight steps that return to the first step after reaching the last step; in contrast, the Neurohedron is a three dimensional sequencer with twelve nodes arranged as a dodecahedron.  With this structure, there is no clear or de facto path that the progression from one node to the next may take, unlike the linear and prescribed nature of a traditional sequencer.

Each face on the device is both a touch-switch that sends input triggers to software as well as an electroluminescent panel that displays feedback from the software.  While the platform could be used in a huge variety of ways, my first inclination was to use the inputs as stimuli in a 12-node neural network, where each face is connected in a graph to the 5 faces adjacent to it.  Each of the edges between these faces is then weighted between 0 and 1, and the progression from one face to another follows the highest weighted edges.

Each node (face) in the network is then mapped to a MIDI note.  These mappings are described by a variety of musical modes.  Thus, a sequence on the Neurohedron will send patterns of MIDI notes to a synthesizer platform (Ableton Live, in this case).  This platform is wide open, of course, so anything could be triggered in principal.

Here’s a demonstration of the software model:

The interface is multi-modal, meaning that there are 5 states the interface can be in that determine what it does:

• Input: Pressing a face triggers an input stimulus to the neural network at that node
• Stop: Holding a face down stops any circuits that encounter that node
• Live-play: Sends MIDI notes pursuant to that face’s current mode-mapping
• (Musical) Mode: Changes the mode-mapping of the network
• Randomize: Randomizes all of the neural network’s weights

Here’s a little demo of Matt Boyle & I playing the Neurohedron in live-play mode.

Process & Production

The Neurohedron’s hardware consists of:

• Plexi & aluminum skeleton
• Laser-etched faces
  • Momentary switches
  • EL panels
• Logic Board
  • ATMEGA328PU Microcontroller
  • 2x shift-in registers
  • 2x shift-out registers
  • FTDI USB Interface
• AC Switching array
  • 12x Optoisolated triac & low-current triac pairs

This video demonstrates the shift-in and -out registers controlling switches and panels:

And this one detailing the AC switching:

Here’s a link to a large version of the Neurohedron for print & publication.

And finally, a gallery of production photos: