I've designed an inrush current limiting circuit and would welcome a review from experts with more experience with this sort of thing. The load will be around 50 x 19 VDC @ 3.5 A linear power supplies connected in parallel with steady state load up to around 20 A @ 230 VAC.

There are a few specific requirements that motivated my own design rather than an off-the-shelf version:

  • I need 25 A @ 230 VAC, and most commercial varieties seem to be 16 A (too low) or 30 A (too high).
  • A 25 A inrush current limiter I did find and test didn't actually limit to 25 Apk, but more (beyond 35 A where my fuse trips).
  • The place where this will go has considerable mechanical constraints; most commercial packages just don't fit.

Here is my schematic: Draft inrush current limiter design by Sean Leavey

As is typical with inrush current limiting circuits, to avoid the power loss in the NTCs during steady state (which at 25 A is significant - around 200 W) I short out the NTCs with a relay after a short delay (~0.6 seconds, depending on tolerance).

There is also a thermal trip switch so that if the NTCs get too hot the current will trip and stay off until the power is cycled (so that the circuit does not yo-yo between on and off as the NTCs get hot and cold). This is my own invention (as far as I know) but I have not tested it before. I did simulate it in SPICE and it seems to work.

The intended sequence when power is first supplied to the circuit is as follows:

  1. With no power, the relays are initially open.
  2. When mains power is initially provided, the transformer drops it down to some manageable level (20 VACpp). The capacitors and LM317 regulator convert this AC into 12 VDC.
  3. Initially the non-inverting inputs to N2A, N2B and N2C are 0 V due to capacitors C8 and C9 (current via R7 then R9/R11 has not yet had time to charge C8 and C9). This means these comparators short their output to ground (they are open collector output type), so the gate-source voltage at T1 and T2 is 0 V and the relays remain off.
  4. The inverting input to N2D is initially around 2.4 VDC. The non-inverting input is 4.2 VDC. The LM339 is therefore initially in high-Z output mode so the pull-up via R7/R8 sets the output to 4.2 VDC. After 0.1 seconds N2B's non-inverting input will exceed the 3 VDC inverting input and go into high-Z output mode, allowing 12 V to develop at T1's gate-source and switching on the left relay. This allows current to flow to the load via NTCs R14 and R15.
  5. After 0.6 seconds, N2A and N2C also go into high-Z mode, switching on the right relay and shorting the NTCs.
  6. The circuit operates in steady state until such time as the temperature sensor measures above about 150°C. At this point, the inverting input exceeds the non-inverting input and so N2D's output gets shorted to ground, which in turn switches off the left relay after around 0.1 seconds, removing the load. Because the shorted output is also connected to the non-inverting input, the inverting input's voltage always exceeds the non-inverting one and so the circuit remains off until the power is cycled.

Some notes:

  • I went for a linear regulated supply over switch mode for simplicity and maximum longevity.
  • The transient voltage suppressor diodes in parallel to the relays (in addition to the normal diodes to snub the induced back emf) are there to shorten the time the relay contacts arc as they close and open, to prolong the contacts' lives. This idea I took from the new Art of Electronics x-chapters book. This is a trade-off between time-spent-arcing and maximum induced back emf.
  • There are lots of timings to pay attention to here: how quickly the 12 VDC rail comes up in comparison to the comparator inputs, etc. I only need three of these circuits for my application and I can make adjustments to capacitor values etc. to try to get it to work reliably.

I can't share the PSU datasheet since it's confidential but it contains the following information:

  • Max input current @ 230 VAC: 0.65 A
  • Worst case inrush current: ≤ 0.25 A²s ( ∫ i² dt ) / ≤ 12 A

I can open up a PSU to look at the input stage but I suspect it'll be a fuse, toroidal transformer then smoothing capacitors and regulation (somewhat similar to what I have in my design). I expect the transformer is what causes the large inrush current on switch-on, since at that point it has no magnetic field and thus initially acts like a low value resistor. Unfortunately the datasheet doesn't state the input capacitance/inductance directly, but perhaps this can be worked out from the values above?

Does anyone spot any issues? Do people think my thermal latch, timings, etc. will work?

  • \$\begingroup\$ As I see it, it looks fine, but don't take my word for it (that's why this is a comment, not an answer). However, you should know that relays have vibrating contacts and with such a high current load, the effects may be felt. An alternative would be a solid-state relay, but there will be some dissipated power. \$\endgroup\$ Jul 14, 2020 at 19:42
  • \$\begingroup\$ @Justin, this is for an academic project and I would probably in any case build it and test it even without help here since we can't really justify hiring a consultant. I only need a few for my needs, I'm not selling them or anything. That said, I hope that sharing my design in public is suitable "payment" for expert input here, so others can benefit too. \$\endgroup\$
    – Sean
    Jul 14, 2020 at 21:41
  • \$\begingroup\$ @aconcernedcitizen, thanks, I tried to add some protection for the contacts but indeed they will be degraded due to the high current over time. This only happens during switching though, and in my application the load won't be switched very often (mainly just during power cuts - as I said, it powers a UPS!). \$\endgroup\$
    – Sean
    Jul 14, 2020 at 21:41
  • \$\begingroup\$ What is your UPS inrush spec? Imax, t duration. What problem are you trying to solve? Breakers or fuses are I vs t rated.with a curve. Always provide design specs to solve a problem. Don't guess. \$\endgroup\$ Jul 17, 2020 at 19:05
  • 1
    \$\begingroup\$ There's an issue with shorting pin 11 of the LM339 to ground. It's recommended to use a 1k - 10k series resistor on the inputs. See Section 5.1.1 of Application Design Guidelines for LM339. Also wondering about pin 14 which floats high under normal operation. Should it be connected to the cathode of D10 to switch the LED on during an over temperature condition? \$\endgroup\$
    – tim
    Jul 19, 2020 at 19:50

2 Answers 2


I've designed an inrush current limiting circuit and would welcome a review from experts with more experience with this sort of thing.

Depending on how the UPS is designed, it may not "play-ball" with the series resistors because they may cause the UPS to try and take a really excessive starting current (which it can't take due to the resistors). The upshot of all of this is that the UPS never really kicks into action until the relay closes (shorting the current limiting resistors) and then, you have the same inrush problem just delayed in time.

So, to design this we really need to know what the front-end circuitry in the UPS is like.

Regarding the contact closure that short the resistors, I'd be much more inclined to activate that contact when the AC supply output has risen to the point when the UPS (if it plays ball) is e.g. 75% of the input voltage. A fixed time delay produced by R11 and C9 is too "open-loop" to be effective.

You also need an input fuse on transformer L1 because most magnetic components like this are not rated to be connected directly across a very resilient mains AC supply. Fuse F1 on the output of L1 won't cut-the-mustard in this respect. Ditto the input varistor U1.

Why are your flyback components two series diodes in series with a zener. I can understand one diode and a zener but two diodes and a zener seems like you may be misunderstanding something.

Does C1 really need to be 1000 uF (1 mF)?

  • \$\begingroup\$ Thanks Andy, much appreciated! I'll split my reply between this and the next comment. UPS front end: unsure, I've asked the manufacturer and will update here if I get a response. The datasheet does say it has a power factor of 0.9 if that helps. Sensing AC supply output: interesting idea! Could this be a comparator comparing stepped down versions of the input and output waveforms? Should they be converted to DC / rms first? If so, how best to do that? Do you know of any existing designs that do this that I could use as a reference? \$\endgroup\$
    – Sean
    Jul 16, 2020 at 12:38
  • \$\begingroup\$ Fuse: good catch, I'll move it before L1. What do you mean about varistor V1 - must this be on the low voltage side of L1? Flyback components: good catch - D5 and D8 not needed. C1: I simulated the circuit with LTspice and found this had to be pretty big to allow the circuit to work. \$\endgroup\$
    – Sean
    Jul 16, 2020 at 12:39
  • \$\begingroup\$ @Sean output sensing: you have your DC voltage for the input - after the bridge rectifier and that 1 mF capacitor. For the output you could use a small capacitive divider (small enough not to present a hazard i.e. 1 nF circa) and rectify it with a single diode to give you a DC voltage that is referenced to your GND node.$$ $$ The varistor ought not be connected directly to line i.e. put it after the fuse that I spoke about. \$\endgroup\$
    – Andy aka
    Jul 16, 2020 at 13:23
  • \$\begingroup\$ I corrected the post: I was wrong to say the UPS is the load - it's not - the load is a bunch of PSUs. The inrush limiter would be the load for the UPS (probably unimportant for the purposes of this exercise). \$\endgroup\$
    – Sean
    Jul 18, 2020 at 8:14
  • \$\begingroup\$ I don't think that affects things - to do the job properly requires knowledge of the output. \$\endgroup\$
    – Andy aka
    Jul 18, 2020 at 13:00

20V TVS in relays coil flyback makes no sense, Are 21A 200V mosfets in(relatively) bulky packages to drive them really necessary? In space critical application I would rather expect bc817.

LM317 - why not LM7812? same result with much less components.

  • \$\begingroup\$ Thanks for your input! I got the relay flyback diodes idea from the new Art of Electronics x-chapters book, as noted in the schematic. Table 1x.7 there shows that a 24 V zener + normal diode combination gives 5 ms release time and only 47 V spike (for a similar relay to mine). I substituted the 24 V zener for a TVS because I figured they were more robust - do you think I should actually use proper zeners here? \$\endgroup\$
    – Sean
    Jul 19, 2020 at 20:59
  • \$\begingroup\$ The MOSFET is indeed way overspecified. I don't need to keep the BOM cost low here - I only need to build 3 of these limiters eventually. Those FETs happen to be ones I have in stock and have footprints for! Same goes for LM317 - that's our standard jellybean part for situations where exact regulator choice isn't important (like in this case). And the space constraint in this case is not so much physical size, rather height: most inrush current limiter modules I've seen are in DIN-mount enclosures that are slightly too wide for the crate I need to fit these into. \$\endgroup\$
    – Sean
    Jul 19, 2020 at 21:00

Your Answer

By clicking “Post Your Answer”, you agree to our terms of service and acknowledge that you have read and understand our privacy policy and code of conduct.

Not the answer you're looking for? Browse other questions tagged or ask your own question.