Systematic Troubleshooting: the short answer
Check power and grounds first, then interface configuration, hardware identity, observed symptoms and log output. Make one controlled change at a time. Example: Fixed in four checks. Each step ruled one thing out instead of rewiring at random.
- Level
- Beginner
- Time
- About 5 min
- Topic
- Tools and measurement
Find faults in a fixed order and change one thing at a time.
Before this: Using a Multimeter
Check power and grounds first, then interface configuration, hardware identity, observed symptoms and log output. Make one controlled change at a time.
A reproducible investigation records the symptom, last known good state, wiring and software revision, measured voltages and a single controlled change. Random rewiring destroys evidence and adds new failure modes.
Follow a safe fault-isolation order: power off visual inspection, supply and grounding review, interface identity check, firmware/log review, and controlled functional validation. Stop at any sign of overheating or unsafe behavior.
Why order matters
Most faults in a beginner project are boring: a missing ground, a loose jumper, the wrong pin number, a cable that only charges. Checking those in a fixed order finds them faster than a clever guess, and it stops you rewiring a working circuit while the real fault sits untouched.
Change one thing at a time and write down what you changed and what happened. If two changes go in together and the circuit works, you never learn which one mattered.
A routine that works
Power off first. Look: any part hot, burnt, backwards or touching its neighbour, any wire in the wrong row? Then power on and measure, rather than assume. Is the supply voltage at the chip's own pins, not just at the battery? Is ground connected everywhere, including between separate boards?
Then check identity: is this the board and port you selected, the right pin number for this board, the right I²C address? Then look at the logs: serial output at the correct baud rate shows far more than blinking lights do. Only then change code.
Split the problem in half
When the system has many parts, test the middle. Does the sensor give a valid reading on the serial monitor? If yes, the fault is downstream (network, dashboard); if not, upstream (sensor, wiring, power). Each test removes half the possibilities. Replace a doubtful part with a known-good one rather than reasoning forever about it.
Print values at each stage: the raw ADC number, the converted voltage, the final reading. A bug is usually where two stages disagree.
Signs that mean stop
Stop and disconnect power if anything gets hot enough to be uncomfortable to touch, smells burnt, or a battery swells. Never leave a lithium cell charging unwatched while debugging. If a part is hot, find out why before powering it again, because the cause is usually still there.
Keep a photo of the last working state. It is the cheapest way to get back to something known.
Write it down
Note the symptom, the last time it worked, what changed, what you measured, and what fixed it. A short fault log builds into the most useful reference you will own, and it makes asking for help easy: people can help when you can say what you tried and what you saw.
- Rewiring several things at once, then not knowing what fixed or broke it.
- Assuming the power is fine without measuring it at the part.
- Skipping the serial monitor and guessing from blinking lights.
The BME280 says 'not found'
- Power: 3.3 V measured between VCC and GND pins on the module.
- Wiring: SDA to GPIO21, SCL to GPIO22, checked against the photo.
- Scanner: finds 0x77, but the code calls begin(0x76).
- Change to begin(0x77): readings appear.
Result. Fixed in four checks. Each step ruled one thing out instead of rewiring at random.
3 quick questions
What should you check first when nothing works?
Why change only one thing at a time?
A part becomes hot. What do you do?
