Definition
One of the meanings of the word synthesis is making a whole out of parts. Creating some new whole by combining distinct components.
When applied to artificial sound generation, it is about creating new sounds or processing existing sounds to create new sounds.
In nature, a combination of various factors (or components) will create a certain sound (or timbre). For example, the human speech system: the combination of the vocal cords, throat, tongue, teeth, mouth and lips to produce sounds: a biological synthesiser.
The same principles are applied to electronics synthesis.
Synthesiser Structure
A synthesiser can basically be composed of two parts:
sound engine: the synthesiser itself, where the components are connected and adjusted in order to produce a range of sounds. It is composed of a sound generator and modifiers. The combination of these components can be fixed (most common) or modular (most flexible).
control interface: used to trig and adjust parameters of the sound generator. The most common type of controller is the MIDI keyboard. Other examples are drum pads, breath controllers, MIDI guitars, etc.
Most Common Synthesis Types
There are several different types of synthesis, each with their advantages and characteristic sounds. Among them we could cite:
Subtractive Synthesis
Additive Synthesis
Frequency Modulation (FM) Synthesis
Wave Shaping
Ring Modulation
Linear Arithmetic Synthesis
Vector Synthesis
Wave Sequencing
Wavetable Synthesis
Physical Modelling
Granular Synthesis
Subtractive Synthesis
This is just the classical method of synthesis used in most analogue synths and in most sample playback synths and samplers.
Subtractive synthesis means that you take a sound (preferably a spectral rich one like a sawtooth or a square/pulse wave, or a sample of a grand piano) and route it trough a filter and amplifier to change its timbre. This way you reduce the level of some partials of the original spectrum and hence the term.
The filter and amplifier section can use modulators to have a variation over time, producing sounds with a more lively feel. Modulation is using a time-varying controller signal to change one or more characteristics of another signal.
Frequency Modulation (FM, AFM)
Frequency Modulation is usually abbreviated FM or AFM (for Advanced Frequency Modulation). This is the family of synthesis methods that brought a breakthrough for commercial digital instruments in the eighties. Basically it means that you control the frequency of an audio oscillator by the frequency of another audio oscillator.
The interesting aspect sound-wise is that you can generate a very wide variety of spectra plus many transient sound characteristics with FM (and not only the never ending variations of electric pianos and bells).
Additive Synthesis, Fourier Synthesis
Any sound - how complex it may be - can be decried as a mixture of a number of sine wave components with different phases and amplitudes. These are the partials of a sound, which are also called harmonics if their frequencies are an integer multiple of the fundamental frequency.
The method to generate a complex sound spectrum as the sum of (many) simple sine waves is called Fourier synthesis - after Joseph Fourier who found its mathematical basis. The more general term additive synthesis can also be used if the waveforms added are not sine waves.
Ideally, a lot of sine oscillators are needed for Fourier synthesis. How many depends on the required range and brightness: a bright bass note - think of a slap bass - may need more than hundred, while a high pitched harmonic sound will probably need only a dozen.
Wave Shaping
Wave shaping refers to a sound manipulation (not generation) technique which applies a (non-linear) function on the original signal (i.e. the output of an oscillator).
This scheme is similar in principle to analogue distortion in a guitar amp or fuzz unit, but offers much more sound variation possibilities including resonance-like effects. Wave shaping can be used as an advanced synthesis method in a way similar to FM.
Ring Modulation, Amplitude Modulation
Ring modulation and amplitude modulation are not complete synthesis methods, but rather processing techniques that are quite common on advanced analogue and digital synths. Sometimes these features are wrongly named or used when the actual implementation is quite different.
Manufacturer specific terminology for similar schemes includes the terms cross modulation and FXM (frequency cross modulation).
Ring modulation is the multiplication of two signals. The output of a ring modulator will contain the sum and the difference of all available input frequency pairs.
Amplitude modulation is the multiplication of two signals, where one signal is always positive.
This buzzword was used by Roland to describe their approach to digital sound synthesis in the eighties. It is based on the observation that the attack transient of a sound is its most important part with respect to human perception.
Vector Synthesis
The first synth to implement this paradigm was the SCI Prophet VS. The VS can mix four oscillators with different waveforms in realtime via a joystick controller and a multistage envelope. While this is a really simple concept, it is effective for expressive play and nice evolving sounds.
The Korg Wavestations and the Yamaha SY22, TG33 and SY35 are other "vectorized" synths. Yamaha keyboards can mix up to two FM and two sampled elements, while the Wavestations mix up to four sample based wave sequencing oscillators.
In principle most synths can do realtime vector synthesis, when fed with MIDI joystick data to crossfade oscillators. If you like to try that you can rewire a PC game joystick to fit your synths pedal jacks.
Wave Sequencing
This term means that a sequence of different sample segments can be used to generate a sound. Korg implemented this on their famous Wavestation synths. The Wavestations oscillators can sequence through programmable patterns of samples.
Each of the patterns consists of a number of individually tuneable sample snippets and each sample in the sequence is assigned its own level and duration. Typical for the Wavestation (and rather easy to program) are "rhythmic" wave sequences in which an oscillator steps through a number of samples in a predefined periodic rhythm.
The Wavestations also combine this with vector synthesis capabilities.
Wavetable Synthesis
This term is used for two completely different things:
Many sound card companies call their RAM based sample playback capabilities like this (because the samples are stored in a table in RAM).
For the PPG Wave and the Waldorf Microwave and Wave synths this term is used to describe the ability to produce a sound by sequencing through a table of different waveforms during the duration of a single note. For the wavetables and waves there is a preset ROM area as well as a user loadable RAM area provided.
Which entry of the wavetable is selected may be controlled by an envelope, LFO or any other modulation source in realtime. Also these synths can interpolate between subsequent waveforms in the wavetable thus smoothing the timbral change.
The waveforms are single cycle ones, so realistic acoustic emulation is out of reach for this technique, but the vastly improved modulation capabilities - compared to sample playback - more than make up for this.
Physical Modelling Synthesis
Physical modelling (PM) is a whole class of synthesis methods that do not synthesise sound based on an abstract mathematical description, like the Fourier transform for additive synthesis or by classical signal processing means like filtering for subtractive synthesis, but rather tries to model the diverse instruments themselves: e.g. the bow, string and resonance corpus of a cello or the plucking finger, string and body for an acoustical guitar.
There are many different physical modelling algorithms including relatively simple ones like Karplus-Strong synthesis and rather complex ones like the waveguide approach which uses multiple delay lines to model strings or air columns.
The biggest advantage of physical modelling is the realtime control it offers. While other synthesis methods offer some algorithm specific and rather arbitrary control parameters like filter cut-off or modulation index, physical modelling enables the use of control parameters that are more musical and have a more complex influence on the timbre and phrasing.
Another advantage of PM is that the sound generation is context sensitive: a note on a clarinet model will sound different if it is played with legato binding to its predecessor or with a little pause in between.
The dependency is much more complex than with the traditional synthesiser portamento or glide function. Another example: the pitch bend of a clarinet patch will not just linearly shift the frequency of the note, but the synth will respond in a similar way to a real clarinet, i.e. it will shift the frequency and timbre for a while but then jump to the octave.
PM has its disadvantages though: Instrument models have to be designed with great care and a lot of knowledge of both instrument acoustics and the necessary math.
On the available PM instruments by Korg and Yamaha, editing is only possible via macro parameters of the otherwise hard-coded instrument models - probably the only practicable way to provide sound programming to the "normal" user.
Granular Synthesis
Granular Synthesis means sequencing through very many very short sound (sample) snippets. The difference to wave sequencing is that the single samples are played for such a short time, that the sequencing is heard more as a timbre than as a rhythm.
Comments