Controlling BIG Power - PCB Layout Challenges

Using Arduino Motors, Mechanics, Power and CNC
#1

Hey all,

I've got an interesting problem here, and was wondering if anyone wanted to throw in suggestions about how to best implement the solution.

I'm trying to build a PWM controller for a 12v @ 100a motor (Yep). I've found a MOSFET that should do the job:

http://www.mouser.com/ProductDetail/International-Rectifier/AUIRFB8409/?qs=sGAEpiMZZMshyDBzk1%2fWi7r8jvGh7mrdq5g7UCjvCmQ%3d

A rheostat isn't a workable solution, weird stuff seems to happen to the torque profile on the low end of the speed range.

The question is, how do I go about designing a PCB around this thing that'll take the current while sticking with FR4 and standard manufacturing processes (ie. a two layer board via OSH Park)?

Some calculations show I would either need 23cm trace widths, or a PCB with 100oz copper layers.

My immediate thought was to add a couple standoff holes at both ends of the traces and screw down bus bars - pretty sure that'd be a workable solution, but I'm not sure how wide the traces should be in the small area between the MOSFET's legs and the beginning of the bus bar.

The second thought, would be punching vias along the entire length of the trace, with an identical trace on the bottom layer of the board. Fill the vias with solder, and that should emulate a very thick power line.

As well, it doesn't seem like the legs on a TO-220 package should be able to take that kind of power without melting, but I guess with a big enough heatsink it should be fine... I'm thinking an enthusiast CPU cooler and big 12v fan should do the job.

So, how crazy am I?

-apt

#2

Add a "keepout" area on top of the traces so that it doesn't get covered with solder mask and anticipate soldering a copper strip or copper wire onto the trace to increase its ampacity.

#3

Buy a 100A motor controller, take it apart and see how they do it. Or just use the controller you bought.

Use several MOSFETs in parallel, and use very short, wide traces. The battery and motor connection terminals can and should be within a fraction of an inch of the MOSFETs on the board.

The trick mentioned in the previous reply of soldering a copper wire to the PCB traces is feasible, and is done in industry regularly.

#4

I’d parallel the power traces with external heavy copper wire.

#5

For your edification, here are photos of the PC board of the TI Jaguar motor controller. It’s only 40 amps, but you get the idea.

1000×747 260 KB

1000×747 249 KB

#6

To avoid any limitations in current, you will need to keep the PWM pulse width below 60 microseconds.

Note from fig 8 (max safe operating area), the approx. max current is

DC:             [b]0.25 A
[/b]10 ms pulse:    [b]0.5 A
[/b]1 ms pulse:     [b]6 A[/b]
100 us pulse: [b]105 A[/b]

From datasheet: Notes on Repetitive Avalanche Curves , Figures 14, 15 (For further info, see AN-1005 at www.irf.com)

This mosfet is best suited for high frequency switching applications ... no need to worry about 23cm trace widths.

Yes, the legs on the TO-220 package seem small, however they can handle 105A for 100 microseconds. Pulse repetition rate would depend on thermal characteristics and heatsink/fan.

#7

Note that this device has a 195A continuous rating. The avalanche rating is the current/time window you must keep within when conduction is caused by over-voltage breakdown, i.e. avalanche, it is NOT the safe operating area of the device in normal operation. If your circuit and layout is carefully designed (basically no parasitic inductance between the drain and any clamping diode), the FET may never avalanche at all!

Vishay AN1005 (pdf) explains it nicely.

If you're going to design against the avalanche guidelines (necessary if you're pushing up against the device's voltage limits), you just need to make sure that the worst-case current pulse though the FET due to reverse breakdown of the body diode fits within the avalanche SOA.

Since this project has 12V on tap and the FET is rated for 40V, it's not in much danger at all unless the design and/or PCB layout are truly horrible. Certainly you don't need to keep the PWM pulses under 100us!

#8

For 100A I'd use a decent package that can handle that current and screws directly to copper or aluminium busbars.

See for instance: http://uk.farnell.com/ixys-semiconductor/ixfn140n20p/mosfet-n-sot-227b/dp/1427319

I've used 4 similar MOSFETs to make a bridge, screwed down to 500g of 1/2" aluminium slab. Aluminium strip forms the high current conductors and the gate/source connections come down from a stripboard above this block with short twisted pairs. Protection 15V zeners are mounted direct to gate/source terminals:

There's a world of difference between theoretical amps (the limit of the silicon chip itself) and actual current handling (depends on the package and the thin gold wires that connect the die to the package terminals). There are TO220 MOSFETs claiming 170A or so, but they will not carry that current in the real world, or anything like it.

[ BTW as a rough guess I'd say the TO220 package can handle 20 to 30A if directly soldered to adequately thick wires ]

#9

Hi, having worked with machines with this level of power, forget about a PCB for carrying the power, PCB for the control circuit OK. As in post #7, MarkT has shown you a typical high current assembly, it is a very mechanical construction but reliable. TO-220 and 195A, you have got to be joking, as explained in earlier posts, theory clashing with reality. You need much bigger package with good mounting and terminating hardware. Also you need to research current protection devices, fuses and isolators, especially when using DC current, different world to AC and this level of power.

Most controllers have flying wires and connectors to connect the low current control to the main current control device.

Tom..... :)

#10

Talking of big power, some nice teardown pictures from a hybrid car power converter module:

http://techno-fandom.org/~hobbit/cars/ginv/i2compon.html

[ which incidentally is the kind of technology needed to handle 100A happily, not a single TO220! ]