Sensing 300A on a SMD PCB

There's magnetic sensors you can glomp up against the conductor and they won't change as much with temperature. To get that much current through a PCB 'trace' without it evaporating is going to require 'xtreme' copper thickness or some kind of bus bar stuck on the board, I guess.

--sp

We did a paper design (still waiting on the customer to get real) for a 240 amp polyphase switcher. We decided to split the current into 8 paths, each with its own sense resistor, and monitor/sum the sense outputs. The board connectors would be several multi-pin Amp things, with 12 wires per connector. So there would never be 240 amps at any one place on the board.

Hundreds of amps aren't compatible with conventional PCB construction. Super-heavy (8 to 20 oz) PCBs are available, expensive and exotic.

Cooling benefits for being physically dispersed, too.

Den tirsdag den 9. februar 2016 kl. 18.16.46 UTC+1 skrev Spehro Pefhany:

all wrapped up in a nice package:

I imagine you could use the same etching process used to make screening cans to make metal parts the shape of high current tracks and then just reflow solder them on top of the pcb

-Lasse

T/620-1537-5-ND/4473976

I was going to suggest those people. I didn't realize they made a part up to 200A -- that's impressive.

Without looking yet, I'm concerned magnetic sensor's phase delay might destabilize a fast switcher's control loop. Maybe not, especially if it uses the current-transformer principle, and not Hall-type. Dunno. If I can use the inductors' DCR, that's attractive, since it's already there, and saves adding more loss.

I think it pretty well has to be buss-bars. I'll consider extreme copper too, but you just can't mount 25-mil pitch parts on heavy copper AFAIK, and I don't want to go to multiple separate assemblies unless really needed.

I was up late last night looking at polyphase switchers, for splitting up the load to help with much of this.

Cheers, James Arthur

I'm thinking along the same lines on many fronts. I really only need 300A at one place, the output.

It's insanely awkward to need 300A even there, but to make things work you've gotta work with what the world has, not what you wish it had.

Cheers, James Arthur

The part that Lasse tossed up uses Hall-effect, but the data sheet doesn't say a lot about AC performance -- it lists an "internal bandwidth" but doesn't give Bode plots.

I'm not sure how DCR sensing is supposed to work, but maybe if you used a bunch of Linear parts in parallel in a polyphase arrangement you could more or less use them as designed, with some clever cascading-sync to make them pop off one after the other.

Lasse's part has a 4uS response time. That's a bit pokey for controlling the switching cycle in a 400kHz current-mode switcher, but more than fast enough for output overload sensing.

Here's one of Linear's 'DCR' (d.c.r.-sensing) parts:

I just discovered them last night. It's a pretty cool idea.

Cheers, James Arthur

So do we, but the multi-pin, multi-wire connectors are nice. The wire resistances equalize the pin currents. Our load, a laser, would be a few feet away. Multiwire cables are cheap and flexible.

One could fab some right-angle things, from stock aluminum extrusions, as input and output bus bars, screwed to pems pressed into the PCB. Still distribute the fets and shunts maybe.

High current on PCBs is nasty. You get hot spots, oxidation, cold flow, burnt craters.

PCB-mount 300 amp hall current sensors are cheap, under $20. Some have multiple solder-down loops for the primary current, some just a hole to pass a bussbar through.

High current in aluminininininium scares me. Maybe it shoudn't, but I'd use copper.

Cheers, James

I can see the bones of the DCR sensing from the schematic. But if you look you'll see that they're filtering the snot out of the inductors' voltage to do it -- their DCR sense pins are getting filtered with a 1st- order LP filter with a time constant of around 200us, which -- barring fancy stuff inside the chip -- is slower than the Allegro part that Lasse was showing.

Copper can be soldered, just to make sure. For production stuff, heavy tinned steel seems to be used as buss bars too, all soldered to the board with as many feet or screws. They still use copper for the more expensive stuff (top of the line servers for example).

What little I skimmed said the R-C filter is supposed to match the L-R time constant of the inductor's L x d.c.r. I've never looked at inductors that way, so I have little feel...

If a 1uH inductor had 5 milliohms d.c.r., the time constant is

That said, you *could* feedback a current-mode switcher based on a switch-current averaged over several cycles, but that could get squirrely. The nice thing about limiting current cycle-by-cycle is that it can't ever get away from you.

Cheers, James Arthur

Aren't those tinned copper?

Anyway, if I have to have something made custom, might as well use the best material, here.

Cheers, James Arthur

I think time constant is L/R so 1uH and 5mohm is 200us?

piglet

An IR detecting photodiode should pick up when it glows red.

NT

I worked on a 200A design a couple of years back. We never launched it other than some initial design work. Wurth has some press fit screw connectors for high current. You will not solder anything unless you do some serious pre-heat. I forget what sense resistors we were trying to use but it posed a problem with heat. We had a rather small, completely closed box, no fans allowed. Ambient was data center cooling so that made it a little better.

I'd have to dig back through my emails and such but it was possible at

This is the high end - used by the MRI guys with 3kA+ supplies driving gradient coils with fast transconductance amplifiers:

No worries about mounting, soldering, connecting, parasitic inductance and resistance - just run the cable through the hole.