the development of synth electronics
Do synths make sense to you? One Two presents a keyboard special with a difference. Not the normal heap of reviews but some writings on where synths stand today — how they were born, who makes the best noises, the success of the science and what the funny words mean.
What has a three octave touch-sensitive mini keyboard, Low Frequency Oscillator, sawtooth and pulse waveforms plus sub-octaves, AD envelope generator with LFO triggered reiteration and 15 voice presets?
Not, as you might think, last year's offering from Casio, but a valve-design mono-synth called a Univox, circa sometime-before-most-of-us-were-born. This device was to be found clipped beneath a piano (an ancient percussive device comprising 6000 parts largely fashioned from wood) for the dual function of providing timbral interest and warming the knees.
This article is concerned with the history of synthesiser electronics, and the conclusion is that fundamental design objectives do not change, but advances in technology allow their realisation with reduced cost and size. In other words, if we are prepared to wait, there may come a touch-sensitive, programmable, 16 voice polyphonic, MIDI equipped synthesiser weighing less than one kilogram and costing ninety nine pounds ninety five. Of course by that time we'll all be drawing our pensions, but at least the MIDI might be debugged.
History has a habit of going back a long way, so where do we start? The first synthesiser is credited to Thaddeus Cahil (1906) who refused to be put off by the fact that no one had invented the amplifier. So he built a set of turbine generators to create musical waveforms at current so great that direct connection to the public phone network was possible. A similar principle but in miniature was used by Laurens Hammond in 1929 — the first electric organ. Mechanical generators are with us to this day buried within Wurlitzer, Rhodes and Hohner 'pianos', but let's get on to the electronics.
Oscillators are the tone source of all synthesisers and owe their existence to the invention of the triode valve in 1906 by Lee de Forest. An oscillator is simply an amplifier with positive feedback via a frequency selective path. Imagine an oscillator as a PA amplifier "howling" because a mike is too close to a speaker. The frequency selective path in the analogy is the characteristics of the room combined with speaker resonances. In an electronic oscillator the frequency is selected by a feedback path through capacitors or inductors which offer an impedance which varies with frequency. This reactive path therefore introduces a time constant whose value alters the frequency of operation.
This simple scheme for an oscillator is OK for generating sinewaves of fixed pitch, but for musical use a wide range is needed and an integrator/comparator circuit is used. A simple time constant is incorporated by placing a resistor in the charging path of a capacitor — the smaller the resistor the faster the capacitor 'fills up'. If the waveform on the capacitor is observed, an exponential (logarithmic) rise is noticed which is great for ADSR type envelope generators (the same principle is used) but undesirable for a voice oscillator.
The solution is to feed the capacitor from a constant current source which ignores the capacitor's initial hunger and dwindling appetite thereafter. By this technique a linear ramp is generated which forms the back of a sawtooth waveform. The remaining oscillator circuitry has to sense when the capacitor is 'full' and arrange for its discharge, which may be instantaneous (to give a sawtooth) or a mirror image of its upward slope (to give a triangle).
Before transistors were available the niceties described above would have appeared to be pipe dreams and the first oscillators used simple gas filled tubes. A capacitor was strapped across a neon lamp and a resistor was put in the lead connecting the lamp to a high voltage supply. The voltage on the capacitor would increase to the point where the inert gas in the lamp would strike and so discharge the capacitor. The cycle would then repeat to give a rather curved sawtooth which was very rich in harmonics. In semiconductor form the relaxation oscillator, as it was called, found its way into Moog's original circuitry of 1964.
With the invention of the oscillator and amplifier, synthesisers sprang out of the woodwork with names like Theremin (1920), Ondes Martenot and Trautonium (1928). Each of these pioneering instruments specialised in gliding tones — in fact it was fixed pitches that they had trouble with. The avant-garde composers of the day had been liberated from semitone quantisation but were champing at the bit for someone to invent the tape recorder (let alone the personal multitracker).
1950 saw their medium in the shops and they duly went wild recording everything in hearing range and generally got electronic music the bad name it is still trying to recover from.
The transistor may have been invented in 1948 but it was no competition for the valve which at that time was near the peak of its development. Silicon transistors had to wait until 1952 and the first integrated circuits appeared in the early 1960s. And then in the mid-sixties, bang! Dr R. A. Moog's papers on the theory of voltage control.
Scarcely had Moog's designs for the world's first musical VCO and VCA appeared than Walter Carlos was roaring up the charts with 'Switched-On Bach'. The electronics were fairly standard analogue computer circuitry and the modules were well established electronic music building blocks, but the big difference was in their control.
Instead of the various parameters of sound generation and control being altered by mechanical devices like potentiometers, they were now under voltage control. Because the only difference between a signal voltage and a control voltage is their relative rate of change, inputs and outputs could be freely connected to allow modulation and other dynamic effects.
Moog's VCO was based on a unijunction relaxation oscillator which produced a sawtooth waveform. Subsequent circuitry produced triangular and pulse outputs by modifiction of the sawtooth. The heart of the oscillator was current controlled and the sum of the voltages presented to the oscillator were converted to a current with an exponential transfer function. This gave the colloquially termed 'log VCO'.
The conversion utilised the natural characteristic of the voltage/current relationship of a semiconductor diode, and matched chains of selected devices were driven from the input voltage summer.
The 'log' function is very important for musical applications. Firstly, the ear's response to frequency is such that octaves require an exact doubling of pitch — an oscillator which will double its frequency for each additional equal increment of input voltage will produce an equally tempered scale when incremented by one twelfth fractions of the octave voltage.
In other words, take a keyboard wired to generate an output voltage which rises by 83.3 millivolts (one volt divided by twelve) per semitone. When connected to an oscillator with a one volt per octave v/f characteristic, it will produce a musical scale. Not only that but the musical intervals will be correct regardless of any offsets like a master tune which may displace the range high or low. Transposition is just a matter of turning a dial.
The other advantage of a log VCO is that frequency modulation such as vibrato appears even throughout the frequency range — not excessive in the bass and barely noticeable in the treble. Both of the above benefits are not to be enjoyed on the alternative linear oscillator which has only the advantages of simplicity and inherent stability.
Discrete components were used in Moog's circuitry (no integrated circuits) and although it was a classic design it suffered from poor long-term stability and a narrow working range. The short term stability was claimed to be 0.1% over a five octave range, and by manually switching in different value capacitors a fair span of frequencies could be generated.
The problems with log VCOs are temperature/bulk resistance effects of the logging components and errors introduced by the finite time required to discharge the oscillator timing capacitor. The temperature effects are due to electrons within the semiconductor material acquiring enough thermal energy to get into the conduction band and so change the nature of the substance. This results in chronic drift which is unacceptable.
If uncorrected the control current will double for a 10 degree centigrade rise. There are two ways around the problem: one is to balance the effect with another which is both equal and opposite; the alternative is to thermostat all the sensitive components in an "oven" well above room temperature. The first solution is the most widely used and requires no calibration. A 'magic' resistor with a selected temperature co-efficient ( +3,300 ppm) forms the lower arm of a potential divider which corrects the voltages from the input summer and the drift disappears. Ovening is more painful because it has to warm up to equilibrium and the heater switching noise can get onto the output waveform making it sound fuzzy.
Because there is a capacitor which has to be discharged in every oscillator, any delay in the process will become more noticeable at higher frequencies where definite flattening occurs. Another effect which deteriorates high frequency tracking is voltage dropped across the bulk of the logging semiconductor at high (relative) currents. Both these effects can be cancelled by devious feedback methods.
If you choose a linear oscillator system you are freed from the problems of exponential conversion for the oscillator, but you have somehow to generate keyboard voltages which will put the oscillator through its musical paces. This can be achieved by calculating fixed value resistors which have poor long-term stability, fitting preset potentiometer trimmers which require each note to be tuned or, by far the best method, using a crafty ladder network as found on the Korg 700.
Modulation is also a problem and really requires the modulating signal to be multiplied by the keyboard voltage to maintain an apparently constant depth effect. A low cost get-out is to use a ring modulator.
Coming up to date with oscillators, the most recent revolution (apart from complete software emulation) has been the musical VCO on a chip. Purpose built integrated circuits by Curtis and SSM appeared around 1979 and the same chips are to be found in Sequential Circuits, Oberheim and Roland gear to name but a few.
These chips were absolutely essential for the evolution of the polyphonic synthesiser because they reduce the component count considerably. The best selling and most sought after mono synth was the Minimoog, so why was it discontinued? The answer would appear to be that there were so many components and setting up adjustments involved that it would have to be sold at the asking price of a modern, chip-based poly. Much sales talk surrounded the Polymoog with the idea that each note had its own Minimoog on a chip. In fact the Poly was little more than an organ with two sets of layered tone generators.
Well, that's surveyed oscillators but how about amplifiers and filters? — back to the Theremin. This strange device had two plates which reacted to the proximity of the hands of the performer. One plate was for frequency and the other for volume. A foot-switch was also fitted to short out the loudspeaker when the delicate tone was no longer required. Neat.
The Trautonium had two 'ribbon controllers', again one for pitch and the other for volume. Filters were rather static things at this period and the idea of tracking did not emerge until the introduction of the Bode Melochord. The so-called travelling formant of this instrument used extra keyboard contacts to switch filters to give the same tone colour over the entire range.
The Univox described in the opening paragraph used a thyratron valve for its envelope modulation and although that may sound made up, be assured that it's true. The middle period before Moog saw only the development of the electronic organ and the occasional World War.
The percussion effect on Hammond organs (don't confuse that with autorhythm) punches a mutation stop (usually five-and-a-third foot pitch) into the output mix for just the beginning of a note. Electronically, there was not much to get excited about, just a bit of playing around with the bias of a signal amplifier and some elementary pulse shaping.
Lamp bulbs plus light dependant resistors were also popular for amplitude modulation but the slow speed of response made the scheme unsuitable for much other than tremolo.
The Voltage Controlled Amplifier design by Moog was a fully balanced affair with a buffered transconductance gain block at its centre. As to be expected with a discrete component design, there were a fair number of adjustments to be made to prevent the control signal breaking through into the audio path. Nevertheless the VCA had a wide operating range, good frequency response and the facility to switch between linear and log control functions. Log is necessary to cover the vast dynamic range of the ear whereas linear control is useful for modulation patches.
By 1974 integrated circuits were cheap and reliable enough to replace many of the discrete component designs and many of the adjustments associated with DC amplifiers were eliminated. Apart from the operational amplifier, the most useful development was the transconductance amplifier which was essentially an off-the-shelf VCA.
Filters have been the subject of extensive research over many years but their use in electronic music requires the facility to vary their centre frequency and resonance over a wide range. From the electronic point of view, the problem is to make the resistors of the classical circuits voltage-controllable. Lamp bulbs and light-dependent-resistors can be used but are slow and non-linear. Other methods include the use of junction field effect transistors (JFETs) which are voltage controlled resistors but are again non-linear and distort all but the smallest signals. Solid state devices which switch between two resistors have also been applied. By altering the mark/space ratio of the switching pulse the effective time-averaged resistance value can be altered. It was a nice idea but needed a range switch to cover the audio spectrum.
In effect a low pass filter transforms mathematically into the equation for a rusty pendulum and this can be patched up using integrators and multipliers. The most universal VCF design uses transconductance amps and op amps in a 'state variable' configuration which gives simultaneous low pass, high pass and band pass outputs. It is very low cost and sounds like it, having a puny 12 dB/octave cutoff slope. The Moog filter is much talked about in synthesiser folk law and started life as a discrete component 24 dB/octave ladder design. As everyone else now provides a similar filter, any remaining audible difference must be due to what is being processed by the circuit.
One of the first voice programmable monophonic synthesisers was the 1977-78 Oberheim OB-1 which boasted an eight voice memory bank. The design architecture separated the front panel controls from their associated voltage control inputs and interposed a card packed with CMOS hard wired logic and an analog to digital converter. Settings held in memory could then be read out via a digital to analogue convertor and held on sample and hold units connected to the voltage controlled circuitry.
The sound producing circuitry was dwarfed by the number of logic chips and analogue switches and like so many other first arrivals it was made obsolete by the microprocessor. The micro could handle the programming side with half its pins snapped off so the spare capacity was drafted in to scan a keyboard and the polyphonic synthesiser became possible.
More and more was then expected of the micro such as arpeggiation, sequencing and finally (for the moment) a speedy digital interface (MIDI). It would seem that a synthesiser with only one micro within is becoming old hat.
A final word about connection. In the beginning there was no such thing because there was no need. Instruments like the Univox came with their own amplifiers and that was that. As electronic music took off and the tools of the trade proliferated, an interconnection standard became necessary and voltage control provided it with its one volt per octave CV and positive going gate and trigger. The height of the interconnection fad seemed to be the patch cord studio Moogs which have since proved to have been development tools for the pre-patched models we have today.
With the advent of MIDI the old problems of CVs, gates, log and linear are all irrelevant. It is now even possible to connect "physical" to "logical" oscillators, and this is because MIDI passes the information of intention rather than the goods to carry the operation out. Perhaps we are at long last at the stage where the music comes before hardware.
Feature by Andy Honeybone
mu:zines is the result of thousands of hours of effort, and will require many thousands more going forward to reach our goals of getting all this content online.
If you value this resource, you can support this project - it really helps!