A complete synthesizer from a single chip — Robert Penfold investigates the possibilities using CEM devices.

While many of the semiconductor devices used in electronic instruments and effects units are general purpose types (such as the ubiquitous 741C), there are an increasing number that are designed specifically for electronic music applications. Possibly the best known range are the Curtis Electromusic Specialties CEM33- devices. These can be found in, amongst others, the Sequential Circuits range, including the Pro 1 and the more recent Prophet 600. They are not unknown in designs for the home-constructor either.

We took a brief look at one of these (the CEM3320 voltage controlled filter) in an earlier Chip Parade article, and here we will consider some of the other devices in this family.

CEM VCO

One of the most impressive devices in the Curtis range is the CEM3340 voltage controlled oscillator. This uses the basic method of connection shown in Fig 1. There is insufficient space here to go into detail about the operation of this device, but as will be apparent from a brief study of the circuit, this device has a comprehensive range of facilities such as three output waveforms (triangle, sawtooth and pulse), hard and soft synchronisation, linear frequency modulation and pulse width modulation. Possibly of greater importance, it has an integral logarithmic converter. Modern synthesizers almost invariably use a one volt per octave control characteristic. In other words, raising the control voltage by one volt doubles the output frequency, another one volt increase doubles the frequency again, and so on. This is very convenient, as it enables notes of the equally tempered musical scale to be obtained using a simple potential divider circuit which provides a linear series of voltages — in other words, just a chain of equal value resistors connected across a voltage source. This also makes digital control very straightforward!

The problem with this exponential control characteristic is that it is not the natural one of any VCO. These have a linear control characteristic with an output frequency that is proportional to the control voltage. Thus a doubling of control voltage from (say) one to two volts would give a doubling of frequency, but a further doubling of pitch would require a further doubling of control voltage from two to four volts. A logarithmic converter therefore has to be used ahead of the control input to compress the control characteristic into the required exponential one. This may not seem to be too difficult, but in practice a logarithmic converter having good temperature stability is difficult to realise.

The built-in converter of the CEM3340, aided by having all the components on the one chip and at the same temperature, achieves what for practical purposes can be regarded as perfect temperature compensation. As the +15 volt supply is used as a voltage reference, this must be well regulated if the potential stability of the device is to be attained in practice. VR1 enables the circuit to be set for a one volt per octave characteristic, and VR2 is adjusted for optimum high frequency tracking accuracy. The amplitudes of the three output waveforms are substantially different, but this can easily be rectified by including a switched gain op-amp stage in the output selector circuit.

Envelope Generation

The CEM3310 is a voltage controlled envelope generator, but it is important to note that this device only generates a control voltage to give ADSR envelope shaping from a suitable voltage controlled amplifier. The signal from an oscillator or other source cannot be directly controlled by the CEM3310.

Figure 2 shows the basic method of connection for the CEM3310. The internal circuit of the device is quite complex, but in common with most ADSR voltage generators, it is built around three electronic switches. These switches are arranged so that a capacitor is first charged (attack), then partially discharged (decay), held at this partially discharged level until the gate pulse ends (sustain), and then fully discharged (release). This gives the classic ADSR envelope shape of Fig 3. In order to have proper control of the envelope shape it is necessary for the sustain level plus the attack, decay, and release items to all be adjustable (to have the standard ADSR synthesizer controls in other words). With some envelope voltage generators, potentiometers are used directly to set these parameters, but with the CEM3310 they are set using control voltages. However, in most cases the control voltages would simply be derived from potentiometers, although voltage control of the device obviously gives the option of something more elaborate if desired. For example, having a release time that decreased with increasing keyboard voltage (pitch) would be quite feasible.

The charge/discharge capacitor, C1, has a high impedance signal across it. The output from pin two is buffered and has an output impedance of 200 ohms with a sink current capability of around 560uA. The ADSR times are controlled by a voltage in the range 0 to —5 volts, giving a wide time range of 2ms to 20s. An attenuator is used at each of these inputs to increase the control voltage range from 60mV per decade to these (more usable) levels. The sustain control voltage is in the range 0 to +5 volts, giving a control range of 0 to 100%. This control directly sets the sustain voltage.

Although +15 and -5 volt supplies are needed, an internal zener diode is connected across the negative supply and earth terminals, so that a +15 volt and -15 volt supply can be used. A 750 ohm dropper resistor must then be included in the negative supply lead to the device or the zener will almost certainly be destroyed. This method of negative supply is a feature of the other Curtis devices mentioned in this article.

The gate input has a typical threshold voltage of 2V3 and its compatibility with TTL and CMOS logic devices is a useful feature. A trigger pulse must be supplied to pin five at the beginning of the gate signal, and this is provided by C2.

Further Amplification

The CEM3330 is a dual voltage controlled amplifier which is based on two operational transconductance amplifiers. An advantage of this device (for electronic music applications) over a dual transconductance amplifier such as the LM13700N is that it also incorporates a logarithmic converter at the control input of each amplifier. This enables linear or exponential control of gain to be achieved (a similar device, the CEM3335, allows only exponential control).

Figure 4 shows the basic application circuit for the CEM3330. This only gives the circuit for one half of the device — equivalent pin numbers for the other section are shown in brackets. An important point to bear in mind if you experiment with this device is that both sections must be connected or the entire device may fail to operate properly. The CEM3330 specification is impressive, with a minimum control range of some 120dB, a noise level of better than -100dB, low distortion, and better than -60dB breakthrough of the control voltage at the output even without any trimming. Also, very accurate gain responses are obtained from both the linear and exponential control inputs.

Compared with a circuit using (say) an LM13700N there is a noticeable lack of bias components, and this is due to the internal biasing. On the other hand, no internal output buffer amplifiers are included, and the output signal is actually in the form of a varying current. An external buffer stage and current-to-voltage converter based on IC2 is therefore needed. The signal input is a virtual earth type, and this enables several inputs to be applied to the device and mixed at the input. It is merely necessary to use an extra 100k series resistor for each additional input. The specified input resistor value of 100k assumes that an input signal of several volts peak-to-peak will be present (which it will in most practical applications), but for smaller input levels a lower value might be necessary in order to take full advantage of the excellent signal-to-noise ratio offered by this device. The linear control inputs are also virtual earth types, and can similarly be fed with more than one control signal.

The quiescent bias current is set by R8 through the signal handling transistors of the device, and the specified value gives a fairly high current for low distortion and wide bandwidth. In some applications, such as where the degree of low frequency modulation to a VCF is being controlled, a higher value (lower current) is preferable, as it gives lower breakthrough of the control voltage to the output, and the reduced bandwidth is of no consequence.

Curtis devices are available in the UK from Digisound Ltd., (Contact Details).