Re: nice polyfuse

On 30-03-2024 17:22, John Larkin wrote:

On Sat, 30 Mar 2024 12:14:47 +0100, Klaus Vestergaard Kragelund
<klauskvik@hotmail.com> wrote:
 

On 22-03-2024 02:16, John Larkin wrote:

On Thu, 21 Mar 2024 16:27:13 -0700, boB <boB@K7IQ.com> wrote:
>

On Wed, 20 Mar 2024 19:38:38 -0700, John Larkin <jl@997PotHill.com>
wrote:
>

I've been designing relay-matrix switch modules (how the mighty have
fallen) and I don't want the customers igniting my PC boards or
welding my relays by ignoring our 2 amp max current spec.
>
Polyfuses are usually terrible, but this Yageo part is pretty nice.
>
https://www.dropbox.com/scl/fi/w7x8rvqgrdua8boqmxg7y/BK60_1-1955033.pdf?rlkey=xpavzb8b8movr2xd4o5amkvx9&dl=0
>
https://www.dropbox.com/scl/fi/czk9ids5bj68ytcimcrb5/BK60.jpg?rlkey=77lrxc915it7y083quld9ectb&dl=0
>
It (slowly) trips at 2.5 amps in still air at room temp, 3.2 amps with
some air flow. It survived 120 volts DC, which is all I had available
on my bench, pulling about 25 mA.
>
The good part is that its cold resistance is only about 0.07 ohms.
>
The next question is, if I put it in series with a 1 ohm 5 watt WW
shunt resistor, does the poly protect it from, say, a stiff 60 or so
volt source?
>
And does it absolutely protect an inner-layer 50 mil wide 2 oz PCB
trace? I need to do a multilayer board that's 1 oz on the outsides,
for BGAs and stuff, but the board houses don't mind making all the
inner layers 2 oz copper.
>
>

>
The series power resistor idea is interesting.  Are you wanting to put
the PTC in a circuit with voltage above its rating ?

>
I plan to spec the instrument for 2 amps and 60 volts max, which is
the poly rating, but I did verify that the Yageo part survives 120
volts.
>

>
If  1 Ohm  5 watts  limits the voltage across the PTC then it's
probably good.
>

>
My intent was to have the polyfuse protect the 1 ohm current shunt
resistor, not the opposite.
>
https://www.dropbox.com/scl/fi/7efsvz7ba7wq4ebdxmpcp/P948A4_Shunts.jpg?rlkey=3sw5o5j2uxjnmog4md8lrgisu&raw=1
>
>
I connected the poly and the 1 ohm 5 watt WW in series and then
connected them to a 60 volt, 5 amp power supply. The resistor smoked
and then unsoldered itself and survived. The next idea might be to put
a couple of giant diodes across the resistor. Maybe the poly will blow
them up too.
>
I will have a series relay to engage the 1 ohm shunt, and an ADC
across the shunt to measure current, so we could software protect it,
open the relay before the resistor falls off the board.
>

>
SW protection is nice, but what happens during power up or when the
customer only applies power to the relay section, not the unit power.
Then there's no protection.

 I'm designing a FITS module, a fault insertion box, aka guillotine
box. If someone has two boxes that are connected by a cable, they can
chop the cable in half and run the hacked ends through the FITS
module. Now they can route all the signals through, open any, or short
any to any other. That's the classic function. I'm adding shorts to
ground, soft ground faults, and current measurement through any wire,
and voltage measurement/waveform acquisition between any two wires or
any wire to ground.
 So I have two current shunts that can be inserted anywhere. I'm
protecting every one of my connector pins with a polyfuse, so nobody
blows traces off my board, but the polys work too slow to protect my 1
ohm 5 watt WW shunt resistor.
 We have an isolated ADC to measure the voltage across the selected
shunt, an ADUM7703. So our FPGA can sense overload on the shunt and
open the series relay before the resistor fries. The SMW51 5 watt
wirewound is rated for 8 kilowatts for 1 millisecond. Some FPGA
algorithm should mostly protect the resistor and various relay
contacts.
 The classic FITS module, designed by my customer, has become
impossible to make now, and had no protections hence lots of relay
failures. We'll include BIST.
 Big companies used to design their own test gear but the guys who did
that have retired and weren't replaced, and kids these days know how
to type but not solder.
 

I was thinking, that you could do a solid state switch with current sensing for each wire, but that would probably be a nightmare in time and parts.
A PTC should work, if you have a trace which is a couple of mm wide, it can handle more than 10A at 60degrees delta. So you should be able to find a PTC that will protect the trace (I am guessing this is low voltage DC)

>
What do you need the series resistor for, maybe I missed the point?
>

Or an optocoupler across the shunt to sense too much voltage.
>
>

0.07 Ohms, cold, is good.
>
We use a 250V 1/2 amp PTC as well as the 1206 size ones.
>
One problem we had with a 1206 15V PTC was that sometimes (rarely) it
would burn through and short to the next layer down which was 5V.  A
thicker PCB should fix that as well as getting rid of copper just
below the PTC on the next layer down.
>
boB

>
We found the surfmount polyfuses to be really bad.
>
Polys are interesting. Given constant current, at some current they
begin a slow self-heat thermal runaway and (eventually) go hi-z into a
basically constant-power mode with surface temp around 100c.
>
I think that with a constant voltage drive, they become a sort of
constant-temperature regulator.
>
I did find that if you run them hot for a while, their cold resistance
goes up, permanently.
>
I wish there was a really good 2-terminal current-limiting device. A
real fuse does that, once.
>
The real surface mount fuses are bad too.
>
If I want to keep blowing things up, which I do, I need a giant power
supply. This looks good:
>
https://siglentna.com/product/sps5082x/
>

>
Looks nice. Now I have to buy one, hoarding instruments you know.
>

That ain't cheap, but we can expense it and get basically a 45%
discount from tax savings.
>
The user interface looks typically bizarre.

>
I have a lot of Siglent stuff.
>
https://www.electronicsdesign.dk/tmp/Lab2023.jpg
>
Good instruments, but horrible PC software.

 We use their dummy load boxes a lot, and my kids have written a Python
library to control them in test racks. I'll probably use the giant
power supply manually, mostly to blow things up. I could add an
outboard mosfet switch to apply programmable pulses to victims.