The Most Dangerous Second in a PCB Project: A Safe First Power-On
Series navigation: Episode 1: The Roadmap · Episode 7: Reading Arduino Pin-Current Specifications · Episode 15
The problem: Connecting full power immediately can turn one assembly error into several damaged parts.
The goal: Make first power-on a sequence of reversible tests with stop conditions.
The result: A bring-up checklist from visual inspection to firmware and interfaces.
AI-generated illustration of preparation for first power-on. No measurement shown is evidence from a real board.
I wanted to plug in USB and see the LED blink
After waiting for a new PCB, the natural impulse is to connect everything and run the full program. That single action combines too many unknowns: assembly, polarity, shorts, regulators, clocks, reset, programming, firmware, and external devices.
If smoke appears, the evidence has already been destroyed.
First power-on should be deliberately boring. The objective is not to demonstrate the whole product. It is to discover the first incorrect condition while energy is limited.
Define success and stop conditions first
For each stage, write expected voltage, current, temperature, logic state, and time. Also define when to stop:
- supply enters current limit unexpectedly;
- input current exceeds the calculated range;
- a rail is outside tolerance;
- resistance to ground is suspiciously low;
- a polarized part is reversed;
- any component heats rapidly;
- smell, sound, discoloration, or unstable cycling appears.
Power off first. Diagnose second. Repeatedly applying power to “see what happens” can turn a recoverable short into a failed board.
Stage 0: inspect without power
Under magnification, compare the board with assembly drawings and the BOM. Look for solder bridges, unsoldered pins, tombstoned parts, reversed diodes or capacitors, wrong resistor values, shifted fine-pitch packages, debris, and connector orientation.
Use continuity or resistance checks with power removed:
- input power to ground;
- each regulated rail to ground;
- adjacent fine-pitch pins where bridges are plausible;
- fuse and protection paths;
- ground continuity between expected points.
A low resistance is not always a short because capacitors charge and real loads may be low impedance. Compare with the design and observe whether the reading changes rather than applying a universal threshold.
Stage 1: power only what must be powered
Disconnect motors, actuators, optional modules, batteries, and external cables. If the architecture permits, isolate downstream rails or leave optional loads unpopulated.
Use a bench supply set to the correct voltage with a conservative current limit based on the expected idle consumption. Start at zero output, connect with polarity verified, enable power briefly, and watch current immediately.
A current limit is protection, not proof of safety. Set too high, it cannot protect the board; set too low, it may prevent normal startup and create misleading resets.
Stage 2: verify the power tree
Measure in dependency order:
input connector
→ protection device
→ pre-regulator node
→ each regulated rail
→ reset and enable rails
→ current at steady state
Check polarity and tolerance against the component data sheets. Use an oscilloscope when startup ramp, oscillation, ripple, or sequencing matters. A multimeter value alone may hide repeated startup-and-shutdown cycles.
After a short powered interval, remove power and check for unexpected heating carefully. A thermal camera can help, but component temperature limits and safe handling procedures still apply.
Stage 3: prove reset, clock, and programming
Before running the application:
- verify reset is at the expected level and timing;
- confirm required oscillators or clocks;
- connect the programming/debug interface;
- read the device identity if supported;
- save option bytes, fuses, or configuration before changing them;
- program a minimal known firmware image.
The first firmware should exercise one observable function, such as a slow LED heartbeat or serial message. Avoid enabling every peripheral and load at once.
Stage 4: add one subsystem at a time
Bring up GPIO, serial output, ADC reference, I2C, SPI, sensors, storage, radio, and power loads separately. Record expected and observed results at each test point.
For a motor stage, begin without the motor where possible, verify gate default state and PWM, then attach a current-limited load appropriate to the design. For a communication bus, confirm idle voltage before sending transactions.
This produces a dependency tree. If I2C fails after its rail, reset, and clock already passed, the search space is much smaller.
A bring-up record
| Stage | Expected | Observed | Pass/stop | Evidence |
|---|---|---|---|---|
| Unpowered input resistance | Design-specific, no hard short | Not yet measured | Pending | Meter setup and board ID |
| Current-limited input | Within calculated idle range | Not yet measured | Pending | Supply settings |
| Main rail | Within regulator tolerance | Not yet measured | Pending | Test-point reading |
| Reset and clock | Match timing requirements | Not yet measured | Pending | Scope capture |
| Programmer connection | Device identified | Not yet tested | Pending | Tool log |
The “not yet” entries are intentional. This guide does not invent a board or readings. Use the table as a template for real evidence.
Completion checklist
- [ ] Expected values and stop conditions are written before power is applied.
- [ ] The board was visually and electrically inspected unpowered.
- [ ] Optional loads are disconnected for initial testing.
- [ ] Supply voltage, polarity, and current limit are verified.
- [ ] Rails are measured in dependency order.
- [ ] Reset, clock, and programming work before full firmware runs.
- [ ] One subsystem is added at a time.
- [ ] Every anomaly causes a stop and written investigation.
- [ ] Measurements include settings, test point, board revision, and units.
Next experiment
Bring-up does not end when a defect is found. The final step is to convert every workaround and failure into an explicit Revision B change that can be reviewed and retested.
Comments
Post a Comment