Evan

An Adom molecule LED nameplate. Evan's name is spelled out by 66 green 0603 LEDs on a 5 × 7 dot-matrix grid, powered from a USB-C receptacle on the east edge. Plug it into a phone charger and it lights up.

Load interactive 3D viewerOrbit, inspect, measure, x-ray the copper. About 15 MB.

The viewer loads when it scrolls into view, about 15 MB. The guided walkthrough starts on its own, seven steps, and the PCB / Schematic / Parts tabs are live. Open it full screen.

The board

Bill of materials

Every line is an Adom Basic Part, so the whole board builds off reels already loaded on the PnP line.

How the letters are generated

FONT in src/lib/index.tsx holds one 5 × 7 glyph per character and TEXT = "EVAN" selects them. A generator walks each glyph's lit pixels and emits one <led> plus one <resistor> per pixel. Changing the name is a one-line change; a longer name needs BOARD_W grown to keep the 16 mm USB-C reservation on the east edge.

Power

IN-S63BTG is an InGaN part: Vf 3.2 V typ at 20 mA, dropping to about 3.0 V at single-digit mA. From a 5 V rail:

(5 V − 3.0 V) / 330 Ω  =  6.1 mA per LED
66 LEDs × 6.1 mA        =  ~400 mA, 2.0 W

That is inside what a source advertising Default USB Power will hand out, so it runs off a phone charger without negotiation. At 560 mcd @ 20 mA these are bright parts, so 6 mA is a comfortable desk brightness rather than a dim one.

There is deliberately no power LED. The letters are the power indicator: lit means 5 V is present. And because each LED is fused by its own resistor, one LED dying does not affect the others. Replace the LED and its 330 Ω pair.

The USB-C body offset fix

Worth recording, because it bit this build and will bite the next one.

The wrapper this board inherited placed the receptacle's 3D body 2.09 mm forward of its own pad row, hanging 2.5 mm off the board edge. The cause is that tscircuit anchors cadModel.positionOffset at pcb_component.center, which is the bounding box of pads and holes, not the pad row. Every hole in this footprint sits on the mouth side, so that centre is 2.7366 mm forward of the pads and the offset inherits the error.

adom-tsci lint does not catch it. It derives the mouth vector from cad.rotation.z, so the error cancels out of its own check and it reports a healthy body_fwd.

The constants are re-derived in src/lib/UsbCReceptacle.tsx as

BODY_FORWARD_MAG      = MODEL_ORIGIN_TO_BACK − FOOTPRINT_CENTER_FROM_PADS
USB_C_MOUTH_FROM_PADS = MODEL_ORIGIN_TO_BACK + MODEL_ORIGIN_TO_MOUTH

with the model numbers measured off the J1 node of a built 3d.glb, the artifact that actually gets rendered, so it cannot disagree with what you see. The result: shell back face exactly on the pad row, mouth 0.4 mm proud, easternmost copper 1.68 mm inside the edge.

src/scripts/check-usbc.mjs re-runs that measurement against the GLB and exits non-zero if either lands outside tolerance. Step 2 of the walkthrough above flies the camera to the receptacle so you can see the seating for yourself.

Schematic and layout

Full vector versions: schematic.svg (141 symbols) and pcb.svg. Both are also tabs inside the interactive viewer above.

Mechanical and molecule compliance

• Absolute sizing is deliberate. A molecule's wings grow the board past its nominal size, and the USB-C mouth has to be positioned against an edge whose coordinate is a number you can trust. size="96x44 absolute" gives a real outer edge of 96.4 × 44.4 mm.
• Four corner machine pins, MachinePinMediumStandard, fixture the board into the standard Adom workcell. No electrical function.
• No SMT pad or anchor hole overhangs. Easternmost copper sits 1.68 mm inside the edge; only the connector shell cantilevers.
• The bottom silkscreen is a documentation panel. Flip the board over and it explains how to power it, how it works, the full BOM, how it was built, the workcell fixturing, and what to do if an LED goes dark.

Known issues and lessons learned

• The 1.65 mm resistor offset is load bearing. Resistors sit 1.65 mm below their LED. At 1.5, 1.55 or 1.6 mm the autorouter drops a via inside a pad's clearance, 0.053 to 0.083 mm against a 0.1 mm minimum. If you change the pitch, the board size, or the part placement, sweep this again and re-check circuit.json for pcb_pad_pad_clearance_error.
• traces: 208 routed / 210 declared is not a failure. The autorouter merges some net members into shared routes without emitting a per-source-trace row. Every port still lands on copper of the right net, so verify geometrically rather than by counting.
• Do not set schX/schY on only some parts. Giving J1 explicit schematic coordinates disabled tscircuit's auto-layout for everything else and stacked all 141 symbols at the origin. Either place every part or place none.
• The whole 3D viewer hung on one localStorage read. The wiki serves this content under a CSP sandbox with no allow-same-origin, so the document has a null origin and any storage access throws a SecurityError. The viewer reads storage during init, and because Adom3DViewer.init() is not awaited, that throw became an unhandled rejection that abandoned everything after it: no light rig, no .env IBL, no render loop, no camera framing, and a getCamera() returning null, which in turn made every walkthrough camera fly bail on its first guard. Nothing was logged and engine.getFps() still reported 60. An in-memory storage shim ahead of the bundle fixes all of it at once. Two genuine CSP limits remain and are handled separately: GLB textures ship as data: URIs because blob: is refused, and the 13 MB bundle ships as parts because render/ 413s over ~4 MB. Full writeup in the export skill.
• MP1–MP4 share one schematic symbol, and cannot be separated. Every other part here has an explicit schX/schY slot, but the machine pins are emitted by <Molecule> as jsx(MachinePin, {type, name, pcbX, pcbY}), and MachinePinOrContact in @tsci/adom-inc.library destructures only {portHints, pcbX, pcbY, name}, dropping any schematic coordinates passed in. So they stack at the origin and adom-tsci lint reports four overlapping symbols that no caller can fix. Cosmetic on a board where the pins carry no net. Filed against the molecule and library packages.
• Mixing tscircuit versions breaks the build. A stale nested circuit-json under @tscircuit/cli produced Export named 'getRotationBetweenPcbPin1Locations' not found. The fix is a clean rm -rf node_modules bun.lock && bun install, not a patch.

Variants

• Different name. TEXT plus the glyphs in FONT are the whole design. Another four-letter name is a one-line change.
• Different colour. Swap LED_MPN for another Adom 0603 LED: IN-S63BTA amber, IN-S63BT5B blue, IN-S63BTR red, IN-S63BT5UW white. Re-derive R_VALUE for that part's forward voltage. Red and amber are roughly 2.0 V parts and want about 470 Ω, not 330 Ω.

Reproducing it

tar xzf evan-led-nameplate-source.tar.gz
cd src && bun install

bunx tsci build lib/index.tsx --glbs --svgs --3d-png --pcb-png
node scripts/check-usbc.mjs        # USB-C geometry, measured off the GLB
adom-tsci lint                     # DRC, routing, schematic overlap

adom-tsci start . --texture-resolution 4096
adom-tsci stop                     # when done, so tsci dev processes do not leak

Files

Source, to recreate the molecule

• plan.md, the original brief, every refinement, and what was deliberately left out
• evan-led-nameplate-source.tar.gz, full project source: lib/, scripts/, configs, walkthrough

Build outputs

• board.glb, 3D model with a 4096 px baked board texture
• pcb.svg, copper, silkscreen and drill layers
• schematic.svg, full schematic, 141 symbols
• viewer/, the self-contained interactive viewer embedded above

Provenance

Authored in tscircuit (TSX) with adom-tsci, autorouted by the tscircuit autorouter. @tsci/adom-inc.molecule 1.1.15, @tsci/adom-inc.library 1.2.1, tscircuit 0.0.2375. Walkthrough generated by adom-tsci walkthrough-gen from circuit.json. Part of the Adom LED nameplate family alongside brianna-led-nameplate and alex-nameplate. Designed by Adom in Fort Worth, Texas. 2026-08-20.