Non-Concordant Tone Generation
A fascinating area of synthesis with practical circuits to start you off.
Most non-electronic musical instruments of western origin produce notes by resonating a string, tube, or skin. This gives a spectrum of output frequencies that are largely harmonically related, and most western music is based on complex harmonies.
Of course, there are musical instruments which give a spectrum of output frequencies which are non-concordant (a bell being perhaps the most obvious example), and in some musical cultures instruments of this type predominate. These produce a very complex spectrum of output frequencies, many of which are not harmonically related to the fundamental output frequency.
A complex sound of this type cannot be generated electronically using a single tone generator plus envelope shaping. It is necessary to use two tone generators plus a double balanced modulator (or ring modulator as it is more commonly called in electromusic applications) to produce the non-harmonically related output signals. A ring modulator has two inputs and a single output, and signals applied at either input do not appear at the output. What do appear at the output are the sum and difference frequencies. To take a simple example, if 200Hz and 800Hz sinewaves are applied to the inputs of a ring modulator, there are just two output frequencies; the 600Hz difference frequency (800Hz - 200Hz = 600Hz), and the 1000Hz sum frequency (800Hz + 200Hz = 1000Hz or 1kHz). In practice one or both of the inputs would be a tone having a large number of harmonics such as sawtooth or square wave, or some other fairly complex signal, so that the output signal would contain a vast number of frequencies — not just two frequencies which could be produced direct from the tone generators without using a ring modulator. Each harmonic of one input reacts with every harmonic at the second input of the modulator to produce sum and-difference frequencies.
A ring modulator can be based on a voltage controlled attenuator using the arrangement shown in the block diagram of Figure 1. Some of the input signal is fed to a mixer via a gain control, and the rest of the input signal is fed to the mixer by way of a voltage controlled amplifier and an inverter. The balance control is adjusted so that the two inputs to the mixer are at precisely the same level, but due to the inclusion of the inverter stage they are out-of-phase. They thus cancel each other out exactly, and a signal applied to the input does not appear at the output. With a good quality voltage controlled amplifier the breakthrough from the control input to the output should be very small as well.
If a signal is applied to the control input, the signal balance at the mixer is lost except when the control signal passes through the quiescent control voltage, and this gives the required intermodulation of the two input signals and the sum and difference signals at the output.
There are a number of double balanced modulator integrated circuits available, and although these are mostly intended for use in communications equipment with high input frequencies, they can be used at audio frequencies with suitably modified component values. This mainly entails an increase in the value of input and output coupling capacitors so that efficient coupling is provided even at the lowest audio frequencies. One of the least expensive double balanced modulator devices is the MC1496 (or equivalent), and a simple ring modulator circuit which employs this device is shown in Figure 2.
This circuit has a simple tone generator built around IC1 which feeds one input of the ring modulator, and this has a frequency range of approximately 50Hz with VR1 at maximum value to 5kHz with it at minimum value. However, reducing the values of C3 and C4 (which should have the same value) gives an inversely proportional increase in the output frequency range, and increasing the values of these components similarly gives a reduction in the output frequency range. For example, changing C3 and C4 from 10nF to 100nF gives an output frequency range of about 5Hz to 500Hz.
The output waveform of the oscillator is a sinewave, but since diode waveshaping is used, a small amount of distortion is evident on the output signal.
Obviously it is not essential to use the built-in oscillator circuit, and any desired signal can be coupled to pin 10 of IC2, but the input signal level should only be about 50mV RMS if overloading is to be avoided.
The other input signal is applied to JK1, and needs to have an amplitude of a few hundred millivolts RMS with the maximum level prior to clipping being approximately 1 volt RMS. The voltage gain of IC2 is set by R13 and using the specified value and input signal to pin 10 the output signal level is comparable to that supplied to JK1. The MC1496 has differential outputs at pins 6 and 12, but in this application it is only necessary to use one of these outputs.
Balance control is set by RV2, and this is carefully adjusted to minimise breakthrough of the oscillator signal with no input applied to JK1. The fundamental oscillator frequency can be completely eliminated, but harmonics generated by IC2 will be left. Nevertheless, a high degree of suppression is achieved with the total noise output of the circuit being typically less than a millivolt, and this is more than adequate for any normal application of the circuit. The circuit requires dual 9 volt supplies, current consumption being only about 3 milliamps from each rail.
An alternative ring modulator circuit is shown in Figure 3, and this requires only a single 15 volt supply rail, but the circuit will work quite well from a supply potential of around 9 volts. The current consumption of the circuit is about 17 milliamps using a 15 volt supply, or about 12 milliamps using a 9 volt supply.
The oscillator section of the circuit is essentially the same as the one used in the earlier circuit, but R1, R2 and C3 are needed to effectively give a central 0V supply rail for the oscillator circuitry.
The ring modulator is based on the SG3402 integrated circuit, and this is designed to handle input signal levels of only about 50 millivolts peak to peak. R7 and R8 are therefore used to attenuate the oscillator signal to the appropriate level, and R9 plus R10 attenuate the signal at the other input of the ring modulator so that an input level of up to about 500 millivolts RMS can be handled without overloading. The overload margin can be further increased if necessary by making R9 higher in value. Tr1 is used as a low gain common emitter amplifier at the output of the circuit, and this simply restores the signal level to one comparable to the input level.
Balance control is provided by RV2, and this can be adjusted to give a very high degree of suppression so that there is no significant breakthrough of the oscillator signal at the output.
Ring modulators can be used in a number of ways, and the circuits of Figures 2 and 3 can be used to process signals from practically any electronic instrument, or even a voice signal. However, the effect obtained is a very extreme one, bearing in mind that input signal only appears at the output at an insignificant level, and that the new frequencies will not (except by chance) be harmonically related to the input frequencies. Therefore, when using a ring modulator and tone generator to process the output of an instrument, it is normally necessary to mix a certain amount of processed signal with the unprocessed one. Figure 4 shows the circuit diagram of a simple and quite conventional mixer circuit which can be used to do this. There are three inputs available so that it is in fact possible to mix both input signals and the output signal in any desired proportions.
Ring modulation can be used in the production of bell-like sounds, and this is achieved using two reasonably pure tones as the input signals to the ring modulator. The two tones are adjusted to slightly different frequencies to give a beat note of a few Hertz (the difference frequency), plus the main bell tone which will of course be at double the pitch of the input tones (the sum frequency). Harmonics of the input tones will react with one another to give a richer more realistic sound, but in order to obtain a really good effect it is necessary to have proper envelope shaping. Bell type tones can also be produced using two oscillators some musical interval apart, but they should be slightly off-tune in order to give a realistic metallic 'clanging' sound.
By using a tone generator and a white noise source as the input signals to a ring modulator, it is possible to obtain a sort of filtered noise, with a high frequency tone tending to give a high pitched noise output. This can be used in the generation of cymbal type sounds, and gives better results than using direct white noise.
Another interesting use of a ring modulator is as a frequency doubler, and in this application it is merely necessary to feed the same signal to both inputs of the modulator, provided the input is sinusoidal. The difference signal is obviously zero, and the sum signal gives the required frequency doubling. By mixing the input signal and the output signal (or by simply unbalancing the ring modulator slightly), a tone having a strong second harmonic can be produced. Note that processing a complex signal using this frequency doubling technique also results in strong intermodulation, and not just the frequency doubling.