Parallel Mosfets and Heat Sink 2

Well, 33A through one MOSFET with an rds of 5.6m ohm is about 500W. So that's 1500W total. That copper will get quite toasty quickly. I'd say you probably can't do that for very long before the temperature of the MOSFETs is exceeded. Is that really going to be that much current all the time or is there some duty cycle?

Glenn

I^2*R = 1089 * 5.6mohm = 6.1 W

Nonetheless, the datasheet max current >30 A ONLY applies if you have a gigantic heat sink AND forced air cooling on the heatsink. Otherwise, the rather limpy 1.87 W max power limit is more likely the case. And, even then, you'd need a REAL heatsink and some amount of air flow.

TTFN

FAQ731-376

Yikes! I don't know what I was thinking with my calculation. Off by two orders of magnitude. I think I need a vacation.

Glenn

With your application being up in the 100A range, you are at the point where you need to think about scaling up the power switching components and the heat sink assembly. Of course this precludes using a PCB based design, but in reality your design looks to be way beyond PCB level.

Various manufacturers make assemblies that are combinations of heat sink and components that mount as part of the mechanical design of your system. Depending on the heat dissipation and temperature requirements you could also be looking at using fans for cooling. One place that I can think of to start looking for this sort of thing would be Darrah Electric (assuming your in the USA).

As far as soldering the PCB component to bus bar, well, I can think of all sorts of mechanical problems with this approach, not the least of which is getting everything up to soldering temperature.

I would also suggest taking a look at IGBTs instead of mosfets given your power requirements. IGBTs are readilly available in "brick" packages which are designed for use with heat sink assemblies.

You will generally need a fan for that kind of power. Fans make a gigantic impact on the product design. I find they can increase the power rating about 4~6x over the same footprint. There is always concern about fan failure. It is over-feared. Fans can be very dependable. They can be motion monitored cheaply. They can be turned ON only when needed.

I would worry a bit about your copper verse board expansion coefficients.

Keith Cress
kcress -

You can solder unusual items to a circuit board. I once had a design where a several ounce piece of machined brass had to be soldered to the back-side of a SMT board - pie-pan heat shield along with solder paste and a MAPP torch borrowed from maintainence did the job just fine for a few dozen prototypes. With the heat you need to dissipate, you might try getting a square copper rod, drilling a hole through it, soldering your devices to it, and cooling with water.

Since you are using 25 Volt MOSFETS, I don't think you need a half-bridge like the FAN7382 which is rated up to 600V and is generally used for florescent lights, motor controls, and inverters (don't be fooled by the ratings of the logic supply). There are other MOSFET drivers out there for lower voltage applications.

Ok, so I'm a little off base, but how else to learn than to talk with people here.

I'm reviving my EE education from 30 years ago (worked as a computer scientist instead). In this project, I''m building a multilevel converter. This is one module. It switches a battery (eventually in parallel with a solar panel) into a circuit with a bunch of other modules to generate a 120Vrms 60Hz power source. The maximum continuous operation will be more like 30A, but once in a while I might have 100A load for a minute or two. I may eventually make a duplicate and operate it 180 degrees out of phase to make a 240Vrms system with both A and B phases.

So far, I have this working using single mosfets with 12 modules driving small loads like light bulbs and portable fans. The logic is basically correct, although I haven't yet implemented synchronizing with another 120Vrms 60Hz source to charge the batteries.

I was planning on blowing air over the copper busbar and I have a thermistor there so I can shut things down if it gets too hot. I was also thinking of dipping the copper busbar in a solder bath to tin it, then with some surface mount solder paste, stick the mosfets to the busbar and use a heat gun to solder them.

I guess I'll go back to the drawing board. It is interesting that I was unable to find an app note by on-semiconductor on how to use this part in a high current application. I did find one with surface mount parameters using the board as the heat sink. I don't think that could possible carry the current without melting the traces.

Comcokid, what would you suggest for the mosfet driver? The FAN7382 is quite easy to use. I need something with a floating supply that can handle double the voltage of my battery safely. I have a voltage doubler circuit that turns on when the high side mosfet turns on so the mosfet can be held on for a long enough time.

Thanks to all for the feedback.

Ed

The reason there is an app note is that there are plenty of application information from heatsink suppliers, like:

TTFN

FAQ731-376

Soldering components to a piece of copper just is not typically done. Also, having different leads soldered to different pieces will likely lead to a failure due to the thermal movement.

I've seen DC motor packages using surface mounted devices that could handle much more than 100A. However, I didn't pay enough attention to how they were transferring the heat. It's possible though to use a whole bunch of vias (no thermal fingers) to a bottom layer that is bolted to a heatsink. It's also possible to use a big copper pad area on one side of the board which has a bare copper strip at the edge that is then clamped to a heatsink. There are manufacturers that will make boards with thicker copper layers.

You could also just go more conventional and use tab devices and hang them off the edge and clamp them to a heatsink.

Veryuniqueid - the FAN7382 part is fine for your design if it works and cost is OK.

Sounds like you're in a similar situation to my current design project. 28 years of analog small-signal and rf, and my employer assigned me to design a 1/2 kW pure sine inverter with features specific to our market. Neither my employer or me had experience in this area. Ive done plenty of buck-boost regulators, but never more than 15 watts before.

Some of the consumer-grade inverters I took apart for this project beef-up their input traces by soldering 14 gauge buss-bar wire along the traces, or with a tooled 1/8" thick x 1/4" tall w/thru-hole legs, standing buss-bar to handle the 20 to 40 amp input currents. All of these design parallel 2 to three MOSFETS on each leg of the first push-pull switch-mode step-up stage.

All good new designs are always a little "off-base". If they weren't, no progress would ever be made.

You need to ditch that mosfet and change to a thru hole tabbed part. Then use an effective heat sink.


Here's one I just did after an extensive heat sink search. You won't find more cooling bang for the buck than this particular extrusion. And it is a stocked model at that!

The clips are totally the way to go too. Never, ever, consider screws and nuts torture. Let me know if you want me to hunt up the part number.



If you're interested there are more pictures at the bottom of this page:




Keith Cress
kcress -

That heat sink sure looks impressive. I'd appreciate a pointer to it. Thanks.

It looks like you have four devices attached to each one. How much power are you sinking with this? What is the temperature difference?

Everything I used was on this page.
See here:

I used item 619-0068 as the heatsink.

I have some you can have. Make sure it will fit your design before I go to the effort though.

Power? I can't remember off hand. It's a 280W battery charger that had to run with no fan in a hot enclosure. About 56W I'd guess.

Got to Aavid and watch the video on the Max Clips. You will never consider screws again.

Keith Cress
kcress -

The clips seem like a good idea, but only if you can ensure that the tab and the heat sink are correctly mating. How do you ensure that the tab surfaces are parallel to the heatsink?

It also looks like you're using something like the In-Sil-8 pads described on the bottom of the catalog page?

TTFN

FAQ731-376

Turns out that the clips are superior to screws because of their pressing on the center of the case instead of the tab. This increases the flatness of the device onto the heatsink.

Using just the tab hole can actually make a gap under the device in some instances.

Yeah SIL pads go hand in hand with the max-clip scheme.

I was majorly impressed when I put those devices onto the sink. Without a jig it took me about 50 seconds to do them all. I would guess screws and hardware would take about 4 minutes. Two a minute with no fumbles.

Keith Cress
kcress -

Hey smoked,

That heat sink application looks strangely familiar. Where did you come up with that one?

Heh heh heh.

Hey! Maybe you remember the power on the hottest device? But likely, probably just the rise.

I'm trying to remember if the reverse protection diode was the hottest device as that would allow a straight V[sub]drop[/sub] x I power number.

Keith Cress
kcress -

If I remember correctly, the output diode was a problem. You were using a dual diode that I think was a 30A and it was getting the hotest. Eventually you changed it to a single diode with a heat sink and it worked quite well. I think the second hotest device was one of the switching transistors (FDP8441)and which one it was changed depened on whether it was running in buck mode or boost mode (Q7, Q8,Q9,Q10). If I recall the inductor got pretty warm too.

The unit with the TO-220 packages and the massive heat sinks ran extremely cool, even under full capacity. How I remember the difficulties, though, in turning the board inside out so that the transistors were on the outside edges.

If I were to go through the old email logs, I might be able to tell you which one it was and what the temperature it ran at.

I do know that I pushed that little charger up close to the limits in test. I had the thing putting out about 300W at an 80C ambient. When I told you this, you started calling me the Dr of Death.

Too bad that they ultimately decided to not go with the product. They are still messing around and scrapped the whole design including the charger and are now focusing on an entirely different processor platform (this is the 2nd time that was changed).

That company takes the concept of the bumbling method of engineering to an entirely new level. One like I have never seen.

Dr Death

Yep the output diode was the biggest heater.

So there you have it veryuniqueid. That diode has a V[sub]F[/sub] of 0.54V

P = V x I
P = 0.54 x 11A
P = 5.94W (convection only 80C ambient)

I believe we saw numbers like 130C

Keith Cress
kcress -