I finally got them !

These are negative temperature coefficient resistors, or NTC. I measured about 104 Ohms at 28°C and the resistance should drop a lot under heat. Which is going to happen because I'll put one in series with the isolation transformer, itself a mere 3 ohms. Under 220Vac, that means > 2A ! And 220V×2A=400W...

So it's going to get pretty hot.

BUT this is orders of magnitude lower than the naked transformer and the inrush current should be palatable to the 1A-rated input circuit breaker.

And as it heats, the NTC's resistance drops and the power too... It should reach some kind of equilibrium

More information is explained at https://www.tdk-electronics.tdk.com/blob/528070/download/4/pdf-inrushcurrentlimiting-an2.pdf

The exact model I have selected is B57364-s121-m, a 21mm disk that should not be touched by naked fingers, and I found the datasheet there : https://www.mouser.fr/Datasheets/_/?Keyword=EPCOS+B57364S121M&FS=True  (I put the datasheet in the files section as well)

Here are the ratings :

Pmax	5W
R @25°C	120Ω
I max (0..-65°C)	3.5A
Rmin (at Imax, 25°C)	0.4Ω

That thing can absorb 1000µF loaded at 230Vac and the resistance has decreased by a factor of 300... The power loss is in the range of 1 to 2W but the part's body temperature could reach 250°C under high load !

Some useful curves have been extracted from the datasheet :

Under a 1A load, the resistance is about 3 ohm, or 3W. There is enough margin to sustain some shorts and give enough time for the 1A circuit breaker to trip. The behaviour under low or light load is however not covered...

---

This solution is the best compromise I have found because an electromechanical delayed switch or a static relay would create too many design challenges. The small power loss is not critical in this application but convenience and repeatability matter : if the protection trips, the only thing to take care of is to not restart the circuit too early, as the NTC needs "some time" to cool enough so it won't trip again.

I'm a total noob with this specific type of part but I learned something new, YAY ! They are as cool as the "polyswitch"/"polyfuse" I have used in other systems in the past.

---

I did a quick test :

That thing can get quite HOT so I was wondering if a thermal switch could help and take some of the burden away...

---

@Morning.Star found this page: https://www.ametherm.com/blog/inrush-current/transformer-inrush-current-40va-transformer

I get some interesting values:

• Stady State current : 1.8A (nothing surprising here)
• Inrush current : 54A (I expected the double: 308V/3 ohms=100A)
• Inductive impedance : 5,7 ohms
• Inductance : 0.9 H (I measured 3,3H so what happened ?)
• Energy : 1300J ? (This rating is not mentioned in the datasheet, I only see a capacity of 2.4J/K )

The NTC cools in about 100s so this is something else to consider. Maybe a light bulb would be simpler in the end ?