You can get mosfets with 80A capability with ~2mOhm quite easily. (and I'm sure you can find others with even better)
Parallel 10 of these and you have sub-mOhm resistance. Since you don't seem to be using this for PWM you shouldn't have to worry about switching syncronicity. Hence the limiting factor is gate drive. Since you don't care about switching speed(within reason) it shouldn't be difficult. (I assume by protection you just want a switch to disconnect the circuit from power)
So it's not going to be the mosfets that are the problem(except, of course, it requires board area). You will need to use at least 4oz copper(thats the highest I've seen) I'd imagine else your traces will need to be quite thick.
For example, a going from 0.5 to 4oz would allow you to reduce your trace size by a factor of 8(or slightly more). So, if your the power, hypothetically, required 8 in thick traces @ 0.5oz then it could be reduced to 1 @ 4oz. Pretty significant.
I think your board, at 4oz, has a total resistance of about 20uOhms. If you route it properly then it shouldn't be an issue. That is about 0.2W total dissipation if 90A were to run through the solid copper.
I think the goal would be to divide the board in such a way that you can parallel the most number of mosfets and limit the gaps. One easy way, would be to simply divide the board into 6 "slots" analogous to 6 wires in parallel:
+-+-+-+ | | | | M M M M... | | | |
+-+-+-+
The traces though are quite large with the gaps between different branches being minimal.
Considering the mosfets are actually quite small you could probably do 5 or
6 in a 6x2 board without to many issues. (I assume it is more than one layer so you can route the gates to another layer)
Of course some calculation may be wrong but it doens't seem implausable at all just based on these estimates.
I would e very wary of using the pcb for this. Thicker copper = more expensive pcb, when it may be cheaper to use formed copper or ali bussbars between the device and edge of the board, mounted to pcb. Gets rid of the conductor heat problem as well.
Many high current psu designs use this sort of thing, bussbars coming out of the end of the case...
I've seen solid copper bus bars, ~6x12mm cross section, sweated to PCBs. Worked great. High production? Commercial, solderable, bus bars might be the thing.
It can be done. On the very first design in my career the system was fed
5V at a whopping 100 amps. Good old 74AS, tons of them. At some point when they figured out that I know a thing or two about noise I became the official master of ceremonies for the motherboard (where the 100 amps when into).
You'll end up spending considerable time calculating planes, vias, thermal reliefs and (very important) the contact areas to the outside world. Then reflow temperature profiles and such because often you can't afford the amount of thermal relief that you are used to.
Don't try this with a 1oz copper four-layer, I've got some scary photo of boards where people did :-)
--
Regards, Joerg
http://www.analogconsultants.com/
"gmail" domain blocked because of excessive spam.
Use another domain or send PM.
Use enough fets in parallel to keep the dissipation down. Multiple fets scatter the heat, too.
Fastons make nice high-current connectors, females soldered on the board, males crimped on wires. If you go for, say, 15 or 20 amps per faston, and use separate wires for each (instead of one #6) you'll get good current sharing. 90 amps in one place can have current crowding problems.
DO NOT bolt lugs to the pcb. The FR4 will cold flow and the thing may catch fire.
1 oz copper is about 500 uohms per square, if you can get actual 1 oz.
From basics, I calculate (surprise) 246uohms per cm for a 2cm wide 1oz trace, at 20c. So, that's 1W per cm^2 dissipation @ 90A. Top-and- bottom traces cut that to 250mW/cm^2, and wider traces, proportionally.
I can imagine a super-short current path, with i/o wires running virtually to the FETs--that'd be pretty decent.
Depends on whether there's airflow or not. I'd assumed a closed battery case before, but that may not be.
Yes. A 20 amps per wire+connection, and one square of 1 oz copper per, that's just 0.2 watts per connection. A fraction of a square should be feasible.
One problem with dumping 90 amps into a single spot will be current crowding. A small diameter contact will have very high current densities close into the contact. The bigger the diameter, the better. More low-current connections is yet better.
Also most circuits will have the current coming into the contact from one direction, so the copper "on the other side" doesn't help... and current density is that much higher.
Just keep the actual current tracing low L/W ratio (on as many layers as possible) and the interconnect to battery/socket routing (formed metal?) multi-point, to reduce loss and spread heat.
Your 'back-to-back' fets will literally be just that - likely on opposite sides of the same real estate, if pass-through impedances can be kept low enough.. Logic and control lines have to work around the periphery of conducting paths, in the cracks and out of the way. Current sensing may be a major issue.
Regardless of what's on paper, your permissible losses are thermally limited. I also strongly suspect that the 90A is an intermittent or limit condition - not one for which normal rises for continuous operation will apply.
When doing opposite structures mind the thermal reliefs that CAD programs put in by default. You probably don't want any for this purpose. Another trick is to stuff dummy parts in there so you get a through-hole soldered connection as well. But the wave-solder temp profile will cause some cussing among the production guys.
--
Regards, Joerg
http://www.analogconsultants.com/
"gmail" domain blocked because of excessive spam.
Use another domain or send PM.
We do "flood over" (no thermals) on lots of boards, for better heatsinking to planes or to reduce via inductance. Production doesn't mind. The reflow profile doesn't change, and a beefy Metcal can handle the hand-soldered thru-hole stuff.
ElectronDepot website is not affiliated with any of the manufacturers or service providers discussed here.
All logos and trade names are the property of their respective owners.