The dev kit got you to a demo. It also hid a set of decisions that the kit’s designers made for you (radio, power, connectors, debug access) and that now have to be made again, for your product. Each one is cheap to make early and expensive to make late, because each one reaches into certification, the enclosure or the factory floor. Here they are at a glance, then one at a time.
| Decision | Make it before | What it locks in |
|---|---|---|
| Module or chip-down radio | Schematic | Certification path and RF layout |
| Parts and alternates | Layout | Supply risk and board revisions |
| Power architecture | Schematic | Battery life and battery size |
| Board outline and antenna | Enclosure CAD | Mechanical fit and radio range |
| Test and debug access | Layout | Factory throughput and field diagnosis |
1. A certified radio module, or a radio on your board?
This decision sets your certification path, so make it first. The FCC lets a pre-certified module be used in multiple host products without approving each one, which means you are not certifying a new transmitter. Your board still has to meet Part 15 Subpart B for its own circuitry, the module has to run with the antennas its grant allows, and the module’s FCC ID needs to be shown on your product for its grant to carry over.
Put the radio chip directly on your board instead and the product is an intentional radiator in its own right, which must be certified before you can market it. A chip-down radio can win on unit cost and size at volume. It costs you schedule, RF layout expertise and a full transmitter test campaign.
2. Parts you can buy, twice
The dev kit’s parts were chosen for a kit, not for your supply chain. Before layout, check every active part’s stock and lifecycle status at more than one distributor, and name an alternate for anything single-sourced. A part that goes on allocation after layout can cost you a board revision, and a board revision costs you a build.
3. The power architecture
On the kit, power came from USB and nobody counted microamps. On your board, the regulator, the battery chemistry, the charger and every pull-up resistor decide battery life together. For a device that mostly sleeps, sleep current often matters more than the radio, and the arithmetic is unforgiving (we worked it through here). Choose the regulator by its quiescent current, give sensors a switched supply rail, and decide now whether the product runs on a coin cell, a larger primary cell or a rechargeable pack.
4. The board and the enclosure, together
The board outline, connector positions, mounting holes and antenna are mechanical decisions as much as electrical ones. For modules with a PCB antenna, Espressif’s design guidelines place the antenna feed point near the board edge, ask you to consider how the housing affects the antenna, and recommend at least 15 mm of clearance in all directions inside it. Put a dummy board in the enclosure CAD with that keep-out drawn in before anyone routes a trace.
Harnesses belong in this conversation too. On a core-sampling drill we converted, the harnesses moved to heavier-gauge wire with modular connectors, for higher current and easy reassembly. Pick connectors for the current, the assembly sequence and the repair, not just the pinout.
5. Programming, test and debug access
Every board that leaves a factory has to be programmed and tested by a fixture, not by an engineer with a USB cable. That means test points on the nets that matter, a programming and debug header the fixture can reach, a factory-test firmware image, and a unique ID for each unit. On a modular smart-home program, prefabricated panels were assembled and quality-checked in custom factory jigs with purpose-built test equipment. The jig is a product in its own right, and it needs the board to cooperate.
After the factory, design for visibility. On an industrial gas system we built, a web dashboard exposes every API endpoint, parameter, data stream, event log and safety alert. The telemetry you design in is the debugger you will have in the field.
What to do this week
- Write a one-paragraph decision record for each of the five, with the reason and the date, even if it says “undecided; need X by Y”.
- Run your BOM through two distributors and flag every single-source part and every part marked not recommended for new designs.
- Put a dummy board and battery in your enclosure CAD now, with the antenna keep-out drawn in.
If your product has a radio, our certification map shows how the module decision changes your path. If you want a second pair of eyes on the architecture, send us a brief.