AM8111 VCO & VCF (100M)


Background In 1978 Roland developed and launched the System 100M with an initial set of 8 modules, including a dual VCO (112) and a single VCO, with a VCF and VCA (110). Roland developed a third VCO module (the 111) as a combination of the VCO and VCF. However whilst a couple of examples exist, it never went into production. The 112 module really makes the 111 obsolete by adding a VCA.

Roland 111 Module Copyright Alex Ball

The AM8111 Module Whilst developing my range of 100M modules in late 2022 I came across the prototype 111 module at Alex Balls website – see photo.  Initially I considered building it as the SH-101 VCO and VCF. But that delivers the same 24dB filter as in the original 121.

Instead the design is based on the Coolaudio V3340 triangle core VCO chip and the Coolaudio V2044A filter chip, a nod to Behringer and the Korg Monopoly. These chips are only available in SMD but that helps with squeezing the circuits onto a 16HP wide PCB! The filter chip is a replica of the SSM2044 which is a transistor ladder filter, with in built Q compensation. This combination provides a different sound to the standard 100M VCO/VCF combination, but it also required some development effort!.

The AM8111 panel is laid out in the same way as the Behringer 112 and 121. The VCO has the same three selectable waveforms (triangle, sawtooth, pulse), and the selected waveform appears at the VCO OUT jack. The VCO also has a sub oscillator at one octave lower than the selected octave range, and it has its own dedicated output. The sawtooth, square and sub oscillator outputs are normalised to the three VCF signal inputs.

The VCO has both hard and soft sync (selected by a slide switch) and tall trimmers for PW and PWM. There are rotary controls for five octaves of range and 12 semitones pof manual pitch variation. There are three CV inputs; the first is dedicated to a keyboard signal and has no level adjustment to ensure precise 1V/octave control. The second and third inputs do have level adjustment with tall trimmers, and are designed for modulation from a LFO, VCO or envelope generator.

The VCF has the usual set of three audio signal inputs (normalised to the VCO outputs), and three frequency cutoff CV inputs. There are illuminated slider potentiometers for frequency cutoff and resonance. The resonance slider is a log taper but inverted to give a reverse log taper, a trick Roland used in the System 100 filter. The slide switch selects the polarity of the the third CV input (for example the output of an ADSR) either inverted or not. The VCO keyboard CV input is normalised to the first CV input on the filter to enable key follow.

Coolaudio V3340 This is the first time I have used this chip, and the implementation uses the Alfa Rpar tuning approach, which includes more trimmers but is easier to setup and more accurate. Various problems were encountered and solved.

Pulse Width Issues The external 0 to +10V control voltage needs to vary the pulse width width from 50% to 5%. The voltage at pin 5 of the V3340 needed to increase for 5% width, compared with the AS3340A, a resistor reduction from 24K to 18K in the Op Amp buffer did the job. The voltage at pin 5 varies from +1.86V to +3.66V. More of a problem was a glitch in the pulse waveform at around 40% width where the leading edge of the waveform breaks down. Higher and lower widths are okay, although the forward edge does start to slope at low frequencies.

Looking at the V3340 datasheet they show a hysteresis circuit with a 12M resistor for feedback (pin 4 to pin 5) and a 10k resistor between pin 5 and the incoming voltage control. I added a 1M2 hysteresis resistor (the CEM3340 datasheet suggests 1M) and added the 10k resistor before pin 5. This solved the glitch and sharpened up the leading edge of the pulse width.

The Alfa Rpar tempco circuit needed adjustment to reduce the 24K resistor to 20K, to give the correct adjustment of 0 mV but the stability of this voltage is significantly larger than in the AS3340A.

AM8111 Module

VCO Output Levels The AM8111 needs to play nicely with the 100M audio signal levels which are typically 10V p-t-p. In theory using a +12V power supply to the V3340 will achieve a triangle of 4V p-t-p, a sawtooth of 8V p-t-p and a pulse of 10.5V p-t-p. Actual levels were slightly lower with a power rail at +11.73V and even lower with the +10V precision rail. Op Amps boost and shift the waveform levels to +/-5V.

LED Driver Problems The original Roland 100M modules (VCA and VCF) use a red and green LED to indicate overload and normal signal respectively. It is quite a complex transistor circuit which does not fit on the AMSynths PCB’s. A simpler circuit with a SMD Op Amp and SMD diodes has been used. This reduces the space needed, but the LED’s pull an additional 10mA of power which reduces the VCO frequency. The LED current is causing the +12V rail to fall by 30 mV which flattens the pitch of the VCO.

The Pitch control and the initial frequency trimmer of the prototype use the +12V power rail, so these were switched to the +10V precision rail. Some resistor changes are needed to get the same base frequency and the correct variation of pitch on the rotary control. This significantly improved the pitch stability but the LED’s were still causing voltage rail changes. Low current LED’s have been used instead (x10 less current at 1 – 2mA), which put less load on the +12V power rail and therefore less voltage rail drift to the V3340 chip.

Coolaudio V2044A I decided to use this chip because of its low cost and easy to solder SOIC package. I have used the SSI2144 in the past but was not impressed with how it sounded and its rather small in a TSOP package. The Coolaudio chip worked well in the prototype, with a nice sounding resonance, helped along by a 3n3F capacitor between the two signal input pins. The frequency control voltage range was set at ±90mV (20Hz – 20kHz cutoff), which is the same as the datasheet suggested values of ±90mV.

The filter provides self oscillation from 16 Hz to 20kHz (as setup above) and the signal output has a DC offset voltage. I have added an offset trimmer to the production module and moved the decoupling capacitor after the output buffer Op Amp rather than before (where it interfered with the signal).

The second prototype has a Q compensation circuit, taken from the SSI2144 datasheet. This reduces the drop in level as resonance increases and it uses a V13700 chip in the external resonance circuit. This works really well and creates a great low pass filter – different to the Roland 24dB OTA.

Module Outcomes & Availability The AM8111 module was designed in January 2023 and a prototype tested in February. A second prototype was ordered in March and tested during May/June. The prototype was demonstrated at SynthFest in October. The module was perfected in early November by adding a bi-polar capacitor into the signal input to ensure no DC offset and a wider headroom through the filter. The module is now in production, check the webstore for details.

Copyright AMSynths 2022