Theo Verelst Synthesize Laboratory page

After having mastered or understood at least some of the principles from my Wave Laboratory page The next step is to gain and use understanding about sound synthesis techniques and some of the theory that underlies the practical applications of the principles in synthesizers and well used sound software, to arrive at musical or soundeffect results.

Musical foundations

Since this page seeks to make musical sounds by applying mathematical and technical knowledge, it is imperative to understand what music is actually made of, before we go into the how to do this with synthesizers.

We all know about orchestral instruments, violins, wind instruments, the piano, and percussive instruments, and can discern these kinds easily enough. Somehow we acquire knowledge about these types of instruments as a child, and our ears can understand and discriminate between a string section and a drum set being played.

Mainly we learn to do so by example and repetition, which is not to hard for these types of disctinctions, and many will have no problem discerning between quite some instruments, even of the same kind, many can in seconds hear the difference between a grand piano, a home upright or a detuned bar barrel, and we can discerns a piccolo froma clarinet and a trumpet quite easily, just as an accordion and a acoustical guitar.

Also an electrical guitar is recogniseable, although I remember I wasn't aware of the idea of instance funky percussive guitar sounds listening to various pop music, tough I would have known the difference between a classical and lets say country and western guitar, which are soundwise maybe not that far apart.

There are various generally recogniseable criteria for telling instruments apart. Lets have a look at them.

The first is their pitch, a bass sound is very different from a flute, to start with because they are tunes and produce tones of different frequencies.

Timbre is one of the most important ones, it makes one instruments inherent tone quality different from others. A piano can be dull, bright, dark, detuned, toy-ish, electrical, strong, fragile, even for the same note being played, lets say a middle 'C'.

The envelope of a sound is the graph of how the sound comes up and dies away or quits. For instance, the envelope of an flute is a pretty steep start, maybe a little peak in sound volume, and a constant sustained note, and when it ends the end is almost abrubt. A saxophone starts a note a bit slower in general, and a horn even slower. A piano tone slowly dies away, while a harp tome dies away quite quickly. A piano note can be damped to stop almost immedeately, or continue pretty long. An electronic organ is a non physical instrument in in basis generates notes with almost completely instantaneous 'attack' portion, completely constant 'sustain' portion, and immedeately falls away when the key is 'release'-d. A percussive element can be brought in that 'decay's in a little while before the sustained volume is reached.

The character of a tone being played is made of these components, and of course can be complex to describe, though are ears are capable of dealing with amazing complexity in this sense. In a orchestra, many of the instruments may sound together, yet can clearly be told apart by many listeners.

The sound basics that cover the sounds of an orchestra are not just the course considerations made above, but also cover the ideas of instruments sounding together, their distance and position, and the reverberation characteristics of the concert hall.

Finally, the idea of the music comes in the picture, working our way up: what do the instruments play. Individually, what melodies, and together: what chords and harmonic progressions. And how is that held together rythmically, and what does expression add to that picture, and timbre changes, and subtle effects such as bending a note a little flat or sharp.

And finally, what does a performance aim for, and how does it deal with musical language the audience may have associations with.

Instrument basics

Before looking at more complicated problems and producer kind of considerations, we'll go into the very many possibilities of recognizing intrument properties, and mimicing them on a synthesizer. This is very usefull for getting an idea of various instruments, their characteristics and uses, and various ways a synthesizer can be put to use, and its limitations.

Also, before in a next chapter looking at more advanced instrument issues, the specific sounds of synthesizers will receive attention, since they have a place of their own in contemporary music, and various sounds they can do are easily recognized.

Introduction: instruments with clear pitch

A musical instrument of any kind is made to produce a tone of some characteristic kind, usually can be played in a certain tune, and is sensitive to the players interaction with is somehow.

First, I will assume we deal with intruments that are tonal, that is we could hum or whistle along with what they play, because they form notes with a certain pitch.

Taking a flute as an example, the combination of the fingers on the holes in the flute determine which frequency the basic tone of the flute has. For a picollo, that is a high frequency, say a few thousands of cycles per second or Herz, for a contrabass, that frequency may be 50 Herz, and the lowest pipe organ bass pipes may produce thundering frequencies with a fundamental as low as 16 herz. Such notes may be 'felt' because 16 vibrations per second at sufficient volume can be picked up by normal nerves as well as the ear.

We've seen that sounds can be characterized in various ways, lets see what these examples are like. Evidently, the picollo has an waveform with a high frequency, and also the timbre is not to hard to unserstand, it is a fairly simple type of note, in fact it is almost sinusodial as we've seen a principle wave form. Also the envelope of the piccolo or another flute is not so hard to understand, the note comes up a but gradually, probably has a little peak when the blowing starts or the lips are opened, is more or less constant for the rest, and simply stops when the player stops pumping air in the mouthpiece and bore.

Is that all? No, in fact certain aspects of such instruments, just ike most others, are hard to capture exactly, thogh they can be described good enough. For instance the air produces noise in the sound that a bit louder when the blowing starts. Often, the note being produced by a skilled flute player shows little variations or even intended modulations in pitch and volume to make the sound more insteresting, and when playing together with others, little tuning differences will be compensated for.

Lets first stick to the basics and see if we can make a flute sound out of sort of mathematical basic building blocks. In fact, the infrastructure, of 'signal path' of many synthesizers can be explained by using this example to start with.

Basic sound and envelopes

As we've seen on the wave laboratory page, the fundamental waveform for frequency analysis is the sine wave, and every other signal can be though of as consisting of a combination of a number of sine waves, ech with their own frequency, amplitude and phase. Normally, we like to start with signal sources, called oscilators, that repeat a certain wave pattern all the time, with a frequency determined by the keyboard of the synthesizer. Though this is not what most natural instruments are like, it makes it easier when an oscilator has a fixed, nown waveform that simply is reapeted, and where we can get used to soundwise. The basic classic synthesis waveforms have infininitely many harmonics in principles, though they become weaker for the higher frequencies, and have strong and memorisable sounds: the sawtooth, the squarewave and the triangle. When a squarewave has an on off ratio of not 50 %, it is called a pulse, which is a reedy type of sound also recogniseable.

Usually, synthesizers have at least two generators, which may have different waveforms, and are mixed together. The frequency and detune knobs determine their relative frequencies, in course and fine tuning manner.

Timbre changes

Modulation and sound combinations

Bringing the parts together

The various parts of the sound as they are formed by the generators, the envelopes, the filter and modulation sections together become the audible note when a key is pressed. They are all dynamic, that is they change all the time, the waveforms from the oscilators are continuously being produced according to the selected type of wave, the envelope will make the sound come up and ddie away, maybe the filter is stable for certain portions of a sound, and the modulators are intended to change the sound.

All these waves are running in the circuitry, together making a note of certain character, and it is up to the ear to discern the result as just these elements in some combination, or as a pleasant or well formed sound because they are made to work well together. How this works is often complicated though explicable. A few examples clarify why the experience needed to make this coming together work right is probably one of the main chalenges in various senses of sound and music producing.

As an example, consider a bass note. The waveform may be a sawtooth to start with, and the envelope may start with a fast attack for not softly played bass. When a key is pressed on the keyboard, the sawtooth will start to run at the pitch going with the key, lets say 40 Herz, making a wave with a cycle duration of .025 seconds or 25 milli seconds. One thing to realize is that that wave may start at the moment the key is pressed at a fixed phase, for instance starting to rise in voltage at the lowest point, or maybe halfway the ramp. For another synth programmer, it may be that the wave continues where it was when the previous key was released, and for yet another, the wave may have continued to cycle in the meanwhile.

When the filter is open, and high frequencies may pass though easily, the opening of the Voltage Controlled Amplifier by the attack portion of the envelope generator when done quicly, may almost instantaneously put the generators output on the filter, and to the speakers, causing the speakers' cone to 'jump' when the wave wasn't at zero when that happens.

To prevent this, the attack control may be put down a little bit, to make a gradual change happen.

Now suppose we use a triangle instead, and the attack is opened fast: the same effect would make the attack portion of the envelope make a clear click, that for instance for a filter with a high resonance value make the resonance happen more than to be expected from the triangles' quite dull spectrum. This could be on purpose, but also the reason for unsatisfactory sounds.

When two detuned oscilators are mixed together, they produce wave patterns with knots and throughs are produced, sometimes taking seconds to pass for a small detune value. The timing of this natural modulation can be combined with other modulation sources and the envelope for the sound for very pleasant sounds, but evidently on need to somehow make the bows and thoughs happen when they're wanted, and make sure that when a fast attack portion is desired, the detuned oscilators may be in counterphase the moment a key is pressed, making no resulting noise at all to start with, no matter what the envelope generator tells the VCA.

When the filter is set pick a certain part from the spectrum by having a noticable resonance, the same principle holds for the interference of every harmonic in the spectum of the detuned oscilators in that part of the spectrum.

Even though sound synthesis this way is comparatively crude, the number of variations of these types of effects is almost infinite. The amount of detail the ear picks up from a sound is so much that complicated, high quality sounds require attention for many, many of such effects on a row and combined, which is why the idea of physical modeling as synthesis method is an interesting alternative to simply sampling and/or analysing such complex sounds, that are not easily produced with standard synthesizer algorithms and signal processing building blocks.

Beware though that their use is absolutely not limited to non-complex instruments. Very amost anchient synthesizer records exist to prove that with modular systems and lots of programminfg skills and diligence (and tape recorders) complicated orchestra registrations can be made which are probably even more pleasant than the real thing at times, see for instance the Walter Carlos 'switched on Bach' and Tomita's 'Pictures at an exhibition' lp's.

Playing the sound

Especially when using basic instrument simulations, the impression the sound makes on the listener is impacted considerably by the way it is actually played. Suppose we make a flute sound, if we then play the bass end of the keyboard with a boogy-woogy bass line melody, not many will recognize the idea of our programming efforts to make the piccolo.

Once and during the sound is programmed on the synthesizer or in the software, the programmer uses the results obtained until that point as feedback in the sound design process. For the piccolo, he might have certain melodic line in mind, which should sound good in certain key, and the sound can be made with that idea in mind.

Making a characteristic flute melody work is a way to focus on a good imitation of this acoustical instrument. While trying the sound on the keyboard, the programmer thinks about improvements and appropriate additions to the sound.

Skills in this process may even be used live, although that is not easy, considering the number of parameters to deal with for even simple instruments. Mastering this type of programming will enable a player to make more variation and more lively playing on stage possible, and use the expressive and sonic range of synthesizer more effectively, both as solo or effect instrument or as versatile backing and supportive instrument.

Why bother about synthesis

Below, I'll on various places go into the answer to the same question: why not just get a general midi instrument on maybe a PC soundcard and play a desired instrument from a large collection of available sounds? Piano 1, piano bright, etc, take a pick and play away!

There are two major reasons not to. First, those instruments may not be much to your liking, and not fit the musical piece you want to make. Second, it is not half as fun as seeing a sound being formed by you own efforts, or at least improved because you play a synthesizer and not a sound producing machine that only has fixed sounds. The general midi instruments are fun enough in themselves, they have variation in instrumens enough to play around with, and some sound decent enough to listen to, though often limited.

The idea of a synthesizer being used instead of an instrument that cannot change its its sounds, makes it possible to try different sounds for a certain piece of music or sound track or effect, which is often rewarding, and for serious productions usually inevitable, and makes a good production possible.

Synthesis kinds

Traditionally, two types of sound synthesis are distinquished: additive and subtractive. In additive instrument, principal waves are added to form a sound, like the registers on an organ, subtractive synthesis such as in many synthesizers starts with a sound source rich with harmonics, and filters away the unwanted ones.

The two possibilities can be combined as is often done in modern instruments, leading to a variety of sound generator algorithms.

A modern and fashionable method to reproduce sounds is sample replay. A very early version is the mellotron, which has little tape loops for every key, starting playback when that key is pressed, containing a recording of an instrument. Modern samplers contain lots of computer memory as tape, and can do quite some signal processing on the recordings, too.

Additionally, there are kinds of sound generators that are a bit harder to capture in terms of the spectrum they generate and how they do so, such as fm synthesis, which deserves attention.

Finally, there are algorithms to generate complicated sounds by doing a form of computer simulation which cannot be seen as an oscilator, but which simulates the operation of a physical instrument, and as such makes sounds all by itself that need not much more treatment by the rest of the familiar synthesizer modules, such as the envelope generator. Generate complex waveforms, like a sampler can reproduce, too, but make those waves themselves, with the possibility to change parameters to have different sounds for different keys or striking force, unlike samples.

Traditional analog subtractive synthesis

As we've seen on the wave laboratory page the traditional wave-forms are sawtooth, square wave, triangle, pulse, and sine. The sine wave is the basis of harmonic analysis: every other sound wave can be expressed as a sum of sine waves, so that the sine waves indicate which frequencies together form the spectrum of that wave. If we assume that the wave repeats itself all the time, than the lowest sine wave in the spectrum, which could be as long as the lenght of the wave, seen as a little graph with the time horizontal and the amplitude vertical. The second harmonic, an octave higher, is the second sine, with a certain amplitude, usually smaller than the fundamental, possibily zero, etc. The sine waves could be shifted to the left or right, which means changing the initial phase, and than a list of all harmonics with their relative phase and amplitude together define a certain waveform, this can be proven mathematically.

The point of such an excercise is that the ear does not so much detect waveforms as frequencies, and that the frequency graph on the wave laboratory page, which is a frequency analysis like I described, except the phase numbers aren't visible, which isn't so bad, gives a better idea of what a certain wave sounds like than the signal form.

One wave is hard to hear, a few are needed to make the little hairs in the inner ear vibrate like little tuning forks, many thousands of them covering the whole spectrum, to indicate to the nerves that they picked up vibrations of a certain frequency. So it makes sense to repeat a waveform and take that idea as an 'oscilator' making a certain type of sound.

Of course many natural sounds change in character with the time they resound, but considering that the frequencies of musical instruments are such that normally hundreds or thousands of wave pieces pass by per second, those changes can be brought about slower, while the oscilator keeps repeating the wave it is instructed to generate.

The subtrative idea is that a filter, which is nearly always present in traditional and also many modern synthesizers, is used to filter of a certain part of the spectrum, like the treble control on an amplifier, except stronger, and with the possibility to change the frequency where the filtering starts.

The point in the frequency spectrum where the filter starts reducing the frequency components is called the cutoff frequency, for instance when the cutoff frequency is set to 440 herz, and an A2 of 440 is played and generated by the oscilator, the fundamental sine component of 440 herz is passed almost unchanged, but the second harmonic would already be attenuated (weaker) by a factor of 16, or to about 6 % . In audio language, decibels, after Bell, the fellow phone inventor, are used to indicate sound levels, normally from lets say 20 dB (very quite surroundings) to 120 dB (jet sound or so, pain threshold), and every dB is a factor of two.

Most analog, familiar filters have a filter characteristic of 24dB per octave, that means that for every doubling of the frequency above the cutoff, doubling the frequency means the filter cuts off another 24 dB or factor of 16.

So starting from the available waveforms, the spectrum can be limited by the filter. Another feature of most filters is that they can be made to resonate, which is like making the part of the spectrum around the cutoff frequency amplified, even to the point where the filter itself becomes a generator of a sine wave at that frequency.

As a another type of 'oscilator', a noise generator unit may be available, which generates random signals, sounding like the noise an FM or TV receiver makes when no station is picked up. Such a signal has a more or less even distribution of frequencies over the spectrum, and a 'pink' noise generator has a spectrum compensated for the lack of lower frequencies in a white noise pattern, hearingwise, sounding warmer. Noise can be used as the basis of drum instruments and effects, or to add breathing for instance to instrument imitations.

To make the volume of a sound build up and cease again, an envelope generator (EG) and voltage controlled amplifier (VCA) are used. The envelope generator makes a curve composed of an attact portion, which rises with a certain rate, a decay portion, which falls to the sustain level with a certain rate, and a release portion, which is the rate at which the sound falls away after a key is released.

A voltage controlled amplifier is like a voltage controlled filter, except that instead of using an input voltage to change the cutoff frequency, it uses an input voltage to change its amplification. It is like having an amplifier with a special input turning the volume control.

When the envelope generator generates an envelope voltage, starting with the attack portion at 0 volt, it feeds it to the VCA to control the amount of signal being fed through at each point in time. The audio input of the vca is connected to the output of the oscilator.

So if we'd want to make a sound that starts slowly, we'd adjust the attack control to a low value, making the envelope generator for the sound produce a slowly rising voltage for the VCA, as it were slowly opening the volume control.

The oscilators for tradional synthesizers, which are completely fashionable after 20 years or so, for good reasons, are adjusted to the right waveform: sawtooth for lots of every integer harmonics (1,2,3,4, etc., strong synthesizer like sound with lots of high frequencies), square for all odd harmonics (typical 'hollow' sound), or triange for not so many odd harmonics (duller sound, bit like a clarinet).

It is prefered to have more than one oscilator driving the voltage controlled amplifier with its waves, because that makes for richer and more varied sounds. Often two audio oscilators are present, with their outputs mmixed together. Both can be adjusted to produce a certain waveform, and the second can be changed to a certain relative frequency to the first, with a course control for lets say a number of tones from an octave, and a fine tune knob.

Fine tuning is important because when the oscilators have a little different frequency, their waves start interacting, which makes for a smooth, changing, richer sound, because the result has knots and throughs in the combination wave when the phases of the signals slowly combine from in phase (signal adds up to times two), to counterphase (signals cancel eachother), see the wave laboratory page for examples.

The oscillators have octave or course frequency controls to change their use from bass over mid range to piccolo sound basis. Some oscilators have pulse with control, which can make a square wave assymetrical, up to very small pulses, which is good for reedy sounds.

The filter usually has its own envelope generator, to dynamically change its cutoff frequency. So if we'd want a sound that starts bright and becomes duller a bit and than stays the same, we adjust the attack to maximum, the decay to a nice value, lets say 1/2 second, and the sustain to the desired constant brightness value, and make the cutoff frequency dependent on this envelope with the appropriate button.

The final basic module in the traditional analog synthesizer is another type of waveform generator, the low frequency oscilator (LFO). It generates the same waveforms an audio generator can do, except at low frequencies, lets say from a cycle per 10 seconds to a few tens or hundred cycles per second, and it may have a real sine wave as output as well, which is desirable.

Such an LFO can be used for modulation purposes, for instance when the control input of the VCA is given a sine wave of a few cycles per second, a tremolo, or amplitude modulation of a sound results. By feeding an oscilator frequency control input a little bit of LFO output, a frequency vibration occurs, such as vibrato on an electronic organ, or lets say like a guitarist bending a note up and down quickly for a changing note. Also the filter cutoff frequency can be given an amount of low frequency oscilator output, to change the brightness of a sound periodically. In general, such modulation makes a sound much more alive and varied, when using a sine wave for small enough frequency and amplitude modulation, the most natural vibrato and tremolo effects results.

A triangle works too, but not so pleasing, it gives more synth type of modulation feel, a square wave of course just jumps between two voltages, which is excellent for police horns or so, and a sawtooth with falling ramp can be used as a repeating envelope pattern, up, slow down, up, etc, for all kinds of effects. Just like the audio oscilators, the LFO may be started from the same point in the wave for every keypress, or simply continue to run.

Basic additive synthesis with sine waves

Sample reproduction

Since computer memories have become available with significant storage capacity, samping has become quite a popular branch of sport in the synthesizer or electronic/digital instrument world.

A sample is a signal stored in a computer that contains like a little tape a recording that can be played back by pressing a key on the keyboard. Also, such a sample can be played back at varying rates, to change its pitch (and length, of course), for instance when an adjacent key on the keyboard is played.

How does this work? In short, first an instrument is recorded in the sampler. Lets say we have a saxophone, and we'd like to use the sound of it in a sampler, we'd fetch a microphone, plug it in the sampler input, tell the saxophone player to get ready, set the sampler up with clear memory to record a sample of certain quality and lenght, tell the player to play a note, lets say a middle a (could be a e flat just the same), and at the same time press the sampler record button. After the sampling time is over, the memory of the sampler would have a recording of the note played on the saxophone.

Assuming the volume and 'take' were ok, the sample can be called back by pressing a note on the samplers keyboard, and probably the a key would play back exactly what is recorded, which on a modern sampler would be up to CD quality, and when using to microphones even stereo, or on an older one starting from maybe a AM radio quality.

The main point of using such a sampler is that now the note can be played also with a C key on the keyboard, and than the pitch of the sample is changed to reflect the change in frequency between a and C. In fact we can now play the saxophone on the sampler keyboard, which is interesting enough, though limited. The main limitations are that a saxophone does not just change pitch for variousnotes on a scale, they also have a little different timbre all the time, a high note is not a low note slower, like playing a 78 rpm record at 33 rpm, that doesn't work. Second, the sampler simply repeats the same sample every time a key is pressed, while every time a player plays the real thing, little variations are audible, and intentional phrasing is present.

Both limitations are serious, and a major reason even relatively old and old fashioned structure synths are in demand even though samplers sell well.

The experiment is also clear recording the human voice: change the pitch on a record only a bit, lets say half an octave, and immedeately it is unrecogniseable, sort of like Mickey Mouse.

For various instruments, piano, reed, brass, to some extend guitar, drums and percussion, the sampling idea of using a single sample (recording) for a reasonable key range on the sampler keyboard works good enough to use. At least when a sample is played, the audience immedeately will recognize the instrument, and a lot more nuance is in the natural sound than a synthesizer will usually be able to generate. Just a pity the nuanec is always the same on playback, but for short lines and with some more work a sampler can be quite a usefull and versatile instrument, with almost unlimted potential sound palette: as long as an original is available uncluttered, it can be sampled.

For new sounds, a sampler cannot be used, but there are very many samples available of may kinds, commercially and in club circuits. A sampler usually has a disc drive, allowing samples to be put on disc and called into the samplers' memory from disc. And there are libraries, for instance on audio CD's or in computer form with thousands of sounds to chose from, which is good when making a soundtrack, record or other production, or to play around with for fun.

The idea of a multisample is to use more than one recording of for instance the saxophone. We could record the sound of it for every note on a scale, in the same volume and playing style, which can sound quite convincing on playback. Crossfading between samples is to make two or more samples for a note, and jumping or gradually going from one sample to the other, depending on how hard a key is played on the keyboard. A loud recording sounds when the key is hit hard, a recording of a gently played sax when a key is played softly.

Making such multisamples, possible also gradually fading from on sample into the next, and crossfade samples is very non-trivial. It is a lot of work to make the boundaries of one sample to the next on the keyboard of the sampler not clearly audible and annoying. It is even harder to make a gradual mix from one sample to another work right. It is easy to imagine that mixing a soft and loud sax sound together is not the same as the sound of a normally played one.

The waveforms of the two samples are probably not of the same frequency even, and certainly the wave forms or harmonics spectrums are quite different and can not just be averaged.

There are many factory preset sample instruments on the market which contain samples which are made to do these things right, or at least reasonable, for a variety of instruments both for natural and synthesizer sounds. Such sample playback machines can give very realistic instrument reproductions of acceptable instrument quality, and the more memory, the more variation. Some have slots for more samples on memory cards, to expand the sound set, others have a limited set of high quality samples that are intended to use as they are, without variation.

Many modern digital piano's are such intruments they contain a number of good multisample sets to reproduce a few kinds of piano sounds, and as such become an instument on their own. Allways these instruments have an element of boredom in their sounds, because the samples are always the same, while a piano has many variations in it sounds, and coupling between strings, for instance. However, with many multisamples, and some additional techniques to bring in harmonic variation, such instruments can do quite convincing piano sounds, without having to make and load in samples yourself.

A synthesizer type of sampler, intended to do your own samples also has a envelope generator and VCA in it (often in practice a Digitally Controlled Amplifier), to change the volume contour of a sample. Aften a filter with its own envelope generator is also present, for the same purpose as in an analog synthesizer, to dynamically change the spectrum of the sample.

Care should be taken that such filters engineering wise do not always receive the same attention as their original analog counterparts, so that their effect may not be as good or strong. Otherwise it would in principle be possible to start with a few samples of the basic waveforms and produce the same sounds as with an analog synth, also because lfo's and multiple sample generators are not uncommon. In practice, this may well not work satisfactory for at least a few objective reasons: the digital audio signal path often has limitations, the control signals may not be accurate enough, and not change fast enough, and the idea of sampling requires care to be taken when a digital sound in the end is converted to an actual signal, as we've seen in the wave laboratory page, such conversion has various pitfalls to work around.

Programming a sampler is usually not with a lot of knobs on the front pannel, both more like a computer synthesizer.

FM synthesis

Hybrid instruments

Other forms of synthesis

PM, 2 kinds

Phrase synthesis


Most modern synthesizers, countrary to their older counterparts, offer a number of built in effects to enhance sounds with. Without any additional effects, the strength of the synthesizers hart becomes clear, but even when the basic synthesis is strong, effects are very often desired to enrich the sound.

When not playing in a big enough, reverberating place, a touch of reverb does a great job to make playing more pleasurable, also with headphones. Completely dry sounds can be annoying and without the natural feel of ambiance quite unnatural and not effective when used with modulation such as for instance vibrato.

Effects are for normal productions and various playing conditions in general to be used sparingly first, as a rule of thunb turn the reverb control a bit down rather than up, unless of course the effect is needed as a major sound effect.

We'll look at the major kinds of effects, of which many variations and combinations exist, but the categories are quite well established and understandable.

There are modulating effects, such as an electronic organs' lesley unit, which includes phasing, chorus, wha-wha like filtering effects and accurate amplitude and frequency modulation, and reverberation type of effects such as a hall simulation, a simple delay line for echo effect, and all kinds of in betweens.


Signal conditioning

The main signal conditioning takes place either as a time depending volume control, or as band filtering. The latter is known in consumer audio equipment as tone controls or equalizer units.

An equalizer offers sliders to suppress or amplify a certain part of the audio spectrum, usually split up in one, half or even 1/3 octave pieces. Every piece of the spectrum may have its own control with maybe + and - 12 or 18 dB of cut or boost for the center frequency of that piece of the spectrum. Every 6 dB means the sound volume doubles, so often the equalizer controls may be opened or closed just a bit, for instance to kill certain resonances, like 50/60 Hz hum, or a microphone feedback frequency.

For such purposes, instead of a graphic equalizer with all the controls on a row for quick graphic overview, a parametric equalizer may be used, where the spectrum is split is a few bands of which the centre frequency can be adjusted.

An important producing, microphone and power amp and tape recording control tool is called a compressor or a limiter. It mainly changes the volume of a sound, and basically it turns its own volume know when it detects a higher volume, usually down when the sounds gets louder, hence the name: compression.

Of course this works electronically, accurately and very fast, so the effect when used right does not have the 'pumping' type of sound results of simply turning a volume control, but the idea is the same: when the sound is loud, it is made softer, when it is soft, it is made louder. Thus the dynamic range of the material fed through a compressor is reduced, which is desired when for instance recording a signal, not to make the tape saturate (meters in the red), though in that case just like for power amplification a limiter may be in place, that only turns back the volume over a specified threshold, and then quite fast.

A compressor has a response characteristic that usually can be adjusted with a envelope generator resembling a part of a synthesizers' unit for the volume shaping the signal: an attack control determines how rapid the compressor will turn back the volume when the signal passing through it gets louder, a decay control sets the rate at which the volume is turned up again when the signal gets softer.

A knee point may be adjustable to determine at which signal strenght the compressor starts to work. A soft knee means that at first it will not compress as much as with higher volumes, a hard knee makes compression work almost straight away when the threshold volume is reached. The amount of compression is expressed in decibel per decibel, which is the ratio of a dynamic change per volume difference.

A limiter is a compressor with a hard knee, fast attack, and quite a compression depth, though practically, more electronical parameters maybe at stake.

Because of the compression of the dynamic range, the signal to noise ratio of the signal is negatively influenced by the compressio operation. Simply put: because in soft passages the volume is turned up, all background noises such a shum and noise are more audible in the end result. To prevent this, two approaches are possible. When the compression effect itself is wanted, it is possible to detect when no input signal is present, and than shut the compressors Voltage Controlled Amplifier off, so that the input noise is stopped from being maximally amplified and annoying in the mix. The effect of shutting all signal under a certain threshold off is called a noise gate. It, too, can have attack, decay (hold) and threshold parameters.

To prevent the adverse effect on the dynamic range of compression or limiting for instance on a recorder tape, it is possible to reverse the effect of the compression on playback, which is a well know principle. The operation is called expansion, and when the tape in between responds well, and the parameters are mirrored accurately, the original signal may be obtained back after the use of compression, tape record, tape playback and expansion. Professional units for dolby X work this way, and well know dolby B and C work the same way, but than for only higher frequencies, leaving the lower ones uncompressed and expanded. Dolby A does companding (the name for the combination) for analog studio recording equipment in 4 frequency bands seperated.

It is important to realize that though companding may be a good idea for some more effects or signal processing, and definately for sampling, that any volume change in between the compressor and the expander will have its effect multiplied at the output. So a delay line maybe be companded, but a chorus unit would not sound the same anymore: the boughs and knots of the phasing effects would influence the expanding, and sound different, though of course that could be tried intentionally..

Combinations of effects are also common in studios, such as as 'exiter' units. It is a form of harmonic enrichening which for a certain band of frequencies adds harmonics by noise, distortion and or (ring) modulation effects, and possible some more specific effects know to certain manufacturers.

Signal distortion

This is historically the domain of electrical guitar players, though a modular synth can also be abused to arrive at similar effects. Mainly the idea is overdriving an electronical circuit so that the signal doesn't pass through unchanged but is distorted because the circuit can't handle it properly anymore, in its basic form by limiting the amplitude of the signal.

When this is done like many transistor amplifier circuits do by nature, the signal is 'hard clipped'. Clipping means the portion above a certain signal level is simply clipped of, and hard is when that happens very unsmooth, which sounds harsh and rarely pleasant.

Fuzz and scream boxes for guitars contain the same idea, but with some way of making the signal limiting work more gradually, and with filtering effects. A guitar amplifier, preferably a radio valve based one has the property of rounding signals off when overdriven, adding mainly second harmonics instead of the thirds from the 'square waving' hard clipping ones, which in general is considered pleasing. Also, such an instrument amplifier is not at all straight in terms of frequency spectrum and all kinds of non-linear and dynamic effects inherent in the circuits.

Thats why they can a lot to the character and playability of the electrical guitar, being mainly intended for such instruments. A synthesizer will generally not sound good when played over them, that instruments needs the frequency range to be good, and solid, distortion free amplification. A bass amplifier may do. But for solo lines, and special effects, it may well be worth connecting a synth up with fuzz boxes, guitar amplifiers, and everything else making distorted noises.

The mix between the direct instrument signal and the distorted signal may be used to add only an amount of distortion to a signal, apart from controlling how much the distorting signal is distorted. To make the amount of distortion similar for soft and hard notes, a compressor may be used ahead of the distortion unit.

Because of the signal limiting, a distortion unit itself also achieves signal compression, and therefore a dynamic range reduction, causing noise and distortion from the input to become more audible. A noise gate may be used after the unit to make the signal path quiet in a mix when there is no signal.


A delay line is just that: the signal goes into a box, and comes out a little later, maybe 1/10 of a second, maybe 2 seconds, depending on the purpose. If we mix the delayed signal together with the original signal and say something in a microphone or play a note, the effect is that of a single echo.

Early delay units were tape based, that is a tape is looped, and a seperate recording head puts the signal on it, while a playback head picks up the signal from the tape a little later, while the eraser head makes the tape clear before the same piece of tape passes the recorder head again.

Such mechanical units wear out, and have fairly poor signal quality, but were well known, and could be made into more complicated echo machines by having more than one playback head in the tape path.

Electronics came to aid later on to make analog delay lines with chips to electronically sample and delay a signal. No wear, but quality still not very good would be the result.

Only with the use of digital memories and computer circuits quality delay lines for longer times could be made. In fact it is like a sampler: analog to digital conversion, storing of signal in memory, and playing back the signal though a digital to analog converter. A knob to control the delay time enables setting of how far the echo is away, and usually there is a 'feedback' knob, which feeds an amount of the echo signal back to the input, so a repeating, fading echo can be made.

The feedback is tricky, and a possible source of problems, because every time the signal is recorded, it is distorted a bit, and the background noise accumulates. Suppose certain frequencies are amplified more or suppressed by the delay unit, then every time a signal passes though it, those frequencies are affected the same way, making them even stronger or weaker, so the sound becomes different. This may be intentional, but usually is something to be aware of.

Ensemble effects: chorus, lesly, phasing

Many effects are based on the idea of having a delay line with variable delay time, modulated by a low frequency oscilator. When a signal is delayed and mixed with the original, filtering occurs because certain frequencies will be cancelled out and others emphasized by the two signals combined, because of the phase change trough the delay.

When the delay time is varied, the effect is that the rate at which a signal goes in and comes out of the delay line is a bit different, so the momenteous frequency of the signal is a bit different. Suppose the delay time is 0.12 second when a wave is put in the delay line, and 0.14 second when it comes out again aobut .13 second later, than the effective frequency when it comes out is .12/.14 x 100 % is a few percent lower, because the wave is made longer by slowing down the virtual tape, in comparison with the recording time.

Both effects are not very easy, but the result can be thought of as a combination of two times the same instrument, with a little different pitch, mixed together. As can be understood, such an ensemble sounds richer, when mixed in stereo wider, and more vibrant than just the input signal.

A chorus is made to beef up, thicken the sound, for instance to make up for the lack of a second oscilator in a synthesizer, or simply to make a smooth 'moving' sound. It uses a lfo with sine or (asymeetrical) triangle wave, and on or a few delay lines modulated with it or more than one oscilator, each for each delay line.

On an electronic organ, chorus is the slow motion of the lesley drum or speaker or horn, which makes a nice, slowly varying motion in the sound, to add a little vibration to the static organ notes. The echos and interfering pipes of a pipe organ can make similar things happen, thats a reason for the effect.

Ensemble effects do something similar, but aimed at stronger effect, and with less periodic modulation, even a noise signal could be used to influence the LFO frequency and delay variation randomly. Pushing the modulation depth and the basic delay length until just before the point where dissonance would occur, the impression of more instruments playing at the same time is enlarged.

The phasing effect comes from the time where magnetic tapes were used for recordings, where two tapes with the same track would be played back, with a slight timing difference. When the tapes run exactly synchronous, the material sounds normal, but when by slowing down one reel a bit with a finger a slight difference in timing starts to appear, the phasing affect starts to appear. By using a short delay time, changing slowly, and maybe adding filtering to the signal, and mixing with the original in 50 % / 50 % ratio, the interfearence like with tape phasing can be imitaded, creating a sort of whirling sound.

Lesley is in fact a complex effect, though the idea can be approximated by a delay line of maybe 20 mS (depending on the frequency spectrum present in the signal to be modulated) with a sine wave of about 6 herz modulating it to about +/- 50 cents (percent of half a chromatic note distance) at target frequencies.

A mechanical lesley box contains a rotating drum in front of a speaker, with an opening on a part of the circumference where the sound comes from, or in some versions a rotating speaker. As a result the sound comes from varying direction, with varying sound path and path length, filtering because of the direction effect, and doppler effect because of the changing distance of the opening to the listener. For an electronical organ an almost indispenseble strong and pleasing effect, and quite hard to adequayely model electronically and even digitally. Many effects units that offer the effect do a fair job but don't get close to the real thing, simply because of the complexity of modeling the effect right.


An important effect for general listening comfort and pleasant ambiance is the reverb unit. Many different types and qualities exist, though there are clearly distinguished categories. The main idea is to simulate the effect of a sound echoing back and forth through a space, for instance a concert hall. By snapping ones' fingers in a reverberating space, it is possible to hear quite some distinct and blurred reflections from a single snap or handclap. When an instrument sounds in a space, the first path to the listener is usually the direct path of the line of sight, which is the shortest.

Then after maybe a tenth of a second the first reflection can be heard, for instance from a wall or the ceiling. After that, a lot of reflections will pass by the listeners' ears, coming from all directions, slowly or quickly fading away, depending on the reflectiveness and size of the hall or room. A few reflections probably will be dominant, and their repetition times audible. The decay of the reflections is a power sequence, basically like an exponential pattern. The way the reflections change depends on the diffuseness of the surfaces, which determines how a sound wave impinging on it scatters. A brick wall is different than varnished wood or plaster.

Many, many different types of concert halls exist, and they all could act as example for a reverb unit. Basically, the reflection paths are modeled in the unit by observing the impulse response of the space: the finger snap echos can be recorded, and used to make delay lines with with the same spacing between the delay times.

A more thorough analysis may include the main paths and model them with a freedom of certain parameters for each main reflection path, and may take into account that the responses of the space are different for different frequencies. The parameters for each type of reverberation offered could be decay time (defined as the time in second until the reverb is damped 60 dB or a factor of a 1/1000 ), amount of spread of the sound or 'fuzziness' the smoothness of the reflections, and the frequency dependence of the reflections.

There are various types of reverb, mainly the simulation of actual spaces, varying from concert hall or church over live stage to rooms, the simulation of well known reverb units, for instance based on springs (older organs and amplifier systems), or metal plates or echo cellar, and special types of effects, not based on existing or known places or examples.

A relatively simple example of a reverb is called a tapped delay, which is a number of delay lines on a row, with adjustable delay time and output mix ratio. When these are given feedback, reverberation effect start to occur, and when the feedbacks are giveng more complicated filtering and the structure is made more complicated, wonderfull and pleasantly full reverb effects can be achieved. In principle, a frequency band spit up inpulse response implementation oculd suffice to make a good impression of a reverb, but the complexity of implementing the effect of such a impulse response accurately enough is considerable, and the disadvantage is that it may not be possible to change certain parameters of the reverb to suit specific needs.

It is also possible to use the idea of reverb on a smaller time scale to do enclosure simulations, like the cabinet of an organ, a guitar amplifier, an instrument even. Also in this case, the impulse response could be used to implement an imitation of the vibrations and standing waves in such enclosures. When used right, such effects are seemingly subtle, but can add significant body and warmth, resonance and realism to for instance synthesizer sounds imitating the instrument with such an enclosure.

Special effects

In the earlier days of mechanical, analog and digital effects, it has been productionwise fashionable to use a well known variation of reverb, called gated reverb. It is born out of the idea of wanting a very thick and rich reverb, lets say a church type of effect, while at the same time not hearing all snare drum beats from half a minute ago still resounding.

The reverb unit is fitted with a noise gate that shuts off the signal when it gets softer than lets say medium volume, so that when a snare frum for instance is hit, massive church reverb is added, and as soon as it dies away about, is suppressed. A strong effect, which can be heard in various variations on numerous records.

Digital reverbs often contain a version of the same idea which is called reversed gated reverb, which makes the reverb effect built up instead of decay.

A complicated effect is the pitch shifter, that generates one or more extra notes based on the signal on the input at some frequency interval, for instance an octave. To do so, it usually must sample the signal, and in digital processing make sample with higher or lower frequency, possibly by a fourier transform, possibly by making sample loops. The effect is that when one note is played, a chord sounds, or a few players seem to be playing at the same time.

The controls on a real life synthesizer

Over the last few decades, the synthesizer has not become on standard instrument at all, there are quite distinct branches on the electronical and digital instrument tree, but there are various recogniseable and quite general facilties most share.

The keyboard

A keyboard for a synthesizer obviously commonly has the same black and white, usually plastic, sometimes wooden, keys that a piano has. To make the electronics respond, such keys have contacts in them, that respond by making a signal for the electronics when a key is pressed. In fact, many keyboard have two contacts for each key, one opening when the key is pressed just a bit, the other when the key is almost down in rest position.

The computer circuitry in the synth measures the time between the opening of the first and the closing of the second contact to determine how fast a key was pressed, when the instrument has such facility. Many older ones don't, but most modern synth do, which is a major improvement.

Many keyboards additionally sense the pressure of the fingers on the keys, and generate a signal which can be used for modulation purposes indicating how hard the keys are pressed down, for istance for adding vibrato to notes when pressing harder.

Some, usually expensive, keyboards even have this facility per key, meaning that every note can receive distinct amounts of modulation. The latter assumes that the synthesizer is polyphonic, which is at present common, but wasn't at all for the 'original' analog synth designs. It requires that the electronics to produce a tone are to be repeated for each possible simultaneous note, costing a lot of circuitry.

Modern synths often have digital signal processing computer circuits in them that allow them to juggle the signal path simulation of 16, 64 and even 128 oscilators, vca's, filters, etc simulataneously. The electronical machines often have advantages over them still, in the departments of accuracy, signal strenght and interest and subtle effect, livelyness, possibly speed of keyboard response and know response, and that often various signals are available for and from the outside world to use the electronics for various tasks. And they go out of tune, and noise, and hum too, of course. Since a keyboard is continuously and intensively used it pays to watch for good quality and playability. Flimpsey, sloppy, weak, or regularly failing (for instance because of dusty or worn out contacts) keyboards are continuous cause of annoyance, and cannot easily be replaced.

Many modern synths also are offered as 'rack' modules, that is without a keyboard, and are then played over the MIDI connection by seperate keyboard. This has one major disadvantage: the response of the midi line and protocol, which is like a computer network connection for synthesizers, may be on the slow side compared to a built in keyboard of a good synth. For experienced players, and certain purposes, and when multi channel pressure data is desired, which causes a lot of data to flow over the midi network connection, such delay may be undesirable. Just like a keyboard that breaks under the conditions of enthousiastic live playing.

Modulation and expression controls

Since long, the standard modulation controls for synthesizers are wheels, usually one for pitch bending a note, and another for activating some modulation source, such as vibrato.

Placed on the left side of the instrument, the pitch bend wheel lets a player bent a note over for instance a few semitones or a fifth, or an octave by turning the upright, vertical wheel up or down. A spring makes the wheel snap back to normal tuning when released, like a tremolo arm on an electrical guitar.

The modulation wheel can be set to control the amount of lfo signal routed to various destinations such as the frequency of the oscilators, the filter cutoff frequency, or the VCA control input, and usually stays in position when released, and can be turned down to make the sound not modulated.

Various alternatives exist, such as a joystick, using simulaneous X and Y motion for two functions, a ribbon, which can be pressed at a certain point to control some kind of modulation or control voltage, a light controller, where a hand covers a light source which is electronically detected, and the simple but effective and accurate idea of the volume or modulation pedal, like the throtle control of a car.

Like on a piano, one or more pedals may be available for isntance to act as sustain pedal, or to advance one sound in the memory for each pedal push in performance.

Programming the darned machine

Electronical synths are like older and many modern radios and stereo systems: each control simply has a knob, button or slider. Therefore the whole front panel usually is crowded with controls, which looks high tech, and is practical because every control can easily be foundin a fixed place and quickly.

Modern synths are based on digital computer circuits, where parameters are held in computer type of memories and even discs and flash cards, without the need for a know for each function. Like a computer, menus are used, and few controls like a mouse or up/down control, while parameter values can be seen in a display window ny names and numbers.

That idea is not bad, but in practice it takes more effort to get good oversight, ease of operation and skill to be able to adjust every parameter in a short time.

To not lose the idea of controlling certain parameters while playing, a set of knobs of sliders may be available that can be programmed to correspond to a certain parameter, for instance cutoff frequency.

A computer can be used to get a better overview of all sound parameters on its screen, and maybe to make programming a sound easier by mouse and keyboard control. The disadvantage is that the plug in, switch on and play and edit idea of a synth is lost: the computer must be brought along (could be a portable), started up, the program started up, communication between the synth established, etc.

For a studio this may work after some getting used to, but still many prefer to have a front panel with direct control knobs for ease of programming. When sounds have many parameters, compromises must be found.

No need for lonely synths: MIDI

The Musical Instrument Digital Interface was made to connect synthesizers, computers, drum computers, and effect equipment together over a standardized connection. It's sort of a synthesizer network standard. It consists of a din plug, two shielded wire interface with In, Out, and Thru connectors for every instrument or machine. The in is repeated at the thru for connecting more than one instrument up in a chain, and the first is friven by a midi controller keyboard or a computer.

The idea is that the interface carries messages for every event happening on a keyboard or controller, split up in 16 channels for different listeners.

General considerations

Sound levels

Sample rate conversion