Some Thoughts on Modular Analog Synthesis and Interface

As a preface, I'd like to admit that I made no effort to search the human factors or electronic music literature to see if I was restating someone else's work, so these ideas may be completely unoriginal. I'd also like to thank Pete Rice & Patrick Pelletier, who put up with my initial ravings on the subject a couple of weeks ago and provided useful feedback.

Before reading it, those of you who don't have a clear picture in your mind of what a modular analog synthesizer is may want to check out some pictures of them. Here is a site containing many analog synthesizer pictures.

An important analogy modular synth in my life has been the Serge, an example of which is (somewhat crappily) pictured below.

(Here are even more Serges: 1, 2, 3) In contrast, here is a classic non-modular analog synthesizer, the MiniMoog (the big differece is hardwired patching and an integral, not optional, organ-style keyboard).

For another contrast, here is a modern rack-mount digital synthesizer, the Kurzweil K2500R.

It is often claimed that there are positive characteristics of modular analog synthesizers (MA-synths) that modern digital synthesizers do not possess. What is so special about them? Is this pure nostalgia or is there something to be learned here?

There are claims about "the analog sound" being this or that ("sweeter", "fatter", "warmer" etc.). (These claims pertain to all analog synthesizers, modular or not.) These claims are a mix of truth and mystique. To the extent that "the analog sound" refers to phenomenon such as circuit non-linearities, it is possible to see how "the analog sound" might be digitally emulated in realtime (with adequate computational power). The important point here is that these differences are not inherent to digital and analog synthesis, rather, they are a reflection of the state of the art in digital synthesis. Another way of stating this is that sounds produced by analog synthesis are a subset of those producible by digital synthesis.

There are also claims about the interfaces to MA-synths being this or that ("more intuitive", "more physical", etc.). It is primarily these claims that I am interested in investigating here. As with differences in sound quality, these differences in interface are not inherent to MA-synths and digital synths. Although not inherent, they are more profound than the differences in sound quality; perhaps this is why I choose to devote my attention to them here.

What is the interface to a digital synthesizer? Typically, it physically consists of momentary buttons and "continuous" controllers (knobs) for input and an LCD display for output. But digital synths can also be programmed via MIDI, which makes the interface abstract and therefore able to take on a wide variety of forms. The only major alternative interface that is used today is the patch editor, a computer program that controls the synth via MIDI. In this case the I/O devices are the familiar mouse/keyboard and monitor respectively.

Below I will compare the interfaces of MA-synths, digital synths, and patch editors in terms of some useful properties for interface characterization. Then I will suggest some ways in which new interfaces to digital synths can be positioned with respect to these properties.

First of all, it is necessary to define some terms. A control is a visible, manipulable part of the interface that affects one or more parameters of synthesis. The state of the device is all of its parameter values. A knob is used here in a general sense to include rotary and straight controls, i.e.a slider is a knob in this context.

The interface to an MA-synth is non-modal: the meaning of a given control is always the same. A consequence of non-modality is that all controls are available to the user in parallel. In contrast, digital synth interfaces are highly modal: a single knob can adjust many different parameters according to context (mode). For example, "soft buttons" are as modal as sin. Patch editors can be non-modal, especially if designed to be able to take advantage of a large monitor. Non-modality is nice in that it gives parameters a unique spatial location. This unique location can serve as a mnemonic and also make access time quicker. One of the prices of non-modality is a physically large interface.

The interface to an MA-synth is state-synchronous: its controls are guaranteed to be in sync with the state of the device. A nice property of state-synchronicity is that the controls themselves form a visual representation of whatever part of the state of the device they control. The values of all of the parameters under control can be "read" off the controls. Digital synths are non-state-synchronous. Patch editors are state-synchronous.

The interface to an MA-synth is passive: there are no controls that can change other controls. For instance, you cannot "gang" knobs together. This is also true of digital synths. Patch editors are active; for example, when you activate a control to change the patch, other lower-level controls will change to reflect that patch.

The interface to an MA-synth is non-global: there are no controls that change the entire state of the instrument. Digital synths and patch editors have global interfaces.

The table below summarizes the characterization of the synth interfaces according to the above-described parameters.

It is important to note that the four properties I have introduced are not independent of each other. For instance, a global, state-synchronous interface must be active. I'm not sure if there are any other dependencies between these properties.

So, in order to make an MA-synth global, it needs to be active. In fact, it needs to be completely active, i.e. it needs to have all knobs motorized and a robotic patcher working with it. In order to make a digital synth state-synchronous, it needs to be active, i.e. it needs motorized knobs. An interesting way to speed up an passive interface would be to visually indicate target control positions and perhaps auditorily report errors in controls. For example, an MA-synth might have a ring of LEDs surrounding every knob and the one closest to the desired angle would be lit. It might also beep at you if you connected a patch cord incorrectly.

A very profound way in which digital synths are modal is their distinction between parameters available during performance and editing. MA-synths have no such distinction. A very profound musical consequence of this is that with digital synths, sound design is viewed as a separate activity from composition/performance (I owe this observation to Tod Machover). First you make your sounds, then you decide when to use them & at what transposition. MA-synths encourage a philosophy in which sound design is composition and composition is sound design.

A neat machine-to-machine interface feature of MA-synths is that you can use different manufacturer's modules together. (Strictly speaking this is not true, but differences in connector type and voltage range are not difficult to overcome.) In contrast, there is no way to get access to the internals of a digital synthesizer. An important consequence of this is that whereas an MA-synth is also an analog signal processor, there is a strict division between digital signal processors and digital synthesizers. (One exception is the Korg Wavestation A/D.) The fact that a whole bunch of interesting signals are completely stuck inside the box in digital synthesis limits the creativity of someone lucky enough to own devices made by several different manufacturers.

A simple example is if one of your synths has a sound that you would love to process through a great filter in another of your synths, your state could be likened to that of a canoe paddler who has wandered far up a body of water known as Shit Creek. The only way in which you may have half a paddle is if the nice filter is part of a sampler. Then you can sample the sound from synth1 and play it back through the nice filter on your sampler. The reason why you only have half a paddle in this case is that if synth1 was indeed a synthesizer, you have lost the parametric control over the sound that you had when it was played in its original instrument.

What is to be done? One option is to return to the use of MA-synths. Another option is to produce non-modal digital synths. Roland's {JD-800? JV-800?} was a move in this direction; to my knowledge no digital synth has more knobs. How about non-modal, state-synchronous digital synths?

How about modular digital synths that talk AES/EBU or SP/DIF or Ethernet or IP Lite? How about modular-like interfaces to nonmodular (monolithic) digital synthesizers? How about modular-like interfaces to modular software synthesizers such as Csound? This has already been done in non-realtime with Digidesign's TurboSynth and in realtime with John Bates Unison and Symbolic Sound's Kyma. But what if the modular-like interface is a physical interface, i.e. not on a computer screen? In other words, what about the following plan. Make a replica of a Buchla front panel. Attach each of its pots to +5V, GND, and an ADC (analog to digital converter). Make every output give +5V. Make every input GND without a cord inserted (use a switching jack) & attach a comparator to the input. "Glue" the ADCs and comparators to a realtime implementation of Csound. Or somehow use Maggie's triangles instead of a Buchla front panel. In any case, after you follow these "simple" steps, you are the proud owner of a digital synthesizer with an MA-synth-like interface!

I'll end these notes somewhat abruptly here. I hope they have defined some useful properties for interface characterization and stimulated some thinking about the design of new digital synthesizers as it might be informed by the experience of modular analog synthesis.