Learn the true meaning of the word "analog" and other '70s arcana.
Analog synthesizers, real and emulated, are enjoying renewed popularity today. Indeed, analog devices of all types might be more numerous now than they were during the heyday of ARP, Moog, and Buchla. At the dawn of the 21st century, there is steady demand for the technologies of yesteryear. This article will help you understand some basic analog-synthesis concepts and introduce you to some of the key features of analog synthesizers. Then you'll be able to better evaluate the analog and pseudoanalog products flooding music stores.
The term analog has nothing to do with resonant filters, ring modulators, or any other widely advertised "analog" features. Vintage analog machines did have these features, but so do digital devices. Analog applies properly to signals and the devices that generate or process them, not to particular synthesizer options.
FIG. 1: A microphone converts a sound wave into an electrical signal that is analogous to the original sound wave. A graph of air pressure over time for the sound wave looks similar to a graph of voltage over time for the electrical signal.
Consider the quintessential analog device: the microphone. A mic responds to fluctuations in air pressure (that is, sound). It puts out corresponding fluctuations in electrical "pressure" (that is, voltage). If you were to plot both air pressure and voltage variations over time, the graphs would look very similar (see Fig 1.).
The fluctuating voltage at the mic output is a signal, not a sound. The signal is an electronic representation, or analog, of the original sound. That's all the word analog means - there's no mystery to it.
Whether it originates in a mic or electronic circuitry, an analog signal represents sound directly and continuously. The voltage does not move in discrete steps from one level to another; it flows smoothly in an infinite continuum of voltage levels. In the analog universe, even stepped waveforms such as square waves move continuously. A square wave's leading and trailing edges aren't truly "square." Examine a square wave closely on an oscilloscope, and you'll see slightly rounded edges, because the voltage takes a finite time to rise and fall. No physical oscillator can achieve the infinitesimal rise time of the ideal square wave.
Analog signals have two key properties: (1) they are continuous, and (2) their parameters - frequency, amplitude, and phase - are continuously (and infinitely) variable. These properties make analog-synthesizer signals profoundly different from digital synthesizer signals. By definition, digital synths represent signals as numbers. Digital signals are quantized into a finite number of discrete steps, and there are no levels between steps. Likewise, parameter values on a digital synthesizer are quantized into a finite number of steps. Smaller step sizes give a digital synthesizer higher resolution; the higher the resolution, the better the synth can approximate the infinite resolution of analog devices.
It follows from properties 1 and 2 that between any two positions on an analog oscillator's frequency knob, there are an infinite number of other settings. As a result, tuning two analog oscillators to the same frequency is virtually impossible.
This statement is not as absurd as it seems. Oscillators resemble the strings that produce individual piano notes. Any piano tuner knows that two or three strings can be tuned to approximately, but not exactly, the same frequency. This slight detuning between strings produces an amplitude/ timbral fluctuation called beating, which is considered musically desirable.
You tune analog oscillators as you do piano strings, listening for beats. Yet even if you did tune two analog oscillators to an exact unison or octave, they'd soon drift out of tune. Even the best analog oscillators are inherently unstable, unlike digital oscillators, which are referenced to a highly accurate system clock. Over time, analog oscillators will drift randomly up or down in frequency. Beating is practically unavoidable on true analog synthesizers, unless the oscillators are synchronized or otherwise artificially stabilized.
The warm sound of beating oscillators with complex waveforms is a popular characteristic of analog synthesizers. To get a similar effect on a digital synthesizer, you must detune the oscillators, preferably with a little randomized frequency variation.
These concepts also apply to filter cutoff and Q, signal gain, and all other parameters: if you want analog sounds from a digital machine, you need the highest resolution you can get. Manufacturers seldom reveal how they quantize parameter values. When you check out an analog-modeling synth, turn the knobs slowly and listen carefully for discontinuities.
Just as with audio signals, control signals determine how an analog synthesizer sounds. To understand this, you need to understand the concept of voltage control. Synthesizer modules such as voltage-controlled oscillators (VCOs), filters (VCFs), and amplifiers (VCAs) have one or more control-signal inputs. A varying voltage (signal) applied to a module's control input causes a particular parameter - a VCO's frequency, a VCF's cutoff frequency, or a VCA's amplitude - to vary in a similar manner. Analog control signals are just as continuous as audio signals. Any parameter under voltage control fluctuates continuously in proportion to the control signal.
Take amplitude envelopes, for instance. An analog envelope generator outputs a smoothly varying voltage. If the gain of a VCA is controlled by this envelope, it rises and falls just as smoothly.
Digital control signals work quite differently. Again, I'll use envelopes as an example. The function of a digital envelope generator is to generate a stream of numbers at some periodic rate. These numbers are used to scale the instantaneous amplitude values of audio samples. The rate at which the envelope generator produces its values is called the control rate. It's usually a fraction of the audio sample rate. For example, if the audio sample rate is 44,100 Hz, the envelope generator might produce values at a control rate of 441 Hz. In this case, there would be a level change every 100 samples. If the control rate is too slow, there can be an audible "staircase" effect.
A modular synthesizer is constructed from an expandable number of functionally separate components called modules. Interconnections between modules are pretty much unrestricted. Many famous analog machines (such as the Moog 55 and ARP 2500) were modular, so the term is often associated with analog synthesizers. However, a digital synthesizer can be modular as well. (Plug-ins are the "modules" of today.)
A well-designed modular synthesizer makes almost no electrical or logical distinction between audio and control signals. Analog modular synthesizers let you control almost any module with the output of any other module. For example, you can "play" a VCA's gain with a keyboard voltage (analog synths use keyboards that generate different constant voltages for each note). You can also control a VCO's waveform with a pink-noise generator's filtered output or create complex control-signal hierarchies with feedback loops.
Most hardware synthesizers today are constrained by preset signal paths. If you want a real modular architecture, you'll have to look among the many software synthesizers currently available (see "Going Soft" in the July 2000 issue of EM).
Several categories of synthesizers contend in the analog marketplace. A true analog synthesizer fully lives up to properties 1 and 2: it boasts continuous signals and infinitely variable parameters. Companies such as Serge Modular and Technosaurus build these system types, and a new, all-analog synthesizer has been announced by Big Briar, Bob Moog's company.
A hybrid synthesizer employs some combination of digital and analog techniques. For example, it might have true analog oscillators, filters, and envelope generators controlled with quantized knobs. Quantization makes it possible to store parameters in memory, but it also sacrifices some degree of analog resolution. The famed Sequential Circuits Prophet-5 was an early hybrid synthesizer, and several products from Studio Electronics fall into this category.
An analog-modeling synthesizer is all digital. To emulate the characteristics of true analog synthesizers, it implements mathematical models of analog circuitry. Analog modeling is a type of physical modeling, but it imitates electronic hardware instead of mechanical or acoustic systems. Software synths that employ analog modeling include BitHeadz's Retro AS-1, TC Electronic's Spark Modular, and Native Instruments' Reaktor. Hardware devices offering analog modeling include the Clavia Nord Lead and Korg's OASYS PCI system and MS2000 synth. Guitarists are mad for devices that model tube amps, speaker cabinets, and spring reverbs, such as the Line 6 Pod.
The term virtual analog seems to refer to any digital synthesizer that imitates analog features. These days, any machine with a resonant filter or a bunch of knobs is marketed as "virtual analog," a term I'll avoid.
Now that I've covered some of the general characteristics of analog synthesizers, I'll turn to some specific features and synthesis techniques associated with analog machines. You'll find that many of these features are implemented on analog-modeling machines. If you need a quick review of modulation synthesis basics, refer to "Square One: Modulation Synthesis Methods" in the March 1999 issue of EM and "Square One: FM Basic Training" in the April 1999 issue.
Analog distortion. The way in which analog synthesizers distort a signal contributes significantly to their sound. A good analog model offers a choice of distortion characteristics, reflecting the operational differences among different brands of analog synthesizers.
FIG. 2: This simple patch can sound noticeably different from one analog synthesizer to the next. This is because each synthesizer''s voltage-controlled amplifier has different distortion characteristics.
For example, if you set up the simple patch in Fig. 2 on an ARP 2600 and a Moog 55 modular system, you'd be surprised at how different they sound. On the ARP, you'd hear a shrill sound with audible harmonic and intermodulation distortion. The Moog would sound warmer and less piercing.
FIG. 3: A sawtooth wave distorted by “soft” clipping. The waveshaping effect of the distortion causes the sawtooth ramp to be rounded. The harmonic content of the signal is somewhat reduced in comparison with an undistorted sawtooth.
The difference stems from the VCAs' behavior. On both machines, the summed sawtooth signals exceed the maximum input level of the VCA. This produces severe clipping in the ARP VCA. But the Moog VCA's "soft" distortion characteristics somewhat reduce the signal's harmonic content. Fig. 3 shows a soft-distorted sawtooth wave; notice the rounding of the sawtooth ramp.
The ARP 2600's high internal signal levels and its clipping characteristics made it difficult to create any patch that was free of hard clipping. This is why many '70s records made with the 2600 sound cheesy and edgy, whereas Moog recordings usually don't.
Oscillator synchronization. Designed to lock VCOs in tune, this feature soon became a desired timbral effect in its own right. In hard sync, a slave VCO is forced to conform to the frequency of a master VCO. The slave resets to the beginning of its cycle at each period of the master. This distorts the slave waveform, usually producing a brilliant, overdriven effect. Soft sync uses a phase-locked loop technique to lock oscillators into harmonic frequency ratios without distortion.
Pulse-width modulation (PWM). A pulse wave alternates periodically between a high-voltage value and a low one. Pulse width is defined as the percentage of each cycle that the pulse wave stays at the high value. For example, a square wave is a pulse wave with a width of 50 percent. In PWM, an audio-frequency modulator varies the pulse width, producing many harmonic components that sound like a complex form of amplitude modula-tion. With a low-frequency (below 20 Hz) modulator, PWM sounds a bit like phasing. PWM is often implemented in analog-modeling synthesizers.
Filter modulation. This type of synthesis requires a filter with a continuously variable cutoff frequency that can be controlled by an audio-frequency modulator. Most analog VCFs meet these requirements, but digital implementations of audio-rate filter modulation are rare. This is too bad; it's an interesting technique, often sounding like amplitude modulation with strong peaks in the spectrum.
Step sequencers. Analog sequencers were inspired by tape loops. They were designed to produce short, looping patterns of melodies, rhythms, and accents. An analog sequencer is driven by a clock signal and repeatedly steps through a series of individually tuned voltages. Generating pitch sequences was the most common application of analog sequencers. The stepped voltages were applied to the pitch-control input of a VCO, producing a repeating melodic sequence that was typically about 16 notes long. Most users of step sequencers never figured out how to vary the steps' duration, although it was quite possible to do so.
A few analog masters, such as Morton Subotnick and Roger Powell, devised amazing variants of this essentially boring procedure. Digitally implementing step sequencing is trivial, and it's making a comeback due to its kinship with fashionable looping devices such as samplers and drum machines.
Sample and hold. Also driven by a clock, a sample-and-hold circuit periodically takes a "snapshot" of its input signal and holds that value until the next clock pulse. This process is similar to that of a sampler's analog-to-digital (A/D) converter, which takes a snapshot of the incoming analog signal's instantaneous level many times per second and stores that value until the next snapshot. However, a sample-and-hold circuit takes its snapshots much more slowly than an A/D converter.
FIG. 4: In this patch, the output of an analog sequencer is sampled by a sample-and-hold circuit and patched to the frequency-control input of a voltage-controlled oscillator (VCO). The sequencer and sample-and-hold module have independent clocks. As the clocks go in and out of phase, an ever-changing pattern of pitches is produced.
On many vintage machines, such as the ARP 2600, the preset input to the sample and hold was noise. Thus, the output was a series of random voltage values, the forerunner of the random LFO waveforms seen on almost every modern synthesizer. Fortunately, you can do many more things with a sample and hold. For example, the patch in Fig. 4 uses a step sequencer with a sample and hold, each independently clocked. A sequence of five voltages is sampled, and the output is applied to a VCO. This produces a pattern of five pitches that varies continuously as the two clocks go in and out of phase.
If you want to delve into analog lore, you'll have to hunt for old, out-of-print publications. Check out these titles:
- James Michmerhuizen's user manual for the ARP synthesizer (originally published in 1971; reprinted in 2000 by the author). You can order it at world.std.com/~jamzen/ theBachWorks/arpman.html.
- Analog Electronic Music Techniques (Schirmer, 1985) by James Wagoner and Joel Naumann.
- The Technique of Electronic Music (Schirmer, 1981) by Thomas Wells.
- Electric Sound (Prentice Hall, 1977) by Joel Chadabe. A historical overview of electronic musical instruments.
Frankly, the analog synthesizers of the '70s were a headache to work with, compared with today's imitations. If you find a vintage machine, play with it for fun, but don't buy it unless you're a historian or nostalgia addict. If you seek analog or analog-like sounds, you're much better off with a modern hybrid or analog-modeling synthesizer. Although it might not meet true analog specs, it will be pitch-stable and have patch memory and MIDI compatibility.