MOSFET Power Dissipation

[Mike Warren]

In the datasheet for a MOSFET I'm looking at is a listing for power dissipation.

Code: [Select]

Total Power Dissipation @ TA = 25°C          PD   62.5  W
Derate above 25°C                                  0.42 W/°C
Total Power Dissipation @ TA= 25°C (Note 1)        1.88 W
Total Power Dissipation @ TA= 25°C (Note 2)        1.36 W

1. When surface mounted to an FR4 board using 0.5 sq. in. pad size.
2. When surface mounted to an FR4 board using minimum recommended pad size.

Why is there a major difference for the first value and the last 2? I understand the difference on the last 2 is because of the heatsinking caused by the PCB, but why is the first rating so much higher?

This is a 24A MOSFET with an on resistance of 36 mOhms

So, at 24A, there is going to be about 860 mV across the device. That means over 20W. If it's really only going to be able to dissipate 1.8W then I can't see any way of getting more than about 7A through the device.

So, why the 2 massively different values under the same (supposed) conditions? (Total Power Dissipation @ TA = 25°C)

[mazurov]

The first number is given for a good heatsink, likely the one which is infinitely good (thermal resistance case-heatsink = 0 so all heat rise is due to the thermal resistance between the die and the case). Other two are for PCB mount, and the smaller one can be used to estimate the power rating with no heatsink at all.

[Mike Warren]

Hmm, that's disappointing, considering there is no way to really properly heatsink a D-PAK device.

Thanks for your reply.

[mazurov]

You can solder it to a copper heatsink. Or you can solder a piece of copper pipe to the tab and water cool it. Look at CPU coolers - many are made of copper and some have (copper) heatpipes. On the other hand, if you need to dissipate 60W DPAK is not a good choice anyway.

[amspire]

So, at 24A, there is going to be about 860 mV across the device. That means over 20W. If it's really only going to be able to dissipate 1.8W then I can't see any way of getting more than about 7A through the device.

You could easily have pulsed currents of 24A, as long as the average power doesn't get up too high.

But also a high current rating goes with a low resistance, so it is common to put high current chips in small packages to minimize the resistance. If they used a chip that could only manage 7A, the resistance would over 0.1 ohms, and so you could only manage about 3A continuous.

If you go for a chip with under 0.01 ohms and increase the amount of PCB heatsink copper, then you might make the 24A.

This is why devices like the IRFR7440PBF are available in D-PAK - 90A and 2.5mOhms.

[Mike Warren]

I don't need to dissipate 60W, but I do need 5W. And adding a heatsink is prohibitively expensive.

Curiously, the couple of D2-PAK devices I've looked at so far only give one (high) dissipation figure.

I guess my point is that a device that's not designed to have a heatsink shouldn't be advertised to be able to handle a current several times more than it's really capable of without an expensive  custom addition.

[mazurov]

5W could be possible on a PCB pad of sufficient size. What is the max. die temperature - newer devices are specified to 175C?

[amspire]

I guess my point is that a device that's not designed to have a heatsink shouldn't be advertised to be able to handle a current several times more than it's really capable of without an expensive  custom addition.

It can handle the rated current - just not continuously. If you were designing a switching power supply with a 10% duty factor, these chips would genuinely handle 24A ON current if you give it a decent amount of copper heatsink, and perhaps use 2oz copper.

The rating is correct. The device has a geniune 24A max rated current and so that data sheet has to say that. It is just not a package designed for continuous high dissipation.

[Mike Warren]

You could easily have pulsed currents of 24A, as long as the average power doesn't get up too high.

The 24A figure is listed for continuous current. To me, this is typical of manufacturers trying to get th most impressive specification in the headlines, rather than what's actually practical.

It's all academic at this point anyway. I now understand what wasn't spelled out in the datasheet and know I'm going to have to go for a much larger and more expensive device.

Thank you both.

[Mike Warren]

5W could be possible on a PCB pad of sufficient size. What is the max. die temperature - newer devices are specified to 175C?

I have quite tight space constraints. There's a plane on the bottom of the board in an area of about 1 sq in and I can surround the device with vias, but the real answer is to use a larger device with a lower on resistance. It'll run cooler anyway. I'll just have to accept the extra cost.

[Mike Warren]

Just as a matter of interest, here's the prototype. It works fine with a D2-PAK MOSFET. I was just trying to get the cost down.

[mazurov]

Looks already pretty cheap to me. How much are you going to save with DPAK?

[Harvs]

Well whether it's right or wrong the way they write this spec, you're just going to have to get use to it.  All switching mosfets are rated like this, you can happily go and buy a 220A continuous TO-220 mosfet.  But the reality is it assumes you can keep the temp of the tab down to 25C and the die will run at max temp. 


But, you just use it as an indicator.  If a modern TO-220 mosfet specs a current of 220A, you know it's going to have a very low Rds(on), and with that a huge gate capacitance.  That's usually the trade off, a 220A mosfet doesn't cost any more than a 60A, but it has a gate capacitance several times the size you need to overcome with high gate drive currents.


BTW, what is it?

[Mike Warren]

Looks already pretty cheap to me. How much are you going to save with DPAK?

It's the most expensive part on there. I was hoping to save about 50c.

[Mike Warren]

BTW, what is it?

It's a controller for a wormhole generator.

Actually, it's just a customisable helper circuit for larger installations. Basically a solid state 24V relay with some logic.

[mazurov]

I would go like that:

1. Estimate heatsink area needed. First, cut a piece of FR-4 the same size as your board. Solder DPAK in the middle, dissipate your watts, check the temperature. If the thing overheats, abandon the idea. If no, repeat with 1/2 of the area, then 1/3 of the area.
2.  If the DPAK temperature is tolerable at 1/3, reroute the rest of the board to have that much copper for the heatsink. Should be doable.

All this can be done in less than a day for the price of US$5 for FR-4 plus your daily salary. BTW, is bottom side available? If yes, you can place vias under the tab, fill them with solder and have very good heat transfer to the bottom side.

[lewis]

5W is a hell of a lot of power to put into a pad that size, even if you have thermal vias to a bigger plane underneath it's still a lot. And if you want to save cost it's going to be 1oz copper which doesn't help. You definitely need a lower RDSon.

[jeroen74]

Maybe you could add something like this:


But they are silly expensive 

[ee851]

If your electronics are running  off mains electricity, you could mount a Peltier junction on the chip, cold side down, to keep it cold at high current.     Peltier junctions suck down tens of amperes at a time, though.

[jeroen74]

You still would have to get rid of the heat in some way, so basically you gain nothing. Better use a heatpipe and an heatsink where the heat can be more easily dissipated.