# Can unwanted coupling drive signal lower than intended?

Let's consider the following schematics- a switch connected to microcontroller's input with internal 50kOhm pullup resistor.

simulate this circuit – Schematic created using CircuitLab

Now, the following oscilloscope trace shows voltage at TP (CH1) and VCC (CH2, AC coupled) at the time, when CCFL lights are switched off in the room.

I do understand that two conductors can become inductively coupled, and that current change in one conductor induces voltage in other conductor. By looking at the oscilloscope trace, I see that signal voltage drops- my assumption was, that only more positive voltage can be induced.

Is my assumption wrong, and negative voltage (i.e. signal voltage drop) can be induced in a conductor which is close to other conductor?

Or this oscilloscope trace does not show inductive coupling, but something else I'm not aware of?

After I added an external pullup of 3.3kOhm, flicking light switch does not result in LOW level input signal seen by microcontroller anymore- I just wanted to understand what exactly caused it.

• In short, yes. Negative voltages are just as possible as positive voltages. It's hard to say what noise you are seeing, or if this is noise actually in the circuit, or due to your measurement technique. Mar 5, 2014 at 20:48
• I'm not sure about magnitude of noise introduced due to my measurement technique, but I'm fairly sure the noise is there- I'm seeing LOW level in code that runs on microcontroller at the time lights are switched off.
– fest
Mar 5, 2014 at 21:10

I do understand that two conductors can become inductively coupled, and that current change in one conductor induces voltage in other conductor. By looking at the oscilloscope trace, I see that signal voltage drops- my assumption was, that only more positive voltage can be induced.

As you suspected this is a wrong assumption.

If two conductors are inductively coupled and a (change in) current in one direction of conductor "A" induces a positive voltage in conductor "B", then a (change in) current in "A" in the opposite direction is going to induce a negative voltage in conductor "B".

In any case where at some instant in time you see a positive induced voltage in some conductor, you're likely at some other time to see negative induced voltages, just because currents don't tend to increase or decrease monotonically toward infinity.

Note: This does not mean I think that the problem you are seeing has anything to do with inductive coupling.

Have you considered that the possible reason you are seeing this is because your "imperfect" oscilloscope (with not really isolated-from-earth input leads) is getting a "jolt" along the AC wiring when the lights turn off. Even if the scope probes are not connected to anything (or even unplugged) you might also see an effect however, if the scope probes are connected to something more substantial (and that substantial thing has capacitance to earth) you could see the same picture. Anything with significant surface area will form a capacitance to earth of a few pF.

The waveform you are seeing is at 2us per division and if you think how long it takes to debounce a switch on an MCU (5 to 50 milli seconds typically) would it surprise you to see an AC power switch exhibiting similar behaviour? The AC switch is also carrying much more current and this could lead to a small but robust arc when the switch opens. There is also the inductance of your power cables to consider - this could easily be in the region (locally) or several micro-henries.

Any amount of capacitance to ground (local earth wire) could cause all manner of spurious resonant spikes and these could dance up and down the cable for a several tens of microseconds even if the switch was totally perfect at opening its contact.

• The circuit exhibits very similar symptoms (microcontroller input goes low) when oscilloscope is not connected. I also might have been vague on swithes - switch that is connected to microcontroller is not being activated at the time of waveform in question - AC switch for lights in the same room is.
– fest
Mar 5, 2014 at 21:25
• How long are the wires to the switch from the micro dude? My answer was really about not believing everything your scope tells you. I'd be interested to see if your scope (unconnected) glitches on the screen? Mar 5, 2014 at 21:49
• Wires to microcontroller are about 10 cm long. The light switch is about 1 meter away, and I don't think the AC wiring is closer to microcontroller than the light switch. I'm also not too much concerned about few tens of mV of noise on VCC - I am concerned about a volt or so of noise on input. I'll arrange an experiment with unconnected scope tomorrow.
– fest
Mar 5, 2014 at 21:52
• If you want to do another experiment (more of a fix really), put a 100nF capacitor across the input to the MCU but make sure your switch is wired in series with (say) 100 ohms to prevent the same effect! Mar 5, 2014 at 22:14
• Yeah, I do get the basics about RC filters (and since it's a user operated switch, I ought to debounce it in code anyway). This was more like an exploratory question- I had "fixed" that problem with a stronger pullup resistor on that input.
– fest
Mar 5, 2014 at 22:35