I have a circuit in which I am trying to use an microcontroller to activate a car horn. I've had a lot of trouble trying to isolate the horn from the rest of the circuit. I have the microcontroller toggling an LED on and off once a second or so, just as a status LED that everything is going good. Then when the horn goes on, the LED stops blinking and just stays on, and the horn isn't being switched off. It's like the microcontroller is being wedged. As I write this maybe the external clock is receiving some feedback somehow? Even if I unplug the 12v car battery from the circuit the microcontroller doesn't recover until I power cycle it. I have tested the circuit with power resistors and everything works as expected. The difficult thing is that I feel like I've done everything possible to isolate the microcontroller from the horn.

I've also tried adding a flyback diode (edit* added to schematic) with no better results (I left it out for this run because one of the guys who helped designed this thing said that a car battery can handle it and a flyback diode isn't needed). I've even tried wrapping the microcontroller board in aluminum foil. That didn't seem to help either.

Any ideas as to what could be going on, or what to try next?


Thanks for all the suggestions. I ended up adding the flyback diode across the coil of the relay and I also simplified a bunch of my arduino code. I then found that if I commented out everything and just had a button as an input to the microcontroller and a single output to the mosfets and horn, then things worked! The problem shows up when I introduce the code that reads from the gyroscope/accelerometer (MPU6050). I think it's over SPI? Anyway, if I include that code in my loop thread, then everything goes awry.

Wire.write(0x3B);  // starting with register 0x3B (ACCEL_XOUT_H)
Wire.requestFrom(MPU_addr,6,true);  // request a total of 6 registers
accX=Wire.read()<<8|Wire.read();  // 0x3B (ACCEL_XOUT_H) & 0x3C (ACCEL_XOUT_L)    
accY=Wire.read()<<8|Wire.read();  // 0x3D (ACCEL_YOUT_H) & 0x3E (ACCEL_YOUT_L)
accZ=Wire.read()<<8|Wire.read();  // 0x3F (ACCEL_ZOUT_H) & 0x40 (ACCEL_ZOUT_L)


simulate this circuit – Schematic created using CircuitLab

picture of circuit microcontroller breadboard picture of horn breadboard wrapped in aluminum foil

  • 1
    \$\begingroup\$ What does "being wedged" mean? Also, it looks like you're probably driving too much current through your optocoupler's LED, though that wouldn't cause this problem (I think). \$\endgroup\$
    – Hearth
    Feb 13, 2019 at 4:56
  • 1
    \$\begingroup\$ "I've also tried adding a flyback diode..." - where? The relay should have a flyback diode across its coil to protect the driver transistor. What resistance is the relay coil? M1 seems to be redundant. Can you move the optocoupler output further away from the MCU circuit? \$\endgroup\$ Feb 13, 2019 at 7:20
  • 2
    \$\begingroup\$ Also, you need a current limit resistor for the LED. Try 1k for both opto-coupler and LED. \$\endgroup\$
    – Steve G
    Feb 13, 2019 at 11:58
  • \$\begingroup\$ I forgot to include it in the schematic, but the resister for the status LED and the optocoupler are both there (1k) each. I'm updating the schematic now. \$\endgroup\$ Feb 13, 2019 at 15:29
  • 1
    \$\begingroup\$ The diode should be across the coil of the relay, not the speaker. \$\endgroup\$
    – CrossRoads
    Feb 13, 2019 at 16:33

1 Answer 1


Connect a capacitor across the microcontroller Vcc to GND directly on top of the chip (leads short as possible). The capacitor should be a ceramic type eg. 100nF to 1uF.

If you want to try shielding, the shield should be connected to GND.

Your open breadboard construction is not very conducive to achieving low EMI susceptibility, to say the least. To try to figure out what is causing it, you can actuate the horn manually with and without the optoisolator in place and see if your uC still hangs. You know the EMI is coming from the horn, but you have not really established how it is getting into the MCU.

  • \$\begingroup\$ @ChaseRoberts You may also want to consider running your MCU on the internal RC oscillator (if available) while you prototype on a breadboard (there is a lot of loop area there). \$\endgroup\$
    – sstobbe
    Feb 13, 2019 at 17:22
  • \$\begingroup\$ Sorry, I am so slow to build this and accept an answer. This is very much a hobby project and I don't have as much time for hobbies as I would like. Stupid question, but I feel like I need to clarify the shield should be connected to the microcontroller ground correct? Not the car battery ground. \$\endgroup\$ Feb 16, 2019 at 0:01
  • \$\begingroup\$ Yes, the micro controller ground, and nothing but the opto output should pass through the shield if at all possible. \$\endgroup\$ Feb 16, 2019 at 0:03
  • \$\begingroup\$ So the optocoupler should go on the same board as the microcontroller? And not the same board as the switches. \$\endgroup\$ Feb 16, 2019 at 0:06
  • \$\begingroup\$ Yes, correct.... \$\endgroup\$ Feb 16, 2019 at 0:06

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.