Chris Collins

Tools · electronics

Fault-finding trainer

A repair bench for the circuit library's circuits: one secretly gets a fault — an open resistor, a dried-out capacitor, an LED in backwards, a solder bridge between neighbouring pads, a broken track — and you find it with a virtual multimeter and scope on the PCB, breadboard or schematic, in as few measurements as you can.

Added

Works with

How to use it

Fault finding is a skill you only learn on broken things, and broken things are hard to come by on purpose. So this bench breaks them for you. It takes a circuit from the circuit library, secretly injects one realistic fault, and simulates both the good and the faulty circuit with circuit-sim's solver. You get the customer’s complaint — "the LED doesn’t light", "the flasher runs too fast", "the output is stuck at 9 V" — and a multimeter.

How to use it

  1. Pick a mode. Practice shows what a good board reads beside every reading, and the table of expected voltages from the start. Challenge shows nothing but your readings until you call it. Teach me walks through three hand-written jobs step by step, the way an expert would take them.

  2. Pick a circuit, or leave it on Any circuit and choose a difficulty: easy faults are opens, shorts, parts in backwards and dead supplies; medium adds wrong values, transistors, dead chip outputs, broken tracks and missing jumpers; hard adds leaky and dried-out capacitors and solder bridges.

  3. Pick the build — a routed PCB or a breadboard. Some faults belong to the build: a solder bridge joins two pads that are really next to each other on that board, and on a breadboard two leads in neighbouring holes can touch, or a jumper wire can be missing (look closely — it is not drawn).

  4. Read the complaint, then measure. Tap a pad (or a lead, or a pin on the schematic) to put the red probe on it; the black probe stays on 0 V unless you untick it, and then the next tap places it. The ranges:

    • DC volts (1 time unit) — what a DMM shows, which for a varying signal is its average.

    • Scope (2) — the waveform at a point, against ground; in practice mode the good board’s trace is drawn dashed behind it.

    • Resistance (2) — with the power off, in circuit.

    • Diode test (2) — with the power off: about 0.6–0.7 V across a silicon junction, 1.7–2 V across a red LED, OL for an open or a reversed junction.

    • Continuity (1) — a beep below 30 Ω.

  5. Call it. Pick the part and what is wrong with it (for a bridge, the two nets; for a broken track or a missing jumper, the net). A right call at par (3 time units) or under scores 1000; each time unit over par takes 8 % off, each wrong call 40 %; practice scores half.

  6. Read the explanation: the fault, what the readings should have been against what they were, the expected-voltage table with the wrong nets marked, and the path an expert would have taken — supply first, half-split, compare with the expected voltages, then confirm with the power off.

Your progress — jobs solved, the measurements you needed per fix, your points — stays in this browser.

The faults

  • Resistors: open, shorted, or ten times too high or low — the multiplier band read one colour off (orange for red, say).

  • Capacitors: open, shorted, leaky (conducting like 2.2 kΩ), and for electrolytics dried out (a tenth of the capacitance with 22 Ω of ESR) or fitted backwards.

  • Diodes and LEDs: open or fitted backwards (and shorted, for diodes).

  • Transistors: an open base junction, or a collector–emitter short (a MOSFET: an open gate or a drain–source short).

  • Chips: a 555, op-amp or LM386 with a dead output stuck low; a regulator with no output; an optocoupler with an open LED or a collector–emitter short.

  • Relays, buttons, inductors and sensors: a relay with an open coil or contacts welded shut; a push button that never makes contact or is stuck closed; an inductor open or shorted; an LDR or thermistor cracked open or shorted out.

  • Supplies: dead (0 V), a flat battery (60 % of its voltage behind 20 Ω), or no input signal.

  • The board: a solder bridge between adjacent pads (pad centres within 2.6 mm, one 0.1″ pitch), a hairline crack in a track — found by taking each copper segment out of the routed board in turn and keeping those that cut a net in two — or, on a breadboard, two leads touching in neighbouring holes or a missing jumper.

Every fault is simulated before it is dealt, and only offered if the readings somewhere differ from the good circuit’s and it produces a symptom you could notice — an LED, one of the circuit’s probes, or one of its stated expectations failing — and only if the voltages really move: somewhere on the board, at some moment of the run, by more than 0.1 V. In a circuit worked by a button, a light or a temperature change, the readings are also compared at the end of each step of that test, so a button that never makes contact shows once it has been pressed. A fault the simulation cannot see (the 555 flasher’s supply decoupling capacitor going open, for instance — the solver’s ideal battery does not care) is never dealt, and a fault that stops the solver converging or makes it oscillate faster than the time step can follow (more than five times the good circuit’s time points) is left out too.

How the readings are made

The circuit is turned into circuit-sim parts on labels: each part pin has a pad (its hole on the board), each pad sits on a copper island, and normally a part pin connects to its pad’s island. A fault inside a part (an open junction, an LED in backwards) moves the part’s pin; a fault in the wiring (a broken track, a missing jumper) moves the pads; shorts and bridges add a small resistor (0.02–0.5 Ω). The labels become a circuit-sim schematic directly — every pin wired diagonally to one hub point per label, so the simulator’s connectivity is exactly the labels — and circuit-sim runs its DC operating point, transient and AC analyses. The circuit library’s own expectations are checked against the faulty run to say which ones fail.

  • Volts are the average over the settled part of the transient run, like a DMM’s reading of a varying signal.

  • Resistance with the power off is a nodal analysis of the resistive network with the sources disconnected: resistors, pots and closed switches conduct; charged capacitors are open; and diodes, transistors and chips do not conduct at an ohmmeter’s low test voltage (about 0.2 V on most DMMs' resistance ranges), so they are left out. Above 40 MΩ reads OL.

  • The diode test runs circuit-sim with the power off and a 3 V source behind 2.2 kΩ between the probes — about 1 mA into a junction, as a DMM’s diode range does — and reads OL above 2 V (so a blue or white LED reads OL on many meters, as it does here).

Assumptions to keep in mind: a chip with no power is treated as open on the ohms and diode ranges (a real one has protection diodes that sometimes show up); supplies and signal sources are ideal, except that a fault adding a low resistance (a short, a bridge, a leaky capacitor) sees them behind 0.5 Ω (a supply) or 50 Ω (a signal source) — across an ideal source a short would change nothing, where a real one pulls the rail or the signal down; and a "dried-out" capacitor is one simple model of age, not a measured one. The expert’s path is generated from the actual readings: the supply first, a half-split of the nets from the supply outwards, the nets that read wrong and the parts on them, and a confirming measurement with the power off.

Runs entirely in your browser; nothing you type leaves this page.