Reverse-Polarity Damage in ICs: The Hot Chip and the Diode-Asymmetry Test

A board that used to work now does this: within seconds of applying power, one integrated circuit — usually the largest, often the microcontroller — becomes too hot to touch, and nothing else on the board comes to life. That combination is diagnostic. Heat means current, and a part that heats while everything else stays dead is the part the current is flowing *through* instead of into the rest of the circuit. Disconnect power first. Every extra second at fault current drives the damage further. ## Why reverse polarity kills a chip that has no obvious protection diode CMOS integrated circuits are full of parasitic diode junctions — every well-to-substrate boundary forms one — plus deliberate ESD clamp diodes on each I/O pin, tying the pin to the supply rail and to ground. In normal operation these are all reverse-biased and electrically invisible. Connect the supply backwards — battery positive on GND, negative on VCC — and the entire set forward-biases at once. Nothing in that path limits current except the source's own ability to deliver it. The chip's absolute-maximum ratings, which typically forbid any pin going more than ~0.3 V below its ground reference, are violated across the whole die simultaneously. What happens next depends entirely on the supply. A weak source sags and browns out harmlessly. A stiff one pushes enough current to melt a junction — see Battery Fault Current: Why the Cell You Pick Decides Whether a Mistake Is Survivable. Once a junction is punched through it stays conductive, so the chip is left with a permanent low-resistance path from VCC to GND that draws hard even after the polarity is corrected. ## The diagnostic: measure resistance in both directions With power removed, put a multimeter across VCC and GND, then swap the probes and measure again. The comparison, not the single number, is what identifies the fault: - **Low in both directions** (a few ohms each way) — a metallic short. Solder bridge, damaged trace, or a shorted ceramic capacitor. Copper doesn't care which way the probes point. This class of fault is findable under magnification and often repairable. - **Low one way, high the other** — junction behaviour. Something semiconducting is conducting when it shouldn't be. Combined with a specific chip heating, this localises the damage inside that chip. - **Hundreds of ohms to megohms both ways**, often climbing as the decoupling capacitors charge off the meter's own test current — a healthy rail. A worked example from a real repair attempt: ~20 Ω one direction, ~2.46 MΩ the other. Twenty ohms across a 3 V rail is a 150 mA load — more than enough to make a small QFP package too hot to hold, and enough to collapse the rail so the display never gets a working supply. The five-orders-of-magnitude asymmetry rules out a solder bridge and points squarely at a broken-down junction. The meter's diode-test mode gives the same information in a different unit, reading a forward voltage one way and open-circuit the other. ## Why this is usually terminal The failed junction sits inside a sealed package. There is nothing to scrape away, reflow, or bridge. The only real fix is replacing the chip, which requires the identical part carrying identical firmware. For a mask-programmed or factory-flashed consumer microcontroller, that part is almost never sold separately, and swapping a fine-pitch package calls for a Hot Air Rework Station: The Essential Tool for SMD Soldering and PCB Repair and a steady hand for a low-probability outcome. Sourcing a replacement board is usually the rational path — one more instance of the economics the Right to Repair Movement (2018-2026) pushes against. If you do pull the chip, re-measure VCC to GND with it off the board. If resistance returns to a sane value, the chip was the entire fault. If the short persists, something else on the rail — a failed capacitor, a regulator — is damaged too, and the board is further gone than you thought.