Quite early in the history of electronics, it was realized that audio systems could become the basis for the generation of music rather than simply its reproduction. Almost as early as triode valves became available, the Theremin was demonstrated. This consisted of a valve beat-frequency oscillator system coupled to a loudspeaker and amplifier . By waving his hands near a metal bar, the player could alter the frequency of one oscillator, so that the beat note, which was normally almost zero, changed so as to give a 'musical' note . Early devices like the Theremin were exhibited widely, and the principles of synthesized music were established long before there was any significant demand for its use . The real start of synthesized music, however, begins with the second series of Hammond organs, introduced in 1939, which used valve oscillators as the source of audio tones which could be switched by a keyboard to an output amplifier and loudspeakers . The synthesizer, so long regarded as the province of the pop-musician, thus had its roots in an instrument that was very widely used in church halls and also in churches.
Synthesizers in general use two routes to the final sound. One route is the voltage controlled oscillator, as pioneered by Moog, the other is the noise source with controllable filters . The voltage controlled oscillator is the logical development of the Theremin principle, and uses an oscillator or set of oscillators which can have the frequency controlled by a DC voltage applied from voltage dividers . Each key of a keyboard instrument operating on this principle therefore need only operate a simple DC voltage divider rather than alter a frequency-determining component such as a ...
Figure 8.1 Wave envelopes, which are the shape formed by the peak amplitudes
of the waves in a note. The sound of a note is affected by both the shape
of the waves and also by the shape of the envelope.
... capacitor or inductor. The noise-source method starts with a generator of wide-band noise, and uses the keys to control filters which pass a narrow bandwidth . In this case, the action of the key is more complex, and often consists of gating the noise waveform to an active filter . The two methods produce very different sounds. The voltage controlled oscillator can produce a waveshape that depends on how the oscillator is biased, so that shapes ranging from sine to square can be produced. The noise filter method inherently produces a more complex waveform, but in some cases not a particularly musical note. Both systems are capable of an almost infinite variation of effect, and can have envelope shapes (Figure 8.1) imposed which will determine whether the note sounds like a specific musical instrument or something quite different.
Both methods are analog, and the introduction of digital methods into synthesis has taken place step by step, and is not yet complete . This, however, is one field in music where digital methods could eventually be totally dominant, with no analog steps of any kind present . Much synthesized music is not 'performed' as such in the sense that the instrument makes a sound which is picked up by a microphone. Synthesizers create an electrical waveform, and whether this is converted into a sound or not is just one option among many. We could, therefore, imagine a system in which the only sounds were heard from headphones run from a small digital to analog converter and decoder, and the whole recording path from instrument to tape was digital . Just as analog synthesizers can make use of the two main sources of the voltage-controlled oscillator or the filtered noise source, so a digital system can, in theory, start at either point . The obvious difference lies in the way that the signal is to be generated, because a digital generator has an output which is a stream of numbers rather than an audio waveform. Since there is no 'original' signal involved, there need be no A-D conversion stage, and the equipment can deal with digital signals until almost loudspeaker level, or directly to a digital recorder . Synthesis methods A digital audio generator consists of a unit in which a series of numbers can be generated whose value corresponds to the amplitude samples of an audio wave. The simple and obvious way to do this, of course, is to use a voltage-controlled oscillator along with an A-D converter, but this is an analog conversion rather than a true digital generator. The alternatives are the use of a microprocessor system to generate numbers that follow a mathematical pattern (simple like that of a sinewave or complex like a sine/cosine series), or the use of ROMs that contain sequences of numbers that can be used to build up waves of specified shapes.
The use of ROMs can be very similar to the way that tape has been used in synthesizers to hold a basic sound pattern which can then be altered by the action of the synthesizer.
The other starting point, of using a noise signal, is less promising. Though a random (or pseudo-random) number generator can be used to provide a digital 'noise' signal, digital methods that provide the equivalent of filtering are not quite so simple. A low-pass filter, for example, implies that numbers are reduced on the basis of the difference between numbers, the equivalent of smoothing steeply rising or falling waveforms. Digital noise generators are already being used to some extent, but coupled to D-A converters and analog filters to achieve synthesis by this type of action. Digital filters are, however, available and their use is increasing.
The main extent of the penetration of digital techniques into synthesis has been concerned with time delays and control techniques. The introduction of controlled reverberation into synthesized signals was at one time carried out only the tape loops and acoustic delays, but units based on charge-coupled devices (CCD' s) have become more common as the price of IC's has fallen.
The next logical step to fully digital delay systems has not been long delayed, and the rapid fall in prices that has affected all digital devices from watches to computers has made its mark here too.
The action of a digital delay circuit for existing analog signals makes use of sampling and an AD converter in order to transform the analog signal into digital form. The digital samples are then stored in memory until the time comes to read them. At one time, serial memory (a shift register) would have been used, but modern techniques demand such luxuries as variable delay time and programmable delay time, so that random access is used greater extent now. The sampling rate is comparatively low on the more modestly priced delay units, and only the top-price units use sampling rates that approach the 44.1 kHz of the CD system. Such units can offer delays of up to a minute that require huge memory capacity - the use of high-speed magnetic discs like the computer hard-disc is often a more useful technique than solid-state memory for such long term delays . MIDI action The most noticeable impact of digital methods upon sound synthesis has been the adoption of the MIDI (musical instrument digital interface) standard both by the manufacturers of musical instruments and the manufacturers of some home computers . The requirement for some standard arose from the increasingly complicated setting procedures that synthesizers required. Since it would be unacceptable to take an interval of a minute or so between notes in order to reset a machine, the principle of using a microprocessor to control the settings has been very attractive for the users of synthesizers . This was done, notably on the Roland Jupiter 8 and the Sequential Circuits Prophet 5, but as other manufacturers rushed to devise sequential microprocessor controls for their own equipment, the lack of standardization became a handicap. It was soon obvious that without standardization, users of microprocessor-based equipment would be confined to the products of manufacturers who made a full range of equipment that used the same methods.
The system, originally proposed by Dave Smith of Sequential Circuits, was modified in the light of suggestions from Roland, Yamaha and others to emerge as the MIDI standard. Though this uses a serial interface with the inevitable limitations of speed, the interface is very cheap to implement and cable length is less of a problem than is the case when parallel-signal systems are used . The standard speed of 31.25 k-baud is well above the speeds used by most computer serial links, and is adequate for most purposes . In this respect, one baud means one signal unit per second and though baud rate does not necessarily mean bits per second, it is usually regarded as having this meaning.
The MIDI system allows codes for notes that span 10 1/2 octaves, and also permits the signaling of changes of memory or for information that is specific to an instrument . The system, though originally devised for controlling keyboard synthesizers can also be used for timing drum and percussion synthesizers, and as a result, MIDI interfaces can be found fitted to a considerable variety of devices - including several models of home computers, notably the Atari ST models . The MIDI system allows up to 16 separate channels to be used, so that the normal fitting of a MIDI interface allows for signals both in and out, except for a device that originates or receives signals only. The system uses 8-bit groups (bytes) of data, and in each byte the individual bits will carry their own information. For example, the channel number can be carried by four bits of one byte, and in this same byte another bit can indicate the type of byte, which would be a status byte if it carried channel number codes and whether or not the following bytes refer to note information. The decoder thus can separate a byte that is used for channel selection from one which is used to carry tone information.
These MIDI codes allow information on note (pitch), volume level on each note, and the type of pitch and amplitude envelopes of each note to be sent both between instruments and from a computer to an instrument . The instruments that are used in a modern MIDI link can operate in any one of four different modes, replacing the older three-mode system ( termed omni, poly and mono), and the mode selection governs the way that the transmitted signals affect the individual instruments.
Mode 1 is the most basic, and MIDI systems in general will run in this mode at switch-on. This mode ignores MIDI channel information so that the equipment will respond to signals on any channel - the instrument will try to play each note that is received in turn, no matter on what channel. The receiving instrument is assumed to be polyphonic - capable of playing more than one note at a time. This mode was incorporated with the needs of beginners in mind and is intended for a very simple system in which a master instrument is connected to a single slave instrument.
Mode 2 is one that is not used to any great extent, because it assumes that the receiving instrument is monophonic, one note at a time, an obsolescent breed. Using Mode 2 with a polyphonic master instrument risks causing confusion because the receiving instrument will not be able to cope with the number of note codes; the usual solution is for the receiving machine to select the last received note, or the lowest or highest pitch, depending on settings . In Mode 3, each instrument is allocated its own channel out of the sixteen that are available, and note information is ignored unless it is on a specified channel . This allows the transmitter to take individual control over each instrument so that each can play an individual part. The notes are, of course, actually sounded in sequence, but because of the comparatively high rate of the MIDI transmission the delay is not audible, and even with sixteen different parts being played, the sounds appear to be simultaneous. The MIDI system needs to be set up so as to ensure that the same channels are being used by both transmitter and receivers . Mode 4 allows for the use of modem synthesizers which can play chords. The older type of synthesizer had one 'voice' meaning that only one note could be played at a time, though special chord keys could be used to provide harmony like the chord buttons on a piano-accordion. More modern units allow multi-voice operation, so that chords can be played in the same way as they are on a piano. This allows much more control over the type of harmonies, since the fixed chord scheme allows only the main major and minor chords to be used.
Mode 4 uses channel 1 of the MIDI transmission to determine the note and other features of voice 1 of the synthesizer, and the other voices are controlled by the other channels . In addition, this mode allows the control of features which may be unique to an instrument, like variation of key pressure affecting loudness, or in the reproduction of glissando. The snag is the many expensive instruments which could exploit mode 4 very effectively are not equipped to do so.