Physical connection, cable to high-current T0220

[CarbonWerkes]

Hi

Sorry if this has been previously addressed- search terms are not always appropriate.

I have an application where I have several power Mosfets (IRL1104) which are capable of 100A. I think the leads would melt at 40, but that is the spec.

OK, so the question for the experts is- how does one go about physically connecting a supply/ground/switch lead to a T0220 when you are dealing with these high currents? Assuming that solder is 1/10 the carrier of copper, just doing a solder bridge from a leg to a throughhole on a local pad is probably not going to work. So, if I am space constrained on the backside, are there any tricks/adapters out there I can leverage? Im going to be running 8-10 of these in a 2x3" board with not a lot of clearance above or below (call it 3/4" top, 3/8" bottom clearance for adapters/cables/bars/whatever).

Thanks all-
Rob

 

[itsmoked]

Hi Rob.

I am working on that at the very moment.

On these designs you have to work the design with the heat dissipation as the top of the design pyramid. You often can't trot out "This is what it can look like - now how do we fit it all in so it works." Which is more typical of board layouts.

25C ambient, 75A will raise the internal bond wires to 220C The package material has problems starting at 190C.

I have my doubts about continuous +100A too.

The data sheet limits the device to 75A.

Your best bet is to make sure your devices are set inboard from the board edge at least 0.250" You need this so you can bring large traces to the pins.

Use at least a four layer board. Run your traces as large as possible on all layers.

Do not take "no" for an answer with regard of using a fan. A fan is hugely successful about removing the heat from the board and leads.

I would go for large meniscuses on the soldering.

Keith Cress
Flamin Systems, Inc.-

http://www.flaminsystems.com

 

[CarbonWerkes]

Hi Keith

Thanks for your reply. In my situation, I have a billet machined Al chassis, so I can mount the T0220s to that, and then use an external fan on the chassis radiator if required. With the 1104, we did some calculations (IR and me), and we did not see the heat you suggest- but then again, there are a lot of assumptions made about gate drive current and frequency.

Anyway, it sounds like you feel it is possible to carry 75 amps on the board copper? My basic calcs show that for 2oz copper and a 150F temp rise for exposed layers (2 layer board), I still need 320 mil traces (630 for single side)- and internals would need to be probably 3x that. Since the legs on a T0220 are 100mil centered, it seems to me pretty much impossible, even with a 4 layer, since you would need to neck the traces down to less than 100mil. And, even with all the trace width in the world, you still are limited by the throughole-copper surface area I believe?

Im wondering if it would be better to just make small bus bars from copper, or use a blank throughhole to position a cable and then direct connect that termination to a leg (crimp/solder)- maybe bend the power legs opposite sides and mount the cables opposite also?

Seems like it would be a common problem, but not much info on it...

Rob

 

[itsmoked]

Outta the box. Can you use a IGBT module instead? They're made for those currents and lug terminals.

http://rocky.digikey.com/WebLib/IXYS/Web Data/MDI(MID,MII)100-12 A3.pdf

Is not your rise for static/convection? Run air over that board and that should drop dramatically.

Next.. You're worried about 300mil traces but are not worried about the tiny wirebonds?

You can bend the leads of TO-220s so they are far apart.(It's a standard bend.) You can run the tab as the other connection to your machined heat sinks. (Keep'em isolated.)

You can keep paralleling the FETs till the currents are reasonable. Tie those tabs together.

You can not worry about the traces too much because the length that needs to be reduced to meet the leads is very short.

Yes you still want to use the inner traces! They still help to reduce resistance and hence I²R losses.

You can talk to your board-house, they may be able to provide 4oz on the outer layers.

With wide traces your board guy can skip the solder mask on the big traces then you reflow the board piling up Big Solder.

You can lay bus wire on un-soldermasked traces to increase the cross section.

Have you considered how you're going to get this current onto and off the board?

I've seen hot tub controllers that were running Big Power across pc boards.

What is this a motor controller?

Keith Cress
Flamin Systems, Inc.-

http://www.flaminsystems.com

 

[CarbonWerkes]

Hi Keith

Thanks again for the thoughts-

Well, this board contains a few different power switch types- the 1104 for low side, and then the IRF331x type intelligent power switches for some high side requirements (which need configurable shutdown points).

Im working against spec sheets here- I dont know if their wire bonds are rhenium, all I know is that the package is rated to a given current at a given temperature; I have to assume they know more about their chip that I do. What I do know is that the system I have is very low profile, and there is not enough volume/air mass to make convective cooling feasible.

Getting current on and off the board is really the primary problem I posed in my note- how to get AWG12-class wires to and from these chips given a constrained volume. My thought was to use the PCB as just a fixture for the Mosfets and the wire, and then physically bond the wire to the Mosfet legs (or to a bus bar/heatsink for leveraging the tab as a path). There is a little bit of work in all that bending and crimping, etc, so I wondered if there are bus bars designed for this type of problem. Worst case I can mill them myself.

Yea, the application is motor control, but as stated, the volume I have to work with is not what I would prefer

Best,
Rob

 

[fsmyth]

Crude ASCII art follows. Adjust appropriately.
On a couple of layers, run your busses thus:
_______________________________ _____
_________ __________ ______ ...._ \
O--O _ \_/ O--O _ \_/ \___|
______/ \__________/ \_________ ...._____
________________________________ _ \
\___|
with holes for screws in the end pads.
Width of busses determined by number of layers.
Use an inner layer for the gates, or bridge
the bus with a resistor and use all the
layers for bus.
<als>

 

[sreid]

For power electronics, PCB is available with much thicker copper clading.