A few general reasons to keep an inrush resistor lower rather than higher in value:
If it’s too high, the input fuse (hopefully you’ve wired one in) may not blow in case of a failure in the power supply downstream. The resistor may limit the short circuit current to less than the fuse rating, preventing it from ever blowing. Consider this possibility too if you’re shorting across it with a relay, and the relay (or driving circuit) also fails, leaving the resistor in while the power supply is shorted.
If it’s so high that it’s comparable to the impedance of your load (3k5 might be?), you get a voltage divider with the inrush resistor, reducing and introducing variability on your input voltage and source impedance.
Of course there are other factors driving the inrush resistance value higher rather than lower in value …
I want to take my load up to 1kW at 50V, or 20A. (Not when I turn it on but later when I turn it up/it turns up.)
That would mean I couldn’t use even 5 ohms.
But then again, the resistor (or NTC) would never see 20A normally. That would flow through the relay.
In case of a dead short, if I want the fuse to blow before the relay closes, then I would have to use a 2 ohm inrush limiter resistor. The inrush current would then be “limited” to 25A.
OTOH with a 10 ohm resistor and a dead short, the fuse would only blow through the relay contacts.
How bad would that actually be?
I don’t really mind having to replace the relay every time together with the fuse. I’ll buy a set of them (they’re sold in sets of 5 anyway, and I’m building a number of these setups). I’ll make sure to attach a big warning. (Any other things to watch out for?)
Which leads to my next question, which I was also wondering about from the beginning:
How much do you need to limit the inrush current to, in order not to get sparks?
I was surprised that a 5 ohm resistor already avoided sparks. That’s still 10 amps. Which is a lot.
When I see those big sparks (bigger than the flint on a cigarette lighter), what kind of current am I seeing?
Frankenstein, umm, doing it the correct way, the resistor size barely matters.
The procedure is : you have two relays. If you want to support 20 amp, the main relay needs to be rated for, well, 20 amps. The second relay can be smaller.
The sequence is :
You power up the device with the smaller relay, which needs to be normally closed, connected to the rest your your device. The smaller relay connects through the resistor. Since your device is drawing no or little power at this point, the voltage on either side of the resistor will be the same, and the capacitors in your device will all charge up.
You need a timing controller. Back in the past, people built them with discrete IC timers, including something called a 555 timer IC. These days it’s easier to use a programmable microcontroller. You may dislike programming but you need about 10 lines of code, you can get through it.
The timing controller should use a linear voltage regulator that the high side is hooked to the power rails for your main bus, after the relay. So it only powers up once your device has power.
It then counts time until enough time has passed, then sends a signal that will close the main relay ,which is normally open, then after that, opens the small relay.
Easy peasy. I bet whatever the rest of your device is, you will want a microcontroller to make stuff happen, so it’s easiest to make this power sequencing happen first and then do whatever main device operation is.
Also if you want to really be sophisticated, what you do is put a resistor divider on the main power bus, and feed a divided voltage into an analog input pin on the microcontroller. (I recommend dividing it down from 0- 50v down to 0-1.5v, leaving you some margin). Then instead of measuring time, you measure voltage directly, and sequence the relays based on actual voltage change. Once the voltage stops changing, you switch in the main relay and switch out the small inrush limiting relay.
This is faster and what you would do in a more robust product.
Another thing I recommend - use a proper schematics tool and just get a PCB made. PCBway can get you a board with a turnaround time of a couple weeks for $30, a week for $70. I would use the tool “EasyEDA” to design the board, it’s good. Better than Eagle and most other hobbyist tools. Once the complexity of a device starts to go up, having a PCB dramatically reduces the assembly errors and just general trouble you will have. I would print onto the mask the component values, which makes assembly faster as well.
Have to read that again, slowly, and think about it carefully…
Actually, there’s a lot that I could do with a microcontroller in addition to sequencing the startup. I could even use a RaspPI or even an Intel NUC with Windows 10! (In fact, already have one that I eventually wanted to use. That little gizmo that I linked to already has control over USB or Bluetooth on Windows.)
Thanks for the write-up! Gonna give that some thought!
OK I read your post again, slowly. I get it now. Bit of a culture shock to consider a microcontroller but doesn’t really sound so complicated actually.
I do wonder: what’s the reason for / advantage of disconnecting the resistor with the smaller relay? The thing is already bridged with the large relay, why have the small relay?