A few years ago, we have bought an IKEA Glimt hanging lamp that operates with halogen lamps. It has a "touch dim" functionality, so when a person touches the metal bar of the lamp, it dims (or when touching for a short time, it switches on/off).

After having replaced the halogen lamps with LEDs and replacing the AC/DC converter, I have noticed that my new LED dimmer has an input that when connected to an external button, it does the same.

My goal now is to build a circuit and connect it to the metal bar such that when touched, it closes the input of the dimmer (pulls down one input to ground).

I have tried a plain darlington pair (with no further components) - indeed on a test board, it is able to close the input as expected similar to what can be seen in tutorial videos on youtube (switching an LED via touch). But the problem is that the LED dimmer emits some electric signal during operation such that it triggers the base of the darlington pair itself via cross-talk, so it keeps being "touched". I have also tried a capacitance sensor (using the touch pins of an ESP32), but due to the large metal parts, there is no numeric change while touching it, it cannot be detected.

So I am a bit lost... what I do remember is that the old halogen AC/DC converter (that supported touch via a single "sense" line) only worked reliably when connecting all three lines (phase and neutral as well as ground).

Any advise/hints? Do I need to equip the darlington pair with more components? The aim again is to pull the switch input of the LED dimmer to ground when touching the lamp's metal components.

  • 2
    \$\begingroup\$ Welcome to EE.SE. A schematic is better than words. You can add one in using the CircuitLab button on the editor toolbar. Double-click a component to edit its properties. 'R' = rotate, 'H' = horizontal flip. 'V' = vertical flip. Note that when you use the CircuitLab button on the editor toolbar and "Save and Insert" on the editor an editable schematic is saved in your post. That makes it easy for us to copy and edit in our answers. You don't need a CircuitLab account, no screengrabs, no image uploads, no background grid. \$\endgroup\$
    – Transistor
    Apr 4, 2021 at 12:54

1 Answer 1


Rather than use DC leakage with a strong CM line AC signal present for touch sensing, this EMI problem may be solved better ways.

It seems your SNR <1 (unless poorly biased, filtered)

Put a shunt R of 1M to earth gnd (across Vbe) and also a >10nF across Vbe with a line f pulse stretcher cap across Vce or across Rc so that RC>20 ms..This will also suppress ESD if low ESR plastic by 100pF/10nF ..Otherwise.... plan B.


Normally for touch switches a frequency in the x or xx kHz range is sensed by amplitude across the contacts. But here there is only 1 contact (bar) and the other is internal Earth-bonded ground.

So how to detect injected common mode (CM) signal at some frequency on an “earthed bar”?


The impedance of the bar is unknown but assumed to provide a safe contact in case of a lightning strike or wiring fault. The human’s impedance for a finger using the HBM (model) is 100 pF + ? R but also the parasitic insulation R (wide range !) and RF characteristic Impedance (377 ohms ) in series with a high C , low R body plasma to a distant earth ground.

I assume you know how to compute impedance for L or C (f).

Thought experiment

Which would work better?

  1. Use an Inductor to bar instead of a wire of impedance X(f) and detect 50V E-field (seen) on 10Meg probe) but now attenuated to xx ? mV line f by ground shunt series L and detect 50,60 Hz low level above e-gnd.
  2. Inject a 100kHz to 500kHz signal into “bar” and suppress with finger capacitance and detect drop in carrier with a BPF and Detector/comparator,
  3. Assume one-shot timer comparator in both to select dim or toggle power.

Your Answer

By clicking “Post Your Answer”, you agree to our terms of service and acknowledge you have read our privacy policy.

Not the answer you're looking for? Browse other questions tagged or ask your own question.