DRAFT - nobody has built this one on a breadboard yet, me included. What it does have going for it is unusual: it is the stripboard Cicada NAND translated hole for hole, and that board is built and working. Every net on the board below was checked against it and all twenty-four match. The one thing added is the volume knob, and it is entirely off the board - so the copper is still the proven circuit. Build it once, fix anything that fought back, then delete this banner.

Cicada NAND on a Breadboard

Four CD4093 voices, each with its own pitch knob and its own mute · By Onion Madder · Mess O' Pedals

One chip, four oscillators, five knobs, four switches. Every gate in a CD4093 is a Schmitt trigger, which means every gate is an oscillator with nothing but a resistor and a capacitor hung off it - so a quad NAND is four voices in a fourteen-pin package. Each one gets its own timing capacitor, so they sit in four different registers: shrill, bright, middle and a slow click. They sum through a resistor each onto one bus and leave through a volume knob on one jack. The four switches are the instrument: it is a drone until you start muting voices in and out of it.

This is the tenth build in the breadboard series, and it is the solderless version of a board that already exists. The stripboard Cicada NAND has been built four times and every one worked, so the circuit here is not a paper design that might be right - it is a working instrument with its copper unrolled onto a solderless board. Power is a 9V battery, and the chip is happy anywhere from 3V to 15V.

One thing here that the stripboard board does not have: a volume knob. It sits off the board between the output capacitor and the jack, wired as a divider, and it earns its hole twice over - four CMOS outputs summed and coupled straight to a jack is a hot signal, and the pot's leg to ground also gives that output node a path to ground it did not otherwise have.

Reading the breadboard

This guide names exact holes, like E20 or C44. Letters run across the board and numbers run along it, and both are printed on the board itself, so every instruction below points at one hole you can put a finger on.

The center channel splits the letters into two banks: A to E on one side and F to J on the other, with E and F being the two rows either side of the channel. The five holes in a bank that share a number are all the same electrical point, which is why a step sometimes names a different row than you would expect - any free hole in that column and bank will do.

Turn the board so the numbers climb left to right. Then A to E is the lower bank and F to J the upper bank, the top rails are the pair above F to J and the bottom rails are the pair below A to E. This build uses both banks and both rail pairs. Before you wire anything, meter each rail end to end; if yours is split, bridge each half with a jumper first.

The chip does the work of a wire here. Voices 1 and 2 live entirely in the lower bank, voices 3 and 4 entirely in the upper. That is not tidiness - it falls out of the pinout, because a CD4093's top half mirrors its bottom half. It also means voice 1 and voice 3 can both use column 10 and voice 2 and voice 4 can both use column 14 without ever touching, since the center channel keeps the banks apart. Do not "fix" one of them onto a spare column.

How it works, in one breath

A Schmitt trigger flips its output when its input crosses one threshold and does not flip back until the input crosses a different, lower one - and that gap is all you need to make an oscillator. Tie the output back to the input through a resistor and hang a capacitor from the input to ground: the output charges the cap through the resistor until it hits the upper threshold, flips, discharges it until it hits the lower one, flips back, forever. The resistor is the pot with its 1K floor, so the knob is the pitch. The capacitor is the other half of the same equation, which is why four different caps give four different registers from one chip. The second input of each NAND is the mute: a NAND with one input held low has its output stuck high no matter what the other input does, so grounding it stops the voice dead - and the 1M resistor is what holds it high the rest of the time. Each output then goes through 10K onto one shared bus, which is the entire mixer, and the 10 µF at the end blocks the DC so your amp only sees the wiggle. The volume pot is the last thing in the chain and the simplest: the whole signal sits across it from top to ground, and the wiper picks off however much of it you want.

Things to try

All single parts, and none of them have been tried on this build.

• Try an A100K for the volume instead of the B100K. An audio taper spreads the useful part of a volume control across the whole sweep; a linear one does most of its work in the last quarter turn. The B is in the parts list because it is the same part as the other four and that keeps the list short, not because it is better here.
• Swap the timing caps. Anything from 100pF to 10 µF does something. Four caps of the same value is a different instrument entirely: four voices that nearly agree, beating against each other.
• Make a mute switch a momentary instead of a toggle, and that voice becomes a key you play rather than a drone you allow.
• Feed one voice into another's mute. Run a wire from a gate output into another voice's enable hole instead of its switch, and the slow voice starts chopping the fast one. Voice 4 at 1 µF is the obvious one to chop with. This is the trick the Locust is built on.
• Give it a proper output. An LM386 Hornet stage and a speaker and it needs nothing plugged into it at all.
• Solder it down. The stripboard Cicada NAND is the same board on 18 by 20 holes, built and working, with a drill template for a 125B box. It has no volume pot - add this one to it the same way, off the board between the output cap and the jack.

This one is mine, and it is the breadboard translation of a board I have built four times. Nobody has built this version yet - when I have, this line will say so. Build it, bend it, make it yours.