Box Dulcimer (Version 1.1)

# Printed Box Dulcimer

A fretted diatonic box zither, played on the lap with a noter. Four

strings — a doubled melody course and two drones — tuned D A d d, on a

550mm scale.

Mostly printed. The parts that carry load are wood: two 12mm oak

dowels run the whole length and take all the string tension, so no

printed part is under sustained pull. A few small pieces of metal sit

wherever a string bears.

Sound. The printed body is fully acoustic. A PLA top doesn't resonate as much as a wooden one, but it's loud enough to hear the melody and drones together, clearly and pleasantly, when playing at home. A stick-on contact pickup can be added for recording or playing through an amp, but it isn't needed to play.

**The body is three closed printed tubes**, each with its top, floor and

both walls in one part, joined by printed sleeves and threaded onto the

spine. There is no separate soundboard, no shell, no lid and no rebate —

they merged. Each tube prints standing on one end, so every surface that

runs the length is vertical and nothing needs support.

It descends from a scheitholt, which is why the body tapers — 118mm at

the head to 148 at the tail — and why two of the four strings drone. The

frets run the full width of the board, so all four courses can be

stopped.

**The tubes are the answer to a version that failed.** The design before

this one was a tray with a lid: the shell was glued up early and the top

only met it near the end, so there was no stage at which the whole box

could be offered up and taken apart again. A complete shell was printed

and never assembled — not because a number was wrong, but because parts

that were individually correct were collectively too inexact and too

hard to reach to bring together. Everything below about sleeves bonded

on one half, witness grooves and a dry run exists because of that.

---

Version 1.1: There was an issue with the action being too low on the dulcimer, so I increased the height of the epoxy+fretwire capped nut. Also if you have issues still, check that all fretboard screws are fully screwed down.

I have also published a Stick Dulcimer, which you may also enjoy making: https://www.printables.com/model/1873318-stick-dulcimer

## Read this before you print anything

Two things about your own materials decide whether the STLs will work

as they are.

### 1. Your tuners

Six individual open-gear tuners, three left-handed and three

right-handed. Four are used and two are spare.

**The printed head fits tuners that satisfy all of this:**

| | |

|---|---|

| Plate length, along the instrument | under **34mm** |

| Plate reaching above the axle | under **12mm** |

| Plate reaching below the axle | under **29mm** |

| Pin length, from the plate's face | under **29mm** |

| Pin diameter | under **20mm** |

| Axle bore in the wall | **6.2mm** |

The pin passes through a 13mm wall before it reaches the slot, so 29mm

from the plate face leaves it 1mm short of the far side. The set this

was designed around has a 30mm plate and a 28 x 5mm pin.

**And the strings have to clear the pins they pass over.** Every string

but the first runs down to its own pin above the pins nearer the nut.

That is always close on a slotted head — a classical guitar clears by

about half a millimetre, and this design by about a third of one. Fatter

pins than 5mm eat that margin directly, and the SCAD will refuse to

render if they eat all of it.

**If yours are different, print the head from the SCAD instead.**

acoustic_head.scad is published for exactly this. Measure six

numbers, put them in the block at the top of the file, and the whole

pegbox re-derives itself — post spacing, post positions, slot length

and the head's overall length. A tuner that cannot be made to fit fails

the render with a sentence telling you which way to move. A larger one

simply produces a slightly longer head.

**Use OpenSCAD to do it. Not Fusion, not FreeCAD, not an online

converter.** A .scad file is not a model — it is a short program that

draws one, written in OpenSCAD's own language, and nothing else runs it.

It is free, it runs on Windows, macOS and Linux, and the stable release

is all this file needs.

**This is not theoretical.** Somebody edited the SCAD for one of my

other designs in FreeCAD. It imported, it looked correct, it exported an

STL that looked correct — and it printed wrong. Other programs read a

.scad by importing the finished shape as a mesh rather than running

the program, so curves get faceted at whatever resolution the importer

picked instead of the one the file asks for, and small holes and thin

walls are where that lands worst. It is a difference you cannot see on

screen and can see on the bed.

The second reason is the asserts, and they are the whole point of

publishing this source rather than an STL. This head is full of them:

they turn "measure six numbers and edit them" into something safe to do,

because a tuner dimension that will not work **stops the render** and

tells you which way to move. An importer that only reads geometry throws

those away. It fails silently, which is worse than failing loudly.

And plainly: if it came out of anything but OpenSCAD, I cannot help you

with it. I cannot reproduce a mesh I did not generate, on a toolchain I

do not run, and from the outside a fault introduced by an importer looks

exactly like a fault in my file.

Two practical notes. **Press F6 (Render) before exporting**, not F5

(Preview) — preview is a fast approximation and exporting from it gives

you a mesh you should not print. And **change only the numbers in the

tuner block at the top.** Everything downstream derives from them; there

is nothing further down that needs adjusting to match, and editing it

will only break the derivation.

Nothing else in the design changes when the tuners do.

### 2. There is no soundboard to buy

Earlier versions of this instrument had a 3mm basswood top, ripped from

A3 model-aircraft sheet against its factory edge and spliced to reach

590mm. **That is gone.** The top is printed, continuous with the walls,

and arrives as part of the body.

What that removes, and it is most of the woodwork in the build: the

sheet, the rip, the splice, the two straps, the caul, the flat 600mm

surface to glue on, the bridge plates and their crossed grain, and the

1500mm rebate glue line that could only be reached from one side.

**It costs radiating efficiency**, and that is a real cost, taken

knowingly. PLA is roughly 4.4 times the mass of basswood for equal

stiffness, so a printed top radiates worse into the air than a wooden

one. The instrument answers that with a contact pickup rather than with

material: a piezo reads the panel's MOTION, and motion is what a

compliant panel has plenty of.

If you want a wooden top, this is not the design for it — the top is

structural here, carrying shear between the two dowels the way an

aircraft wing does, and it cannot simply be swapped.

## What you need besides the prints

**Wood — and it is now two dowels and a noter, nothing else**

- two 12mm oak dowels, **one from each of two 1000mm sticks**. Neither

is spliced, and **they are cut at different times for different

reasons** — see "Cutting the two dowels" below. Both figures moved

when the tail lost its socket, so neither is a number to take off a

drawing.

- a 10mm dowel for the noter, about 150mm

*Gone from this list: the soundboard sheet, the splice pieces and the

two bridge-plate offcuts. The plates went with the material that

justified them — they were crosswise basswood because a wooden panel

bends about a lengthwise axis and resists in its weak direction, so a

crossed-grain plate is stiff exactly where the panel is soft. Printed

PLA has no weak direction to shore up, and the bridge feet bear directly

on the top at 0.30 MPa, a seventh of the creep threshold.*

**Metal**

- four tuners (see above)

- **no tailpiece.** The anchor is four nails driven into a hardwood

block, loop-end strings over the heads. A banjo clamshell was

specified for a session and abandoned: it hooked over a printed lip,

squeezed rather than fastened, at the highest-tension point in the

instrument

- **no hardwood block.** The tail block is printed, with its nail bores

and string spacing already in it. It was hardwood because nails in PLA

were rejected at 2mm x 10mm deep; the anchors are 25mm deep now, which

is several times the same threshold. **Let the printer do the

precision, and stop asking the builder to** — the anchor spacing

arrives as a side effect of printing rather than as four holes drilled

to a template

- **five nails, anywhere from about 2 to 2.9mm, at least 35mm long**:

four anchors, and one bedded at the tail plate's top rear corner as

the tail nut. One 2.5mm nail about 40mm for the nut at the other end.

**The anchor bores are 3.0mm and that is deliberate — a bonded

clearance, not a press fit.** The nail is meant to be smaller than the

bore, with epoxy filling the annulus along the full 25mm. That is a

chemical anchor: load spread down the whole length instead of

concentrated at the mouth, and no wedging. An interference fit in PLA

does the opposite and splits along the layer lines. It also means the

nail size is not critical, which is why the range above is a range.

*Use liquid epoxy for this, not putty — putty will not flow 25mm down

a blind bore and will pack solid at the mouth with a dry void behind

it. Butter the nail rather than filling the hole, and push it home

with a twist so the air can escape past the shank.*

**The tail nut nail is the exception, because it is a bearing surface

rather than a fastener.** The string breaks over it and the height of

its top face sets the action at that end, so do not simply lay it in

the groove and let it fall to the bottom — a nail narrower than the

3.3mm groove will sit low and can roll under tension. Bed it at a

chosen height and let the epoxy hold it there, the same way the capped

nut sets its wire on putty.

- **one 5 cent coin** for the head. At the tail, a pad of epoxy putty

rather than a coin — it spreads the dowel's end bearing the same way

and has no diameter to fit

- **screws — see below.** 14 for the fretboard, 4 for the tail

sandwich, 8 for the tuner plates

**The screws, which are not "small screws"**

Every hole for them is dimensioned in the files. The object that goes

in the holes was not, until it had to be bought.

| Where | Size | Count |

|---|---|---|

| Fretboard into head and riser | **3.0 x 16mm** countersunk | 14 |

| Tail sandwich | **3.0 x 35mm** countersunk | 4 |

| Tuner plates | whatever came with the tuners | 8 |

**Self-tapping woodscrews, not M3 machine screws.** They share a

diameter and nothing else. screw_pilot = 2.5 is sized for a coarse

asymmetric thread cutting its own way into PLA; an M3 machine thread

engages almost nothing in the same hole and strips under a hand

driver. Listings confuse the two constantly — a great many parts sold

as "M3 self tapping" or even "M3.5" are woodscrews with a metric label,

which is fine, but read the photograph for a sharp point and a coarse

thread rather than trusting the name.

**Fully threaded.** A plain shank under the head jams in the pilot and

holds the joint apart while the screw feels tight.

**Not 3.5mm and not M4.** The fretboard screws go into the **4mm wall

lands** at the top of the riser's side walls, not into solid material.

A 3.5mm screw cutting its own thread there leaves about 0.25mm each

side and bursts the boss. 3.0 leaves 0.5. For the same reason there is

nowhere to put a heat-set insert: an M3 brass insert is about 4.0mm

across and the wall is 4.0mm wide.

**Lengths are stack heights, and countersunk screws are measured

overall.** 16 gives 8mm of board plus 8mm into a 10mm pilot — 2mm of

air at the bottom, which is what you want. A 20 bottoms out and jacks

the board off the riser, which is the one thing those screws exist to

prevent. At the tail, 18mm of upper piece plus 3mm of top pass before

the screw touches the lower piece at all, so 20 does not arrive and 25

bites 4.

**Check the countersink angle against what you buy.** The fretboard's

is cut 3.4 to 6.4 over 2.0mm deep — about 74 degrees included.

Woodscrew heads are usually 82 or 90, which bear on the rim rather than

across the cone. For a 90-degree head set screw_csk = 1.5; for 82,

1.72. Both follow from the same 1.5mm radial step. The head must also

be **6.4mm or under** to drop in at all.

Head diameter is unconstrained at the tail: those holes are plain 3.4mm

bores with no countersink, and the heads are meant to bear on the

printed face and spread the load across it.

**Strings**

D'Addario EJ64 or equivalent: .012 / .012 / .014 / .022 wound.

**Adhesive**

- **Slow-set epoxy** for every joint in the instrument. PVA wood glue

will not bond to PLA — it needs a porous surface to key into — and

with the wooden top gone there are no wood-to-wood joints left. PVA is

out of this build entirely.

- **Cyanoacrylate**, a few drops, to tack a joint square before the

epoxy goes on. It is not holding anything together; it is a clamp made

of glue, and it turns a timed operation into a positioning one

followed by as long as you like.

- **Epoxy putty** for one pocket: the pad in the tail block's front

face, which the upper dowel end-bears on. Earlier versions also used

it to fill gaps at the body's butt joints; the sleeves make that

unnecessary.

- **Not silicone.** It stays compliant and turns string energy into

heat. Never in the drive path — and with a contact pickup the drive

path is the whole top.

- **Not PVC cement** — it warps PLA.

- **Epoxy putty** is fine anywhere; it cures rigid.

---

## Print this first

Five small parts, about 136g and an hour between them, before anything

long goes on the plate. They are named so they sort to the top of the

folder:

fitcheck_section.stl 42g

fitcheck_ring.stl 7g

fitcheck_web.stl 18g

fitcheck_nail.stl 13.5g

fitcheck_riser_jig.stl 56g at 100%, far less at 25%

The first four are not part of the instrument, and they go in the bin

when you are done. Between them they answer every question that gets

expensive later. **The fifth is a jig** — it is not a test, it is a tool

you need during assembly, and it is described where it is used rather

than here. Print it at 20-30% infill; it only ever carries its own

weight.

*fitcheck_section.stl** — a 25mm slice off a segment's joint end. It

shows the whole section at once, carries the dowel boss, and is cut at

the joint so a ring seats in it.

*fitcheck_ring.stl** — a 10mm slice of the j0 coupler. It goes into

the section's cavity, and the pair together proves the section, the

boss, the dowel and the joint fit in one sitting. *The two only mate

because they are cut at the same station: an earlier version of the

section was cut at the head end, where the cavity is 10mm narrower, and

the ring could not go in at all.*

*fitcheck_web.stl** — the front cap's web. The cap's spigot is the

coupler's ring at a different width and needs no separate trial; what is

untested is a 4mm web standing on end with a boss-shaped bite out of its

top.

*fitcheck_nail.stl** — six bores at 0.1mm steps. **Print it flat,

bores vertical, at 100% infill.** The orientation is not optional: in

the tailpiece the nails run along the print's layer axis, and a bore

through layers is tighter and rounder than one across them, so a gauge

printed on edge measures the wrong thing.

Push your own nail into each bore and find the first one it enters with

firm hand pressure and no twisting. The labels read the deviation from

nominal — -2, -1, 0, +1 — so the answer means the same whatever

nail you are holding. That number is your printer's small-hole error,

and it is a property of your machine and filament rather than of this

design: 0.1 to 0.3mm undersize is ordinary, because the extrudate pulls

inward off the wall it was laid against.

**Then push your actual dowel through the section's boss.**

The dowel is the only component in this build that is not made to a

drawing. It is sold as 12mm and arrives anywhere from 11.8 to 12.3,

often slightly oval from drying, and every bore in the instrument is cut

to suit it. **Oak is worse than birch on both counts** — birch is the

standard dowel wood because it turns clean and stays round, while oak is

ring-porous and moves more — so measure yours at several points AND

several rotations before trusting the 0.5mm clearance.

It should slide with light finger pressure and no twisting. If it binds,

**sand the dowel, not the bore.** The dowel is cheap, replaceable and

the variable part; a reamed PLA bore goes oval and oversize and you

cannot put it back. Sand the whole length in one go with the dowel

turning, not the tight spot alone, or you will make a waist.

## Printing

Roughly 720g of filament at 25% infill, 1.3kg if you print it solid.

**The filenames here are the STL names in the download.** They are not

the SCAD names, and they sort into build order rather than source order.

| File | Off | Size | Orientation |

|---|---|---|---|

| body_1_segment_g0 | 1 | 128 x 200 x 44 | **STANDING ON AN END FACE** |

| body_2_segment_g1 | 1 | 138 x 200 x 44 | **STANDING ON AN END FACE** |

| body_3_segment_g2 | 1 | 148 x 200 x 44 | **STANDING ON AN END FACE** |

| body_4_coupler_j0 | 1 | 117.6 x 42 x 37.6 | standing on end |

| body_5_coupler_j1 | 1 | 127.6 x 56 x 37.6 | standing on end |

| body_6_cap_front | 1 | 107.6 x 18 x 37.6 | **web down** |

| tailpiece | 1 | 148 x 76 x 64 | **standing on end, 100% infill** |

| head | 1 | 43 x 233 x 40 | **on its side, slot up, DIAGONAL on the plate** |

| riser_s0 riser_s1 riser_s2 | 1 each | 24 x 177 x 18, s2 is 19 wide | flat, channel upward |

| fretboard_s0 fretboard_s1 | 1 each | 38 x 197 and 38 x 212, both x 9 | flat, **frets upward** |

| fretboard_pins | 1 plate of 4 | 21 x 12 x 3 | flat |

| nut_plain OR nut_capped | 1 | 31 x 6 x 6.5 / 31 x 6 x 4 | flat |

| bridge OR bridge_capped | 1 | 50 x 29 x 6 | on its large face |

| string_comb | 1, optional | 33 x 8 x 8.4 | **fins upward** |

| shim_s0_* shim_s1_* | print all six | 25.5 x 197 / 212 x 0.5-1.5 | flat |

*head is 233mm long and will not fit square on a 220 bed.** Turn it

about 45 degrees. It must stay slot-up.

**The two nuts and the two bridges are ALTERNATIVES, not both.**

nut_plain is a sleeve for a nail; nut_capped takes fret wire bedded

in putty and lets you set the action rather than accept it. The two

bridges are different profiles — print both if you like and compare,

they are 7g each, but only one goes on the instrument.

**The tail is one part now.** Earlier versions had a separate tail cap,

tail block and cover. They merged into tailpiece, which is why there

is no cap_tail in the download.

**The segments stand on end, and that is the whole fork.** A tube stood

on one open end has every wall vertical and no overhang anywhere — no

supports, no bridging, and no roof to hold up. A tray has to print

open-side-up, which is exactly why the top used to be a separate part:

you cannot roof a box in FDM. Close the box and the problem disappears

instead of getting worse.

What it costs is aspect ratio. 200mm tall on a 148 x 44 footprint is

4.5:1, and the deep body is what makes that tolerable — at the original

30mm depth it was 6.7:1.

**Expect some corner lift on the bed face** and do not panic about it.

The first g0 rocked slightly. That face is a joint face, the sleeves are

relieved at the corners for exactly this, and 42-56mm of sleeve is what

holds the joint square rather than the faces. **A brim is worth using.**

**The tail block prints rear face down.** The nails enter from the rear

and the pad pocket opens from the front, and they are 3mm apart at the

bores. Turn it round and the nails go into the pad. The index notch is

on the melody side.

**Orientation is not a suggestion.** The dowel bores run horizontally

and are teardropped so their crowns do not droop; stand a part on its

side and the bore becomes vertical, prints beautifully, and the

finished instrument does not go together. The head is the one to be

careful with — it needs turning about 45° on the plate to fit, which is

fine, but it must stay slot-up.

**No supports anywhere.** Every overhang is either self-supporting or

opens upward. If your slicer wants supports, the part is the wrong way

up.

**Infill.** 25% is plenty for most of it, and the segments, sleeves and

riser are thin-walled enough that infill barely changes their weight.

**On a segment, infill barely applies at all.** Stood on end, the floor

and the top are both VERTICAL walls, so the slicer makes them solid —

which is why a segment weighs what it does and why fitcheck_section

is worth measuring with calipers. Wall 5, floor and top 3: if your

slicer is hollowing a 3mm vertical wall, that is the whole box affected

and the coupon is where you find out.

Two exceptions:

- **the head and the tail want 40–50%**, or solid regions around the

coin seats. The coins spread the dowel's thrust below PLA's creep

threshold, and that argument fails if the material behind them is

mostly air. Get this wrong and the action drifts flat over months.

- **the bridge is better light** — 20%. It sits on the driving point

and its mass loads the top.

**Perimeters matter more than infill on the segments**, and the reason

has changed. It used to be about holding air so the sound holes were

where the air moved. With a contact pickup that argument is gone — and

the box is deliberately NOT sealed, because a sealed box is an air

spring that stiffens the very panel the pickup reads.

What perimeters buy now is a wall that is actually 5mm of plastic, on a

part whose walls are its structure. Three perimeters and generous solid

layers at both ends, since those ends are the joint faces.

### Optional: frets in a second colour

Printed in one colour, the frets are the same shade as the board they

stand on, and in white they nearly vanish under stage light. There is a

free way to fix that.

**Above z = 8.00mm, the only thing on the fretboard is the frets.** The

board is 8mm thick and the crowns are 1.2mm proud of it. Everything

else — the countersinks, the nut pocket, the side markers, the assembly

numbers — sits at or below the top face. So a single filament change at

that height colours every fret and nothing else.

Set it by **height**, not by layer number, since that depends on your

first layer. If you are watching the print rather than scripting it,

**err late**: one layer late leaves a 0.2mm base under each crown that

nobody will ever see, while one layer early puts a band of the new

colour across the whole top face of the board, which is glaring and

cannot be fixed without reprinting.

**Dark into light is the hard direction.** Purge generously — much more

than feels necessary — before resuming. Each fret layer is only about

250mm² of material across the whole board, so there is very little

geometry for a bad purge to hide in. A light board with dark frets

purges far more easily if you would rather not fight it.

Why not an inlay. A thin sheet in a second colour, slotted for the

frets to poke through, is the obvious alternative and it does not work.

The crown is 1.2mm and a 10mm noter was shown to overcome 1.0mm, so

there is 0.2mm of margin in the whole design. Any overlay spends it,

and raising the crown to compensate takes the same amount straight off

an action that is already 1.77mm falling to 1.60. It can only be done

by recessing the board's top face so the inlay finishes flush — which

is a great deal of work, and several new ways to fail, for what a

filament change gives you for nothing.

### And fill the side markers

**There are six of them, three a side.** They used to be on the melody

edge only, which is the side facing AWAY from the player on a

right-handed lap dulcimer — so the one thing they exist for could not be

done with them. Marking one edge gives the instrument a handedness it

has in no other respect: the frets run the full width, and nothing else

cares which end of your lap the head is at.

The markers are recessed **2.5mm** from each edge, at frets 3, 7 and 10

— 33mm of board left in the middle. That depth was chosen because a shallower one read

half-lit rather than black. Wipe permanent marker or acrylic paint into

them and wipe the surface clean. A recess is the one place ink lasts,

because nothing touches it.

**Do not ink the fret crowns.** The noter rides on them, so it polishes

the ink off and carries it onto your hands while it does. If you want

to mark the board itself, put the line in the gap behind each fret —

just as readable looking down, and never touched. Isopropyl alcohol

lifts marker off PLA without harming it, so this is reversible; keep it

off the butt joint between the two fretboard sections, which is an

epoxy line.

### What is actually visible

Worth knowing before you commit colours: **almost none of the plastic is

hidden any more, and that is new.** The old design had a housing and a

tail piece sealed inside the box where colour did not matter. Both are

gone — the housing became a boss integral to the segment, and the tail

became the tailpiece, wrapping the tube's end. What is left inside is

the two joint sleeves and the front cap's spigot, and even those show at

the seam.

The segments themselves are the instrument's whole outside. Everything

else is on show too, including the riser — which is easy to think of as

internal and is not. Past the fretboard's tip it runs another 188mm to

the tail, standing 18mm proud of the top with the bridge on it, and that

is the part you look at while playing.

Under the fretboard it still shows from the side, since the board is

38mm wide over a 24mm riser.

**And match the putty, on the three parts that carry it.** This is the

one thing I would do differently given the chance, and it costs nothing

but choosing a spool before you print rather than after.

Epoxy putty is grey or off-white and it does not take colour. Three

parts have it on show:

- **the nut**, whose trough is filled with putty to bed the fret wire.

This is the one that matters most — it sits directly under your eye at

the head end, the putty runs the full width of the board, and the

crown standing proud of it means the fill is never hidden

- **the tailpiece**, whose pad fills a pocket in the arm's front face —

visible in the gap between the riser's end and the arm

- **the bridge**, if you bed the cap or fill anywhere on it

Printed in a colour close to the cured putty, those joints read as part

of the part. Printed in a contrasting one, every fill is a grey patch on

a coloured surface, and putty is the one material here that cannot be

tidied after the fact — it does not sand back cleanly and it does not

take paint the way a printed recess does.

The segments, riser and fretboard can be whatever you like: no putty

touches them. It is only the three parts above, and the choice has to be

made before the print rather than at assembly, which is why it belongs

in this section and not in Finishing.

---

## Assembly

Read this whole section before starting.

**The big change from earlier versions: the box can be dry-assembled.**

The joint sleeves are a slip fit until you bond them, so the whole

instrument goes together, gets played with, and comes apart again. The

previous design glued the shell into a tray at step 3 and only met the

top at step 13, which meant the one test that is not the model checked

against itself could never be run on the box.

So the order below is: **build it dry, twice, and only then decide what

to glue.**

### Which glue, and it is simpler than it was

**PVA will not bond to PLA at all** — it needs a porous surface. With no

wood-to-wood joints left in the instrument, PVA is out of the build

entirely.

| | |

|---|---|

| clear epoxy | every printed joint, and the dowels where they are held |

| CA | tacking a joint square before the epoxy goes on |

| epoxy putty | the tail block's pad, and nothing else |

*The putty is one pocket. Earlier versions used it as gap filler along

the shell's butt joints; the sleeves make that unnecessary.*

### Sub-assemblies

1. **Fretboard.** Join the two sections with the printed pins and

epoxy. They butt directly — the butt face is what keeps the fret

spacing continuous, so make sure it closes.

2. **Nut.** A printed sleeve with a length of fret wire bedded in epoxy

putty. The wire is the bearing surface the strings break over; the

putty only holds it at the right height. **Do not cut string slots —

there are none, and there is nothing to file.**

**Cut the wire to length and straighten it first.** These are

pre-radiused guitar frets, curved for a curved fretboard, and this

board is flat. Check the piece against a straightedge before it goes

anywhere near the putty. Do not expect the putty to hold a spring

flat: over the 27mm span, a 12-inch radius lifts the centre about

0.35mm above the ends, which is a fifth of the clearance at fret 1

and lands differently on every string.

**Fill the trough with putty and press the wire in tang-down, crown

up.** The trough is open at both ends so surplus has somewhere to go

instead of jacking the wire up — expect some to squeeze out, and wipe

it before it cures. There is 0.9mm of putty under the tang and that

gap is the entire height adjustment: **the tang must not touch the

printed floor.** If it bottoms out, the print sets the string height

again and the reason this part exists is gone.

**Set the crown 3mm above the board** and check it at both ends and

the middle while the putty is still soft. That is the number to aim

at; everything else about the sleeve is derived from it.

**The sleeve is meant to be slack in its pocket.** It floats about

0.3mm in y, which is under a cent at this scale, and its corners are

scalloped so it seats on its faces and can never hang on a printed

corner radius. If it feels loose, that is the design and not a fault

— the earlier sleeve was tight and rocked on the pocket's corners,

which is what this replaced.

**The keyed end goes to the melody side.** A recess at one end

swallows the pocket's boss, so it only fits one way round.

**The marks on the rim are string stations, not slots.** Once the

putty has cured, cut shallow guides into it either side of the wire

at those marks — just enough to stop a string wandering. The break

angle over this nut is about 15 degrees, which is ample down-force,

and there is only 0.9mm of crown to lose depth in. Do not cut into

the wire.

3. **Tailpiece.** It prints as one part — end cap, tail block and cover

merged — so there is nothing to assemble. What it needs is the nails

bedded, and that is the whole job:

**Butter each anchor nail with liquid epoxy and tap it lightly into

its printed bore.** Four of them, 25mm deep. The bore is wider than

the nail on purpose, so the epoxy fills the annulus down the full

length and carries the load — do not drive them hard, and do not

reach for putty, which will not flow down a blind bore. A twist as

each one goes in lets the air escape past the shank.

**Then the fifth nail, in the top rear corner** — this is the tail

nut, and it is not optional: the string wraps that corner through

about 88 degrees under full tension, and a printed corner beds in

under it. This one is a bearing surface rather than an anchor, so its

height matters. Set it where you want the string to break rather than

letting it drop to the bottom of the groove, and let the epoxy hold

it there.

**Then leave it 24 hours before any string goes near it.** Epoxy

reaches handling strength long before it reaches full strength, and

this joint takes the whole of the string tension. Nothing else in the

build is waiting on it, so there is no reason to rush it.

Then the putty pad in the front pocket, dowel end pressed in to shape

it and withdrawn.

**Fit it notch-up on the melody side.** The index notch tells you

which way round it goes, on the side the close pair of anchors falls.

Rolled 180 degrees the anchors mirror and two strings drag sideways

in their bridge slots.

4. **Riser.** Leave the three sections loose. They get joined **on the

dowel**, which is what keeps them in line.

### The dry run — and do it before anything is bonded

Every part offered up in assembly order, no glue and no strings, with a

timer. GLUING_THE_BOX.pdf is right that this is the only test in the

project that is not the model checked against itself: the head's coin

seat passed every automated check and was physically unassemblable, a

21.6mm disc reachable only through a 12.7mm bore, and it was found by

hand.

1. **Thread the lower dowel through g0's boss.** Then g0 + g1 together on

the same dowel, then all three. A boss that is true over 200mm can

still add up over 600, and this is the only chance to find that out.

Sight along the dowel where it emerges rather than feeling for

resistance — 12mm of oak in a 12.5mm bore has half a millimetre to

hide in per segment.

2. **Sleeves in dry.** j0 between g0 and g1, j1 between g1 and g2. They

are **not interchangeable** — j0 is 42mm long and 117.6 wide, j1 is

56 and 127.6. Mark them as they come off the plate.

3. **Caps in dry**, front and tail. The front cap's spigot is shorter

than the tail's; it is not a mistake.

4. **Riser onto the upper dowel**, three sections, joined on the dowel.

**Drop fitcheck_riser_jig over the riser's free run** — past the

fretboard, over the middle segment — and let it settle. Its angled

legs land on the body's top corners and find the centreline on their

own, whatever the taper is doing at that station.

5. **Fretboard down** over head and riser, screwed at seven stations.

6. **Bridge stood up** at y=550, and the head and tail block offered up.

7. **Then take it all apart.**

**Note every joint that feels snug dry** — a 0.1mm slip fit becomes a jam

fit with an epoxy film in it.

### Cutting the two dowels

**Build long, then trim.** That is the project's one-way rule and it

applies to both dowels — but they are not the same measurement, and it

is worth knowing which way each one wants to be wrong.

**The lower dowel: cut it long on purpose.** It is epoxied into the

head's 223mm bore, and its tail end runs to the tube's end at y=645

where the tailpiece's plate closes it. That plate is a **positive stop**.

Cut the dowel a little long and push until it touches, and you know it

is home — instead of guessing against rising friction in a 600mm blind

bore with epoxy already going off. If it fouls the plate, take a

millimetre off. The arithmetic gives about 823mm; measure it on your own

parts, because the head's bore is printed and the figure assumes the

- 10 at its far end is real.

**The upper dowel: cut it last, and slightly long.** It is a strut

between two bearing faces — the 5c coin in the head and the putty pad in

the tailpiece's arm — so its length depends on where the riser's three

sections actually end up once they are joined **on the dowel**. That is

a stack of three real parts, not a number, so measure it on the

assembled instrument rather than deriving it.

Slightly long is right, and trimming is easy: on a cured pad you can see

exactly where the dowel has been bearing.

**But there is one criterion, and it runs the other way from what you

might expect.** Too long is the bad direction. The tailpiece's plate

bears on the tube's end face, and that is what puts the tail's 206 N

into compression instead of levering a bonded footprint in peel. An

over-long upper dowel pushes the tailpiece rearward and **holds the

plate off the end face** — at which point nothing is bearing and the

spigot is carrying the load in adhesive, which is the arrangement this

design exists to avoid.

It is visible, which is what makes it easy: **sight along the seam

between the plate and the tube's end.** Closed all round and it is

seated. Any light and the dowel is long. Trim and try again.

### The spine

The two dowels do different jobs and are glued differently. **The lower

is in tension. The upper is in compression.** That one fact explains the

whole adhesive scheme.

**The lower dowel is held.** Epoxy it into the head's 223mm bore, which

has glue reservoir grooves cut into it for the purpose — and cut it long,

per "Cutting the two dowels" above. At 205 N over

that length the shear is about 0.03 MPa — nothing — and that margin is

why it can be the anchor at that end.

**Then bond it in g2's boss only, and slide g2 on last.** Not down its

whole length. Between nut and bridge the body carries a constant

moment, and a constant moment means zero shear force along the flange —

so the dowel needs anchoring at its ends, not shear-connecting down its

length. Which is fortunate, because a 600mm closed bore cannot be glued:

you would butter the dowel and the entrance would scrape it clean.

g0 and g1 thread dry.

**The upper dowel is glued nowhere at all.** It end-bears on a 5c coin

at the head and the putty pad in the tailpiece's arm, and a dowel you can

still draw out is a dowel you can recover. Cut it last, on the assembled

instrument, and check the plate's seam is closed.

### Closing the box

**Bond the sleeves with the joint assembled and clamped straight.** This

is the whole point of them. A printed end face is not a machined one, and

the first g0 off the bed rocked slightly on the corners of its joint

face. If you bond a sleeve into one segment on its own, the epoxy sets it

square to that segment's error and then propagates it down the spine. Bond

it in place, aligned, and the epoxy takes the error up instead.

The sleeve becomes the datum; the end faces stop being it. That is a

spigot joint, and it is what the design should always have been once the

faces were known to be printed.

1. Assemble the joint dry, sight down the spine, and clamp it straight.

2. Tack with CA at two or three points to hold the alignment.

3. Then run clear epoxy into the joint and let it cure at its own pace.

**Bond up to the witness groove.** Each sleeve carries a groove round

three faces at the bond line, so you can see how far it has gone in. The

half beyond the groove stays a slip fit.

The caps go in the same way, both from outside — neither needs the box

open.

### Bonding the riser: you cannot clamp it, so do not try

**There is no clamp access, and the body is why.** A clamp has to reach

over the top face and round to the underside, and the underside here is

the floor of a closed 44mm tube. There is nothing down there to bear

against that is not the inside of a sealed box. This is not a matter of

owning the right clamps — closing the tubes removed clamping access as a

side effect, and clamp reach is not a dimension that appears in any file.

**Use weights, and that is the correct method rather than a fallback.**

Epoxy is gap-filling. It wants contact and alignment, not squeeze, and a

clamp on this joint would mostly risk bowing the riser or dishing the

top. Books, tins, anything flat and heavy, spaced along the run.

**Sort the weights out before you mix anything.** The failure here is

not pressure, it is scrambling for something heavy with epoxy already

going off.

**And let fitcheck_riser_jig hold it straight while they hold it

down.** That division is the whole point: alignment and pressure are

wanted at the same moment and you cannot manage both by hand. Drop the

jig on, let it settle, position the riser, weight it, then **lift the

jig off before the epoxy hardens.** Wax or tape its channel if you are

working close to it.

It is a jig. It never becomes part of the instrument, and a permanent

version of it would sit exactly where the strumming hand goes.

### Finishing

14. **Fretboard** onto the head and riser with fourteen 3.0 x 16mm

countersunk screws — seven fret gaps, two rows — into the riser's

two wall lands. The station at 24mm from the nut is drilled in the

head member, so two of the fourteen go there and twelve into the

riser.

15. **Tuners.** Sit each plate against the scribed line on the pegbox's

side, mark through its own holes, drill and fit. The dimples show

roughly where the tuner lands; they are deliberately too shallow to

steer a drill, because the plate is its own template and cannot

disagree with itself.

16. **The vent.** Nothing to cut — it is printed into g2's top at y=500.

There is no plug family and no tuning by ear: with a contact pickup

the box's air resonance is not what you are listening to. The vent

exists so the box is not a sealed air spring stiffening the panel

the pickup reads, and so it can breathe against temperature. Its

size does not matter.

17. **Nut and bridge.** Which nut you fitted decides what happens here,

and the two are different jobs — do not follow the wrong one.

*nut_capped** — nothing to file. The wire is the bearing

surface and its height was set in putty when you bedded it, back

in sub-assembly 2. The only work left is the shallow guides

either side of the wire at the rim marks. **Do not file the wire.**

*nut_plain** — file the slots down until the file reaches the

nail, leave the last 0.2–0.3mm and let the strings bed through.

Bridge is **not glued**; it stands under string pressure so it can

be moved.

**The nut sleeve is not glued either.** Its pocket fixes it in all

three axes and string pressure pushes it down into that pocket, so

glue would only be stopping it falling out while the strings are

off — which it will do, whenever you tip the instrument, right up

until it is strung. Live with that. Nut height is set by the pocket

depth, so an unglued sleeve means action at the nut is a two-gram

reprint of the fretboard rather than surgery.

If it becomes a nuisance, glue **at the ends and never underneath**.

Anything with body under the sleeve is not weak glue, it is a shim

you did not intend, sitting inside the pocket that defines the nut's

height. A dab of PVA at each end is thinnest — but **Titebond III does

not release.** It is the waterproof one, and warm water is Titebond

Original's escape, not its. A dab here is permanent. If you want it

reversible, **museum putty at the ends leaves nothing at all.**

18. **Shellac, last.** Everything that has to bond is already covered by

the part that bonds to it — the riser and the tail's nose from above,

the housing, plates and straps from below, the shell's rebate all

round the edge. So finish after the gluing rather than masking before

it, and there is nothing to tape.

**One joint is deliberately left until after this, and it is not an

oversight.** The string comb, if you fit one, goes on last — its

position depends on where the strings actually run, and that is not

known until the instrument is strung and playing. So it lands on a

shellacked surface.

*Earlier versions left two joints for this stage: the tail cover,

and the chosen rosette plug. Neither part exists now — the cover

merged into the tailpiece and the plug family went with the

Helmholtz tuning.*

**Wipe the footprint with alcohol before bonding it.** Shellac is

the one material here that comes off again, which is why it is the

finish — so this needs no forethought at shellac time and no masking.

Epoxy onto a shellac film bonds to the film, and the film peels off

the wood.

---

## Setting up

**Wear eye protection the first time you bring it up to tension.** Four

strings at pitch put roughly 200N through the tail, and every part in

that load path is printed. A weak layer bond does not announce itself

beforehand — it holds through tuning and lets go without warning, and

a string or an anchor that releases under that much tension goes

somewhere at speed. It costs nothing to put glasses on and it is the one

moment in this build where something can leave the instrument.

The same goes for anyone standing nearby, and for the first few days

afterwards rather than only the first minute: epoxy that has not reached

full strength and a joint that is creeping will both fail later than the

first tuning, not during it.

**String spacing** is 5 / 11 / 11mm at the nut and 5 / 12 / 12 at the

bridge, melody pair first. The pair stays 5mm apart the whole way, so a

noter meets the same two strings at every fret.

**Which string on which tuner.** The four posts sit at one height but

at different distances from the nut, so each string leaves at its own

break angle — about 15, 12.4, 10.6 and 9.2 degrees from the nut

outward. A shallow break rattles a thin plain string soonest, which

argues for the melody pair on the two nearest posts.

**But the posts alternate sides of the pegbox, and that pulls the other

way.** Posts 1 and 3 sit against one wall, 2 and 4 against the other. A

string whose post is on the opposite side from its nut station has to

cross the centreline to get there, and two strings crossing in opposite

directions must intersect. Assigning purely by break angle produces

exactly that.

So there is a real choice here rather than a rule:

- **feed each station from its own side** — melody stations from the

posts on the melody wall, and within each side let the farther post

take the more outboard station. No crossings, at the cost of putting

the melody pair on shallower posts.

- **or assign by break angle** and accept that two strings cross

somewhere in the pegbox.

**The crossing is not an acoustic fault.** Everything forward of the nut

is outside the speaking length, so it cannot affect pitch or intonation.

What it costs is tuning stability — a string arriving at its station

from an angle bears sideways on whatever locates it, binds while you

tune, then releases with a ping and overshoots. And if two strings

actually touch where they cross, steel on steel will buzz.

*string_comb is the other answer, and it does not force the choice.**

### The string comb (optional)

A small four-slotted part that glues to the head's top face 25mm forward

of the nut. Its slots sit at the nut's own stations, so the last 25mm of

each string runs parallel to the centreline and arrives at the nut dead

straight, with nothing sideways to bind on. The crossing still happens —

just forward of the comb, where it does not matter.

**Fit it after the instrument is strung and playing,** not during

assembly. Slide the strings into their slots, check it sits square, then

a dab of epoxy underneath. It carries no load.

**It must not touch the underside of any string.** The slot floors are

cut below every string and the fins stand between them, and that is

deliberate: a comb taking downforce is a comb stealing break angle from

the nut, which is the one thing here with nothing to spare. If a string

rests on it, it is sitting too high or too far forward.

### The shims

shim_s0_0.5mm.stl shim_s1_0.5mm.stl

shim_s0_1.0mm.stl shim_s1_1.0mm.stl <- fit these

shim_s0_1.5mm.stl shim_s1_1.5mm.stl

Six files, three thicknesses per fretboard section, about a gram each.

They go **under the fretboard**, and they are the only action adjustment

that survives on a finished instrument — which is why the fretboard is

screwed rather than glued.

**Fit the 1.0 pair at build time.** Raising the fretboard lowers the

action, so starting from the middle lets you go either way afterwards.

With no shim fitted you can only ever go one way.

**They do nothing at the nut.** The effect is

shim x (distance from nut / scale length) — the nut rides up with the

board, so a 1mm shim is worth about 0.4mm around the middle of the board

and nothing at all at fret 1. If your action is high at the first fret,

the shims cannot help you and the nut is the part to change.

**Bridge position.** Nominally 550mm from the nut face — which, with a

nail nut, means the nail's centreline. Expect to set it back **1.2 to

1.5mm**: pressing a string down onto a fret stretches it slightly

sharp, about 11 cents at the first fret and a flat 4 across the rest,

and that setback puts the middle of the board within a cent.

**The vent is not a tuning control any more, and the pickup is why.**

Earlier versions had two octagonal apertures and a family of eight

interchangeable plugs, stepped roughly a tone apart, with the 11mm bore

deliberately absent because it landed on the bass drone at 148 Hz. You

dropped a plug in, tapped the top, found the pitch and swapped it. It was

the one acoustic variable left open on purpose and the only one you could

change after the box was shut.

**All of that solved a Helmholtz problem the pickup makes moot.** A

contact transducer reads the panel's motion, not the air in the box, so

the cavity's resonance stops being what you are listening to. The plug

family, the deliberate absence, the liner and the ear-tuned aperture are

gone with it.

What is left is one vent of no particular size, printed into g2's top,

and it does two jobs that are not about pitch:

- **it stops the box being an air spring.** Compress a sealed top and the

trapped air pushes back, so the panel stiffens and moves less — and

motion is exactly what the pickup reads. Sealing the box would add

stiffness at the one surface being listened to.

- **it lets the box breathe** against temperature, which a closed

one-and-a-half-litre plastic vessel otherwise does not.

Decorative holes are fine. No holes is a mistake.

*The one thing worth knowing that the plug family taught: an amplified

hollow box feeds back at its air resonance. If you ever hear that, the

answer is damping inside or a different pickup position, not a smaller

vent.*

---

## The fret ladder

Equal temperament on a 550mm scale, measured from the nut face. Given

so you can check a print against a rule before stringing it.

| Fret | Note | mm | | Fret | Note | mm |

|---|---|---|---|---|---|---|

| 1 | E | 60.0 | | 7 | D | 275.0 |

| 2 | F# | 113.5 | | 8 | E | 305.0 |

| 3 | G | 138.0 | | 9 | F# | 331.7 |

| 4 | A | 182.9 | | 10 | G | 344.0 |

| 5 | B | 223.0 | | 11 | A | 366.5 |

| 6 | C | 241.3 | | | | |

| 6½ | C# | 258.6 | | | | |

The 6½ fret is not optional. Without it the instrument plays

permanently in the mode most players are trying to avoid.

Action comes out at about 1.8mm at the first fret and 1.6mm at the

eleventh — level rather than rising. That is better intonated than a

traditional rising action, and the risk it runs is buzz rather than

pitch.

---

## Testing before you commit

**This section used to be four pages about glueing basswood**: a flat

600mm surface, a splice coupon broken by hand, a caul, and the order in

which to check flatness against strength. All of it went with the wooden

top.

What replaces it is shorter and lives in two other places:

**The four fitcheck_ parts, under "Print this first"** — about 80g.

They settle the section, the dowel through the boss, the joint clearance

and your machine's small-hole error before any long print starts.

**The dry run, under "Assembly"** — the whole instrument slip-fitted

together and taken apart again. That is the test that used to be

impossible.

The principle behind both is unchanged and is the reason this section

still exists at all: **nothing gets bonded until it has been done dry.**

A closed tube cannot be reopened, and unlike the old shell it cannot even

be reached into.

### The one thing that has no coupon

Whether a 200mm tube prints square over its full height. No 25mm slice

can answer it and no amount of arithmetic can either — the first article

is the test. Print ONE segment, stand it on a flat surface, and check

that the other end sits without rocking before committing to the other

two.

If it rocks slightly at the corners, that is what the sleeves and the

corner relief are for. If it is bowed along its length, that is a

different fault and no joint fixes it, because the dowel has to thread

all three.

## About the instrument in the photos

The files here are all current. The instrument in the photographs was

not printed entirely from them, and it is worth saying where the

difference is.

**Three parts on it were printed during the previous fork** — the head,

the riser and the fretboard. Everything else was printed from the files

as published. What has changed in those three files since is position

markers, scribe lines and the seat arrangement at the head end: nothing

that touches a joint, a bore, a dowel path or a bearing surface. Where a

marker moved it was marked by hand at the bench, and reprinting a head

for several hours to gain a marker is the wrong trade on an instrument

that already plays.

**This does not affect you.** Print every file in this set and you get a

complete, consistent instrument — in some respects a tidier one, because

those revisions are in the files you have. There is no step where you

need to compensate for something done at a bench you cannot see.

## Licence

CC BY-NC-SA. Remix and rebuild freely, credit the source, share alike,

and do not sell it.

If you build one, post a make — and if you find something wrong, say

so. Several things in this design were only found by someone putting a

real part against a real fitting.