2
\$\begingroup\$

I came across an SMT FET (IPT015N10N5ATMA1) that has an output of 300Amps in a rather small form factor. Are thermal vias enough to get heat away from the 8-PowerSFN package? Or would something like an Aluminum IMS board or something else be required?

I haven't dealt much with these current levels on a board, but if they make parts like this, what PCB design considerations need to be made to pull the most heat from this package?

It seems like a lot of heat to dissipate if operated near the max rating.

\$\endgroup\$
  • 2
    \$\begingroup\$ If you only ever look at first pages of datasheets, you may find you can assume many wondrous things. If, in this specific case you scroll on to page 3 and make good note of all it says and the notes that go with it, you may potentially adjust your assumptions. \$\endgroup\$ – Asmyldof Apr 24 '17 at 18:29
  • 1
    \$\begingroup\$ Why so much hostility dude? Can't we just discuss things? I am sorry thermal and FETs are not my specialty, I have spent many years working on many other aspects of EE... \$\endgroup\$ – radix07 Apr 24 '17 at 19:04
  • 1
    \$\begingroup\$ @Asmyldof I wouldn't say that. Do you mean the note about 6cm^2? \$\endgroup\$ – BeB00 Apr 24 '17 at 19:04
  • 1
    \$\begingroup\$ @radix07 I did not read the comments of Asmyldof as rude - they seem just very dry, and not really directed to a person. And if I'm wrong, you could just pretend he was not rude... \$\endgroup\$ – Volker Siegel Apr 24 '17 at 22:54
  • 1
    \$\begingroup\$ No, @radix07 correctly identified an air of intolerance to a question asked by someone who didn't already know the answer to their own question, an intolerance that pervades SE.EE and drove me away from it nearly 2 years ago, and still seems to be quite present. \$\endgroup\$ – Techydude Apr 27 '17 at 6:18
6
\$\begingroup\$

It is annoying when all the information isn't in one datasheet, but you can typically find a lot more info on the manufacturers website. In this case, everything is on here. Specifically, read this. Basically, thick copper and thermal vias are what you need. Exactly how much extra copper you need to deal with the heat will depend on your application.

If you want to dissapate the maximum 325W, you're going to need some serious active cooling, with heat pipes and fans. You can see that a normal PCB will not be enough for that scenario, because the thermal resistance with 6cm^2 of single layer PCB is 40K/W, more than 2 orders of magnitude too high. The max thermal resistance of the case itself is 0.4 K/W, so it would be dodgy running it at 325W even with amazing cooling.

I would guess that to get rid of 100W of heat, with the typical 0.2K/W case resistance, you would need a solution with a resistance of less than 0.3K/W, which is a pretty amazing heat pipe.

Edit: to answer part of your question, yes an aluminium IMS will be needed at the very least.

\$\endgroup\$
  • 1
    \$\begingroup\$ Till the day I die I will never understand why they specify things that way or even sell such things in packages that are so hard to add the heat-sink you REALLY need.... \$\endgroup\$ – Trevor_G Apr 24 '17 at 18:17
  • \$\begingroup\$ Thank you! I had troubles finding that document digging through Infineon's site for some reason. \$\endgroup\$ – radix07 Apr 24 '17 at 18:40
5
\$\begingroup\$

The device is rated for 300A continuous, and at 150A with a gate-source voltage of 10V, it has a worst-case Rds(on) of 1.5 mΩ. Assuming 1.5 mΩ is still its Rds(on) at 300A (it should be close, at least), that means this little thing is dissipating 135W of heat.

The junction-to-case thermal resistance of the package is rated as a maximum of 0.4°C/W, so the junction will be 54°C above the temperature of the case; there's nothing you can do about this. The device is rated to operate up to 175°C junction temperature, and assuming 25°C ambient, that means you have a margin of 96°C over 135W, so your case-to-ambient thermal resistance has to be no more than 0.7°C/W. While this is quite a heatsink, it's not outside the realm of feasibility.

Of course, when operating in the linear mode... You're not going to get three hundred amps through that thing and have it survive, not without taking extraordinary measures for cooling. And I do mean extraordinary.

\$\endgroup\$
  • 2
    \$\begingroup\$ Can't really use that FET in linear mode due to the SOA of TrenchMOS being restricted at high voltages, they have a phenomenon similar to BJT second breakdown. Look at SOA graph on datasheet, allowed current at high voltages is tiny... \$\endgroup\$ – peufeu Apr 24 '17 at 19:17
  • \$\begingroup\$ Oh, wow, I see your point. Don't think I've ever dealt with trenchfets before. Still, switching losses when trying to switch 300A will likely be extreme! \$\endgroup\$ – Hearth Apr 24 '17 at 22:06
  • \$\begingroup\$ Yeah for linear mode only certain types of FETs will do, for example Fairchild planar stripe DMOS like FQP19N20 do not have this "second breakdown", their SOA is huge, very important for stuff like audio amps. TrenchMOS would go poof in a linear application. Now, to switch 300A... since losses are RI^2, sharing I between several paralleled FETs is desirable, plus it spreads the heat and makes it easier to dissipate... \$\endgroup\$ – peufeu Apr 25 '17 at 20:14
5
\$\begingroup\$

Never assume you can run a FET at its "rated" current. You have to read the conditions under which the part is rated, which are often misleading, then decide. For example, sometimes the rated current is specified with a junction temperature of 25°C.

If one can figure out how to keep the junction at 25°C while 300A passing through it (which may not be impossible - but it's not trivial) then the full rated current can be used. Otherwise you have to do the thermal calculations to see how much current the device can support. An aluminum substrate board can certainly help, as can active cooling, but it's all down to thermal calculations and management.

\$\endgroup\$
1
\$\begingroup\$

300 Watts, not a problem, I like a challenge. Thermal management is almost a hobby of mine. I am running thermal experiments 24/7.

I am a contractor for the University of Florida to do LED grow lights for Horticulture Research. Lots of Red LEDs are used. Red LEDs have poor performance when it comes to temperature.

This Red LED loses almost 50% of its Radiant Flux at 80°C.
A grow light fixture with a Tj of 85°C is considered do have decent thermal management.

thermal flux curve red led

I do not use thermal vias in a traditional way. Just as I do not use metal core PCBs.

The path of least thermal resistance is NOT going through the PCB to the bottom side.

The thermal pad of the SMT FET (IPT015N10N5ATMA1) is soldered to the copper layer on the top side of the PCB. So why would I put the heatsink on the bottom side?

I'm going use a copper layer as thick as I need to get the heat away for the FET.

Good thermal conductivity required as much cross sectional area and as little distance for the thermal flux to travel.

The thermal pad of this FET is soldered to copper so I will bolt down a thick copper bar to the same copper layer, as close I as can get to the FET.

The PCB layout would be similar to this SOIC 8 Power Pad.
It has thermal vias just in case I need to sandwich the PCB between two copper bars. Mostly they give me a good place to place a thermal couple to measure thermal management.

Pad 50 is for a 4-40 machine screw to bolt to the copper bar. The left edge of this footprint would be on the edge of the PCB. I would put a piece of 0.0007 soft annealed copper foil, squeeze it in a brake press then torque the screw so tight the foil fills every little nook and cranny.

Where the copper bar goes depends on the environment. It could be a heat pipe connected to a radiator. If it's no holds barred, I'm going to attach it to a copper water pipe with ice water flow through it.

I'm done with the thermal management when I see condensation on the copper bar.

enter image description here




Here is an example of the PCB attached to a bar soldered to a water pipe.
The FET would be closer to the bar. Here the bar is not as close to not block the photons.

enter image description here




Got a new shipment of copper last week, these are the latest.

  1. is similar to the above board
  2. I use CoBs as heat sources
  3. This is similar to the above but has two boards attached to the pipe. This is testing a board with 32 3 Watt LEDs
  4. This looks like a candidate for the FET. But with a bar like #5.
  5. this is a 0.375 x 0.25 bar with one end bent with a radius of 0.3125" to mate with a 1/2" copper pipe.
  6. a 0.375" x 0.125" x 24" with both ends having a 0.3125" radius bend.
  7. a 0.625" diameter (same as a 1/2" pipe OD) 17-4 Annealed Cold Finish Stainless Steel, strong enough to withstand a sledge hammer beating a copper bar into submission.

Photo taken tonight

enter image description here




That strip with 32 3 Watt LEDs is at 38°C . Considering this is my garage in Florida and the water temp is 27°C that's not too bad. There is two boards too. To give a little more perspective the thermal vias under the LEDs, when first touched, do not feel hot at all. While the Formica surface of the table, 3" below the LEDs, is about 50°C from the photons. There is no heat coming off this board.

Photo taken tonight

enter image description here




My "Water Tower"

enter image description here

\$\endgroup\$

Your Answer

By clicking “Post Your Answer”, you agree to our terms of service, privacy policy and cookie policy

Not the answer you're looking for? Browse other questions tagged or ask your own question.