A working prototype on the bench feels like the finish line. In reality it is closer to the halfway point. Between a board that works once on an engineer's desk and a product that works every time in a customer's hands, there are stages that are easy to underestimate: testing, design for manufacturing, small-batch builds and preparing for certification.
This article walks through the stages most embedded products go through, what each one should produce, and the mistakes that most often cost time and money. Whether you build in-house or with a partner, knowing the path helps you plan budget and timelines realistically.
1. Requirements: write down what "done" means
Most project overruns start here. "A device that measures temperature and sends it to the cloud" is an idea, not a requirement. Before any schematic is drawn, the team should agree on:
- Function — what the product measures, controls or displays, and how accurately.
- Power — mains, 12/24 V industrial supply, battery or solar, and the battery life target.
- Connectivity — WiFi, Ethernet, 4G, LoRa, Bluetooth, RS-485, or none.
- Environment — temperature range, dust, moisture, vibration, electrical noise.
- Size and enclosure — DIN rail, wall mount, handheld, or a board inside someone else's machine.
- Volume and cost — 50 units a year and 50,000 units a year lead to very different designs.
- Compliance — which markets the product will be sold in and which certifications that implies.
A one- or two-page requirements document, reviewed by both the business and engineering sides, prevents the most expensive kind of change: the one discovered after the PCB is made.
2. Architecture and component selection
With requirements fixed, the engineer chooses the building blocks: the microcontroller or processor, the radio, the power supply topology, sensors and interfaces. This is where a microcontroller-based design is weighed against an embedded Linux one, and where long-term availability matters as much as price.
A part that is cheap today but single-source or near end-of-life can force a redesign two years later. Good practice is to prefer components with multiple sources or pin-compatible alternatives, and to check stock and lifecycle status before committing.
3. Schematic design
The schematic turns the architecture into a complete circuit: power rails, protection, the processor and its support circuitry, interfaces, connectors, and test points. A careful schematic review — ideally by a second engineer — catches problems such as missing pull-ups, wrong voltage levels, insufficient protection on field wiring, or a regulator that will run too hot.
For industrial products, protection deserves extra attention: reverse polarity, surges on long cables, ESD on exposed connectors and isolation on RS-485 or analog inputs are often what separates a lab prototype from a product that survives in a panel.
4. PCB layout
Layout is where the circuit becomes a physical board. Decisions on layer count, component placement, ground planes, trace widths and spacing affect reliability, noise, heat and manufacturing cost. Switching regulators, radios and high-speed signals need particular care.
The output is not just a board picture but a manufacturing package: Gerber or ODB++ files, drill files, a bill of materials with manufacturer part numbers, pick-and-place files and assembly drawings. If any of these is incomplete, the fabricator or assembler has to guess — and guesses become defects.
5. Firmware and software
Firmware development usually starts in parallel on a development kit and moves to the real board when it arrives. Typical work includes drivers for sensors and peripherals, communication stacks such as Modbus or MQTT, power management, a bootloader for field updates and the application logic itself.
Two things are often left too late. The first is field updates: if the product cannot be updated safely after shipping, every bug becomes a site visit. The second is logging and diagnostics: when a unit misbehaves at a customer site, you need a way to find out why. If the product talks to the cloud or a phone, the companion web or mobile application is part of the same plan.
6. Prototype bring-up and testing
The first boards are brought up step by step — power rails first, then the processor, then each interface. Problems found here are normal; the point of a prototype is to find them. Once the board works, it needs testing beyond "it powers on":
- Functional tests of every feature against the requirements.
- Accuracy and calibration checks for measurements.
- Temperature, supply-voltage and long-duration (soak) tests.
- Power consumption measurements for battery products.
- Pre-compliance checks for emissions and immunity, if certification is planned.
Expect at least one board revision. Planning for it in the schedule is far better than hoping it will not be needed.
7. Design for manufacturing (DFM) and testing in production
A design that a skilled engineer can assemble by hand may still be hard to build in volume. DFM review looks at component packages, panelisation, fiducials, clearances, and whether parts are available from the assembler's usual suppliers. Design for test (DFT) asks how every unit will be checked quickly at the end of the line — test points, a test fixture, and test firmware that reports pass or fail.
8. Small-batch production
Before ordering hundreds or thousands of units, build a small batch — often 10 to 50. This shakes out assembly issues, validates the test process, and gives you units for field trials, customer demos and certification testing. Feedback from the first field installations frequently leads to small but valuable changes in firmware, labelling or the enclosure.
9. Certification preparation
Depending on the product and the market, certifications may include EMC and safety testing, radio approvals for wireless products, and country-specific registrations. Certification itself is done by accredited labs, but preparation is part of development: designing with compliance in mind, running pre-compliance tests, and assembling the technical documentation labs ask for. Discovering a failing emission test after the design is frozen is one of the most expensive surprises in product development.
10. Documentation and handover
At the end of the project you should own a complete set of design files, source code, build instructions, test procedures and the bill of materials. Without them, every future change depends on the people who built the first version.
How SensorLoop can help
At SensorLoop we work across these stages — from hardware design and PCB layout to firmware, embedded Linux and application development. Our own ModbusSudo gateway went through the same process. If you have a product idea, or an existing product that needs a redesign, tell us about it.