The Spectrum Synthesiser (Part 2)
The final part of our professional quality monophonic instrument
Since publication of the Spectrum articles was delayed earlier this year, many improvements have been made to the original design. The synthesiser can still be built for around £200, plus cabinet, yet offers features found only on expensive commercial instruments.
For the benefit of newcomers to the magazine, and to bring our regular readers up to date with the improvements that have been made, we have reprinted some of the original material. This is the final part of the project which contains sufficient information to enable experienced constructors to build the Spectrum. PCB track layouts and component overlays, cabinet drawings, a wiring chart and more comprehensive circuit descriptions are available in the Spectrum Synthesiser book, available from Maplin Publications for £1 plus 24p postage.
The ring modulator (Figure 15) is based around IC20 and processes the pulse wave of VCO1 and the triangle wave of VCO2 to produce complex non-harmonic sounds. It functions in a similar way to the rampwave shaper of the Spectrum LFO by inverting the triangle wave about its midpoint when the pulse wave is high, and leaving it unchanged when low. This constitutes four quadrant multiplication of the value of the triangle wave by the value of the pulse wave (-1 or +1). When the pulse output is low TR12 is off and the triangle wave is inverted with a gain of 2 by IC20a. The output is mixed with the original triangle wave of half the amplitude and opposite phase by IC20b. With the pulse output high the collector of TR12 is at -15V and the output of IC20a is positive. This reverse biases D32, and no signal reaches IC20b via R221. The original triangle wave is inverted by IC20b and shifted by the current through R220. The output of IC20b is the required product.
The noise generator is quite conventional, using the thermal noise of a semiconductor junction as a source. TR14 amplifies the noise on the emitter of TR13 to about 4mV p-p, which is boosted to +2.5V by IC21. RV31 mixes the noise and RM signals, which are then fed to IC22, a transconductance amplifier which acts as a VCA. S11b selects the appropriate modulation source, which is conditioned by IC23. The LFO signals are symmetrical about 0V, whilst +EG swings from 0V to +5V and -EG goes from 0V to -5V. In order that all these signals have the same effect, therefore, an offset is selected by S11a and added to the modulation so that pin 6 of IC23 always swings between 0V (maximum gain) and about -14 volts. The CA3080 is really a current controlled amplifier, and so R237 converts this voltage swing into a control current. Since IC23 cannot completely cut this current off, R238 and diodes D33-D35 are included to ensure that the amplifier is truly off at the maximum negative control voltage.
The heart of the filter is the CEM 3320 IC from Curtis Electromusic Specialities. Designed especially for use in voltage controlled filters, this IC contains four identical filter elements controlled by a temperature compensated exponential converter. Each element contains a transconductance type amplifier plus a buffer amplifier to avoid loading of the TCA's output. Depending on how the circuit is connected, either low pass or high pass filter sections may be created as in Figure 16; the three modes of the Spectrum's filter are formed by different combinations of these.
The low pass response is obtained with four low pass filter sections; since each section has a roll-off of -6dB/octave, the overall filter slope is -24dB/octave. The band pass response has two low pass sections, preceded by two high pass sections so that only signals in a narrow range of frequencies are allowed through. The low band pass position, as you might expect, is a mixture of the preceding two configurations and consists of only one high pass section followed by three low pass stages. Switch S12 rearranges the signal paths and biasing around the IC to allow the three different configurations to be achieved.
IC24b is a four input mixer, accepting signals from the VCOs, the noise/RM VCA and the external input socket JK7. R242 is included to combat stray capacitance effects caused by the long leads to the VCO waveform selectors.
The CEM 3320 does not have a summing control input as the oscillators do, and so IC26 performs this function. As well as modulation inputs selected by S13, the key CV is fed in via the 'keyboard follow' control RV40. When this control is at maximum, the filter's cut-off frequency has the same 1V/octave law as the oscillators, and hence will track the keyboard so that the notes have a constant timbre. On most acoustic instruments, however, the upper notes have less harmonics than the lower ones, and if the key CV is attenuated by RV40 this effect may be obtained on the Spectrum. RV37 is included to allow setting up of the 1V/octave law, and if required, may be set to give the reverse of the above effect. In this case, setting the 'keyboard follow' control to 10 will cause higher notes to have more harmonics, and true keyboard following will occur at some lower setting.
The last board in the synthesiser, but by no means the least, contains two VCAs and two envelope generators (EGs); the overall circuit is given in Figure 17. Both VCAs are contained in IC28, a CEM 3330.
IC28a performs the envelope shaping function, and is fed with the envelope signal via R274 since this IC works with current inputs and outputs rather than voltages. R273 performs the same function for the audio input, whilst IC29b converts the output current back into a voltage.
Panning and modulation are performed by IC28b, which works in an identical manner to IC28a; audio and control inputs are via R287 and R288 respectively, and output conversion is done by IC29c. When the FUNCTION switch S14 is in one of the MOD positions, both stereo outputs are connected to the second VCA, which then simply modulates the amplitude of the envelope shaper output according to the LFO waveform. IC30 amplifies and level shifts the selected waveforms so that the top end of RV42 always swings between 0 and +12V. Instead of going to 0V, which would cause IC28b to cut off the signal when the DEPTH control was at minimum, the other end of RV42 goes to a reference voltage generated by R292, 293, RV44 and buffered by IC27a.
In the pan mode, only one stereo output comes from the second VCA; the other is fed from the input of this VCA, the envelope shaper's output, via IC29d which subtracts the first channel's signal. This means that as one channel's output becomes louder, the other becomes softer and vice versa, in such a way that the total output is constant; so the volume is unaffected, but panning is achieved. The gain of the various circuits is arranged so that when IC28b is at around unity gain (100uA into pin 12) the output of the two channels is equal; i.e. 3V peak to peak with one VCO on, no filtering and RV45 at maximum. With full modulation, therefore, each output swings between zero and twice this figure.
IC29a combines half of each of the stereo outputs to give a mono signal of the same amplitude, which is affected by modulation but not by panning.
While the Spectrum's output is normally in the region of 3V pk-pk, 1V rms, factors such as modulation, resonance on the filter etc. can increase this to a maximum of 25V pk-pk. If required, the output may be attenuated by inserting resistors in series with the clockwise tags of RV45a and b. The output may be fed into any impedance greater than 25k; below about 10k, loss of bass may become apparent.
Once again, Curtis Electromusic come to the rescue and each envelope generator is built with a CEM 3310. Both circuits are identical in most respects, except that IC32 has an inverter on its output to provide EG+ and EG- signals, plus the circuitry for achieving key repeat.
R309 and 311, C59 and 61 set the speed range of each generator, and have been chosen to facilitate setting very fast attack times whilst allowing slow decay and release. These components affect all three times equally, and if desired, R309 and 311 may be increased to 'slow down' the envelope times.
Sustain level is controlled by RV48 and RV53. It is important that the sustain control voltage at pin 9 of each IC should not exceed the peak level attained during the attack phase; since this level is available on pin 3, the sustain pots are simply run from this voltage. If external modulation of sustain level was required, a more elaborate level sensing circuit would be necessary (as described in the Curtis data sheet).
Pin 4 is the gate input, and the trigger signal for pin 5 on each IC is derived by C57. In addition, IC33a and TR15 are brought into play on the 'repeat' and 'key repeat' functions; IC33a detects when the envelope output has reached the sustain level (i.e. the attack and decay phases are finished) and TR15 briefly pulls the trigger inputs high to restart both envelopes.
IC27b detects the signal at pin 16 of IC32, and lights D38 to indicate when this IC is in its attack phase.
Use the printed circuit board as a template to mark the fixing holes on the underside of the keyboard chassis. Mark them such that the edge of the board holding the bars will be about 5mm from the plungers and then drill for 6BA clearance. Fit the 48 divider resistors on the component side of the board along with the 12 veropins and solder in place. Cut the palladium bars to length and fit them to the track side using small loops of wire passed over the bar, through the mounting holes and twisted on the component side. Make sure each bar is well seated before soldering at each loop position on both sides.
The gate bar should lie flat on the PCB, whilst the S/H bar should be spaced away from the surface slightly by wrapping the mounting wire round the bar before soldering. This gives one wire diameter under the bar, and ensures more reliable contact.
Cut each plunger to length, leaving the nearest slot to the key end for the contact. Tin 5mm of both ends of the contact springs and fit each one by passing the thin end through the detached plunger and soldering it to the pad on the PCB. If you've marked the PCB mounting holes correctly then for proper operation the end of the spring should be about 2mm from the far edge of the pad. The positioning of the PCB and the springs on the PCB is not critical as long as when the PCB is mounted and the plungers clipped on, the springs are under slight tension to ensure positive contact. Mount the PCB to the chassis using 6BA bolts, ½" spacers and nuts, and washers to separate them further. The keys opposite the mounting positions will have to be temporarily removed to fit the bolts, and this should be done before drilling if a hand-held drill is used, to avoid the possibility of damage to the keys. Again, the spacing is not critical so long as all the contacts normally clear both bars and make contact with both when their keys are depressed. A ½" spacer and one nut were found to be about right, though washers could be used if a high or low action to the keys is preferred. Connect the two halves of the board together using short wire links across the Veropin pairs. This completes the keyboard construction.
The power supply should be set up first; none of the other circuits will work without it, of course, and various voltages are derived from the + and -15 volt rails. Adjust the output voltages without the rest of the circuitry connected to begin with; RV1 sets the + 15V output, RV2 the -15V. Use the most accurate voltmeter you can get hold of; a digital multimeter would be best, and an oscilloscope is likely to be more accurate than a cheap mechanical meter. On the prototype, the entire synthesiser consumed around 115mA on the +15V line, and 130mA on the -15V line. If you have a dual bench power supply, you may like to check the consumption of the rest of the synthesiser before connecting it to the PSU. If not, the Spectrum's supply has current limiting to protect it from faults, but it is still worthwhile to insert a current meter in each supply line in turn to check for excessive current drain. Once you are sure there is nothing drastically wrong, the power supply can be connected up to the rest of the circuitry. Connect the output socket(s) to an amplifier, and you should be able to persuade the synthesiser to make some sort of a noise, although it will probably be horribly out of tune. After allowing everything to warm up for as long as possible — 1 hour say — the rest of the circuits can be set up in the following order.
Set the TUNE control to midpoint, and the GLIDE control to zero. Monitor the key CV output from the VCO (pin 99) with the most accurate voltmeter at your disposal. If the Spectrum is to be used with other equipment already calibrated at 1 volt per octave, a digital meter will be essential here; otherwise, this measurement is less critical.
Press middle C on the keyboard. The key CV should be roughly 0 volts; make a note of what it actually is. Now press the next C up from middle C, which should produce a key CV 1 volt above that for middle C. If it is more than this, turn RV3 clockwise and vice versa. The middle C key CV will now have changed, so repeat this procedure as many times as necessary to obtain the correct 1 volt per octave change.
The VCOs are the heart of the synthesiser, and time and trouble taken in setting them up carefully will be directly reflected in the final performance of the instrument. Some way of monitoring the oscillators' frequency and comparing it with a reference will be necessary. The ideal solution is a digital frequency meter, which combines monitor and reference in one.
Set VCO1's range to 8', and sound the first A up the keyboard; note its frequency, which will eventually be 220Hz; don't worry if it isn't.
Press the second A up, and its frequency should be an octave above the first; i.e. exactly twice that of the first.
If it is flat, i.e. lower than it should be, turn RV23 anticlockwise and vice versa.
Now go back to the bottom A, which will also have changed, and repeat the process as many times as is necessary to obtain an exact doubling of frequency when going from the first A to the second.
The upper frequency range needs to be set separately; set VCO1's range to 2', and once again play two notes an octave apart. This time, leave RV23 strictly alone and adjust RV55 to give a doubling in frequency. The VCO will always be flat, so turn RV55 anticlockwise to correct this; this adjustment is not as critical as the basic low frequency one.
No references are required for the rest of the tuning up; VCO2 is best adjusted with reference to VCO1 to ensure the two oscillators track exactly.
Listen to VCO1 and VCO2 together, both on the 8' range and with VCO2's TUNE control central. Press any note low on the keyboard, and tune the VCOs together with RV18. Now press a high note and, by switching VCO1 and VCO2 off alternately, determine whether VCO2 is sharp or flat in relation to VCO1. If it is flat, turn RV24 anticlockwise and vice versa.
Repeat the above paragraph until the oscillators stay in tune over the whole span of the keyboard, but without changing ranges at this point.
Now switch both VCOs to 2' range, and repeat the procedure, tuning RV56. VCO2 will always be flat to begin with, and so RV56 will need to be turned anticlockwise.
Set both VCOs to the 64' range, play a high note, and tune the oscillators together using RV17 or 18. Switch VCO1 to 32' and adjust RV19 for minimum beating; then switch VCO2 to 32' and tune the VCOs together again with RV20. Switch VCO1 to 16' and adjust RV12, then switch VCO2 to 16' and both oscillators should be in tune; if not, trim RV20 very slightly. Switch VCO1 to 8' and adjust RV11; adjust RV10 with VCO1 on 4' and VCO2 on 8', and finally switch VCO1 to 2' and VCO2 to 4' and adjust RV9.
The oscillators should now remain in tune with each other over the whole range of the keyboard and range switches; in practice, slight anomalies in the control characteristics will prevent perfection being achieved, but only the slightest touch of VCO2 TUNE should be necessary to correct any mistracking.
Once the oscillators are tracking satisfactorily, set VCO2 TUNE and the keyboard TUNE to mid position, and tune the second A up the keyboard to middle A, or 440Hz. RV17 tunes VCO1, and RV18 tunes VCO2. If the Spectrum is to be used with another instrument which cannot be tuned, you may prefer to tune up to that instead.
RV27 may be used to set the width of VCO2s pulse output, or simply left midway.
RV29 should be set to give 3.85 volts on its wiper, and RV30 to give 1.6 volts on its wiper.
The final VCO adjustment is to centre the horizontal joystick movement. Loosen RV13's clamp screw, shown in Figure 25. Set controller FUNCTION to VCO1, and DEPTH to 10, whereupon VCO1 will probably go wildly out of tune. Hold the joystick lever and RV13's trim tab central, and rotate the body of RV13 to bring VCO1 back into tune; then do up the clamp screw. Once the joystick is mounted, and after transporting the synthesiser, adjust the trim tab so that when the controller DEPTH control is rotated back and forth, no perceptible pitch change takes place.
RV8 is the only adjustment on the LFO. Set oscillator modulation as follows: SOURCE to LFO MAN, DEPTH to 10 and FUNCTION to VCO 1 + 2. Modulation of the VCOs will now be apparent; with the joystick lever and RV7's trim tab central, adjust RV8 until there is no modulation breakthrough.
Switch off both VCOs, and turn up the NOISE AND RM LEVEL. Select square wave output from the LFO, and turn noise & RM modulation SOURCE to + LFO. Turn RV35 fully anticlockwise, so that noise comes through loudly whilst the LFO LED is off, and quietly when it is on; a fairly slow LFO rate is advisable. Now turn RV35 clockwise until the noise is just cut off during the LED on periods. If any clicking or thumping is apparent as the LFO switches, adjust RV33 to get rid of it.
Now turn the SOURCE switch to + EG, turn the envelope generator SUSTAIN to zero, and turn RV34 fully anticlockwise. Some noise will now be heard on the Spectrum's output; turn RV34 clockwise until it just disappears. Turn down the noise LEVEL, and return SUSTAIN to 10.
RV37 adjusts the filter's volts per octave characteristic, which is not nearly as critical (or difficult) as the adjustment of the VCOs, and may be done most simply by ear. Set the filter controls as follows: RESPONSE to BP, FREQUENCY about midway, KEYBOARD FOLLOW to 10, RESONANCE to 10 and DEPTH to 0. The filter should oscillate with a pure tone which can be played from the keyboard; to avoid confusion, make sure both VCOs and the noise & RM are off. Set RV37 midway, and play a scale on the keyboard; e.g. C major, all the white notes between one C and the next. If the scale sounds 'compressed' — as if it should go on longer to reach the proper note — turn RV37 clockwise, and vice versa.
Altering RV37 will also alter the tuning of the whole scale, but carry on playing and adjusting until the scale 'sounds right'; like the doh, re, mi... etc you learnt in school.
Finally turn the resonance down ready for the final setting up.
With the synthesiser still set to give no sound, turn the GATE MODE switch to LFO, set the envelope shaper SUSTAIN to 10 and ATTACK and RELEASE to 0. Turn up the LEVEL control, and there will be a 'thump' each time the LFO switches (along with some background noise). Adjust RV41 to minimise this thump.
Now switch the GATE MODE back to HOLD, and select either LFO MOD on the OUTPUT FUNCTION selector; the LFO should still be giving a square wave. Turn up the DEPTH control, and the thumping will return, but sharper this time — more of a clicking sound. Adjust RV43 to get rid of this as far as possible. If necessary, keep turning up the amplifier's volume as these adjustments progress to keep the clicking audible.
Turn DEPTH back to minimum, select any 'pan' position on the FUNCTION switch, and monitor the stereo outputs with a dual beam 'scope or well-balanced amplifier and headphones. Turn on one of the VCOs, and adjust RV44 to give equal outputs from each channel.
Finally, adjust RV50 to give -0.24 volts on pin 156 — or the clockwise tag of any ATTACK, DECAY or RELEASE pot — with respect to 0V.
This completes the construction of the Spectrum Synthesiser. Articles on playing technique and details of a demonstration cassette will be published in future issues of E&MM.
Previous article in this issue:
Next article in this issue:
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!
Please note: Our yearly hosting fees are due every March, so monetary donations are especially appreciated to help meet this cost. Thank you for your support!