REV A · PROTO NOTES

Making one PCB at a time, fast

The honest map of how to get a board in hand — from same-afternoon in your garage to a week from Shenzhen — and the setup worth buying.

U0 The ladder that compounds: assembly & test

Before the fab options below, the strategic point: board fabrication is a solved commodity — assembly capability is the asset. JLCPCB will sell anyone a 6-layer board for lunch money, so every dollar you sink into fancier fab buys capability the market already prices at ~$2. Every dollar you sink into assembly and test multiplies across every design you'll ever make — and each rung unlocks part classes and volumes the previous one couldn't touch:

• Rung 1 — Stencil + hotplate ~$100. Laser-cut stencil, MHP-30 or a $60 UYUE plate. Handles 0603/0805 passives and QFPs with care, at tens-of-boards scale. You're already here if you follow the pipeline below.
• Rung 2 — Pick-and-place $1.5–8k. Opulo LumenPnP (~$1.5k, open-source, active community) or a used NeoDen (~$5–8k, faster, real feeders). This is the big unlock: consistent 0402s and QFNs, hundreds of boards per run, placement that doesn't depend on your caffeine level. Budget honestly — feeders quietly become half the machine's real cost.
• Rung 3 — Profiled reflow $300–500. A Controleo3-converted toaster oven or a modded T-962A. A real thermal profile is what makes BGAs and lead-free joints reliable instead of lucky — and BGAs are the gate to modern MCUs, RAM, and anything interesting.
• Rung 4 — Inspection $200–400. A stereo microscope changes your yield overnight; farm out X-ray for the occasional BGA question. AOI can wait until volume justifies it.
• Rung 5 — Functional test mostly labor. Pogo-pin jigs, scripted firmware flash + self-test. "Tested" is the word that separates a product from a bag of boards — it's what lets you charge, and it's nearly pure reusable engineering.

U1 In-house — same-day, the fastest per iteration

Fiber laser + chemical etch ~30–60 min/board

The current hobbyist-pro meta. A 20W fiber laser (~$1.5–2.5k — xTool F1 class or a galvo unit) ablates a spray-paint or dry-film mask off copper clad, then a cupric chloride etch tank finishes the board. Clean 6–8 mil traces, and the laser also cuts solder-paste stencils and does part marking — triple duty. Two-layer works with careful flip alignment plus riveted or soldered-wire vias. No plated through-holes, no solder mask (UV-cure paint one if you care).

CNC isolation milling ~1–2 hr/board

Bantam Tools desktop mill (~$4k) or a $250 Genmitsu 3018 with an engraving bit. Zero chemicals, but traces are rougher — 10 mil practical floor — and fine-pitch QFPs get dicey. Bantam is the "it just works" option; the 3018 is the "$250 and a weekend of tuning" option.

Assembly: hotplate reflow ~$100 total

Either route, add an MHP-30 or a $60 UYUE hotplate plus a laser-cut stencil and you're reflowing SMD boards the same afternoon you design them.

Skip: Voltera-style conductive-ink printers. Slow, resistive traces, dead end.

U2 Outsourced — least effort, 2–7 days

U3 The recommended pipeline

Fiber laser + etch tank + hotplate, ~$2–3k all in. It's the only path measured in hours, not days — every reroll of a footprint mistake is free. Then run both lanes in parallel:

• Iterate in-house, same day. Etched board, wire vias, hotplate reflow. Ugly but truthful.
• Fire the JLCPCB order the moment a rev looks right. The masked, plated-via version is inbound while you keep working.

U4 Can this make money?

What doesn't work: selling bare boards. You will never out-fab JLCPCB at $2 a board; anyone selling bare etched boards is competing against a Shenzhen fab line with a garage chemistry set. What sells:

• Lane 1 — Same-day turnaround for local labs & startups. US fast-turn shops charge $200–400+ for 1–3 day service on the same board JLC sells for $2. A rig that delivers a working prototype same day, locally sells into exactly that gap — and the buyers (university labs, hardware startups, R&D groups) are the ones whose burn rate makes a lost week cost more than any board. You're charging for hours-not-days, not for FR4, and no national player can serve "this afternoon."
• Lane 2 — Small-run assembly. US small-batch PCBA carries $300–500 setup fees, so 10–50 board runs are orphaned. Stencil + hotplate + steady hands does those runs profitably at hourly-labor economics — and the same local labs from Lane 1 are the natural first customers. Rework/microsoldering repair pays real hourly rates on gear you already own.
• Lane 3 — Your own products, sold online. Tindie/Crowd Supply/direct-store niche modules — the proven ceiling for garage PCB operations. Margin lives in the design; the rig compresses the iteration loop and the assembly ladder ships the units. Service lanes 1–2 are the paycheck while lane 3 compounds.

U5 The path to more sophisticated boards

The DIY fab ladder, with honest difficulty:

• 1 — You are here. 1–2 layer, 6–8 mil, wire/rivet vias, no mask. Covers ~80% of prototypes.
• 2 — Solder mask. UV-curable photoimageable mask, exposed with the laser or a UV lamp. A weekend of process tuning. Worth doing.
• 3 — Plated through-holes. Conductive-ink activation + a copper electroplating tank (~$500 in chemistry). Doable, genuinely fiddly — the first rung where most people rationally stop.
• 4 — Multilayer. Prepreg lamination in a heated press, cross-layer registration. Possible at home; miserable, low-yield, and JLC sells 4-layer for ~$8. Don't.
• 5 — Pro gear. LPKF ProtoMat + plating system ($20–50k), or used fab-line equipment from closing US board shops ($100k+). Only makes sense as a business, not a capability.

The money path in one line: rig compresses iteration → sell speed and small-run assembly locally → design something worth owning → outsource the fancy boards, assemble and test in-house, keep the product margin.