Module 1: Bring-Up & First Power
Sequence a safe bring-up, run a smoke test, and localize a fault. Mix of concept and calculation, then a debug scenario.
-
1conceptA DMM reads 5 Ω from a 3.3 V rail to GND on an unpowered, freshly assembled board. What is the right next action?
-
Full power into a 5 Ω rail-to-GND can dump amps into a fault and destroy parts before you learn anything.
-
Correct. A few ohms rail-to-GND on an unpowered board is a low-impedance path: assume a short (bridge, reversed cap, flipped part) and bound it before applying power.
-
A healthy rail reads far higher than 5 Ω to GND unpowered (decoupling is high-impedance at DC); 5 Ω is a red flag.
-
You have not localized the fault yet: swapping the regulator is a guess, and the short may not be the regulator.
-
-
2conceptOrder these bring-up steps correctly: (1) load firmware, (2) check rails, (3) rail-to-GND short check, (4) apply current-limited power.
-
Loading firmware first means powering an unverified board and trusting logic before power, which is backwards.
-
Correct: short-check unpowered, then current-limited power, then verify rails, then bring up the brain. Power before logic, incrementally.
-
Applying power before the short check skips the cheapest, safest test.
-
Checking rails before any power is applied measures nothing meaningful.
-
-
3conceptBest instrument to confirm a crystal oscillator is actually starting?
-
A DMM shows an average, not the oscillation; a running crystal can read a meaningless mid-rail DC.
-
Correct. You need to see frequency and amplitude of a fast, low-amplitude AC signal: a scope with a low-C probe (a 10x passive loads less than 1x) and a short ground lead avoids killing or distorting the oscillation.
-
A logic analyzer only sees digital thresholds, not the analog start-up envelope, and its capacitance can stop a marginal oscillator.
-
Thermal tells you nothing about whether the oscillator is oscillating.
-
-
4calcA board's expected idle current is 80 mA. Where should you set the bench supply's current limit for the first power-on?
-
Setting the limit exactly at idle leaves no headroom: normal inrush or a slightly higher real idle trips it and masks a healthy board.
-
Correct: a limit a little above expected idle lets a healthy board run but turns a short into a gentle, watchable current-limited event instead of a bang.
-
A 3 A limit defeats the purpose: a short would draw amps and damage parts before you react.
-
Below idle, a healthy board can never reach its operating point, so you learn nothing.
-
-
5scenarioOn first power-on the board pulls ~2x expected current and one IC gets hot. Which sequence best localizes the fault?
-
Forcing more current into a fault is how you turn a debuggable board into a dead one.
-
Correct: current-limited re-power makes the fault observable; thermal/freeze-spray points at the part; the DMM confirms the shorted rail; isolating a section (lift a pin / cut a rail) bounds it to a root cause (bridge, reversed electrolytic, shorted decoupling cap, misoriented part).
-
Blind reflow is a shotgun fix that destroys evidence and rarely finds the real cause.
-
Swapping the part skips root-cause; if the real fault is a bridge or reversed cap, the new IC cooks too.
-