100A switch & pcb connection

Need some mechanical boosting just to take the weight and strain of the connector?

Here's a picture of part of a 200A welding MOSFET switch I repaired recently. Six MOSFETs on ~1" square aluminium bar that bolted to the PCB. Connection was via a tapped hole one end of the bar. There were two like this one with N-channel FETs, and a full length one on the PCB with 12 x P-channel FETs.

In the switch I looked at, they had a tiny 1.5A MOSFET driver ('4427) pushing six FET gates, each via 22R series resistor and a SMD cap, unknown value.

Grant.

I like to use nylon-shell AMP or Molex connectors for high current i/o on PC boards. If you use a bundle of wires and share the current, the wire resistance tends to equalize the pin currents.

Bolting current-carrying things to PC boards is bad news. The FR4 will cold-flow, the connection resistance will go up, the thing will get hotter, and it runs away. And you can't assume that if, for example, you bolt a bus bar to a board with 6 bolts, that the current will be shared evenly among those bolts. It won't.

Here's a 120 amp MRI gradient driver board:

ftp://jjlarkin.lmi.net/Amp.jpg

The centrally-located Amp connectors bring in +-200 volt power and export the resulting signal. Multiple PCB layers fan out the currents to the 32 power fets. The shiny bits are nickel-plated copper heat spreaders. Pulsed output into the gradient coils is about 18KW.

Another nice way to do this is to solder faston blades to the board, here and there, and run wires with faston lugs to the board. That way, you can distribute the current any way that works best, and again you get to use the wire resistance to equalize currents.

John

e

Wow, they used the aluminum block for a connection? I wouldn't feel comfortable doing that. Thermally it looked pretty cool, so to speak...

Thanks for the pic.

-- Cheers, James Arthur

makes a quarter milli-ohm, that is.

Splitting up the cable is slick electrically, but it's a nuisance here--I've basically got to interrupt an existing battery cable, one that already has crimped ring terminals.

Multiple bolts would provide mechanical contact of the bus bar to pcb only, not electrical connection. Electrical path is pure copper, from wire to ring to bus bar to pcb trace (with a smidgeon of gold/nickel on the way).

I take your point about the cold-flow potential--don't you think split- washers have enough travel to cover that? How much does FR4 flow, anyhow? More than a split washer, I take it.

I guess my 2nd fave option is to solder the bus bar, hanging off the pcb, then bolt the ring terminal to that. That's a simple, metal-to- metal connection, two bolts. Easy, strong, fast.

That's a pretty board, as usual. I'm not terribly worried about spreading the currents. Maybe I should be. I figured a 1mm thick Cu bus bar, say 1x10mm cross-section by 50mm (.040" x 0.4" x 2"), having a resistance of roughly 80 u-ohms, would more than suffice. After all, it's only 500A. :-)

-- Cheers, James Arthur

metal core pcb's may be an interesting approach for the OP's issue:

google will give you more links/options.

Thanks, Jure Z>

Electrically it was good, see:

for the PCB mating part. Source pattern you can see on PCB top, drains underneath didn't go anywhere, they connected that MOV on RH side to the metal block as well. Not that I was happy seeing a 100V MOV across 6 x 100V rated MOSFETs, but the machine lasted eight years before it blew up.

John L's big amp looks prettier, but he spread out like suburban sprawl, this thing is European high density ;) Fan cooling not so good in a metal workshop environment.

Grant.

I see a lot of that method for DC motor controllers, it's a weak point if you don't put enough copper in, or run too much current per connection, but easy otherwise. I have no idea how much current per lug or blade, but I've seen burnt copper due to fault currents with them. What can happen is the terminal is moved by the wires, the solder joint lets go a bit, then the current heats and eats out the metal from the weakened joint.

For how long do you intend to let this enablement of the short last? Until something melts, or the main pops? Why not negotiate for some kind of slow-blow shutdown? I.e., if your in-rush surge exceeds some certain time, (or X amount of energy), then say, "NO, TOO MUCH" and turn off?

Good Luck! Rich

Split washers are goofy electrically. Stainless screws have a lot of resistance. Phosphor-bronze wave washers are cool. As are Bellevilles. But I just don't like high-current pressure contacts to PC boards... been burned too many times.

One thing that helps is to use copper traces as current-equalizing ballasts. It's just a matter of geometry.

Anyhow, think hard about current sharing. Given parallel paths, 1 milliohm and two milliohms, most of the current takes the 1m path.

Good point.

John

Wow, they used the aluminum block for a connection? I wouldn't feel comfortable doing that. Thermally it looked pretty cool, so to speak...

Thanks for the pic.

-- Cheers, James Arthur

metal core pcb's may be an interesting approach for the OP's issue:

google will give you more links/options.

Thanks, Jure Z.

Indefinitely. Not catching fire only for temporary shorts won't do.

The supply's current-limited and is itself short-circuit protected, but I've got to withstand the full brunt of whatever it puts out, for as long as it puts it out. 120A gives me 50% margin.

If I keep FET resistance to 1 milliohm, that's just 14watts dissipation spread across several devices, with heat sinks and air. Easy.

n

Nope. The equipment has to recover and be ready to go as soon as the fault is cleared. Wires and circuits are designed to withstand whatever they must to achieve that. Nothing melts. That's nicer anyhow--no midnight runs to DigiKey for 120A slo-blows.

-- Cheers, James Arthur

Hi James, I've got some pieces of stamped metal (Tin plated brass I believe.) with a threaded hole and four tabs that get soldered into a pcb. (I can't recall the name, but I got them from Keystone.) Only a

8-32 thread I'm afraid so probably not big enough for you. But perhaps someone makes something similar but bigger. Or could you thread a piece of copper and solder that to the pcb?

George H.

,

le

ed

l

I wouldn't pass any current through the split-washers or any of the hardware, they're strictly mechanical, but you've talked me out of clamping the FR4 in the middle anyhow.

Gut-wise it seems like it could be made to work, but why mess with it? If you say its a pain, that's good enough for me. Metal clamping metal removes the question mark.

Here it'll be 80 micro-ohms end-to-end. Center-fed that's two branches, 40 micro-ohms each. From there I could run separate JJLHT(tm) ballasting traces, that would force more equal currents to the FETs.

One big thing in my favor is that I don't have to run 500A continuously, the really high current lasts just 1mS. On-resistance has to be low to limit voltage drop during that surge. Otherwise under normal steady-state conditions it's only going to pass >I guess my 2nd fave option is to solder the bus bar, hanging off the

I had a bad experience with Fast-ons recently. Older piece of equipment, dead. The female side lost its grip. I'm not sure why/ how, but the result was erosion by arcing, and failure (intermittent). It looked normal, but it was open.

-- Cheers, James Arthur

e:

Thanks Jure, but heat won't be a problem. Lots of copper and wonderful modern FETs will keep dissipation remarkably low.

-- Cheers, James Arthur

Thanks. I'll have a look over at Keystone. That could save me spec'ing a custom part. Brass, however, has more than douvble the resistance of copper, so copper's what I'll be looking for.

I wouldn't send high current through the hardware--that just doesn't seem like a good idea--it's the bus bar's thickness that matters most.

No threads needed (I don't trust them in copper anyhow, it's so soft.) This calls for some ASCII.....(!)

bolt / washer crimp _ _ / wire / _|_\/_|_/ ring terminal / ______ |__|--|__| /

------| |----- |--| ---. copper bus bar, soldered to FR4 .'.'.'| |\ \ \ |--| \ \| /

------|_____|----- |--| ---'______________ FR-4 |@ @|--|@ @ @ @ @ @ @ @ @ @| / |_@_|--|_@_@_@_@_@_@_@_@_@_|___________/ /________ | x |--| x x x x x x x x x x x x x x x/ /x x x x x| |x_x|--|x_x_x_x_x_x_x_x_x_x_x_x_x_x_x/ /x_x_x_x_x_| |__|--|__| / / | |--| | \ |_|--|_| \ / |--| split washer / '--' nut

-- Cheers, James Arthur

Wow! Nice ASCII picture! I applaud you.

John

Brass is more than 4x the resistance per this reference, and it gets worse as you add more copper to it(!).

=A0 =A0 =A0 FR-4

=A0 =A0 =A0 =A0 =A0 =A0 =A0 /

/________

/x x x x x|

x_x_x_x_x_|

=A0 =A0 =A0 =A0 =A0 =A0 / /

Thanks, ASCII rules. Only problem is I got so sucked into the drawing I didn't draw what I meant!

Here, this is what I was talking about last.... (bus bar hanging off the FR-4 for easy access, bolt clamps metal to metal, no FR-4 in the middle.)

bolt / washer crimp _ _ / wire / _|_\/_|_/ ring terminal / ______ |__|--|__| /

------| |----- |--| ---. copper bus bar, soldered to FR4 .'.'.'| |\ \ \ |--| \ \| /

------|_____|----- |--| ---'_________________ FR-4 |@ @|--|@ @ @ @ @ @ @ @ @ @ @ | / |_@_|--|_@_@_@_@_@_@_@_@_@_@_@|_________/ /________ |__|--|__| |x x x x x x x x x x x x x/ /x x x x x| / | |--| | | x x x x x x x x x x x x/ /x x x x x | / |_|--|_| |x_x_x_x_x_x_x_x_x_x_x_x/ /x_x_x_x_x_x| split '--' \ / / washer nut

There, that's better.

-- Cheers, James Arthur

Dang, I was just about to rag you over the FR4 sandwich, but you fixed it first. I must be slowing down or something.

John

But, is the only thing holding the bus bar to the FR-4, the glue between the copper plating and the FR-4? I wanna secure the bar to the board somehow.

Thanks Rich