I've got a circuit that's having problems at startup with inrush current, leading to large spikes on VCC (parasitic inductive of the power supply / leads etc) that sometimes results in smoke coming from chips in my circuit. It's powered with ~12-16V and drives a 5v circuit that draws 40-80mA @ 5v.

The excessively large 82uF input capacitor is an electrolytic, it's there so transient disruptions to VCC won't reset the microprocessor, and the 10uF input capacitor is a ceramic SMD component - as suggested by the DC-DC converter IC datasheet.

I'd like to include a fuse, so failure of this circuit won't effect any other consumers connected to VCC (and fire and heat coming from this circuit is not acceptable under any circumstances) - but this has not worked out so far due to the inrush current.

I'm considering adding a 25 ohm input resistor in series with VCC (ie before fuse) to limit the circuits current. Under worst case normal operating conditions, this would have a drop of 2v, dissipating 0.16 watts. At startup, it would limit inrush current to 0.64A - which would also allow me to use my 1A fuse as well as avoiding any inductance in the supply leading to deadly voltage spikes.

Does this sound like rational circuit design, and an acceptable solution to the problem?

I intend to use 0603 SMD components for this input resistor, and fuse - or should I size them up?

Power supply circuit with DC-DC converter

To demonstrate the source of my pain, I setup this test circuit and apologetically powered it with 16v by touching power to the wires. In the scope trace, you can see a spike in voltage up to 19v coming from my (cheap) bench PSU.

Test circuit


If I add a bunch more parasitic inductance to the power supply to simulate an even worse supply, these spikes quickly become deadly. I have the scope connected across the load resistor, and the inductor inline with the supply leads.

Test circuit with more parasitic inductance

Trace from test circuit with more parasitic inductance

This is definitely worst case scenario, and the offender is my rubbish bench PSU - but one of my design requirements is for my circuit to be near bulletproof to external factors.

  • \$\begingroup\$ Depending on how long the circuit is in the transient state at power on I'd suggest on only using an 0603 if you're in the microseconds. Even then just to be safe a 1206 isn't a big component and is rated at 0.25W \$\endgroup\$
    – ChrisD91
    May 14 at 15:10
  • \$\begingroup\$ Deadly voltage spikes? How are these generated? The 82uF capacitor will resist the change of voltage so you should get a nice charge curve if there is enough inductance. Also note a 25Ohm resistor would mean the fuse would never blow. If you are concerned about voltage spikes then add protection like a varistor or TVS diode. I’d suggest you find out where these spikes are coming from and address that first. \$\endgroup\$
    – Kartman
    May 14 at 18:56
  • \$\begingroup\$ @Kartman Thank you for your input! Indeed the circuit is more reliable with the 82uF in place (although I've still seen it smoke); and without, it vaporizes instantly. I hadn't thought about the fuse never blowing! Thanks for pointing that out, I would be best omitting it entirely if I went this route. I've never used a TVS diode, but that sounds like it could help me achieve my goal here - perhaps even in conjunction with an input resistor. There's really 2 problems, inrush current charging the large capacitors, and protection from VCC spikes (although the former can contribute to the latter) \$\endgroup\$
    – Warrick
    May 15 at 0:15
  • \$\begingroup\$ @ChrisD91 I think you're right, a 1206 wouldn't hurt - especially given the fuse would never blow, so it would need to withstand constant dissipation. \$\endgroup\$
    – Warrick
    May 15 at 0:19
  • \$\begingroup\$ What is your load? Or are you seeing smoke without any load? \$\endgroup\$ May 15 at 1:08


Your Answer

By clicking “Post Your Answer”, you agree to our terms of service, privacy policy and cookie policy

Browse other questions tagged or ask your own question.