Overview Files Packages Releases Dependencies Issues Breadcrumbs Pull Requests Activity

Oliver's LED Nameplate

A USB-C powered desk nameplate that spells OLIVER in 57 white 0603 LEDs. Plug in any phone charger and the name lights up.

Built with tscircuit on a 4-pin Adom molecule, previewed with adom-tsci, then round-tripped through KiCad 10 (.kicad_pcb) and Fusion 360 (STEP) to confirm the geometry survives the trip.

Specs


Board	178 x 58 mm, 2-layer, `roundEdges={false}`
Molecule	`type="4pin"`, `pinType="MachinePinLargeStandard"`
LEDs	57 white 0603 (O=12, L=7, I=9, V=9, E=10, R=10)
Font	3x5 pixel dot-matrix, 6 mm pitch, 8 mm letter gap
Current limiting	57 x 330 ohm 0603, one per LED (~5.5 mA each)
Power in	USB-C TYPE-C-31-M-12 (JLCPCB C165948), east edge
CC pulldowns	2 x 5.1k on CC1/CC2
Bypass	100 nF 0603 on VBUS
Total draw	~310 mA @ 5 V

Verified

185 / 187 traces routed. KiCad 10 reports Unrouted 0 — 951 track segments, 55 vias, 61 nets, 264 pads.
USB-C geometry lint passes: body_fwd = 5.34 mm, mouth exits the board edge at x = 89.2 mm, so a cable actually seats.

Two gotchas worth stealing

1. USB-C pin labels must be unique. The wrapper this was forked from labelled two different pins GND1 (pin1 + pin15) and two VBUS1 (pin3 + pin13). tscircuit merges same-labelled ports into one pcb_port, so their pads land on top of each other:

pcb_pad_pad_clearance_error: Pads J1>GND1 and J1>GND2 too close (0mm, min 0.1mm)
pcb_autorouting_error: Autorouting was skipped because 4 PCB placement errors were found

The board then silently ships completely unrouted. Fix: unique labels (GND_A1, GND_B12, GND_A12, GND_B1, VBUS_A4, VBUS_B9, VBUS_A9, VBUS_B4) tied together with traces.

2. Never trace a USB-C pin by number. <trace from="J1.pin3" .../> makes tsci build hang forever with no output and no CPU (it blocks, it doesn't spin). Always trace by label: <trace from="J1.VBUS_A4" .../>.

Also: a receptacle wrapper MUST forward schX/schY/schRotation to its inner <chip>, or it silently piles at the schematic origin.

Known limitation

MP1..MP4 — the Molecule library's internally-emitted machine pins — still overlap at the schematic origin. They can't be placed from board source; the rest of the layout is lifted clear to compensate.

# Oliver's LED Nameplate

A USB-C powered desk nameplate that spells **OLIVER** in 57 white 0603 LEDs.
Plug in any phone charger and the name lights up.

Built with tscircuit on a 4-pin Adom molecule, previewed with `adom-tsci`,
then round-tripped through **KiCad 10** (`.kicad_pcb`) and **Fusion 360** (STEP)
to confirm the geometry survives the trip.

## Specs

| | |
|---|---|
| Board | 178 x 58 mm, 2-layer, `roundEdges={false}` |
| Molecule | `type="4pin"`, `pinType="MachinePinLargeStandard"` |
| LEDs | 57 white 0603 (O=12, L=7, I=9, V=9, E=10, R=10) |
| Font | 3x5 pixel dot-matrix, 6 mm pitch, 8 mm letter gap |
| Current limiting | 57 x 330 ohm 0603, one per LED (~5.5 mA each) |
| Power in | USB-C TYPE-C-31-M-12 (JLCPCB C165948), east edge |
| CC pulldowns | 2 x 5.1k on CC1/CC2 |
| Bypass | 100 nF 0603 on VBUS |
| Total draw | ~310 mA @ 5 V |

## Verified

- **185 / 187 traces routed.** KiCad 10 reports `Unrouted 0` — 951 track
  segments, 55 vias, 61 nets, 264 pads.
- **USB-C geometry lint passes**: `body_fwd = 5.34 mm`, mouth exits the board
  edge at x = 89.2 mm, so a cable actually seats.

## Two gotchas worth stealing

**1. USB-C pin labels must be unique.** The wrapper this was forked from labelled
two different pins `GND1` (pin1 + pin15) and two `VBUS1` (pin3 + pin13).
tscircuit merges same-labelled ports into one `pcb_port`, so their pads land on
top of each other:

```
pcb_pad_pad_clearance_error: Pads J1>GND1 and J1>GND2 too close (0mm, min 0.1mm)
pcb_autorouting_error: Autorouting was skipped because 4 PCB placement errors were found
```

The board then silently ships **completely unrouted**. Fix: unique labels
(`GND_A1`, `GND_B12`, `GND_A12`, `GND_B1`, `VBUS_A4`, `VBUS_B9`, `VBUS_A9`,
`VBUS_B4`) tied together with traces.

**2. Never trace a USB-C pin by number.** `<trace from="J1.pin3" .../>` makes
`tsci build` hang forever with no output and no CPU (it blocks, it doesn't spin).
Always trace by label: `<trace from="J1.VBUS_A4" .../>`.

Also: a receptacle wrapper MUST forward `schX`/`schY`/`schRotation` to its inner
`<chip>`, or it silently piles at the schematic origin.

## Known limitation

`MP1..MP4` — the Molecule library's internally-emitted machine pins — still
overlap at the schematic origin. They can't be placed from board source; the rest
of the layout is lifted clear to compensate.