Precisely Experimental: A Progression of Analog Synthesizers

By R. Luke DuBois

Luke Dubois walks us through a brief history of Electronic Music Research in New York City and his work to rethink the traditional design paradigms of analog synthesizers. He speaks about the Shelfisizer project built at NYU IDM and has built interactive p5 sketches that allows you to explore how the Shelfisizer works.

When we talk about music, we have a vocabulary for how music is made. We perform, or improvise, or compose, or notate. More recently, we record, overdub, track our sounds. In the 21st Century, more and more musicians do something a little different – they program music. This summer I spent some time thinking about what this means, and how we could revisit the original tool by which music was programmed: the sequencer.

Around twenty-five years ago, long before I’d heard of ITP, I started hanging out at the Electronic Music Center at Columbia University. Founded in the mid-1950s, and now renamed the Computer Music Center, this place is the oldest research facility for electronic music in the Western Hemisphere. The CMC became a pioneer in electronic research, ranging from early experiments with tape delay, basic research in audio mixing and spatialization, and the development of open source computer music languages. In addition, because of the CMC, Columbia has one of the top programs in music composition and sound art in the United States. But despite its over sixty years doing innovative work, the Center is best known for a room-size relic in the form of a piece of vacuum tube-powered equipment called the RCA Mark II Sound Synthesizer, nicknamed “Victor”. Victor, as it’s been often claimed, is not the world’s first synthesizer.

An Older Photo of Faculty Members at the Computer Music Center, standing in front of the RCA Mark 2 Sound Synth

Depending on whether electricity is key to how you define synthesizer and how pedantic you might be, the world’s first synthesizer may, in fact, be the Telharmonium, developed in the 1890s. Alternately, whoever started naming the stops on pipe organs after the instruments they mimicked – most likely in Byzantium in the 8th Century CE – could have been the inventor of the synthesizer.

Similarly, Victor  wasn’t the world’s first sequencer – player pianos and music boxes, developed in the age of mechanical innovation that gave us cuckoo clocks and pocket watches, could reliably play repeated sequences of notes over a century before.

In college and graduate school I was a fairly consistent presence at the CMC, where I would play it, show it off, and sing its praises to visitors. During these demos, I’d generally gloss over the fact that the system was obsolete within five years of its installation because some smart people in New Jersey had just invented the transistor.

Short Audio Piece Composed on the RCA

What Victor did have, as a first, was a combination of synthesizer and sequencer that was robust, reliable, and fully “programmable” by the user. While it wasn’t a computer by any stretch of the imagination, Victor had a lot of things about it that were influenced by early computers, telecommunications infrastructure, and information theory. The sequencer ran on paper tape, similar to a player piano, but with a more open notational schema. Instead of the holes simply representing notes, they could be assigned to signal changes in amplitudes, durations, filter settings, and so on.

Small patchboards allowed the user to rewire the large racks of equipment that made up each “voice” of the synthesizer such that, for example, one voice could be mellow and soft and another could be brassy with sharp attacks. Finally, a four-bit binary encoder system was used for every parameter, assigning the entire vocabulary of musical attributes the RCA Mark II understood to a value from 0 to 15.

The correspondence between Columbia and RCA around the development of the synthesizer is  interesting, in that it highlights two things. One, is that getting the machine built and paid for was a little bit of a grift: Vladimir Ussachevsky and Milton Babbitt, the main faculty instigators at Columbia and Princeton, respectively, promised RCA that the synthesizer, once built, would replace the symphony orchestra. RCA was, after all, the Radio Corporation of America, and had a vested interest in the continuing success of radio stations that, more than likely, were supporting a unionized radio orchestra that was expensive to maintain.

The second point  of interest is that Victor was designed to be (and, once built, praised as) a precision instrument. This machine can play notes faster, more in tune, and with higher accuracy in every sense than the best concert violinist money can buy. The player piano interface, the patching system, and the all-around 4-bit quantization,  meant that music written for Victor would come out as intended every time. Herb Belar and Harry Olson, the RCA engineers who designed the machine, claimed that it reproduced the programmed music with “military-grade” reliability.

The problem was, there was no way to make the machine play inaccurately, or randomly, or in a way that could include improvisation or experimentation as part of the creative process. The machine was quite loud, so loud that a shellac record lathe was hooked up to record an inexpensive run of a composition for listening in the next room. This high modernist riff on the dub plate meant that the RCA workflow was iterative, non-real time, and strangely disconnected from the immediacy of control that one expects modern analog synthesizers to have.

Luke DuBois at the Computer Music Center

Columbia had other synthesizers as well, including a series of modular synthesizers developed a decade after Victor, using transistors that would condemn the large, vacuum tube machine to obsolescence by its 10th birthday.  

Serge Modular System
Serge Modular System

These synthesizers, made by Don Buchla and Serge Tcherepnin, were “West Coast” modular synthesizers, and were in many ways the opposite of the RCA Mark II. Optimized for experimentation, easy to use in live performance, and designed with a certain intentional inability to save your work, they were the machines that I really learned to compose on, making weird fantasias and downtempo dance music with the band I had started called the Freight Elevator Quartet.

The Buchla and Serge machines, like the RCA Mark II, had sequencers – rows of knobs that you could then trigger through stage-by-stage, to control the frequency of an oscillator to make a melody (or any other parameter on the synthesizer). But they also had modules that generated noise, or warped the sounds you could make in unexpected ways. Buchla’s most famous module in this vein was fittingly called the Source of Uncertainty, and if he liked you, he’d give you a red panel with your order, coated with LSD that’s apparently still potent 40 years later, with the idea that the contact high would encourage you to use the equipment in more adventurous ways.

Serge synthesizers were of particular interest to me, because unlike most systems, they were designed to obfuscate (or eliminate entirely) the distinction between voltages that were “making sound” (audio) and voltages that were “controlling sound” (control voltages). This allowed you to take a timing clock and speed it up for use as a square wave, or taking something that would normally filter a sound and using it to warp and distort the signal controlling the volume of something else. 

Last year, we commissioned Darrin Weiner, a synthesizer builder in Berlin, to make us a Serge Modular system at NYU.

Serge Modular System Custom Built for IDM
Serge Modular System Custom Built for IDM

This machine has been an incredible addition to the IDM audio lab, and a lot of fun to work with. I kept thinking, though, about the RCA / Buchla dichotomy of synthesizer systems that are precise, versus systems  that are experimental. This past summer, I invested some time in developing a new set of modules that were, in fact, precisely experimental, to work with our new Serge. These modules would use Arduino-class microcontrollers and CMOS chips to keep the part count low, while moving as much of the logic as possible to software, and keep “analog” only that would eventually make their way to the audio path of the synthesizer. This had the added bonus of making easier to design and more legible to work with, lowering the bar for students at NYU to make more designs in the future. These new modules were designed to be housed in a standard 19-inch rack drawer with holes drilled in the front for the interface. I call this project the Shelfisizer.

Picture of the Shelfisizer Synthesizer
The Shelfisizer Synthesizer
Shot of the Inside of the Shelfisizer Synthesizer
As you can see, it’s a bit of a mess inside, but it gets the work done.

The modules I developed were all attempts to broaden our understanding of what a sequencer does. Three of them are mocked up below as P5 sketches.

The first is fairly straightforward:

It consists of four rows of patterns that can be up to 16 steps long. These patterns go in lock-step, so that when row A is on step 7 of the pattern, so are rows B, C, and D. This is a classic pattern sequencer used in, e.g. drum machines, but has a key-driven interface to cruise around the grid and change patterns.

Next, I thought I’d reimagine the sequence to be less about a specific pattern and more about specific cycles of activity:

In this module, six channels of voltages can be compared to a threshold (and sometimes each other). When a clock signal is applied giving the system a rhythmic pulse, the state of the voltages determine whether a note gets triggered. Different modes change these rules, and a number of the modes apply hysteresis, so that the incoming voltages have to fall below the threshold before they can trigger again. By changing the sliders, you can simulate the musical effects of rising voltage ramps of different speeds against different clocks.

Finally, a third sequencer considers the sequence not as a series of steps but as a table of numbers:

In this module, three voltages scan sixteen possible values, so that value 1 is output when an incoming voltage is low, value 16 when it’s at its maximum, and everything in between. This means that a rising voltage input will play the values in order, like a traditional sequencer. It also means, however, that a noise source will randomly sample the values; and a triangle wave input will scan the sequence back and forth like a palindrome.

The idea here is to break open these systems – musical, electronic, computational – and treat them more metaphorically than literally. You can see the hardware in action here:

One of the things I learned from building the Shelfisizer is that there’s a false dichotomy in so many of the machines we use to make music (and art, for that matter). We are presented with many tools that claim to be precise; other tools promise us room to experiment. The circuits I made this summer were an attempt to have both – to find experimentation in precision, and accurate ways to bring about complex musical interactions.

A playlist of music I’ve been making over the past year with the Serge and the Shelfisizer can be found here:

Hope you enjoy.