Ghost glow in LED bulbs
A Socratic walk-through of the ghost glow in LED bulbs — reasoned out one step at a time, not lectured.
The question we started with
THE QUESTION #Why do some LED bulbs keep glowing faintly after the switch is turned off?
The switch is off. The room is dark. And in the corner of your eye a bulb is faintly alight — or blinking, once every several seconds, with an obstinate regularity that is somehow worse.
The obvious reading is that the switch has failed. But the same switch ran a filament lamp in that fitting for years, and the filament went black the instant it was thrown. So either the switch changed, or what "off" means changed.
Reasoning it through
REASONING #Begin with what "off" is. We imagine a switch producing a perfect nothing: zero current, zero connection. What it produces is a gap of a few millimetres of air, in a house where the switched wire may run many metres alongside a permanently live one inside the same cable.
Two parallel conductors separated by insulation are a capacitor. A small one — of order a hundred picofarads per metre in ordinary flat twin cable, a recalled figure rather than a measured one — but alternating mains needs no connection to push current through a capacitance. Take a ten-metre shared run, so roughly a nanofarad. Its reactance at fifty hertz is one over two-pi-f-C: 1/(6.28 x 50 x 1e-9), about 3.2 megohms. Across 230 volts that passes roughly 70 microamps.
Hold that against the old lamp. A 60-watt filament on 230 volts draws about a quarter of an amp — some three thousand times more. Seventy microamps heats a filament by an amount nothing in your house could detect. It was always there; the filament had no way to report it.
Now ask what an LED bulb does with the same trickle. It is not a lamp but a small power supply with diodes on the output: mains rectified into a capacitor, and a driver running the LEDs from the voltage on it. If that capacitor sits near 300 volts, 70 microamps delivers on the order of 20 milliwatts — and modern LEDs are efficient enough that a few milliwatts is visible in a dark room.
That also explains the blink, the stranger symptom and the better evidence. If the leakage is smaller than the driver needs to keep running, the capacitor charges slowly, reaches start-up voltage, the driver fires and dumps the stored charge into a brief flash, the voltage collapses below the running threshold, and it repeats. That is a relaxation oscillator, its period set by how fast the leak refills the reservoir — which is why the blink is slow, regular, and different in every fitting.
Capacitive coupling is not the only source. An illuminated rocker switch, an electronic timer, a dimmer or an occupancy sensor all need power for their own electronics and, having no neutral at the switch, take a small standby current through the lamp — a deliberate decision, made when the load at the other end was guaranteed to swallow it.
A third source is not benign: if the switch was wired to break the neutral rather than the live, the lamp remains at line potential when "off", and leakage to earth can keep the driver alive. Same symptom, different meaning.
Can we distinguish them? Fit a bleed device across the lamp — a dummy load or "anti-flicker" capacitor, or an incandescent in a second holder on the same switched circuit — so the trickle has somewhere to go that is not the driver. The refuting observation: if the glow survives that, leakage into the driver is not the story, and the next thing to establish is whether the switch breaks neutral instead of live, a wiring fault and a shock hazard that belongs to an electrician. A second discriminator is free: if the glow fades to black within a minute of switch-off, it is charge already stored in the driver, and there is no leak at all.
The analogy
THE ANALOGY #Think of a reservoir fed by a valve that was never quite watertight, with an overflow that tips the contents out whenever the level reaches the top. Under the old arrangement the reservoir supplied a city, and a leak that small was lost in the demand. Replace the city with a single household tap and the same leak fills the reservoir, and every so often the overflow tips.
Water leaks through a physical gap, whereas capacitive coupling passes current through intact insulation with nothing defective; and a reservoir tips by gravity, whereas the driver's threshold is a designed decision about when there is enough voltage to start.
Clarifying the model
THE MODEL #The folk explanation is that the bulb is faulty or cheap, and it is mostly false — though false in a way that keeps proving itself right. Bulbs differ in how much leakage they tolerate before glowing, some carry an internal bleed path, and some drivers start at a lower voltage. So swapping brands really does sometimes cure it, which everybody reads as proof the first bulb was defective. What changed was a pairing. The leak was there before the LED arrived and remains after it leaves.
The second belief worth correcting is that a glowing bulb wastes significant electricity. The power is milliwatts — derived above as roughly twenty for a plausible coupling. I will not attach a cost, since tariffs and the actual coupling vary far too much for a figure to mean anything, but the order of magnitude is the point: this is not an energy problem. It is a signal that something upstream is passing current, and whether that matters depends on which source. Capacitive coupling and an illuminated switch are nuisances. A switched neutral is not.
One honest limit: which source is at work is genuinely hard to settle from outside the wall, since all three produce the same faint light, and published leakage figures for switches and sensors vary so widely between models that quoting one would mislead. The tests above narrow it; they do not close it.
The neighbouring puzzle — an LED flickering on a dimmer — shares this fact and none of this circuit. There the switch is conducting and a TRIAC cannot stay latched on so little current; here the switch is open and a little current still arrives. One fact, two failures: an LED asks for about a thousandth of what a filament asked for, so everything designed around the filament's appetite misbehaves at both ends.
A picture of it
THE PICTURE #How to readStart at the filled circle and follow the arrow into "Leakage current arrives" — everything downstream turns on how big that trickle is relative to what the driver wants. The three arrows leaving it are the three outcomes, labelled by what decides between them: a bleed load diverts it and the lamp stays dark, a large leak runs the driver continuously as a steady glow, and an in-between leak feeds the charging state. The loop between charging and firing is the blink — charge, fire, collapse, charge again — and the other arrow out of charging is the common happy case, where the leak never reaches start-up.
What became clearer
WHAT CLEARED #"Off" was never zero. Wiring has always passed a small current through open switches by capacitive coupling, and switches with their own electronics have always borrowed a little through the lamp — and for a century the load was a filament coarse enough that neither showed. Replacing it with a device running on milliwatts did not create a leak; it installed a meter on one. The glow is a measurement, and the question worth asking is which of three quite different sources it reports.
Where to go next
ONWARD #- How a switch loop without a neutral works, and why newer wiring practice runs a neutral to the switch position.
Key terms
TERMS #| Term | What it means |
|---|---|
| Capacitive coupling | alternating current passing between two conductors separated by insulation, without contact. |
| Driver | the power supply inside an LED lamp that converts mains into a regulated current for the diodes. |
| Bleed load | a resistor or capacitor across a lamp, absorbing leakage that would otherwise reach the driver. |
Every term the collection defines is gathered in the glossary.