Sound Fuckulator

Paul Stevenson's LM567 noise box, traced off his drawing · By Onion Madder · Mess O' Pedals

The original lives here, published July 2012: Sound Fuckulator on Paul In The Lab. Go and read his, it is three sentences and it is better than mine.

An LM567 tone decoder is a chip built to listen for one frequency and close a switch when it hears it - touch-tone dialers, remote controls, that sort of thing. Paul Stevenson pointed one at a guitar instead. It half-hears the note, the switch chatters against what you are playing, and a 2N2222 stage turns the chattering into something you can put in a song. His own verdict, in 2012, was that it makes a guitar sound nasty and horrible, and he meant it as a recommendation. This page is his board, traced hole for hole, with the two connections it needs to actually run - built here, and it does.

The Sound Fuckulator layout drawn as a stripboard diagram, titled Sound Fuckulator, 14 by 11 stripboard. Eleven orange copper strips run left to right, lettered a to k down the left side, with fourteen columns numbered above them. A pale square marked IC1 LM567 sits in the middle of the board across four strips. Around it are small green film capacitors marked C3, C4, C6, C7 and C8, pink electrolytics marked C1, C2 and C5, three outlined resistors standing upright marked R1, R2 and R3, and a small brown transistor outline marked Q1 with its legs lettered C, B and E down the strips beside it. Four dark blue vertical wire links cross the board, two long ones near the middle and right and two shorter ones. Gray X marks show every cut strip. Holes carrying wires off the board are labeled down both margins: VR3 25K, VR12 25K, 9V and IN on the left, and GND, VR3 100K, VR12 100K and OUT on the right. A parts list runs underneath and a credit line reads Made with Copper Bottom. Below it the same board is drawn a second time from the solder side, mirrored, with the column numbers running right to left, every solder point as a dark dot, the cuts as X marks, the chip as a dashed outline with pin 1 boxed and the pads labeled on the margin they now sit nearest. — The board as Copper Bottom draws it: the component side on top, then the same board from the solder side with the columns running the other way, so a cut is the same hole in both pictures. Everything here is Paul's except the two links described below.

The finished board lit up on a dark desk mat, with the rounded edge of a pink keyboard along the top of the frame. A small square of stripboard sits in the middle carrying a black eight-pin chip, a black electrolytic can standing up, two dark green film capacitors and several yellow ceramic capacitors, with the printed row letters and column numbers of the sheet it was cut from still readable on the copper. Wires in orange, black, blue, white, yellow, gray and green run off it in every direction. Two potentiometers sit above the board on flying leads, bare shafts with no knobs, each wired on two lugs with the third empty, and a small blue toggle switch stands between them. A larger toggle with a red body sits lower down beside the power lead. A yellow DC socket at the left has a barrel plug in it, a jack below the board takes a black right-angle plug, and a small black jack at the right has a plug in it. An LED on two long leads lies at the bottom left, lit. Nothing is in a box. — Built and running, with the rate LED lit at the bottom left - that is the one on pin 5, and it blinks rather than sitting on. Two pots, the octave toggle in blue, two jacks and the supply, nothing boxed. **The second, red toggle is not in the layout file** and what it switches is not stated here.

What you'll need

tap to expand / collapse

Fifteen parts on the board, counting the links, and it is a cheap drawer-clearing build. The only thing worth reading twice is the supply, and there is a note about it further down.

1× LM567 tone decoder, DIP-8 (IC1) - spot it: eight legs, LM567 or NE567 printed on top, a notch at one end. The NE567 is the same part; so is the LMC567, but that one runs at double the oscillator frequency, so it will not sit where this one does
1× 2N2222 NPN transistor (Q1) - spot it: three legs, either a small black half-moon or a little metal can. Check your own part's pinout against the board - the drawing letters the legs C, B and E down strips h, i and j, and the TO-18 metal can and the TO-92 plastic ones are not all lettered the same way around. The board's letters are what to trust
3× 100K resistor (R1, R2, R3) - spot it: bands brown-black-yellow. R3 is Q1's collector load, R1 and R2 are the bias pair that also turn out to be the load resistor the chip's output needs
2× 10uF electrolytic (C1, C5) - spot it: a little can with a stripe down the minus side. C1 is the input coupling cap, C5 sets the frequency with whatever resistance ends up on pin 5
1× 100uF electrolytic (C2) - spot it: the biggest can in the bag. This is the chip's output filter and it wants to be large
1× 180nF film capacitor (C4) - spot it: marked 184. An odd value; 150n or 220n will both do something, just not quite the same something
1× 100nF film capacitor (C3) - spot it: marked 104. The loop filter, which is what sets how wide a band the chip will answer to
2× 10nF ceramic (C7, C8) - spot it: marked 103
1× 220pF ceramic (C6) - spot it: marked 221. The smallest cap here and the easiest to grab the wrong one of
Stripboard, 14 holes by 11 strips, and five wire links

Off the board

2× jacks for in and out
A supply between 3.5V and 8.5V. Paul's pad says 9v. Read the supply note below before you wire a fresh battery to it
Two pots and an SPST toggle. Paul prints three pots at 25K and 100K; the build here is 5K and 5K plus a switch where his third pot goes, and the switch is a sub-octave, one or two octaves down depending on what you tie it to. The timing pot's value decides whether this is a pitched oscillator or a gate mangler. See the pots
An LED and its resistor, optional and not on Paul's drawing. It goes on pin 5 and nowhere else - see where the LED goes
An SPST toggle for power, in the supply lead between the DC jack and the 9V pad. Not in the layout file either, but this board never stops making noise, so a way to switch it off is not optional in practice
Wire, and a box if it earns one

The layout file is the guide

Like every layout on this shelf, the build document is the file itself: ↓ Download Sound_Fuckulator.json - import it into Copper Bottom and the editor gives you the board picture, the placement walkthrough, the electrical checks and a printable build sheet, all from the same file. No coordinates are written out on this page on purpose: the editor is the viewer, and a number typed twice is a number that can drift.

There is a second file, and it is there for honesty rather than for building: ↓ Download Sound_Fuckulator_AsPrinted.json is Paul's drawing transcribed with nothing added, so you can see exactly what was published and judge the two additions for yourself. Import it and Copper Bottom will tell you the same thing it told me: IC1.8 is the only thing on its net - connects nowhere. If you would rather just look at it than import it, here is the same file drawn out: ↗ Paul's board as printed, drawn.

The two links, and why

Paul drew this straight onto stripboard in 2012 and was upfront that there was never a schematic behind it - asked for one in the comments two years later, he said he made this from my head and had not bothered. So the drawing is the whole document, and when the drawing is the whole document a missing wire has nowhere to hide. Two nets on his board have no way of reaching the rest of it.

Neither can be fixed off the board, which is what makes this a certainty rather than an opinion: neither net carries a pad. No pot, jack or flying lead can land on them however the pots turn out to be wired. The only thing that can reach them is a link on the board. Both have been added and nothing else was moved.

J4, column 13, strip c to strip g. The chip's output had nowhere to go. Pin 8 sat alone on its own stub of strip c. It is an open-collector output, which means it can only pull down and needs a resistor pulling it up - TI's datasheet says in as many words that it requires an external load resistor. Meanwhile the node at strip g - R1 to the transistor's base, R2 to its collector, C7 to the rail - is a textbook bias network with nothing driving it. Land pin 8 on strip g and both problems answer each other: the output gets its 200K pull-up through R2 and R3, and the transistor gets something to react to
J5, column 2, strip a to strip b. The b and k rail is ground. C2 hangs off pin 1, the output filter, and that cap has to go to ground. C3 is daisy-chained from pin 2 across to pin 1 rather than run to the rail, which works only because a 100uF holds pin 1 at ground as far as any audio signal is concerned - and the datasheet's own rule, that the output filter should be at least twice the loop filter, fits Paul's 100uF against 100nF exactly. C6 then becomes a plain rolloff from the transistor's base to ground. Three separate parts want that rail to be earth, and his drawing never joins it to the earth he already has

None of this is a criticism of the drawing and it should not be read as one. A layout drawn from the head, published free, and left up for fourteen years is a generous thing to do, and the circuit underneath it is his. Two wires is a remarkably short list. If it turns out he wired those two nets some other way on the board he actually built, his answer wins and this page changes.

The pots, settled on the bench

Paul labels four holes for pots: VR3 25k on the ground rail and VR3 100k on pin 6, VR1,2 25k on C4's far end and VR1,2 100k on pin 5. Three pots, four labels, each name paired with two different numbers, and no schematic anywhere to check it against. This page shipped saying nobody had settled it. It is settled now, and the answer is that both readings of those labels are right at the same time.

Pair the holes by value and you get one pot across the two 100k holes, pin 5 to pin 6, and one across the two 25k holes, C4's far end to ground. Pair them by name and VR3's two holes give you a third pot, pin 6 to ground. Those look like competing readings and they are not, because pin 6 and the ground rail each take two pot legs. With that one allowance all three pots land on Paul's four holes, every label is used, and nothing is invented.

Control	Between	What it does
VR1,2 100k printed	pin 5 to pin 6 (PIN5 to PIN6)	The timing resistor, and the one control that has a clean name. At 100k it is a DEPTH knob and at 5k it is a PITCH knob - see below
GRIND Paul's VR3 100k and 25k printed	pin 6 to ground (PIN6 to GND_A)	Sits across the timing capacitor and pulls the timing node's DC level, so it shifts the frequency and how hard the oscillator runs at the same time. It fights the timing pot, on purpose
VR1,2 25k printed fit an SPST toggle	C4's far end to ground (IN_S to GND_A)	A SUB-OCTAVE. What this hole is tied to decides how far down the output sits: ground is normal, the input jack's sleeve is an octave down, the input jack's tip is two octaves down, which takes it from a sort of chirping to a growl. Paul prints a pot here and a pot is the wrong part - see below

The timing pot is two different instruments depending on its value, and that is the most useful thing on this page. Free-running frequency is fo = 1 / (1.1 × R × C), with R from pin 5 to pin 6 and C5's 10uF on pin 6. At Paul's printed 100k that lands between 0.9Hz and 40Hz - every setting is below hearing, so you never hear the oscillator, only what it does to the gate, and the knob reads as depth. Swap it for a 5k and the range becomes 18Hz to about 190Hz, the oscillator itself comes into earshot, and sweeping the same knob gives you a rising tone. Paul's value is the mangler. The small one is the synth.

So the build here is two pots and a toggle: 5k for the timing pot and 5k on pin 6, matched so neither dominates - at 100k against 5k the pin 6 pot wins everywhere on its travel - and a switch on the third hole pair. No knob map is given for the pin 6 pot, because there is not an honest one. It differentiates the crunch rather than owning one parameter, and with three controls leaning on the same loop a knob that does nothing at one setting is the one that matters at another. Fit them, turn them together, and letter your own box.

The third control is a switch here, not a pot, and the arithmetic says why. C4 is 180nF, which is about 900 ohms at 1kHz, and a pot between that hole and ground is competing against it. So the whole crossover happens down in the first kilohm, and on a 25k pot that is the bottom four percent of the rotation - one parked spot that works and three-quarters of a turn that does nothing. The two ends are the two sounds and the middle is just the handover, so an SPST toggle across the same two holes gives you both and throws away nothing.

And there are three states here, not two, because it matters what you tie that hole to. Ground is the ordinary sound. The input jack's sleeve puts it an octave down. The input jack's tip puts it two octaves down, and that is the big one - it takes the box from a sort of chirping to a growl.

IN_S tied to	Output	What it costs you
ground (GND_A)	normal	nothing
input jack sleeve	one octave down	an isolated input jack, and an isolated supply output if anything sits upstream of it
input jack tip	two octaves down	nothing

The sleeve option is the one that constrains the box. A normal metal jack bolted into a metal enclosure has its sleeve on ground already, so the sleeve and ground settings collapse into the same thing and the one-octave step disappears. Getting it back means an isolated input jack - plastic bodied, or nylon shoulder washers - and even then, any pedal in front of this one on a daisy-chained supply re-grounds the guitar's shield through the shared power ground. Two robust settings, or three fragile ones. The tip wiring needs none of that, and it is the bigger effect anyway. Same shape as any bridged amp, where SPK- is not ground and its jack has to be insulated from the box.

If you want all three, the part is a 1-pole 3-position rotary or an SP3T wired ground / sleeve / tip, which is a sub-octave selector on one knob. That is a suggestion and nobody has built it; the two-position version is what is on the bench.

Why octaves and not some other interval is the interesting part. Pin 1 is the output filter, the node that decides how quickly the detector is allowed to change its mind. Push the input into it and you move that threshold far enough that the chip stops completing a decision on every cycle and starts completing one on every other cycle - and an output that fires half as often is, exactly, one octave down. A divide by two that falls out of the topology rather than being designed in.

That started as a reading of the datasheet and then the bench tested it by accident. If the mechanism is cycle-skipping, then pushing harder should skip more cycles and land on the next octave rather than somewhere in between. The tip carries the actual signal where the sleeve only carries return current, so it is a much harder push - and it comes out at two octaves, a divide by four. Exact octaves are the tell: a shifted threshold or a filter would have moved the pitch somewhere arbitrary, and only integer division gives you intervals that land on the nose. It is still a reading rather than a scope trace, but it made a prediction before it was tested and the prediction held.

That is a departure from Paul's drawing, which prints a pot there, and it is worth saying that he got to the same answer first on his other board. His verdict on the Devil's Triangle's channel-level pots was that they are useless and you should fit switches instead. Same builder, same shelf, same call, arrived at twice independently. Fit a pot here if you want to hear the crossover for yourself; you will end up parking it.

Put a small resistor in series with the pin 6 pot. At a hard zero it shorts pin 6 to ground and the oscillator stops dead. A 470R or 1k keeps the end of the sweep alive instead of falling off a cliff. That is ours, from the bench, not his.

Where the LED goes, and why it is the only place

There is no LED on Paul's drawing. If you want one, it goes on the PIN5 pad - pin 5 - with its resistor, down to ground, and it blinks at the oscillator rate rather than sitting on steadily. That makes it a readout of the depth knob instead of a power light, and at the slow end of the sweep you can watch the thing think.

Every other pad was tried and pin 5 is the only one that both lights and costs nothing. That is a bench result, and the reason it comes out that way is worth knowing before you experiment on your own board: pin 5 is the only real output on this chip. TI's own oscillator circuit takes a square wave off it through a load resistor to ground, so it is built to drive something. Everything else here is either a supply rail or a high-impedance timing or filter node - pin 6 holds the timing capacitor, IN_S feeds the output filter through C4, IN and OUT sit behind coupling capacitors - and a few milliamps through any of those does not color what is on them, it erases it.

This is a general rule about indicator LEDs, not a fact about this board. The Analog Bit Crusher on this shelf had exactly this fault the other way around - its LED landed on the sample-and-hold node, where a few milliamps off a 1nF cap deletes the held voltage rather than degrading it. Copper Bottom cannot catch it either, because electrically it is a perfectly good connection. Put an indicator on an output, never on a node that is storing something.

It is a synth that takes a signal, not a pedal

This board makes noise with nothing plugged into it. Turn your source off and it keeps clicking and burbling away to itself. That is not a fault and there is no wiring that fixes it: the LM567 is an oscillator with a detector bolted on, the oscillator free-runs whether or not anything arrives at pin 3, and with no signal to lock to the detector simply flops about. Pin 8 chatters, Q1 passes the chatter on, and that is the output. Nobody asks a drone box to stop droning, and that is the family this is in.

Boxed, it turns out to be two instruments depending on what is in the jack, and that is worth knowing before you decide it is broken. Her words: if theres an input plugged into the jack it works immediately as a signal distortion. if you turn it on with nothing plugged in, it takes forever to start up but it makes noise on its own. Both of those are correct behavior.

Watch it wake up, with nothing plugged into the input. The LED comes on the moment the power toggle goes over, and then nothing happens for a while - the first audible sound arrives about five seconds later and builds from there. Timed off this clip: the LED lights at about 3.5 seconds and the first sound at about 9, so the gap is five and a half. 22 seconds, sound on.

Measured from cold it is about five seconds, and the LED says where those five seconds go. That LED sits on pin 5, the oscillator's own output, and it comes up the instant you switch on - so the VCO, the timing pots and C5 are all working from the start, and the whole wait is on the detector side of the chip. It was fitted as an indicator and turned out to be a test point.

Two capacitors on that side could do it and the page will not pretend to know which. C2, the 100uF on pin 1, is the output filter and has to charge before pin 8 will move at all. C1, the 10uF on pin 3, has its outer plate reaching nothing but the jack's own insulation resistance when the jack is empty. C2 explains the other half better - the output filter charges from the detector's own response, so a real signal drives it hard and crosses in milliseconds while noise alone crawls, which is exactly why it plays at once with something plugged in. One test separates them: fit a 1M from the IN pad to ground and try again from cold. If the five seconds collapses it was C1's floating plate; if it stays, it is C2 and the resistor is just good practice. That resistor is an ordinary input pull-down - the Clean Boost on this shelf has one - and it is not in Paul's drawing. Nobody has run the test yet.

If this feels familiar it is because it is. Her verdict on the wake-up: its the same weird tweak the mfos wsg has. Ray Wilson's Weird Sound Generator is on this site too, and a slow reluctant wake-up is apparently just something a certain kind of noise box does.

There is a second, shorter wait that happens either way. C2 is the reason. That 100uF works against the 4.7k the chip has inside it on pin 1, which is a 0.47 second time constant, and nothing settles in one time constant - four or five of them is two to two and a half seconds before the output filter has charged far enough for the detector to do anything. That is the couple of seconds you get even with a source plugged in. It is the same capacitor that makes the slow end of the depth knob lurch about, so the startup wait and the good sound are one part and you cannot have one without the other. Drop C2 and it will come up faster and settle harder; that trade is untested.

The depth knob is also the how-much-does-it-burble knob, because it is one mechanism doing both jobs. At the slow end the output filter and the center frequency are the same order and the loop can never settle, which is the sound you built this for and the idle noise, at once. Wind it fast enough for the chip to decide cleanly and it does go quiet, and it also stops being interesting.

So you can have off-and-silent, but not on-and-waiting. A true-bypass footswitch disconnects the board from the output jack, so bypassed it really is silent - engaged is just never transparent.

Which makes the honest description of this thing a synth that can take a signal, rather than a pedal. The closest relative on this shelf is the Dragonfly VCO once it has been bent about: it runs on its own, and an input is something it will chase rather than something it processes.

That family resemblance is not a loose one. The Dragonfly is a CD4046 and this is an LM567, and both are phase-locked loops - a free-running oscillator with a detector that will lock onto an incoming signal when it can and hunt around when it cannot. The Dragonfly's guide describes its IN_T pad as what the loop chases in its second mode, and that is this whole circuit's job description. Paul pointed his at a guitar and called the result nasty, and he was right, but what he built is a small synth with an input on it.

Boxing it up

The finished pedal standing upright on a desk, a glossy purple enclosure with SOUND and FUCKULATOR hand-lettered across the face in pink marker, one word above a lit indicator LED and one below it. Two black knurled knobs with white pointer lines are stacked down the left of the face and two metal toggle switches down the right, with three jacks along the bottom edge carrying a right-angle power plug and two cables. Beside it on the left sits a cream-colored box of the same shape lettered opto duet in pink, its own indicator lit, with a toggle on its top edge and a cable running from it to the purple box. A pink keyboard fills the background. — Boxed in a 125B stood on end, drilled from the template above. The face is the sheet exactly: the rate LED between the two words, PITCH and the octave toggle in a row, the second pot and POWER below them, and the three jacks along the bottom. Feeding it here is the Opto Duet, which is what it was tested on all the way through.

The enclosure is a 125B stood upright, which is the format the whole synth board uses here - boxes stand shoulder to shoulder and get rearranged, so everything is on the face and nothing is on a wall. Twelve millimetres of lid lip at each side leaves 42 mm to drill in, which is what makes it two columns and no more.

Don't eyeball the holes. There is a 1:1 drilling template for the 125B - the box drawn unfolded, so you wrap the sheet around it and every hole lands by folding instead of by measuring. Face plus four walls, true-size circles with a punch dot in the middle. The walls are blank; everything is on the face.
↓ Same box, jacks on top (PDF)
The second sheet keeps every control where this one has it and moves the jacks - power and audio - to the top wall, so the cables leave upward. Two ways to drill one box; pick one.
It is drawn as a synth, not a pedal. 3.5 mm in and out like the rest of this shelf, and no footswitch - the board is never silent, so a bypass could only ever be an on switch, and it is not in the layout file either.
Eight holes: the rate LED at the top, PITCH and the octave toggle in a row, the second pot and the power toggle below them, and DC, IN and OUT across the bottom. The octave toggle here is the tip wiring, so both jacks ground to the box in the ordinary way and nothing has to be isolated.
POWER and the LED are off-board, not in the layout file. They are on the sheet because they are in the parts list above, which is the rule this shelf follows - a template is drawn from the layout file plus the page's own off-board list, never from a box somebody built with extras on it.
The second knob is lettered GRIND - it ran as VR3, Paul's own designator, until it was named. It pulls the timing node's DC level against the timing pot, so it shifts the frequency and how hard the oscillator runs at once. Letter it something else if your build says otherwise.
Print at 100% / “Actual size”, never “Fit to page”, and check the scale bar before you drill anything.

The supply, before you plug anything in

The pad says 9v and the LM567's absolute maximum is 9V. TI puts the recommended supply at 3.5V to 8.5V and the absolute maximum - the number past which damage may occur - at 9V flat. A fresh alkaline block sits at 9.5 to 9.6V, and there is no regulator and no series diode anywhere on this board. Paul built his and it worked, so this is margin rather than a wall. Two ordinary silicon diodes in series with the incoming positive drop about 1.4V, land you near 8.2V off a fresh cell, and give you reverse-polarity protection for nothing. The 9V and the 3.5-8.5V are the datasheet; the two diodes are a suggestion.

The netlist

Quoted verbatim from the layout file. It was written from the board, so checking the board against it can only pass - what it is for is carrying the circuit somewhere else, and telling you what each strip is supposed to be doing while you stare at it.

# Sound Fuckulator - Paul Stevenson LM567, traced from his stripboard drawing # 11 strips of 14 holes, 17 parts, 1 chip, 8 cuts # # LM567 pins: 1 OF_CAP, 2 LF_CAP, 3 IN, 4 VCC, 5 T_RES, 6 T_CAP, 7 GND, 8 OUT. # Pin 4 lands on the 9v strip and pin 7 on the GND strip, which is what fixes # the chip the way around it is drawn. # # CORRECTED. J4 carries pin 8 out to the g node so the open-collector output # has somewhere to go and R2 + R3 become its load; J5 ties the b/k rail to the # ground rail, which is what C2, C3 and C6 all need it to be. Nothing else on # the board moved. 9V: @9V IC1.4 R3.A C7.A GND: @GND_A @GND_B IC1.7 C5.B Q1.E C2.B C6.B OF_CAP: IC1.1 C4.A C2.A C3.A LF_CAP: IC1.2 C3.B IC_IN: IC1.3 C1.A T_CAP: @PIN6 IC1.6 C5.A T_RES: @PIN5 IC1.5 SLEEVE: @IN_S C4.B IN: @IN_T C1.B BIAS: IC1.8 C7.B R1.A R2.A COLL: R3.B R2.B C8.A Q1.C BASE: R1.B C6.A Q1.B OUT: @OUT C8.B

How the chip sits is not a judgement call. Paul drew the pin 1 marker at the top left, but you do not have to take that on trust: pin 4 is the supply pin and it lands on the strip carrying the 9v pad, and pin 7 is ground and it lands on the strip carrying the GND pad. Only one orientation does both. If your DIP goes in the other way around you will find out immediately and expensively.

The off-board wiring

Eight pads leave the board: two jacks, the supply and the four control connections.

Pad	Goes to
IN	Input jack tip. It reaches pin 3 through C1. The sleeve does NOT go to GND on this build - it goes to IN_S, which is what the octave toggle switches. See the controls
OUT	Output jack tip. The sleeve goes to GND. It comes off Q1's collector through C8
9V / GND	The supply, positive to 9V and negative to GND. Read the supply note above first - the chip's absolute maximum is 9V and a fresh battery is over it. A power toggle goes in the positive lead, between the DC jack and this pad
PIN5	Pin 5, the timing resistor pin. One leg of the timing pot; its other leg goes to PIN6. An LED and its resistor go here too, to ground, and it blinks at the rate - see below
IN_S	Reaches pin 1 through C4, so whatever lands here is coupled into the output filter rather than connected to it. The input jack's sleeve goes here, and so does one lug of the toggle; the other lug goes to GND_A - see the pots
PIN6	Pin 6, the timing capacitor pin, where C5 10uF already sits. Two pot legs land here: the far end of the timing pot, and one end of the pot that runs down to GND_A
GND_A	This strip is the ground rail, which is worth knowing before you wire anything to it. Two legs land here: the pin 6 pot's far end, and the toggle's other lug from IN_S

Built, and heard

It was built on 18 September 2026 off the corrected file, bare on the bench, fed from an Opto Duet rather than a guitar. It works, and the build answered three things the drawing could not.

The added link to pin 8 is real. That was the bigger of the two corrections on this page and it was an argument until the board existed. On Paul's drawing as printed, pin 8 sits alone on its stub and reaches nothing, so Q1 would sit saturated and no pot anywhere could change it. Sweeping the timing pot changes the sound, which means pin 8 is driving Q1, which means the link belongs there. The as-printed file is still on this page to download and compare, but it is a record of the drawing rather than a thing to build.

J5 is NOT proven the same way, and the difference is worth being straight about. The board could not work at all without J4, so building it settles that one. J5 grounds the b and k rail, and the build is only consistent with it - a board that works does not show that rail had to be earth, only that earthing it did no harm. The argument for it above is read off the datasheet and off what C2, C3 and C6 each need, which is strong but is still a reading. Paul is the one person who could settle it outright.

The controls are answered, above, and all three land on his own four holes. And the input return is a control, which is the surprise. Taking the input jack's sleeve to ground in the ordinary way is one sound; taking it to the IN_S hole instead, so the return runs through C4 into pin 1, is the other, and the output sits an octave apart between them. Take it off the jack's TIP rather than its sleeve and it drops two octaves instead. Both holes are Paul's, so the third control is a switch across them. He prints a pot there and the useful part of a pot there is its two ends, so the switch is what the board wanted.

How it works, in one breath

The LM567 is a phase-locked loop with a detector bolted on. Its oscillator free-runs at whatever R on pin 5 and C on pin 6 tell it to, your guitar arrives at pin 3 through C1, and when the two are close enough the loop locks and pin 8 pulls down. C3 on pin 2 sets how close is close enough; C2 on pin 1 decides how quickly the output is allowed to change its mind. Feed it a guitar rather than a dial tone and it spends its whole life half-locking, losing it and grabbing again, which is the sound.

Pin 8 can only pull down, so it hangs on the bias node of the transistor stage - R1 to Q1's base, R2 up to its collector, R3 from the collector to the supply, C7 holding the middle of that divider steady. Q1 is an ordinary common-emitter amplifier with collector-feedback bias, and the chip yanking on its bias is what turns a lock-and-slip into a gated, spitting output. It leaves through C8.

Those two paragraphs are read off the datasheet and off the copper. The board has been built and heard since, and nothing in them turned out to be wrong - but the mechanisms are still a reading of the chip rather than something measured on it. Paul's own verdict is the only other first-hand account, and he only ever said it was nasty.

Things to try

A toggle on C5 gives you both instruments. Leave C5 at 10uF and put a second 100uF in parallel with a toggle in its ground leg, positive toward pin 6 like the first one. Open, a 5k timing pot sweeps 18Hz to 190Hz and you hear pitch; closed, the same knob sweeps 1.6Hz to 18Hz and you are back in Paul's mangling range. Switching the cap works where switching the resistor does not, because a series resistor big enough to reach the slow end would swamp a 5k pot and leave it almost no travel. Untested - it is arithmetic and two parts
Swap the timing pot and find your own value. 100k is Paul's and it is all gate; 5k is pitch; 25k sits between them. It is the one part on this board that changes what the instrument is
Feed it something that is not a guitar. A tone decoder pointed at a drum machine, or at another oscillator, is a different instrument. The Dragonfly VCO makes a clean square that this thing should lock to properly, which means you can hear the difference between locking and failing to
A Eurorack version. This is a phase-locked loop with a free-running oscillator and an input it will chase, which is a module rather than an effect, and the Eurorack shelf already has the pattern: the built board rides inside the module hole for hole with a power header, a 78L05 and level shifting bolted on. The house rule is that a board earns its rack version once the standalone one is verified, and this one now is

The circuit and the original layout are Paul Stevenson's, published on Paul In The Lab in July 2012 and free to build ever since. The same hands drew the Devil's Triangle, which is also on this shelf and has been built here twice. What is mine is the trace, the two added links, the checking and the words. The name is his and it stays.