Planning a Prototype: the short answer
Good projects start with requirements, BOM, interface map, safety checks, test plan and fallback choices. A requirement should describe what is measurable: sample interval, operating location, expected duration, accuracy class and failure response. Example: A one-page plan that fixes the parts list before buying, and names the risk you will test first.
- Level
- Beginner
- Time
- About 5 min
- Topic
- PCB and making
Write requirements, choose parts and list risks before buying anything.
Good projects start with requirements, BOM, interface map, safety checks, test plan and fallback choices.
A requirement should describe what is measurable: sample interval, operating location, expected duration, accuracy class and failure response. Architecture maps functions into compatible modules before any purchases.
Build a requirements table with must-have versus optional features, then create a BOM whose rows reference exact board and module variants. Track unresolved assumptions as explicit engineering risks.
Start from what must be true
A good requirement can be checked. "Monitor the room" cannot be checked. "Read temperature every 5 minutes to within ±1 °C, store a week of data, and show it on my phone" can. Write down how often, how accurately, how long, where, and what should happen when something fails. The numbers choose your parts for you.
Mark each requirement must-have or nice-to-have. Build the must-haves first, and add the rest only if time and budget remain.
Choose by interface, then by part
Sketch a block diagram first: sensors in, controller in the middle, outputs and communication out. Then check that each arrow is possible: do the voltages match, are there enough pins of the right type, is the data bus shared without an address clash, and is there enough supply current for the worst moment, such as Wi-Fi transmit or a motor starting?
Only then pick specific parts. A bill of materials row should name the exact board or module, its supply voltage, its interface and a link or part number, so a different person could buy the same thing.
Budget power and money together
Add up the worst-case current of everything on a rail, then add 30 % margin. An ESP32 with Wi-Fi may draw 250 mA for a short peak, so a weak USB charger or thin wire can reset the board even though the average is low. For battery projects, the average over a day decides the runtime, and sleeping between readings is what makes weeks possible.
In money, leave room for the parts you break or find you did not need: a spare of anything cheap, extra jumpers, one more sensor. A plan with no spare parts is a plan that stalls on the first mistake.
List the risks before they arrive
A risk is something you do not know yet: whether the sensor is accurate enough, whether the Wi-Fi reaches the shed, whether the enclosure gets hot in sun. For each, write how you will find out and what you will do if it goes badly. Test the riskiest parts first, before you spend time on the easy ones.
Plan a test: how you will know the project works. A simple one is enough: place a reference thermometer next to it for an hour and compare.
Keep it small, then grow it
A prototype that does one thing reliably beats one that does five things badly. Get the sensor reading to the serial monitor, then to the network, then to the dashboard, then add the enclosure and the battery. At each stage you have something that works to go back to.
- Buying parts before checking that voltage, interface and pins are compatible.
- Writing requirements with no numbers, so nobody can say whether the project is done.
- Leaving the riskiest part until last.
Requirements for a room temperature logger
- Must: reading every 5 minutes, ±1 °C, store a week, show on a phone.
- Nice: humidity, battery power.
- Board: ESP32 (Wi-Fi). Sensor: BME280 (temperature and humidity, I²C).
- Risk: Wi-Fi drops. Plan: keep a queue of readings and resend.
Result. A one-page plan that fixes the parts list before buying, and names the risk you will test first.
3 quick questions
Which requirement can you test?
When do you write the parts list?
Why list risks early?
