Soft-start circuits when you want the relay to live

The problem arrives at turn-on

You flip the switch on your bench supply or robot controller, and for a few microseconds the input capacitors look like a dead short. The inrush current spike can exceed fifty times your steady-state draw. Relays chatter or weld shut. Circuit breakers trip. If you're using a wall adapter, it may simply give up and enter hiccup mode until you unplug it.

I've watched a 10 A relay contact pit itself to failure after two dozen power cycles on a prototype that pulled 2 A in normal operation. The inrush was doing the damage, not the load.

How soft-start helps

A soft-start circuit limits inrush by inserting resistance during the first few milliseconds of power-up, then bypassing that resistance once the bulk capacitors have charged. You get a controlled ramp instead of a step function.

The simplest version uses an NTC thermistor in series with your input. Cold resistance is high; as current flows the thermistor heats and its resistance drops. It works, but recovery time is slow—if you power-cycle quickly, the thermistor hasn't cooled and you lose protection.

I prefer an active circuit: a small resistor in series with the input, bypassed by a MOSFET or relay after a delay. The resistor does the current limiting during charge-up. Once the bulk caps hit 90% of input voltage, the bypass closes and you're running at full efficiency.

Circuit you can trust

Use a P-channel MOSFET as the bypass. Its gate is pulled high through a resistor to keep it off at startup. A timing RC network on the gate slowly pulls it low, turning the FET on after your chosen delay—typically 10 to 50 ms.

Pick your series resistor to limit inrush to something reasonable: if you have 2200 µF of bulk capacitance on a 12 V rail, you want maybe 1 Ω to keep the peak under 10 A. Power dissipation is brief, so a 2 W resistor handles it without complaint.

The MOSFET needs low R_DS(on)—under 50 mΩ—so it doesn't waste power once bypassed. A FQP27P06 or similar will do the job for a few amps; scale up if you're building something heavier.

If you're designing a board that already has a microcontroller, skip the RC timing and drive the FET gate from a GPIO after a software delay. You get precise control and can add diagnostics—measure input voltage before closing the bypass, check that it's in range, log the event. I've done this on motor controllers where the µC was already managing everything else; it costs one pin and five lines of startup code.

Relay-based bypass for higher current

Above 10 A continuous, a MOSFET that can handle the current and dissipate the I²R loss during bypass gets expensive or needs a heatsink. A relay becomes simpler: your series resistor limits inrush, the relay coil is powered by a delayed turn-on circuit, and once the relay closes you have milliohms of contact resistance.

Use a time-delay relay module or build one with a 555 and a small signal relay to drive the main power relay coil. The resistor sees current only during the delay period, so even a 5 W part is plenty.

This is the approach I took on a 48 V, 20 A robotics power board. The series resistor was 2.2 Ω / 10 W, bypassed by a 30 A automotive relay after 30 ms. The relay has been power-cycled hundreds of times without contact wear.

Don't skip this on big supplies

If your project draws more than a few amps or has bulk capacitance over 1000 µF, a soft-start circuit is not optional. I've seen people rely on the internal resistance of their power supply or hope the breaker won't trip. It works until it doesn't, and then you're replacing parts or debugging intermittent failures that only happen at power-on.

Size your resistor and timing for the actual capacitance you've placed on the rail—include every bulk cap, every local decoupling cap that's large enough to matter. Measure with a scope if you want to confirm the ramp shape, but the math is straightforward: τ = RC, and you want at least three time constants to reach 95% charge.

When you get it right, the relay lives, the breaker stays closed, and your power supply doesn't spend half its life in hiccup mode. That's worth an extra hour on the schematic and a couple of passive parts in your BOM.

For more on sizing bulk capacitors in the first place, see Bulk capacitor math that fits your drawer. If you're adding this to a board that already has input protection, pair it with Reverse polarity protection you'll thank yourself for so you're covered from both directions.