This series looks inside the 'black boxes' of electronics to provide a greater understanding of their function and application. We also hope that it will stimulate readers to experiment with these new products and share their ideas and designs with others interested in the field of electro-music.

PART 2: Curtis Electromusic Specialities
Charles Blakey, Digisound Limited

Having looked at the CEM 3340 Voltage Controlled Oscillator last month, let us now proceed to the VCF, VCA and EG.

CEM 3320 Voltage Controlled Filter (VCF)

Filters are used to modify the harmonic content of signals and in a synthesiser one requires a VCF which will track the VCO so that the harmonic content remains constant over the range of the keyboard. Thus if the VCO is set to a 1 volt/octave scale then the VCF must have the same scale and so we are back to the exponential generator.

The functional block diagram of the CEM 3320 is shown in Figure 1 and pin 12 is labelled the frequency control input to an internal exponential generator which in turn is connected to the four gain cells in the filter. This generator does not have IC1 and associated scaling components shown in Figure 2 (discussed last month). For a dedicated application one might get away with a simple 100k/1k8 resistor network which will produce a nominal 18mV/volt at the resistor junction which is connected to pin 12. The free end of the 1k8 resistor of course goes to ground and we must also provide some means of keeping the control range to the 100k resistor between -1V4 and +8V6, which is the best range of the exponential generator. Another snag with this simple approach is that an increasing positive voltage will decrease the cut-off frequency which is the opposite response to the VCO. The alternative approach is to add IC1, etc. of Figure 2 and suitable values are R1 = 100k, R3 = 91k, R4 = 20k, R5 = 56k and R6 = 1k0 with the junction of R5/R6 connected to pin 12. One may then add other input resistors for coarse control, initial frequency and so on, as discussed for the CEM 3340 VCO. While the gain cells of the CEM 3320 are temperature compensated there is still the temperature sensitive 1/T term for the exponential generator and this could be compensated, if considered necessary, by using a 3500ppm/°C temperature compensating resistor in place of R6.

The inputs to the four filter stages are at pins 1, 2, 17 and 18. Normally pin 1 will be made the first stage, for reasons that will be apparent later, but the other stages may be used in any order to suit PCB layout. Each filter stage consists of a variable gain cell followed by a high impedance buffer. Note that the buffer outputs are not short circuit protected and care should be exercised so as not to short them to ground or either supply. Because the gain cell is a current-in, current-out device instead of the usual voltage-in, current-out type the circuit configurations may appear unusual. To simplify matters the first two stages of 24dB/octave low pass and high pass filters are both illustrated in Figure 3. The first point to note is that the input to each variable gain cell (shown only for Stage 1 of the low pass filter) has a forward biased diode to ground and so provides a low impedance summing node of 0V65 above ground. The input current, derived from the input signal, may therefore be obtained with a resistor, or resistors, terminating at this node.

For normal operation each stage is set up with a feedback resistor from the buffer output to the input of the variable gain cell to establish a reference current. In the quiescent state (no signal input) the buffer output will adjust itself to maintain this reference current. For lowest voltage control feedthrough, that is breakthrough of any modulating waveform or 'plop' noises with rapid changes in DC control voltages, and for maximum output signal the quiescent output voltage of each buffer should be equal to 0.46 Vcc which for a +15V positive supply equals 6V9. The internal reference current is a nominal 63uA and thus the feedback resistor may be calculated from 6V9 - 0V65 (remember the diode!) divided by 63uA, which gives a nominal 100k. Referring to the low pass filter circuit, Stage 1 will normally have a zero DC quiescent voltage and therefore all of the input current, I_in, equal to I_ref will be provided by the 100k feedback resistor.

For Stage 2, as well as for Stages 3 and 4 not illustrated, I_in will be made up of 63uA from the feed back resistor plus 70uA from the quiescent voltage of 6V9 across the 91k coupling resistor between stages. We therefore have to sink this excess 70uA - and this is done with the 220k bias resistor connected to the -15V supply. For the high pass configuration the input stage and subsequent stages are capacitor coupled, which blocks out the quiescent voltage of the buffer and therefore the required reference current is attained solely with the 100k feedback resistor for all stages.

Hopefully it will now be obvious that if the input signal contains a DC component then the low pass filter should be capacitor coupled at its input to avoid upsetting the desirable current conditions and raising the voltage at the buffer output. A band pass filter is not illustrated since this may be configured simply by using two stages of high pass followed by two stages of low pass, the latter being as Stage 2 of Figure 3a. A capacitor of 270pF will be satisfactory for these circuits.

The signal input should be a maximum consistent with avoiding clipping at output stages and normally 5V p-p will be acceptable with a +15V supply. The input/output structure of any filter is, however, up to the designer so as to suit a particular application. The input signals may be directly connected, as shown in Figure 3, although in this case the signal source should be of low impedance for the high pass filter. Similarly the output may simply have a capacitor in line to block the DC voltage from the buffer stage. An alternative approach is to use op amp inverters for both the input and output of the filter to provide the greatest flexibility in terms of impedance, signal levels and so on. In the latter case the residual DC voltage may be trimmed out using a preset connected from the negative supply and the summing input of the op amp.

The CEM 3320 has a traditional transconductance amplifier, similar in principle to the CA3080, whose output is internally connected to pin 1, which is the input stage of the first filter section. This may be used for voltage control of resonance (Q) by feedback of the output to pin 8 via a blocking capacitor and a resistor, the latter being about 50k for the filters outlined above. Voltage control is applied to pin 9 via a resistor whose value is selected according to the control voltage available and is such that oscillation does, or does not occur, according to preference, at maximum control voltage. If using a +15V supply for the control voltage try a 150k resistor to begin with. Remembering that the output of the transconductance amplifier is connected to the first stage, it may also be used as a VCA to control the amplitude of the signal input and in this mode it may be used to prevent clipping of large signals or perhaps even to introduce signal clipping to add colouration.

For negative supply voltages in excess of -4V, a current limiting resistor must be placed between the negative supply and pin 13 whose value is calculated from (V_EE-2.7)/0.008. For a -15V supply a 1k5 resistor is suitable. Some improvement in control voltage feedthrough may, however, be obtained by using a variable resistor in series with a fixed resistor and by switching back and forth between the extremes of control voltage while adjusting the trimmer to give the same DC voltage at the extremes. A value of 1k0 for the fixed resistor and 1k0 for the trimmer is a practical combination for a -15V supply.

CEM 3330 Dual Voltage Controlled Amplifier (VCA)

Figure 4 illustrates the functional block diagram of the CEM 3330 and some of the features previously discussed are evident. The variable gain cell is a current-in, current-out type and provision is made for both linear and exponential control of gain. Since there are two VCA's in the package we will often have to give two pin numbers during the discussion and the pin numbers for the second VCA are given in brackets. Linear inputs, pins 7(12), are accepted as currents, I_CL, and the on-chip log converter generates the logarithm of this current while exponential control inputs, V_CE at pins 6(14), are transmitted unchanged to the gain cell.

Before discussing selection of component values there are a number of features of the CEM 3330 which should be noted. First, that both the signal inputs, pins 4(13), and the linear control inputs, pins 7(12), are summing nodes which allows any number of signal and linear control voltages to be mixed within the IC. Next that the current output should be converted to a voltage using an op amp configured as a current-to-voltage converter (Figure 5) and use Ohm's Law, V = IxR_F, to calculate the voltage. The op amp should be a low noise high slew rate type such as LF351 or NE5534. Third, referring to Figure 4 one sees that pin 8 is labelled 'Idle Adjust' and one of the novel features of the CEM 3330 is the ability to alter the quiescent standby current of the signal carrying resistors and hence the operation of the gain cells between Class A (say, 100uA) and Class B (say, 1uA). The choice depends on the application and in general terms operating at high standby current will increase slew rate, increase available output current and decrease distortion but at the expense of increased noise and control voltage feedthrough. For typical VCA synthesiser application one would compromise and choose an operating point of about Class AB (about 7uA standby current) which would be achieved with a 6k8 resistor between pins 8 and 5.

Lastly, for most VCA applications the main concern is achieving a wide dynamic range which for a properly configured CEM 3330 is a minimum of 100dB for linear control and 120dB for exponential control. This range must, of course, be consistent with retaining a high signal to noise ratio, low voltage control feedthrough and low distortion.

As a starting point, with minimum distortion in mind, the signal input resistor, R_I (and the signal output resistor, R_F, should be chosen to give input and output currents of 100uA which for a +10V signal requires 100k resistors. The upper current limits are dictated by the choice of standby current and reference to the data sheet for the Class AB chosen above shows a peak cell current (input plus output), C_P, of about +600uA. In no case therefore should R, or R_F be chosen to exceed V_max/½C_P where V_max is the peak input or output voltage. Another restraint is that the sum of the signal voltages should not exceed ±10V. Now if we look at the equation for the total voltage gain of the CEM 3330 we have:

where I_CL is the linear input current into pin 7(12), I_REF a reference current into pin 2(15), V_CE the exponential control voltage to pin 6(14), and V_T which is our old temperature dependent friend and can be taken as 26mv.

Now if we did not require exponential control then pins 6(14) would be grounded and we would only have to juggle with I_CL and I_REF having already fixed R_I and R_F. For most synthesiser VCA applications we do, however, require this control for the reason stated in Part 1's introduction and two features of this exponential control input should be kept in mind. First it will be the lowest control voltage which results in maximum gain and secondly, as is the case with other IC's already described, the control voltage will have to be attenuated. As regards the latter we are no longer restricted to an 18mV change at pin 6(14) for each one volt change at the input resistor and it may be scaled to suit, for example, 27mV/volt using a 36k/1k0 divider (the 1kO being grounded and the resistor junction at pin 6(14)). I_CL can be as high as 300uA but is best restricted to 100uA if linear scale accuracy is essential and I_REF should be set in the range of 50uA to 200uA.

A practical example follows. Suppose I_REF is set to 100uA, the signal input is 10Vp-p and we have 0 to +10V control voltages for both the linear and exponential control inputs. A suitable starting point for a VCA would then be: 150k resistor connected between pin 2(15) and the +15V supply to establish the 100uA reference current; a 100k resistor, R_I to pins 4(13); a 36k/1k0 divider to pin 6(14) to provide 27mV/volt; a 100k resistor to pin 7(12) to develop a maximum of 100uA for I_CL; and an output resistor, R_F, of 51k. It is now relatively simple to find your own starting conditions by substitution, for example, if the signal input is 5V p-p then R_I is halved to 51k.

Compensation components for the CEM 3330 are necessary and it may also be desirable, depending on application or the compromises made in design to achieve certain objectives, to provide trimmers to improve distortion and control voltage feedthrough.

A summary of these is shown in Figure 6. 6a illustrates compensation for the signal input and the diode (IN4148, etc) is to prevent latch-up. With these component values then pins 9(11) must also have 150pF capacitors to ground. Figure 6b shows the compensation components for the linear control inputs. 6c is an arrangement for allowing distortion to be trimmed to a minimum while 6d allows reduction of control voltage feedthrough. Note that for 6d the compensation components shown in 6a must also be added and that the resistor, R_CVF, is calculated from 15V/I_idle which for the 7uA example above gives a value of 2M1 and so 1M8 would be satisfactory.

One major benefit of the simultaneous linear and exponential controls provided is that it is possible to configure each VCA for exponential response and then incorporate linear amplitude modulation (tremolo). It will also be apparent that a negative voltage into the linear control input will gate the VCA off. Do not gate the VCA off by applying a negative voltage to the gain pins 2(15).

Finally, if the negative supply is greater than -7V5 then a current limiting resistor should be placed between the negative supply line and pin 5. This is selected by (V_EE -7.2)/I_EE and in this case I_EE depends on the idle current to pin 8, being: 10mA for idle currents less than 10uA; 12mA when the idle current is between 10uA and 50uA; and 14mA for idle currents in the range 50uA to 200uA. For the 7uA idle current example and a negative supply of -15V then the calculated resistor value is 780R and thus a 750R resistor will be suitable.

CEM 3310 Voltage Controlled Envelope Generator

First let us clarify some terminology. As most readers will know, when one of the keys of a synthesiser is pressed there are usually two output voltages generated. One is the control voltage associated with a particular key and primarily used to control the VCO and VCF and this voltage will (or should) remain constant until a different key is played. The second is a constant voltage change, for example, a step change from 0 to +15V, or from -7V to +7V and so on, and this change is used to initiate the cycle of an envelope generator. This voltage remains at its changed level for the duration the key is held down and determines the sustain period, that is, on releasing the key an ADSR envelope generator will go into its release (R) phase. In many articles this voltage is referred to as a 'TRIGGER' but when discussing purpose designed integrated circuits it is always referred as the 'GATE' voltage.

For the CEM 3310 envelope generator we need two control voltages, namely, the GATE voltage as already defined and a TRIGGER voltage. Normally if only a gate voltage is present at the appropriate input then an AD (attack-decay) envelope will be generated whereas if simultaneous gate and trigger pulses are received then an ADSR (attack-decay-sustain-release) envelope will be produced. Similarly if a second trigger pulse is received while the gate voltage is still present then the generator will recommence the attack cycle and thus allow complex envelopes to be produced. The different envelopes in relation to the status of the gate and trigger pulses is illustrated in Figure 7. Note that while the CEM 3310 may be configured to give the AD envelope with a gate pulse it is somewhat complicated and beyond the scope of this article.

The block diagram for the CEM 3310 is shown in Figure 8. If only a gate voltage is available then the trigger may be derived from it.

Typically the gate voltage goes to pin 4 which has a 10k pull-down resistor to ground and also via a 3n3 capacitor to the trigger input at pin 5. These allow any ground referenced gate pulses up to +18V (V_CC = +15V).

The RC time constants for attack, decay and release are a function of R_XC_X times the exponential multiplier EXP (-V_C/V_T), where R_X is a resistor taken from the output (pin 2) to I_in at pin 10, C_X is the timing capacitor connected from pin 1 to ground, and V_C is the control voltage. A convenient value for C_X is 33nF and R_X may then be set over a wide range to suit the control voltages. In practice, however, it is desirable to keep R_X low, although above 24k, so as to minimise control voltage feedthrough and other errors. A suitable value for R_X is 27k with the 33nF capacitor. The tracking between different CEM 3310's is typically ±15% and therefore if more accurate tracking between a number of devices is required one can use a fixed resistor plus a trimmer for R_X to compensate for differences between devices.

The control scale sensitivity of the CEM 3310 is 60mV/decade (18mV/octave) and so for a four decade range of, say, 2 milliseconds to 20 seconds, one only requires a -240mV voltage excursion at the A, D and R pins which are numbers 15, 12 and 13 respectively. The impedance at these pins should be kept low to maintain the best accuracy and thus 27k/470R resistive dividers with the junctions at their respective inputs and the other end of the 470R resistor grounded will give slightly more than the four decade control range when the input control voltage is varied from 0 to -15V. The higher the negative voltage the longer the time. For the sustain level control a voltage of 0 to +5V at pin 9 will vary the sustain level from 0 to 100%. Again the sustain voltage may be obtained using a resistive divider connected to the positive supply line via a potentiometer. The control inputs to the CEM 3310 are NPN transistors which allows a single attenuator to drive the same parameter on several devices configured in a multiple chip system, for example, a polyphonic synthesiser. The output voltage of the CEM 3310 is nominally 0 to +5V and the impedance of the driven load should be no lower than 20k. If necessary the output may be buffered using an appropriate FET op-amp. A compensation capacitor of 22nF should be tied to ground from pin 8 in all cases.

Two other features of the CEM 3310 may be useful for some applications. First that pin 16 outputs a voltage of between -0V4 to -1V2 only during the attack phase and this may be used to generate a logic signal to indicate the attack phase. Secondly, if the sustain voltage were to exceed the maximum attack voltage of the CEM 3310 then at the end of the attack curve the output voltage will ramp up to the sustain voltage. This feature can be taken care of by trimming the maximum sustain voltage so that it equals the maximum peak attack voltage but another technique is to use the envelope peak output available from pin 3. For the latter a precision rectifier must be connected which will then automatically limit the sustain voltage.

Details are shown on the relevant data sheet and it should be realised that many op amps have protective diodes at their inputs and so cannot be used in the precision rectifier mode.

As usual we have to place a current limiting resistor between pin 6 and the negative supply if the latter exceeds a certain value, which for the CEM 3310 is -7V5. This resistor is calculated from (V_EE-7.2)/0.010 and for a -15V supply gives a calculated value of 780R and a practical value of 750R. While the other CEM devices described will operate with positive voltages down to +9V (+10V for the CEM 3340) the specified limit for the CEM 3310 is +12V5. Thus to simplify power supply design +15V is a good choice for a full complement of CEM devices although +15V, the -15V for external op amps, and a -5V supply has many merits.

This concludes the review of just four of the specialised integrated circuits available for electro-music applications. If at first reading you find it a bit difficult to follow then don't despair since their internal circuitry is probably far more complicated than the overall circuit in which they are used. On the other hand if some readers find it too simple then data sheets are available but it should be stressed that these are for the relatively experienced designer. One thing is certain, such devices will play an increasingly important role in synthesiser and related designs.

REFERENCES: Data sheets for CEM 3310, CEM 3320, CEM 3330/3335 and CEM 3340/3345 prepared by Curtis Electromusic Specialties.