The CubuSynth Engine is a dual analog VCO with wave shaping functions, internal modulations and a chaos generator. It is based around two CEM3340 VCO integrated circuits. This post is about my DIY build and a look at the waveforms. Go to the link above for video demos.

The DIY Kit

I bought this kit from CubuSynth’s Etsy Store. It shipped from Portugal, designed by Ruben Sponar, who kindly sent me the schematics when I emailed him.

The PC boards are well-made. Notice the tiny PC board that holds a 4-position slide switch.

Building the kit

The build instructions are basic, intended for more experienced builders. The order of parts to assemble is described, but not in a step-by-step manner. The board silkscreen identifies the parts by number and value, though the value is covered up once the component is installed. All the needed parts were supplied in the full kit. I probably had all the resistors and most of the capacitors and transistors, and some of the chips. I went with the full kit to get the right switches, pots and headers.

The main PC board is the one in back. It is joined to the control board via three headers and held in place by the nuts on the three toggle switches. A clever design where the switches are on the back board and tall enough to protrude through the front board and front panel. I soldered the ICs, as is my practice with well-made PC boards, but I recommend using the provided IC sockets if you are not prepared to replace a chip if needed.

I found three places where components should be spaced farther apart. Two resistors, positioned at the ends of ICs were a tad too close. I had to resolder to move them out of the way to get the chips in. If you get v1.1 of the PC board, you should hold off on installing R3 (next to U1) and R181 (next to U15) until after soldering the socket (or the chip, if you choose). U4 and U9 are too close together. I carefully filed the ends of these ICs to get U4 seated well (I soldered the two CMOS chips last).

The only parts I substituted were the 0.1uf bypass caps, replacing the supplied discs with small yellow Multilayer Ceramic X7 types that I had in my stock.

I did not install the 10-pin header for the expander that I’m not using. And I substituted three 10uf Nichicon electrolytic capacitors for the three 47uf ones supplied, because 10uf is enough power bypassing and I prefer Nichicon.

When adding all the panel parts, I broke out an extra step. The panel needs to be installed with all nuts in place before soldering. I left the LEDs for a separate step, since the panel has to be removed again anyway. It’s easier to deal with the LEDs after the other panel parts have been soldered.

Calibration

Trimming the sine wave was easy, having the scope that reads min and max voltages. I simply trimmed to center the wave.

The 1V/Octave trimming was straightforward, too. I found I had to turn the trimmers counter-clockwise several turns to reduce the span. But once I got the sweet spot, tracking held at 50-100Hz and also at 2000Hz to 4000Hz. I didn’t even need to move the HF compensation trimmers.

The rainbow of LEDs around the Chaos knob is a nice visual feature. The Chaos circuit is a mini-Turing Machine, clocked by VCO1 or an external clock, and using VCO2 as the data input. Chaos has its own output and is also normalled to the Exponential FM and PWM inputs of both oscillators.

Glancing at this panel, one might wonder what the two waveform selector switches are for, since all of the waveforms have separate outputs. The answer is that the selected waveform is routed to several places internally, including the linear FM input of the opposite oscillator, the wave folder input of its own oscillator, plus the 1v2 (comparator), the ring mod, and the amplitude modulation combined outputs. All of that internal routing affords a lot of cross-modulation capability within this complex oscillator.

I like to substitute knobs, and I did so for Engine. I just like the look of Davies clone knobs and I have inserts that allow them to be used with knurled shaft pots. I also had some nice knurled knobs that I used for the waveform selector switches.

A look at the outputs

With no external FM inputs, the coarse frequency knobs afford a range of about 15Hz to 50Khz! I think that’s a little wide. I might modify the coarse knobs for less range, winding up at 20Khz or so at the top end. The sine wave drops in amplitude much over 10Khz or so. Negative CV can lower the frequency into LFO range, but this oscillator is designed mainly for audio purposes.

The range of the pulse width knobs is just right. Without any PWM CV input, the knobs go down to thin pulses at either extreme and don’t shut off the output. You might ask why such a large knob is used for PW. The answer is all the places that the pulse can be used internally (see above).

UPDATE: See my post about resistor changes to improve the waveforms.

The sine wave has a big amplitude, almost +/-6V. I centered it using the sine trimmer.

VCO 2 pulse wave looks cleaner.

Picture shows the triangle wave being folded. It’s only about 3 volts peak-to-peak, and rides on a DC bias of about +5V. I don’t see why it has to be like this. A couple of resistors set the gain and offset. I’m going to experiment with those values.

I forgot to DC-couple the scope, but the 1v2 output is +/-5V anyway.

Again, I forgot to DC-couple the scope. But a DC-coupled look shows that the AM output is bipolar.

Steps in the chaos output occur at the rate of the clock, which is either VCO1 or external. It looks very spikey, which is probably OK for a chaos output, but I’d rather see it be clean for external use.

Addendum – A few modifications

I changed some resistors to fix the wavefolder amplitude and offset. Also, I added small capacitors to remove spikes from the Ring Mod and Chaos outputs. As for the Compare 1v2 output, I didn’t see a way to fix the spikes.

Wavefolder output changes

R189 and R190, originally 20K, were changed to 47K. The amplitude of the folder outputs is now 8.5V peak-to-peak, instead of 3V P-P. I removed R236 and R237, which were imparting a +5V offset. This nicely centered wavefolder 2, but wavefolder 1 had a little too much negative bias. So I replace R236 with a 360K resistor that brought it closer to center. A 390K would have worked as well.

If you want to do this mod, just omit R236 and 237 (47K) and put those resistors into R189 and R190. You may still have some offset, but the amplitude will be around 8.5V.

Removing spikes from Ring Mod and Chaos outputs

Any time I see a feedback resistor in an op amp I know that I can add a small capacitor to act as a low pass filter to remove spikes. I chose 22pf caps and placed them in parallel with R77 for the Ring Mod and R191 for the the Chaos.