White LED dimly lights by touching it

[jrmymllr]

My son breadboarded a circuit that uses a white LED and is powered by a bench supply that internally uses a switcher.  He came running to me to ask why the LED remains dimly light even after releasing a button he hooked up to control the LED. While demonstrating it to me, and failing to get it to malfunction the way he described, I discovered it was because sometimes he'd touch one of the LED leads, which would make it light up very dimly. Even with one of the power connections from the power supply disconnected, it would still do it.

I thought this was interesting, but not completely unexpected. We all know what happens when you touch the input of an audio amp, scope, or a logic probe (remember those?). So clearly there's some voltage there. I got out the current meter and held one probe while touching the other to the LED and saw around 800nA on AC mode. Sidenote: wow aren't modern LEDs efficient!!

I've never give this phenomenon much thought in all my decades of tinkering, but what is actually going on here? First I propose there's some leakage between the power supply output and one or more of the AC conductors, although this still occurs to a lesser extent when the supply's hard power switch is off, thus breaking the hot/line. But are we picking up AC leakage or fields and returning to earth through the power cord? This was in his shag carpet bedroom so I'd assume that's not very conductive.

[ataradov]

This is capacitive coupling of the mains though the body. The current path is hard to predict but if it remains connected to the outlet, the earth terminal is not going to be interrupted by the power switch.

And yes, LEDs are supper efficient. On my boards I run them at single digit uA to not get blinded by simple status LEDs.

[jrmymllr]

This is capacitive coupling of the mains though the body. The current path is hard to predict but if it remains connected to the outlet, the earth terminal is not going to be interrupted by the power switch.

And yes, LEDs are supper efficient. On my boards I run them at single digit uA to not get blinded by simple status LEDs.

I get the earth always remains connected, as it should be. Capacitive coupling makes sense, but from where? I'm 3 feet from the walls, unless there's a cable under the floor which is quite possible since it's on the 2nd floor. Or maybe this much current can be capacitively coupled from feet away? I don't recall the capacitance formula but I assume distance is squared or cubed.

[ataradov]

From the environment. Your body is one big capacitor plate.

This is the same thing you see on the scope when you touch the probe tip without touching anything else. I just tried and simply sitting in front of a computer, I get a sine with ~5vpp amplitude. It is not going to deliver enough  current, but it can deliver a few uA.

[amyk]

The power levels involved are extremely low. Even RF from cellphones is enough to power LEDs "passively" with only a suitable antenna connected.

"What is the reverse recovery time of an LED" is not something I've given much thought to either...

[tom66]

If the bench power supply is a Class II without earth (uncommon for lab supplies, but perhaps possible for a low power cheap supply) then the output will float at around 60Vac relative to earth (assuming 120Vac mains). This is perfectly safe for humans as the current is very low but if you provide a path to earth then you will be able to get ~0.5mA to flow through an LED, which will easily make it illuminate.  Class II designs are common with lower power switchmode power supplies, usually sub 100W, though it is uncommon for lab power supplies to not use an earth as usually an earth terminal is provided on the front of the unit.

I have seen a similar phenomenon when I used my earthed soldering iron on a board that I had accidentally left connected to the mains via a Class II power supply - LEDs on the board would glow slightly as a path to ground was created.  I suppose it could damage sensitive electronics so I wouldn't recommend doing that on anything important.