A non-contact AC voltage detector is the cheap pen-shaped tool an electrician waves near a wire to check it’s live without touching it. For an electric-circuits project I built one from scratch — an antenna, three transistor gain stages, an LED, and a buzzer. The part worth writing down isn’t the schematic that’s on every datasheet; it’s that I built it three times and it failed three different ways.
The principle is small. A live mains wire radiates a weak 50/60 Hz electric field, and an antenna held near it picks up a tiny voltage by capacitive coupling — far too small to light anything. So the signal runs into three 2N3904 transistors (β ≈ 100) in cascade, biased by a 100k/10k/1k resistor chain, until it’s big enough to drive the LED and a piezo buzzer off a 9 V battery. I sized the bias network in a QUCS simulation first, aiming at ordinary 220 V AC.
Prototype one I soldered point-to-point. Dead. The current through the chain never got high enough to trigger the indicators, and the likeliest culprit was the antenna — too short to couple enough field in the first place. A detector that can’t gather the signal can’t amplify its way out of the problem.
Prototype two went onto a PCB with copper traces, and failed in the exact opposite direction: one transistor burned out from excessive current. So within two builds I’d hit both failure modes of the same circuit — starved in one, cooked in the other — which is a blunt lesson in how completely a three-stage amplifier rides on its bias being right on the bench, not just right in the simulator.
Prototype three went on a breadboard, and it finally lit up. Held near a bare, exposed conductor, it detected the AC cleanly. And then it failed the only test that mattered: held near a normal insulated wire — the kind you’d actually be checking — it read nothing. The field that leaks past insulation at a working distance was weaker than my gain chain could pull out of the noise.
That’s the one that stuck. The demo condition (bare wire, close, strong field) and the real condition (insulated wire, everyday distance) were different problems, and passing the easy one told me almost nothing about the hard one. A non-contact sensor lives or dies on its weakest realistic signal, not its strongest convenient one — so the spec that matters was never “can it detect AC,” it’s “can it detect AC through the insulation, at the distance, the job actually has.” Antenna length, bias, gain — everything upstream is just trying to earn that one margin.