During the early 1960s, the standard model of analog synthesis became subtractive. Audio signals from one or more oscillators generating waveforms containing lots of overtones pass through a filter to shape the sound by suppressing—subtracting—certain frequencies and exaggerating others, then through an amplifier to vary the sound’s volume.
subtractive synthesis: the use of oscillators that generate richly harmonic waveforms and routing their signals through voltage-controlled filters to attenuate specific harmonics and contour the resulting timbre
additive synthesis: creating complex timbres using multiple oscillators that generate less complex waveforms, traditionally sine waves
sine waves: oscillator-generated, periodic voltages that rise and fall smoothly and symmetrically, following the trigonometric formula for the sine function; at audible frequencies a sine wave produces only the fundamental frequency and no harmonics, sounding similar to a flute; at sub-audible frequencies, the sine wave excels at producing vibrato by modulating oscillator pitch and tremolo by modulating amplitude
vibrato: periodic variations in pitch, typically at a rate of around 7Hz
envelope generator (EG): a circuit that simulates the audible progression of an acoustic sound
—which has a beginning, middle, and end—by generating a control voltage or digital signal whose level changes over time according to specific parameters; a common ADSR EG provides time variables called attack, decay, and release and a sustain level between the decay and release stages
linear frequency modulation: a type of audio synthesis in which changes in the amount of
modulation don’t affect the center frequency of the carrier oscillator, which remains constant
The opposite, additive synthesis, actually existed earlier. Although the Hammond tonewheel organ—epitomized by the popular B-3—and its pipe-organ ancestors are beyond the scope of this book, they actually use additive methods. Traditional additive synthesis uses multiple sine waves, which produce only fundamental frequencies and no overtones. All of the sine waves sound at
various harmonic frequencies, and you combine them to generate complex timbres. Depending on the additive synth, there are different techniques to modify the frequency and amplitude of every partial generator as sounds play. In the Digital Keyboards Synergy, for example, each partial generator is paired with a multistage envelope generator that you can loop and synchronize with other envelope generators. As potent as additive synthesis can be, it’s challenging to program and thus hasn’t been as popular as subtractive synthesis—as illustrated by the relative rarity of additive synths on the market.
Synthesists in search of new and unique timbres eagerly awaited digital synthesis during the 1970s and early ’80s, but the technology was very expensive at first. Then in 1983, Yamaha unleashed the DX7 and changed the whole synthesis ballgame. Using the linear frequency-modulation (FM) synthesis technique John Chowning developed at Stanford University in 1973, the company offered the DX7 at a reasonable list price of $1,995. As implemented in the DX7, linear FM uses sine waves
—somewhat like additive synthesis but in a totally different configuration. Inside the DX7 are six operators, each consisting of a digital sine-wave oscillator, an envelope generator, an amplifier, pitch and modulation inputs, and an output. Modulating a carrier operator with a signal generated by a modulator operator results in frequency modulation and the generation of harmonics a sine wave doesn’t produce on its own. Thirty-two algorithms—arrangements of modulator and carrier operators
—allow the generation of complex timbres without the need for all of the oscillators required in an additive synthesizer. Additive synthesis and linear FM do a more credible job of synthesizing
acoustic-sounding timbres than analog synthesis, but their clean, precise, and edgy quality eventually left many synthesists craving the warmth of analog.
Digital sampling and polyphonic sample playback arrived with the original Fairlight CMI at the end of the 1970s, but few musicians could afford its price: up to $36,000. But the technology—
digitally recording acoustic instruments, playing them back at different pitches and editing them, and composing music by sequencing their playback—proved enchanting. As other sampler manufacturers got involved, bit resolution expanded from eight to twelve and then sixteen bits and beyond to
improve the sound quality. Other improvements evolved and instrument prices dropped to ranges that were truly affordable, making sampling dominant. It remains widespread today and, although
considerable progress has made sample playback more expressive, there’s no way to avoid the truth about sampling: it’s essentially the audio equivalent of a snapshot. With filtering and processing you can vary sonic qualities of the sample, but the sample waveform itself is fixed.
aperiodic: irregular and nonrepeating
Somewhat related to sampling is the microsound practice of granular synthesis. Theorized by Dennis Gabor in the 1940s and introduced as a synthesis technique by Iannis Xenakis in the late ’50s, granular synthesis involves very brief snippets—1 to 100 milliseconds (ms) in length—of sound.
Each snippet, or grain, comprises an aperiodic waveform that you contour using differently shaped amplitude envelopes. You arrange grains in sequences to build sound objects, but the grains are so brief it takes thousands of them to create the equivalent of a note. Granular synthesis is rare because it takes a lot of user effort, computer-processing firepower, and storage space to execute. Among the instruments capable of doing it are Symbolic Sound’s Kyma and computer applications including Csound, SuperCollider, Cycling ’74 Max/MSP, and Native Instruments Reaktor. The Malström softsynth in Propellerhead Reason does what’s called Graintable Synthesis, a combination of granular and wavetable synthesis.
With wavetable synthesis, oscillators scan through digitized waveforms stored in ROM.
Wavetables can be samples or generated using mathematical equations, and they’re typically much more complex than average analog waveforms. Early wavetable synths include the Rocky Mount Instruments Keyboard Computer and the PPG Wave 2.2, introduced in 1974 and 1982, respectively.
One dynamic and expressive technique Wave designer Wolfgang Palm developed is the ability to sweep through digital wavetables using envelope generators. The linear-arithmetic (LA) synthesis technique Roland engineers developed for the D-50, released in 1987, combined sample playback, wavetable synthesis, and built-in effects, but it was a joke originally played by one of the Japanese engineers that resulted in some of the D-50’s most memorable sounds. The engineer programmed a digital sequencer that would rhythmically cycle through waveforms in an uncontrolled but hilarious-sounding manner. The D-50 programmers Eric Persing and Adrian Scott enjoyed the results so much that they convinced Roland to make the capability a permanent part of the synth’s factory soundbank.
Kurzweil later incorporated a similar technique in a more controllable fashion in their K2000 and K2500 series of synths.
Malström, Propellerhead’s Graintable Synthesizer, was the second virtual synth introduced to Reason. Thanks to its two flexible modulator sections, Malström patches can evolve and move almost as if an arpeggiator or sequencer were involved. (Courtesy of Propellerhead Software AB)
A second-generation Rocky Mount Instruments Keyboard Computer, officially known as the RMI KC-II, rests on its chrome stand above multiple foot controllers. The wavetable-synthesis soundset loaded at the factory into the original Keyboard Computer—mostly borrowed from Allen’s digital theater organ in what Clark Ferguson, who began working for RMI at about the time of the KC’s introduction, describes as “a shotgun wedding”—didn’t appeal to the customers RMI was after. Although users could load alternative wavetables into the first KC from punched cards via an optical reader, the company later developed the KC-II with new sounds and an extended feature set. (Photo courtesy of Allen Organ Company, LLC)
A few years earlier, Chris Meyer at Sequential Circuits developed vector synthesis, which allows dynamic cross-fading among signals coming from four oscillators using a dedicated envelope and a joystick. Sequential introduced the technique in their Prophet-VS, which could generate sounds Meyer describes as “struck attacks that faded into shimmering flutes, clarinets that opened up into raw sawtooth waves, and much more.” Sadly, the Prophet-VS—which entered the market in 1986—
didn’t sell well, and its production ceased when Yamaha purchased Sequential Circuits in 1988.
Vector synthesis later appeared in Yamaha’s SY22 and TG33 and Korg’s Wavestation, all developed under the direction of Sequential’s founder Dave Smith. Vector synthesis, however, wasn’t the only magic that went into the Wavestation. Korg’s engineers also improved on the wave-cycling
techniques that appeared earlier on Roland and Kurzweil synths to develop a spinoff technique called wave sequencing, which allows the user to select, arrange, and loop through brief sampled
waveforms to create complex rhythmic patterns.
Vector synthesis went with Dave Smith post Sequential Circuits first to Yamaha and later to Korg, represented here by the powerful and popular Wavestation. (Courtesy of Korg)
During the mid-1990s, the next big thing in synthesis became physical modeling (PM), which involves the use of mathematical equations to represent physical laws and actions. When applied to sound synthesis, PM algorithms allow the simulation of acoustic musical instruments. You can use a wave equation to model a vibrating violin string, including the tension it’s under, the density of its linear mass, and its air displacement. To model the entire violin, you also need wave equations for its body, bridge, catgut strings, bow, and the like. The computation of these details is CPU intensive.
Julius O. Smith of Stanford University’s Center for Computer Research in Music and Acoustics
(CCRMA) devised the digital waveguide filter, which significantly decreases the number of functions described in the wave equation and reduces the PM computation load required from computers. The technique has appeared in numerous synths, including the Yamaha VL1, E-mu Morpheus, and Korg Kronos. However, creating great sounds using physical-modeling synthesis isn’t by any means easy.
While all of the details in the wave equation models are ripe for expressive control in performance, they actually demand such control. Otherwise the synthesized sound will suffer—as will listeners. If you decide to go down the physical-modeling synthesis path, be prepared to incorporate as many different controllers as you can and practice, practice, practice.
During all of the initial PM excitement, many musicians lamented the virtual disappearance in new synths of good analog tone and an interface blessed with front-panel knobs as opposed to a few
buttons and a deep menu system. For years manufacturers tried to make sample-playback sound analog, to no avail. It was Hans Nordelius of Clavia in Sweden who initially found the right digital and hardware ingredients for many synthesists. He delivered the first of his distinctive crimson-red synths, the original Nord Lead, in 1995. Not only does it sport more knobs than we’d seen on a new synth in over a decade, but it also generates a convincingly analog sound thanks to the digital signal processing (DSP) synthesis method Nordelius and his team developed.
Introduced at NAMM in January 2011, the Kronos continues Korg’s tradition as a prominent manufacturer of state-of-the-art synth workstations. This is the seventy-three-key model. (Courtesy of Korg)