🔬 The Bring-Up Mindset

Bring-up is the disciplined ritual that turns a bare assembly into a trusted system without destroying evidence along the way. The whole craft rests on one ordering rule: power before logic, and bring power up incrementally, never all at once. Before any voltage touches the board, you inspect; when voltage does arrive, it arrives through a leash you can yank.

By the end, you can

1. Justify the power-before-logic bring-up order and its incremental ramp
2. Run a pre-power inspection: AOI/visual, polarity, keying, and a rail-to-GND short check
3. Read a rail-to-ground resistance and decide whether it is safe to apply power
4. Explain why a current-limited supply is the single most important bring-up tool

Intuition first

Think of a fresh board the way a bomb-disposal tech thinks of a package: assume the worst, move in a fixed order, and never do the irreversible step first. The irreversible step here is applying full power. A short you cannot see will, at a stiff supply, pull tens of amps through a hair-thin trace or a tiny part and convert your evidence into a charred crater in milliseconds.

The fix is almost embarrassingly simple: put a current-limited bench supply between the wall and the board, and set the limit just above what a healthy board should draw. Now a short does not explode. It gently pins the supply at its limit while the voltage sags, and you stand there watching the fault happen safely. The current-limited supply is a circuit breaker slow enough to learn from.

The pre-power inspection

Nothing here needs the board to be alive. Each step is cheap, and each one can save a part.

Look before you leap (AOI / visual)

Automated optical inspection (AOI) is what the assembly house runs; on your bench the human version is a loupe or a microscope and the assembly drawing. You are hunting for the classic defects: solder bridges between fine-pitch pins, tombstoning (a passive stood up on one end like a domino), missing or rotated parts, and the two that bite hardest: polarity on electrolytic capacitors and diodes, and connector keying. A backwards electrolytic is a future bang; a connector you can plug in two ways is a future reversed supply.

Continuity and the rail-to-GND short check

Now the multimeter, board unpowered. Set it to continuity and resistance and check the thing that matters most: each power rail to ground. A healthy rail reads high resistance to ground at DC: the decoupling capacitors look like an open circuit once they finish their tiny initial charge through the meter. A rail that reads a few ohms to ground is almost certainly shorted: a bridge, a reversed part, or a cap that died on the line.

How low is “too low”? Reason it from Ohm’s law. If a rail that should sit at $V$ volts reads $R$ ohms to ground, then at full power it would try to pass

$$I = \frac{V}{R}$$

through that path. A 3.3 V rail reading 5 Ω to ground would draw

$$I = \frac{3.3\ \text{V}}{5\ \Omega} = 0.66\ \text{A}$$

into a dead short, far more than the few-milliamp idle of a sleeping driver board, and concentrated in whatever tiny defect is the short. That is your stop sign.

A worked bring-up order

Put the pieces in sequence and the rest of this module falls out of it. For the robot hand’s finger-driver board:

1. Inspect: loupe the fine-pitch motor-driver IC and the connector, confirm the electrolytic on the motor rail is the right way round, check the BOM.
2. Short-check: DMM each rail (3.3 V logic, 5 V sensor, the motor rail) to GND. All read high; none reads single-digit ohms. Cleared to proceed.
3. Current-limited power: bench supply on the motor rail, limit set just above the board’s expected idle. Ramp $V_{in}$ up; watch the current.
4. Rails / clocks / reset: confirm each rail is at voltage with acceptable ripple and in the right order; confirm the oscillator starts and reset releases cleanly.
5. Brain: attach the debugger over JTAG/SWD, get a heartbeat LED or a debug-UART banner, then load firmware.
6. Buses, one at a time: bring up the encoder SPI, then the sensor I²C, then CAN.
7. Functional test: run the built-in self-test, exercise a finger, document everything you measured.

Skipping straight to step 5 on a board that failed step 2 is how benches catch fire.

?A DMM reads 4 Ω from the 5 V sensor rail to GND on an unpowered, just-assembled board. What now?

• A Apply 5 V at full current to confirm the rail works
• B Treat it as a probable short and localize it before applying power
• C Nothing: 4 Ω is normal once decoupling caps are fitted
• D Bump the supply current limit so the rail can settle

?Why does a current-limited supply make bring-up safer than a stiff wall adapter?

• A It outputs a cleaner, lower-noise voltage
• B It caps the fault current and lets the voltage sag, turning a short into a slow, watchable event
• C It electrically isolates the board from earth ground
• D It automatically sequences the rails in the correct order

Lab: cleared-for-power checklist

On any board you can get your hands on, run the pre-power pass before you ever switch on: visually inspect against the BOM/assembly drawing; with the DMM, log each rail’s resistance to ground; note the board’s expected idle current from the datasheets. You should finish with a one-line verdict per rail (“3.3 V → 1.2 kΩ, OK”) and a current-limit value to dial in next lesson. If any rail reads single-digit ohms, you have found a fault before it found you.

↘ What AOI actually checks, and why electrolytics fail backwards

Automated optical inspection lights the board from several angles and compares each joint and part against a known-good reference: presence, position, polarity marks, and solder-fillet shape. It is fast and catches the population defects (missing, rotated, tombstoned, bridged) but it cannot see electrical faults: a cracked part that still looks fine passes AOI and fails your short check. The two inspections are complementary: AOI for “is the right part there, the right way, soldered well,” DMM for “is anything shorted.”

Reversed electrolytic capacitors deserve their fearsome reputation. The dielectric is a wafer-thin oxide grown electrochemically on the anode foil, and it only stays intact while the correct polarity holds it there. Reverse the voltage and the oxide breaks down, leakage current heats the electrolyte, the can pressurizes, and the vent (or the whole case) lets go. That is why polarity is a pre-power check: once you apply power to a backwards electrolytic, the failure is already in motion.

Grounded in Wikipedia: “Automated optical inspection”, “Electrolytic capacitor”, “Printed circuit board” (CC BY-SA).

Key takeaways

• ▸A fresh board is guilty until proven innocent: inspect first, apply power last.
• ▸The bring-up order is inspect → short-check → current-limited power → rails/clock/reset → brain → buses → functional, documenting throughout.
• ▸Rail-to-GND single-digit ohms on an unpowered board means a probable short: do not apply power.
• ▸A current-limited supply is a circuit breaker slow enough to watch: it turns a short into a safe, diagnosable event.
• ▸Polarity and keying are pre-power checks because reversed electrolytics fail the instant power arrives.

Practice 1 warm-up

Before powering a board, you measure the 3.3 V rail to ground and read 1.8 kΩ, and the motor rail to ground and read 3 Ω. Which rail, if either, blocks power-on, and why?

↘ Show worked solution

The motor rail blocks it. 1.8 kΩ on the 3.3 V rail is a normal high-impedance reading (decoupling looks open at DC), fine. 3 Ω on the motor rail is a near-short: at, say, 12 V it would try to pass $I = 12/3 = 4$ A into a defect. Stop, and localize the motor rail short (look for a reversed bulk electrolytic or a bridge at the driver IC) before applying any power.

Practice 2 core

Put these in the correct bring-up order and say why the first and last belong where they do: load firmware, check rails, rail-to-GND short check, apply current-limited power.

↘ Show worked solution

Order: short-check → apply current-limited power → check rails → load firmware. The short check is first because it is unpowered, free, and catches the faults that would destroy the board the instant power arrives. Loading firmware is last because it trusts that power, clocks, and reset are already proven. Running code on a board whose rails you have not verified just adds an unknown on top of an unknown.

Practice 3 stretch

Your healthy finger-driver board idles at about 70 mA. You set the bench supply’s current limit to 90 mA for first power-on. On a second board the supply immediately pins at 90 mA and the output voltage collapses to 1.5 V. What does this tell you, and what is your next move?

↘ Show worked solution

The board is drawing far more than its 70 mA idle: the supply has hit its 90 mA limit and is holding current constant while voltage sags (exactly the behaviour you set it up to show). That is the signature of a short or a grossly over-drawing fault. Next move: leave it current-limited (it is safe there), and localize: feel/scan for the part that is heating, hit suspects with freeze spray, and DMM the collapsed rail to confirm the low impedance. You are now in the smoke-test localization loop, which is the next lesson.