simulate this circuit – Schematic created using CircuitLab

I have built a circuit, which consists of a 9V AC transformer, feeding a bridge rectifier and the current to the load (a length of resistance wire ~3Ω) controlled by a MOSFET. The gate of the MOSFET receives a PWM signal from an Arduino control module which allows the duty cycle to be varied from 0% - 100%.

NOTE The MOSFET is intended to chop the full wave rectified AC voltage.

The circuit works, and has been in use for a year.

I decided to add a circuit, consisting of a diode and filter capacitor to derive a DC power for the Arduino, but several electrolytics went up in smoke, and the output voltage considerably exceeded the 13V I was expecting (so it ended up being removed).

I recently bought a BitScope and was using this to circuit.

When at 0% to voltage is the expected full wave rectified DC.

As the duty cycle of the PWM control increased there appeared voltage spikes on the rectified DC, and also on the input AC. enter image description here

It appears that when the MOSFET turns off the falling current generates a spike. As there is no other component with any inductance, this seems to be coming from the transformer itself. This was unexpected, as I am working well below the current rating of the transformer, at a much higher frequency than it would normally run. I modified the control circuit to produce 4kHz PWM, which reduced the amplitude of the spikes, but they are still of concern.

I had expected some ripple, but not the massive spikes, which I am concerned may exceed the 60V rating of the MOSFET.

My question is, can anyone suggest some way of suppressing these. None of the normal techniques for suppressing inductive pulses from the load seem applicable.

  • \$\begingroup\$ What PWM frequency were you using previously? What's the input frequency? (you mentioned it's much higher than the transformer would normally run, so not mains) \$\endgroup\$ Nov 24, 2017 at 6:53
  • \$\begingroup\$ @immibis 490Hz (shown on the schematic) - this is the default for Arduino \$\endgroup\$
    – Milliways
    Nov 24, 2017 at 7:20
  • \$\begingroup\$ "a length of resistance wire ~3Ω" - how long is the resistance wire, and what is it used for? \$\endgroup\$ Nov 24, 2017 at 7:34

2 Answers 2


First Note: You're talking about pure DC but I cannot see any high-value electrolytic capacitor (except 100n) for filtering after the bridge. I assume that you've forgotten to add that to the schematic in the question.

It's pretty normal to see spikes because there's switching. But please note that measurement technique is also important. Because when you use the ground clip of the probe, it acts as an antenna. This leads to some unwanted measurement artifacts. So, please make your measurement like shown below:

enter image description here

Note that even if the circuit does not contain any inductances, there can be seen some spikes/ringings caused by:

  • long cables
  • PCB design (i.e. long and thin tracks)
  • internal inductances of the load
  • internal drain inductance of the MOSFET

So, check these first.

As for suppressing, you can put a pi filter between supply and the load. And also you can place a RC snubber across D-S terminals to suppress the spikes right at the switching node.

  • \$\begingroup\$ I never mentioned "Pure DC". The MOSFET was intended to chop the full wave rectified AC. It seemed superfluous to use a large expensive capacitor to produce "Pure DC" just so I could chop it. I am reconsidering this decision, just to smooth the spikes, but a suitable capacitor would dwarf the matchbox sized controller. \$\endgroup\$
    – Milliways
    Nov 24, 2017 at 6:45
  • \$\begingroup\$ Oops, sorry then, I've misread it. I also didn't know the intention of using the rectified AC. Sorry again. A big capacitor after the bridge may not help to smooth the spikes, by the way. As I suggested in my answer, a pi filter will be more effective. \$\endgroup\$ Nov 24, 2017 at 6:56

All of your problems can be explained by a simple application of Ohm's Law.

What is ratio of 3 Ohm load to the ESR of the 820uF cap ? Then you can estimate the ratio rise in peak current. 10 millohm? 100m?

An unregulated supply such as this will rise 50% minimum with a light load.
transformer loss(10%) + Peak/Avg of sine wave (41%) =~ 50%

For low loss the caps need to be at least 2x the transformer voltage and have a dissipation factor of <<1%.

Also consider an approximation that %Ripple Current (pk/avg) = 1/ Ripple voltage (Pk/avg) so 10% ripple voltage implies current spikes are 10x DC average current for 10% of the time.

  • \$\begingroup\$ The 820µF cap is NOT part of the active circuit - this is the add-on mentioned, and is isolated by a diode. It was intended to provide <100mA to the control circuit. The MOSFET was intended to chop the full wave rectified AC, because it seemed pointless to filter to DC, so it could be chopped. \$\endgroup\$
    – Milliways
    Nov 24, 2017 at 7:23
  • \$\begingroup\$ FET spike is from wire inductance and/or probe gnd inductance error. I was trying to explain why your add-on 820uF raised the voltage of 100nF and other "blown caps" \$\endgroup\$ Nov 24, 2017 at 7:37

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.