Parallel Mosfets and Heat Sink 2

[veryuniqueid]

I am using an NTD70N03R mosfet from On Semiconductor in a half bridge configuration and would like to put several of them in parallel to handle more current (i.e. instead of 2 mosfets, I will have 6) Assuming I can generate the drive current to do this (I'm using a FAN7382 to drive it) my concern is dissipating the heat.

The circuit could draw about 100A which should divide about evenly between the three parallel mosfets and be within their ratings. Running 100A through a PC board trace seems excessive, so I'm looking at using 1/8inch by 1/2inch by about 2inches copper busbar. I am planning to solder the mosfets to the busbar so in addition to being a conductor it will also act as a heat sink. The gate pins will be soldered to a trace on the board, and the source pins will be soldered to another copper busbar.

My questions are one: is this a reasonable way to parallel mosfets and two: will the busbar provide enough of a heat sink?

Thanks for your help.

Ed

 

[veryuniqueid]

Itsmoked and all,

Thanks, This has been very helpful to me.

Isn't the diode power dissipation only a percentage of the 5.94W depending on your PWM duty cycle? Was the problem with the dual diode because they weren't quite matched and thus one started hogging the current, got hot and started hogging more current and got hotter ...?

Also, it looks like the power attributed to the switching of your MOSFETs is probably greater than the I*I*Rdson losses because of the PWM frequency.

Is the following analysis correct?

A three inch piece of that heat sink has a thermal resistance of about 3 degrees C per watt (per the data sheet) and it looks like you have two six inch pieces. So you have 1.5 degree C per watt for each heat sink But you have four devices on them, so that should be 6 degrees C per watt per device. Keeping junction temperatures below 175C you get 95 degrees C for each sink. The FDP8441 junction to case is 0.5 degree C per watt and the insulating pads look like they are 1.5 degrees C per watt. It looks like you have four of these on the one heat sink, so that is 0.5 (case) + 1.5 (pad) + 6 (sink) = 8 degrees C per watt total junction to ambient for each device. Using the 95 degree C temperature difference, it can handle 11.9W for each device attached to it. You seem to be comfortably below this especially since you aren't using boost mode and buck mode at the same time.

Ed

 

[itsmoked]

Your numbers seem right.

Your diode assumption is off a bit as it was strictly there as a blocking diode and hence had up to about 11A of DC put across it.

It was also a single not a dual. Duals are very common and not all that useful for blocking due to their lousy sharing.

http://www.st.com/stonline/products/literature/ds/11216.pdf

The MOSFETs have a far far lower on resistance than the diodes. If you want to get carried away you use MOSFETs instead of diodes for reverse blocking. But that adds parts.

You're most welcome.

Keith Cress
kcress -

http://www.flaminsystems.com

 

[EricRICPDAS]

I like Mosfets for less heat dissipation than diodes. I found that when on, they are low resistance in BOTH directions. Even the 'body diode' becomes low resistance.

Is it a good assumption that current really divides equally to three parallel low resistance devices? Note Comcokid's finding of beefed up input leads. The connecting wires would need to be low in resistance compared to the device, or else the connecting wires will direct most current to one device. Considering this, perhaps a single IGBT instead of each parallel assembly is safer to use.

Eric Ratliff

http://www.icpdas-usa.com

 

[IRstuff]

The ON resistance is dictated by the gate voltage, the drain voltage, and the substrate reverse bias. And unlike batteries, the behavior, while non-linear, is not grossly so, so 3 paralleled devices will behave much like 3 separate devices individually supplied by the same drain voltage.

TTFN

FAQ731-376

 

[jimkirk]

The on resistance of MOSFETs increases with die temperature, so if everything is properly handled, they will automatically balance the load. A hot FET will have the resistance go up, effective offloading the current to the other FETs in parallel.

 

[itsmoked]

BTW, Linear makes Ideal Diode Controllers to run MOSFETS like very low loss diodes.

LT4355, 57, 51, 54

Keith Cress
kcress -

http://www.flaminsystems.com

 

[DHambley]

If your going to be paralleling MOSFETs, here's another good tip. Use beads on the gate drives to ward-off nasty 20MHz oscillations on the gate. It's the result of the interplay between the drain-gate capacitance and parasitic source and drain inductance.
I saw an app-note here several years ago:

http://www.microsemi.com/micnotes/APT0402.pdf

I've found that a bead which is about 50ohms at 100MHz has just the right damping at the gate-drive range of freq's that you need.

 

[itsmoked]

Nice tip hgldr. Thanks!

Keith Cress
kcress -

http://www.flaminsystems.com