UNVERIFIED. This module is drawn and checked in Copper Bottom - the copper agrees with the netlist below, and the checks come back clean. The chip, its timing cap and its decoupling are the Dragonfly VCO's, which I have built and heard. The exponential converter is a new circuit, and how well it tracks is exactly what nobody knows yet. The rack block has not been built. Every number on this page is a datasheet number, not a measurement. Build it once, fix whatever fought back, and this banner comes off.

Dragonfly Rack

A 4046 that tracks a keyboard · The Dragonfly VCO, re-wired for 1 V/oct · 6 HP · By Onion Madder · Mess O' Pedals

The Dragonfly VCO is a CD4046 played by a knob. A rack wants it played by a voltage, and not any voltage: one volt per octave, the rule every keyboard and sequencer in a Eurorack case agrees on. A 4046 is linear - twice the volts on its control pin is roughly twice the frequency - so the knob-and-pin-9 way of playing it had to go, and in its place is the classic two-transistor exponential converter feeding the chip's current-setting pin. Pitch knob, 1 V/oct jack, drop switch, out. Six HP.

What is the Dragonfly, and what is new

What is kept, hole for hole: the 4046 at its place, C1 (the 10n timing cap), C2 and C3, and the links J1 to J6. What came off, and all of it is the linear control path that 1 V/oct replaces: the range pot and its 1K on pin 11, the pitch pot's 100K into pin 9, the mode toggle and the PLL chase mode (pin 9 has to be free for that, and it is now tied to the 5 V rail), the signal-in jack, and the volume pot with its output cap. Three links tie up what that leaves - pin 9 to +5, pin 14 to ground so the unused comparator input never floats - and pin 11's strip runs into the block, where the converter sits.

The converter. Half of the first TL072 adds up the 1 V/oct jack and the pitch knob, 100K each, with a 5K trimmer as its feedback, so its output falls by about 18 mV for every volt in. That goes to the base of Q2. Q1's base sits on ground, the two emitters are tied together, and the other half of the chip drives them: it holds Q2's collector at a virtual ground with a 4M7 from +12 setting a reference current of about 2.5 µA. Raise the control voltage by a volt: the summer drops 18 mV, the servo lowers both emitters 18 mV, Q1's base-emitter voltage rises 18 mV, and its collector current doubles. Q1's collector is pin 11. One volt, one octave. The trimmer is the scale adjustment - a tuner and a keyboard set it - and no temperature compensation is drawn, so it will drift with the room the way a first expo VCO does.

The drop bend was a UV LED and a toggle into the range pot's floor node, and there is no range pot now. The DROP_6 and DROP_7 pads sit on the timing cap's two strips: a second 10n and a toggle in series between them drops the whole range an octave. Same switch on the panel, a straighter route.

The top of the range. At the top of the sweep the converter asks pin 11 for more current than the 4046's own internal follower can pass, and the pitch flattens out. That is the ceiling, not a fault. Where it lands is a bench number; so is the bottom, which the 4M7 sets.

The layout file is the guide

As on every stripboard guide here, the build document is the file itself: ↓ Download Dragonfly_Rack.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.

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. The bug board's nets keep the names they have on its own page.

The off-board wiring

20 pads leave the board. Ten of them are the power header.

The panel

6 HP, which is 30.5 mm of a 128.5 mm blank - Amplified Parts sells reversible black-and-aluminum blanks in every width from 2 to 20 HP. The template is drawn for 9 mm Alpha pots (7 mm holes), Thonkiconn jacks (6 mm), sub-mini toggles (6.5 mm) and 5 mm LEDs, with the rack's rails kept clear top and bottom and the four M3 mounting holes where Doepfer puts them. I/O along the bottom, controls above, as everything on this site does it.

The rack block, the same on every module

The right-hand end of every board on this shelf is the same idea drawn six times. The 2×5 header takes the rack's power cable, mounted with its pins running down the strips so each pair of pins lands on one strip: -12 V, then three grounds, then +12 V. That is the only way it can go on stripboard - turned the other way, the -12 and ground pins would share copper. The red stripe on the cable goes to H1 and H2. A 1N5817 stands in series with each rail so a cable put on backwards does nothing instead of everything. A 78L05 then makes a 5 V rail for the chip, and a TL072 on the real ±12 V does the rack-level work.

Five volts for the CMOS, not twelve, is what makes these rack-polite for free. A CD4000-series input switches near half its supply, so a +5 V gate from a sequencer clocks a 5 V chip directly but would never reach the ~7 V a 12 V chip wants; a pot across 5 V makes exactly the 0 to 5 V control voltage every rack sequencer speaks; and a hot or negative rack signal into a CMOS input needs only a 100K in series, the chip's own input diodes doing the clamping at well under a tenth of a milliamp. It is also a change from the 9 V these boards were built at - every chip is rated down to 3 V so the circuits are fine, but the oscillators will sit at slightly different pitches, and that is one of the things the bench is for.

The IDC socket on the power cable overhangs the header by about two strips at each end. Nothing tall stands there - the generator that drew this refuses to put anything standing under the socket - so keep it that way if you move a part.

How it works, in one breath

A 4046's oscillator charges its timing cap with a copy of whatever current flows out of pin 11. On the Dragonfly that current came from a resistor - the range pot - so the knob set the pitch and pin 9 nudged it. Here pin 9 is parked at the rail and the current is sunk by a transistor whose collector current doubles for every 18 mV added to its base: an exponential. A second, matched transistor and an op-amp make that exponential start from a known current instead of whatever the first transistor felt like, which is the whole reason there are two. The output stage is separate and simpler: the square off pin 4 through a cap and a 10K into an inverting stage with 22K feedback, so 0-to-5 becomes about ±5, out through 1K.

Things to try

• Set the scale. Feed the jack 1 V then 2 V from a keyboard or a sequencer and turn the trimmer until the interval reads an octave on a tuner. Do it warm
• Fine tune. A second pot into the summer through a 1M gives a few semitones of fine pitch. The summer does not care how many inputs it has
• Tempco. If the drift bothers you, the 5K trimmer's job can be done by a 1K PTC thermistor glued to the transistor pair, the standard fix
• Triangle-ish. The voltage across the timing cap is a ramp; a buffer on pin 6 or 7 through a cap gives a second, softer waveform. Unmeasured, untried

The oscillator is the Dragonfly VCO layout from this site. The exponential converter is the textbook two-transistor converter with an op-amp reference that every analog VCO since the 1970s carries in some form; feeding it into a CD4046's pin 11 rather than into a discrete integrator is the part that is ours, and unproven. Drawn in Copper Bottom.