Week 8 — Electronics production

Group assignment for this week is on the Chaihuo Week 8 group assignment page. The rest of this page is my individual work.

Individual assignment

Week 8 turned the Week 6 layout into copper, then into a stuffed board on the bench. I already had schematic and routed layout; this week was Gerber export, Mods toolpaths, milling on the lab PCB CNC, soldering the XIAO ESP32-S3, and a first firmware upload.

PCB to PNG

I exported F.Cu and Edge.Cuts from KiCad, then uploaded the Gerbers to Gerber2PNG. Mods expects high-contrast PNGs (white features on a black background), so I generated three 1000 DPI images: traces, drills, and board outline. The preview matched the Week 6 layout at about 45 × 62 mm before I moved on to toolpaths.

Gerber2PNG outputs for traces, drills, and outline

Production workflow checklist

The assessment wording mentions tool-path generation, milling or laser engraving, stuffing, debugging, and programming. My board used the milling route on copper-clad stock, not laser engraving, so I tracked the full process this way.

Workflow step	What I did	Evidence on this page
Tool-path generation	Converted KiCad Gerbers to PNG and generated trace / outline G-code in Mods CE.	Gerber2PNG outputs, Mods parameters, simulation, and saved G-code screenshots.
Milling	Ran trace isolation first, then board outline on the KEXU PCB CNC.	Machine setup photos, zeroing photo, milling video, and final milled board photo.
Stuffing	Soldered the XIAO ESP32-S3 and then added the off-board harness wires.	Soldered XIAO and wire-harness photos.
Debugging	Found a lifted 3.3 V trace and changed the power path for the encoder branch.	Damage photo and explanation of the AMS1117 workaround.
Programming	Uploaded `pcb_testing.ino` and verified the W8 SYSTEM screen.	Firmware test video and downloadable Arduino sketch.

Toolpaths in Mods CE

I used the G-Code Mill 2D PCB workflow at modsproject.org. Trace isolation and board outline are separate programs: different PNG inputs, different passes, and separate .nc files loaded on the machine one at a time.

For traces I loaded traces_toplayer_0.png and chose the 0.40 mm flat preset: 0.102 mm cut depth, four isolation offsets at 0.5 stepover. I simulated the path in Mods before posting; the estimate was about 31 minutes at 2.5 mm/s feed and 11 000 rpm spindle speed. Output: traces_toplayer_0.png.nc.

Mods toolpath simulation for trace isolation

For the outline I loaded outline_toplayer_2.png, ran a contour program with the same flat end mill, and saved outline_toplayer_2.png.nc with matching feeds and spindle speed.

Milling

On the lab KEXU PCB mill I clamped single-sided copper-clad stock on the spoil board, touched off XY/Z zero on the controller, and ran the trace file first. I stayed at the machine for the full isolation pass, swapped to the outline program, and re-checked zero before cutting the board free. Flat stock and a firm clamp matter here; any flex shows up as uneven isolation depth.

Copper clad board clamped on the PCB CNC

KEXU controller with work coordinates zeroed

After vacuuming copper dust off the bed I deburred the edges and checked isolation between traces with the naked eye. The milled board matches the Week 6 layout: XIAO footprint in the centre, 1×3 header rows along the sides, and SW1 at the top edge.

Stuffing and soldering

I soldered in a fixed order: XIAO first, then harness wires to the encoder and LCD nets, then off-board modules on jumpers so I could debug without desoldering the module. The castellated XIAO pads are tight. Flux and a narrow tip helped, but this step still took longer than breadboard wiring.

XIAO ESP32-S3 soldered on the milled carrier board

Problem: lifted 3.3 V trace

While soldering a header I overheated one area and lifted part of the on-board 3.3 V distribution. I could no longer rely on the PCB pad to feed the rotary encoder, so the encoder branch runs from a separate AMS1117 module instead. The wiring change is documented on the input-device page.

Close-up of trace damage from overheating during soldering

Bench bring-up

With the encoder, LCD backpack, and regulator connected, the stack matches the harness plan in pcb_testing. USB powers the XIAO; the encoder uses the external 3.3 V regulator. This is the board I programmed for assessment — milled, stuffed, and running custom firmware.

Full bench wiring with encoder, LCD, and regulator

W8 SYSTEM: firmware test

Once the sketch is uploaded it lands on the W8 SYSTEM page. Line 1 shows the page name; line 2 prints uptime (t…s) and the LCD backpack I²C address in hex. That confirms the XIAO is running, I²C init succeeded, and the display responds.

Download Arduino sketch (pcb_testing.ino)

Reflection

Simulating in Mods before touching copper caught at least one polarity mistake. On the machine, zeroing and clamping weighed as much as the KiCad layout. Bump the origin between trace and outline files and the cuts mis-register. Hand soldering on milled copper is less forgiving than a breadboard: one lifted trace changed how I powered an input device until I could respin.

Following the group PCB milling checklist (clamp clearance, tool diameter match, touch-off, deburr before inspection) made the solo run boring in a good way. Next revision: widen the 3.3 V pour and enlarge pads before milling again.