Skip to content


2011 August 28

Update: The electrical schematic, PCB and software is now on Github.

We humans tend to treat our perceptions as holistic—as seeing things as things, instead of as an accumulation of parts, details and features. This is the great abstracting power of the 3-pound neural network in our heads, and is a major differentiating characteristic from other computational paradigms.

The holistic pattern of Deconspectrum is that of the spectrum analyzer, a tool for understanding the components of a sound signal. You’d recognize the spectrum analyzer as the (often colorful) bar graph showing how loud the bass, mids and treble are on a fancy car radio or web MP3 player (which are often just fake animations). These graphical displays are intended to give an insight into the component parts of a signal that complement our own perceptions: we can hear the loudness of the bass or the absence of midrange, but we couldn’t tell you something as specific as how much energy there is at 250hz (at least, not without practice).

Any given sound can be understood as an accumulation of sine waves, a simple oscillation of air. Breaking apart a sound into these components is known as decomposition, and is the basic principal of a mathematical technique called the Fast-Fourier Transform, which gives us access to the raw numbers of a sound, the exact amounts of energy, that the unaided ear can’t provide. The spectrum analyzer supplements our perception by dividing a signal into “bands,” organized by the frequencies of these component parts. Or, at the very least, they give the manufacturers of audio devices something flashy to add to their product.

These tools are always presented as a single whole with a single function, one complete thing that breaks down the features of another perceptually complete thing. Deconspectrum pulls these things apart, eviscerates them, granting their constituents a new autonomy in the form of small luminous cubes that are each tuned to their own frequency band. These bands are mapped to the color of the cube, from yellow to green to blue to red and all the colors in between. They are conceptually identical to the rising and falling columns of the spectrum analyzer, but now spatially distributed and able to be moved and re-ordered—and crucially, still capable of emerging the holistic experience of the sound and its analysis. In other words, Deconspectrum is both the sum and the parts, the spectrum analyzer and its frequency bands, the pattern and the features.

Deconspectrum is best experienced in quiet, when the viewer can whistle, sing or hum distinct tones and see before them their melody reflected in color.

The individual units are completely autonomous, containing their own microphone, processor and color LED, and are also for sale individually as artist editions for $25. As such, they have an additional “standalone” mode that follows the peak frequency of whatever it hears, creating a colorful reflection of a melody, voice or spontaneous noise. They are both 9V battery and wall powerable with any 9 volt+ power supply.

Related Posts:

4 Responses
  1. October 6, 2011

    Following up on dorkbot-nyc Wed Oct 5th where I talked with you after your presentation. I want to invite you to show your work at Eclectic Chaos IIA, Nov 11-13, at Valley Settlement House in West Orange, New Jersey. You can check Eclectic Chaos out on the website: If you want to explore what it entails, send me an email and my reply will have all my contact info … or you can send me yours and I will contact you directly. Deconspectrum fits perfectly into our developing concept (not really new since da Vinci) of a SMArT education, drawing children into science, math, and technology through art … in a sustainable and replicable program.

Trackbacks and Pingbacks

  1. Programming an ATMega With No Bootloader Using the USBtinyISP | Limina.Log
  2. Arduino: Compensating for a Logarithmic Curve | Limina.Log
  3. Dorkbot Next Wednesday! | Creative DIY Cultures and Participatory Learning

Comments are closed.