Re: programmable circuit breaker

On Wed, 5 Mar 2025 11:01:48 -0000 (UTC), Phil Hobbs
<pcdhSpamMeSenseless@electrooptical.net> wrote:

piglet <erichpwagner@hotmail.com> wrote:

john larkin <jl@650pot.com> wrote:

 
We're designing a modular power system and figured we should have a
relay module, and my PCB layout guy is fast so I decided to whip out a
simple module.
 
Featuritus kicks in. We normally measure voltages and currents, so the
next logical step is to make it a programmable circuit breaker too.
That also protects my relays and PCB traces, to some extent.
 
So how might a user program a circuit breaker? Just RMS current with
some time constant? Allow fast and slow trips?
 
Fuses are usually specified to trip at some I^2*T, but that can't be
the whole story, because 1 mA is a lot of I^2*T in ten years.
 
And my current sensor saturates. If the module is specified for 7.5
amps, and the 10-amp Hall sensor saturates a bit past 12 amps, so a
zillion amps looks like 12 so the I^2*T math doesn't work at, say, 30
amps.
 
Sine waves sort of work if they don't clip too hard. Luckily, sine
waves are kinda flat on top.
 
So I need a trip algorithm. That will be executed in an FPGA that sees
a fast ADC that is digitizing the Hall sensor output.
 
 

 
Aren’t fuses specified by minimum carrying current? So a 1A fuse is
guaranteed to carry 1A forever and reacts I^2t only above that.
 
I guess your current sensor should remain monotonic upto the maximum
possible current your supply can generate?

I'm a relay, and I can't control whatever a customer wants to switch.
He might try to short a monster battery or something by mistake.

 
 

>
Nah, that’s way too normal and prudent. ;)
>
First off, obviously you want a conventional fuse sized to prevent the
thing catching fire if something fails.  Having an upstream switching
supply kick off at about the right point helps a lot too.

I figure that my electronics can open a relay faster than any fuse can
respond, and a user won't have to replace a blown fuse.

Here's my circuit:

https://www.dropbox.com/scl/fi/qxksgybi9fnkbzj7k7k8n/23S946A8_sh9.pdf?rlkey=rmts6e1kz04o2tn48kzeslszg&dl=0

Maybe I shouldn't diode clamp the relay coils. I'll have to test the
dropout time.

>
You can certainly estimate the dissipation when the sensor rails, but it
requires making some assumptions about the current waveform. 
>
For a start, I might try identifying anomalous cases e. g. an overvolted
transformer saturating, and cut those off when the sensor rails, regardless
of I**2t.
>
If the load is linear- looking, it may be reasonable to assume the current
is sinusoidal, and curve-fit the part that stays onscale.
>
I’ll also put in a plug for my second-favorite method for fast temperature
measurement: cut out a little niche inside the thermal pour, right next to
the active device. Put an 0603 thermistor with one end on the pour and the
other on a skinny trace to whatever you’re measuring with.
>
If you avoid the temptation to something more convenient that looks
similar, this trick will give you 100-ms thermal time constants, which is
good enough for many interesting things. (Which possibly includes
protecting power FETs from melting before the fuse blows.)
>
Cheers
>
Phil Hobbs

Somebody should make a part with a resistor/current shunt and a
temperature sensor inside, an electronic resettable fuse. I've used
some IC e-fuses and they are flakey.