From the Archives
Have you ever wondered why Donald Buchla chose to separate audio and control signals in his instruments? Of course you have.
At risk of being overly reductive, we can say that most structurally similar instruments of the era—that is to say, early analog modular synthesizers from the 1960s and 1970s which used patch cables for inter-module signal connections—used a single style of patch cable and relatively uniform voltage ranges for most internal signals. In contemporary Moog systems, for instance, the majority of inter-module signal connections are handled via 1/4" TS cables, regardless of whether the signal carried over the cable is meant to be used as a control voltage or as a part of the instrument's audio path. To this day, most modular synthesizers maintain this one-cable-type-for-all-signals approach. And in fact, many users of such instruments see the resultant blurring of roles between devices intended to act on sound and musical structure as being advantageous. To some, such instruments may seem to present a uniquely “open-ended” approach to musical organization.
As these technologies initially emerged, though, instrument designers such as Moog or Buchla had few prior models to draw upon. It was up to each individual designer to decide how and why they might approach all aspects of system design, including the differentiation versus overlap of tools designed to produce and shape sound itself as opposed to tools designed to shape musical structure.
Even in his earliest electronic instruments, Buchla made the decision to differentiate between audio and control structures, going so far as to use entirely distinct cable types for audio and control/timing signals (3.58mm unbalanced TS connections for audio and 4mm single-conductor “banana” jack connections for control signals and timing pulses).
So, back to our initial question: why? This question has been publicly raised countless times in recent years—having gained new pertinence following the emergence and proliferation of Eurorack modular synthesizers, which present a more homogenized approach to the sound versus structure differentiation. So, from a modern perspective, this differentiation might initially seem peculiar, counter-intuitive, or even musically limiting. Luckily, we don't have to postulate as to why Buchla made the decision to distinguish between these aspects of sonic organization: he published an explanation of his reasoning in the early 1970s.
In 1971, prompted by perceived public confusion surrounding the topic, Buchla wrote an open letter to be published in the ninth issue of the experimental music magazine/journal Source: Music of the Avant-Garde. In this letter, he provides a series of reasons for maintaining the differentiation of sound and structure in his instruments. Some reasons are based on user experience, some are technically pragmatic, and others point to design considerations for augmentation of his instruments via computer control—all points are interesting and well worth considering as you approach any modular electronic musical instrument. The letter in question is transcribed below in its entirety.
-RG
On the Desirability of Distinguishing Between
Sound and Structure
For several years we’ve lived with some fundamental differences between East and West Coast electronic music instrumentation. Discussions in the literature have been minimal and reasonably non-partial, with musicians more or less left to their own means to seek out the system that best suited their needs. But now we find old features and resurrected schemes being touted as vastly superior and even new and revolutionary. It is to one significant difference, the differentiation versus confusion of sound and structure, that I would like to address some remarks.
In the early ‘60s electronic instrumentation included band-switched laboratory oscillators, segment-type waveform generators, miscellaneous surplus electronic gear, keyboards, relays and telegraph keys. Anything with an input was fair game for modulation by anything else; instruments were used indiscriminately to generate sound and/or control functions. The break with this equipment-imposed tradition came in 1964 with the introduction of the Modular Electronic Music System (now marketed by C.B.S. Musical Instruments). This system differentiated between audio signals (the raw material of electronic music), control voltages (defining vertical structure, including pitch, timbre, loudness, location, and other musical parameters), and timing pulses (defining points in time, intervals, and durations). Note that this separation is not musically revolutionary—it is incorporated in virtually every scheme of musical notation used—sounds emerge from instruments; parameters are described by symbols; time is defined by marks and spaces organized along some single dimension (classically horizontal). So why did we isolate the functions and why do we persist in maintaining the distinction? The reasons are numerous, interrelated, and here follow.
1. (conceptual)
A complex system becomes inherently simpler when broken into distinct subsystems. A musical situation is easier to implement and to comprehend if the functions of the elements and their connectives are clearly discernable (sic) by their visual characteristics. This principle is perhaps the most important reason for maintaining a consistent distinction between sound and structure.
2. (electrical)
To achieve the highest accuracy and stability, control voltages must span a relatively large range of values; ranges of from five to twenty volts are employed in various synthesizers—the series 200 Electric Music Box operates on a 15-volt control range. Audio voltages, if adequate “headroom” is to be provided, should nominally be around one volt (peak to peak of three volts), leaving adequate dynamic range for handling live sound and the inevitable peaks contributed by processes such as summation, multiplication, equalization, and reverberation. Thus, the optimal amplitudes of control voltages and audio signals are quite dissimilar. This discrepancy requires accommodations that include the provision of high audio gain to enable audio signals to control parameters. It is this high in-system gain that accounts, at least in part, for the relatively high noise level and signal leakage encountered when attempting complex patching of certain synthesizers.
3. (related electrical consideration)
In systems utilizing interchangeable control voltages and signals, all elements (even gates, mixer, and modulation inputs) must be d.c.-coupled. Offset and drift is amplified instead of blocked, producing patching (or switching) clicks and occasional frequency uncertainties. (In patchcord systems clicks are undesirable, as patches frequently are altered while playing the instrument.)
4. (logical)
The differentiation of function allows for some simple but powerful logic to be implemented at the patching level. Audio outputs have in all current systems low constant impedance, and when connected together result in a mixing with slight attenuation of whatever signals are present. This is an occasionally useful thing to do with audio outputs, but hardly interesting when applied to control voltages. More useful is the “larger than” function made possible by using diode-coupled control voltage outputs and implemented by direct connection of these outputs. The result is that an element controls a particular parameter only when its output voltage exceeds the voltage produced by other elements connected in parallel.
A similar, but even more useful, benefit accrues from the utilization of diode-coupled pulse outputs. The logical OR connection is implemented by connecting any number of outputs together, allowing an event to be initiated by any of several outputs connected to a single input. An additional interesting feature of the 200 series pulse outputs results from the utilization of a three-state logic scheme that permits event times and durations to be transmitted simultaneously without mutual blocking. The above may be summarized with a simple rule for interconnection: freely interconnect inputs and outputs of the same type—the results are always predictable and useful.
5. (functional)
As was observed above, characteristics for control outputs and audio outputs are pragmatically different. Similarly for input characteristics: signal inputs are desirably a.c.-coupled and their attenuation controlled with audio taper controls; we find processing* inputs most useful for control voltages; pulses, of course, need no input controls. So control and signal modules ideally have dissimilar input and output structures. Their internal functional requirements are invariably quite different as well, and those characteristics that require optimizing for audio processing are seldom the same ones that should be maximized for handling control voltages. Examples of this principle of functional optimization include virtually every module I’ve designed.
6. (computers)
Programming giant computers to pump out a million or so bits per second to be translated into sound is undoubtedly a worthwhile enterprise, but it doesn’t seem as though most of us will ever get near one, much less actually play music on one. The problems are many, but a particularly significant factor would seem to be the cost of handling the enormous amount of data necessary to directly specify every instant of sound pressure. Now, if computers dealt only with the structural parameters of sound and left the actual generation of sound to external hardware, we could drop the required information rate by a factor of around ten thousand and let a minicomputer assume the task of generating control voltages and timing pulses. This is still not a trivial task to fully implement (progress has been made here and there), but it is certainly made easier by initial differentiation of signals and control voltages in the hardware to be controlled, and by making all musical parameters voltage (and therefore computer) controllable.
I’ve tried to be brief, but quite a few considerations preceded my decision to isolate signals and control voltages. Let me know if any points need elaboration. And beware of the generalized engineering approach—it seldom produces the most desirable specific operational environment.
Peace and good music,
Donald F. Buchla
First published in issue 9 of Source – Music of the Avant Garde , 1971
* “Processing” inputs, as employed in the series 200 Electric Music Box, are associated with linear, center-off controls that allow control voltage attenuation and inversion in one continuous rotation.