All About Additive (Part 2)
In the final part of a look into the world of additive synthesis, Chris Meyer explains the different approaches to additive synthesis adopted by various manufacturers and which synths have employed them.
We conclude our series on additive synthesis and resynthesis by defining some of the terms involved and surveying a selection of "additive" machines.
DEFINING EXACTLY WHAT an additive machine is, can be tricky. Since "additive" itself is a buzzword these days, certain manufacturers are anxious that you believe that theirs is an additive machine. In truth, there are different levels of additive synthesis. It can be defined as any synthesis method where harmonic spectra are added together to create a new spectrum - essentially, anything with more than one oscillator qualifies. And we're no better off defining it as a system where the user has control over the harmonic spectrum - again, almost any form of synthesis falls into this category.
Instead, let's define three basic classes of additive machine, and accept that the machines will vary within these classes in terms of their power of additive synthesis. The one common denominator among these machines is that they allow you to build the timbre or harmonic spectrum of the final sound by specifying the harmonics and their relative strengths. They vary in how many harmonics you can control, how much control you have over them (pitch and amplitude envelopes along with frequency ratios and detuning), whether you can control the harmonic content in real time (via velocity and so on) and whether you can start with a sound sample and derive a synthesised sound from it.
We'll call the first class wavetable synthesisers. These essentially give the user the ability to build initial waveforms by picking the relative strengths of the harmonics in that wave. A subclass of these is advanced wavetable machines which allow some form of fading between these waves, resulting in predictable harmonic movements.
The second class is additive samplers. These allow you to define the harmonic content and movement (relative strengths of harmonics, their detunings and so on) of the final sound. The result is computed into a sample which is then played back by the sampler. You have no further control over the harmonic content or movement of the sound (aside from processing like subtractive filtering if it's present). Software that creates samples also qualifies (we'll refer to this as additive sampling software).
The third and purest class contains machines that give you control over the harmonic content and movement during the creation of the sound and in real time while the sound is playing back. These are true additive synthesisers - no qualifying remarks necessary. Most commonly, these synthesisers have one oscillator to generate each harmonic, so they can each be manipulated in real time.
In all the above classes, you create the sound from scratch. Resynthesisers start with a real sound and break it down into a set of parameters that one of the above class of machines can regenerate. True resynthesisers (as opposed to common or garden samplers), give you some ability to alter the captured sound - this means either being a class three machine, or allowing you to alter the harmonic content of the wavetable or contour of the sample after the breaking down has taken place.
Below is a summary of some of these machines (and their relatives), in rough chronological order of release:
- Crumar GDS and Synergy
These were perhaps the first commercially available true additive synthesisers. They actually let the user define pitch and amplitude envelopes of sounds. The 32-partial oscillators were set up in a row - each one could either feed the output or frequency modulate (FM) the next oscillator in the row ("FM'ing" adds complexity to the sound). The GDS was the big near-prototype beast (used early on by the likes of Klaus Schulze); the Synergy being the production version. Later in life, the Synergy was available with a Kaypro computer and editing software.
The engineers who developed the Synergy have formed MuLogix in the States that make an updated, rackmounted version of the Synergy called the Slave 32. Unfortunately, these are not widely available.
- Fairlight CMI
Renowned for their sampling, the Fairlights also boast additive synthesis - the Series II in particular attempted to marry sampling and additive synthesis as closely as possible. You were supposed to select a sample rate based on the pitch of the sound so that one wave would fit into 128 samples. Then the Fairlight would make a Fourier analysis of each waveform to determine its harmonic spectrum. You could then alter this harmonic spectrum, or create one from scratch by drawing a wave-to-wave amplitude envelope for each harmonic (up to 64) of the sound. The sound was then "computer crunched" into a sample (or back into one). This made it an early additive sample machine, and also, since you could start with a sample and edit its spectrum, a resynthesis sampler.
- Yamaha DX Series
Believe it or not, Yamaha's FM synthesisers can be used as true additive synthesisers. Of course, they're limited in this regard - in a six-operator synth (DX1, 5, 7, 7II, TX7, 816, 802), you have control over the harmonic tuning of six sine waves and their amplitude envelopes (four-operator machines give you four). Use the upper algorithms that have most or all the operators wired directly to the output (the "organ algorithms", as they're sometimes called). Restrictions do exist - for example, you have far fewer partials than normally desired and only one pitch envelope - but they're enough to get you started.
- PPG Wave and Waveterm
The PPG Wave was one of the first wavetable synthesisers. A number of waveforms were set end to end, and envelopes, LFOs, velocity, pressure and so on, determined which waveform was played at any given moment. The waves could have smooth progressions from one shape to another, or sudden jumps. The Waveterm, which was a sampling and editing terminal for the Wave, allowed you to create waveforms (using additive synthesis) and design progressions. This qualifies it as an advanced additive wavetable synthesiser.
- Kawai K3
This was a wavetable synthesiser that allowed the user to create two waveforms using additive techniques (defining harmonics and their relative strengths), thus making it an additive wavetable synthesiser.
- Kurzweil K150FS
A lot of confusion surrounded this piece of equipment when it was first released. Many thought it was a rackmounted K250 - an impression its early advertising and manual did little to dispel - but it sounded disappointing when compared to the realism of the sample-based 250. In actuality, this is one of the more impressive true additive synthesisers around and Kurzweil invested a lot of effort in its ability to model natural sounds as closely as possible. Finally, the 150 incorporated the letters "FS" in its title which stood for Fourier Synthesiser (after Fourier, the man who did the pioneering work on additive analysis). A sound modelling program also comes with the machine.
The 150 has 240 oscillators that can each produce a sine wave or noise which it dynamically assigns to voices as needed (with a maximum of 64 oscillators per voice). Harmonics can be tagged as being of less importance, and these get stolen first if another voice is needed. Harmonics may be tuned, and may have up to 2400-point amplitude envelopes with some very interesting looping features. Several functions may be edited from the front panel, such as MIDI assignments and chorusing. The sound modelling program allows the user to create sounds. This is a powerful machine hampered by a difficult user interface and initial confusion over exactly what it was.
- Korg DSS1
Korg's first sampler combined a lot of synthesiser-type features with its sampling functions. One of these was intended to encourage people to use it as a wavetable synthesiser - allowing creation of waveforms by drawing them or selecting harmonics. Thus, it was an additive wavetable instrument.
- Sequential Prophet VS
A wavetable synthesiser, this. The VS's voice structure contains four oscillators per voice which can be mixed in real time (with envelopes, aftertouch, a joystick and so on). You can create new waveforms by mixing four existing waveforms (including sine waves) together. Because you can create waveforms (and later dynamically mix them), harmonic progressions could be set up, making this an advanced additive wavetable synthesiser. Each wave can be treated as four "groups" of harmonics that can be faded against each other - this over-generalises the harmonic progression of a real sound a bit, but is still useful. Waveforms may also be sent to the VS using the MIDI Sample Dump Standard.
- Wersi MK1 and EX20
The Wersi is a little-known but powerful wavetable machine from a German manufacturer known primarily for their advanced organs. Reflecting that, it is a cross between organ ideas (you'll find Leslie, key-click, and distortion effects) and some very advanced synthesiser concepts (zones, complex envelopes along with multitimbral operation). It also qualifies as an additive wavetable synthesiser since you can create waveshapes by selecting and mixing harmonics (up to 32) using an organ "drawbar" analogy. There are 16 sliders on the front panel akin to a graphic equaliser (a comfortable, intuitive system) that may be moved in real time with immediate results.
- Digidesign Softsynth
One of the first commercial synthesis programs, Softsynth is additive sample software that runs on either the Apple Macintosh or Atari ST. It allows you to set the frequency ratio, amplitude envelope, and pitch envelope of up to 32 harmonics in two different fashions (harmonic graph and time slice modes) with a good graphic interface. The partials may have sine, triangle, square, or noise waveshapes. The resulting sound may be previewed on the computer's speaker and crunched into a sample for downloading into virtually any currently available sampling instrument.
- Casio FZ1
Like Korg's DSS1, this is a sampler that encourages you to turn it into a wavetable synthesiser. It has a section where you can define the harmonic content of a wave, thus qualifying it as an additive wavetable machine when used in this mode.
- Kawai K5
The voice structure of Kawai's additive machine is related to that of the Prophet VS, in that it has four oscillators per voice. But where the VS mixed the four waveforms through what is in essence a quad panner, each oscillator has its own envelope, modulation and so on. (You can define the waves produced by these oscillators by determining the relative strengths of up to 63 harmonics. Kawai present this system as 63 harmonics that can be fed through one of four envelopes.) This ends up being a simplification of a full additive machine - as opposed to having to construct an envelope for each harmonic, there are only four to deal with. No, this doesn't give the "full" power of additive (and this is still technically an advanced additive wavetable machine); but sometimes all that power can be too laborious to deal with. Like the VS, the K5 may also receive waveforms as samples over MIDI (using a Kawai SysEx message, not a MIDI Sample Dump), making it possible to take snapshots of a sample, modify them, and transfer them to the Kawai - a crude form of resynthesis.
- E-mu Systems Emax SE
E-mu has just freshened up their popular sampling keyboard with a software update that includes a "spectrum interpolation synthesis" module. In essence, this is an additive synthesis software package that resides inside the Emax. Sounds may have up to 24 harmonics, and are divided into 24 time slices (which unfortunately seem to have a fixed position in time). The amplitude and pitch of envelope of each harmonic may be created by either editing its status per time slice or by "drawing" its profile with the data entry slider while the Emax clicks off time. The transition from time slice to time slice may be smooth or abrupt.
The harmonic content of the sound at each time slice is called a "spectrum", and these may be copied to other spectrum locations. Frequency ratios of the partials may also be defined (although they must be integers at the slice if you intend to copy it to another location - as described last month, having something like 3/2 sine waves per wavelength ends up a little distorted). You may also draw the amplitude of each harmonic at a given time slice. The result may then be crunched into a sample that the Emax will play back, which qualifies it as an additive sample machine. Having controls over all the parameters that additive synthesis requires is difficult for any interface; the SE does a commendable job in making it relatively easy.
- Technos Acxel
Whereas the Fairlight is a sampler that borrowed additive techniques (and eventually resolved things into samples), the Technos is a true additive synthesiser/resynthesiser that uses samples as the starting point for its sounds. It has a large number of oscillators that can have any of 32 different preset waveshapes (including sine), or you can draw your own. These may have their frequency ratio set, and multistage (up to 1024 point) envelopes for pitch and amplitude may be constructed. The number of harmonics per voice is adjustable from eight to 128 (32 is the default). Lots of processing is available besides simple additive construction, such as digital filtering, FM, modulation and so on. Since all these parameters are being executed in real time, full real-time control over the sound is available. An additional highlight of the Acxel is The Grapher - a visual interface that you can touch and trace out new curves on.
- Lyre FDSS Studio/FDSoft
One of the Frankfurt conversation pieces, the Lyre is a true additive synthesiser boasting a large number of individual oscillators. The base unit has 128 of these, which can be spread across one to eight voices (the Lyre may be expanded out to 1024 harmonics and 64 voices). Each oscillator produces a sine wave that may be noise-modulated to produce hazier frequency "bands" centred around the harmonic's pitch (like pitched noise, for breaths and scrapes). Multistage (up to 128 segment) envelopes control pitch and amplitude, and the harmonics can be detuned to any desired ratio. Editing and control software runs on either an IBM PC or Apple Macintosh. A stand-alone version of the sound modelling software - FDSoft - is an additive synthesis software package that performs many of the same functions, and produces a resultant sound that can be downloaded to sampling keyboards for playback. Samples for resynthesis may be input over MIDI (Sample Dump Standard) and crunched into a form that FDSoft understands, thus the Lyre and FDSoft/sampler combination is a resynthesiser.
ADDITIVE SYNTHESIS IS ultimately very powerful, but that power only comes with many, many parameters to manage. Implementing all this power is a difficult and expensive proposition in terms of both hardware and software. Consequently, a lot of manufacturers are going to shout "additive" and "resynthesis" when, in fact, they're only calling on part of the power of additive synthesis. But remember, that pigeon-holing a machine doesn't diminish its musical usefulness.