5

I am designing a device based on TO-247 MOSFETs. The power is pretty high - about 50W per package, so it's going to be either water-cooling or something big with fins and fan. All is great, except they have the drain connected to the thermal pad.

So common heatsink should somehow isolate between the different drains.

One way is by the thermal pad, but since I am using relatively high voltage (400V) it scares me that a thin pad may be punctured.

But I am out of good ideas about what else it can be.

enter image description here

TQQQ
  • 831
  • 4
  • 16
  • Not a problem. Those packages are suitable for over 100W, with appropriate cooling. A cost reduction to TO-220 types might even be feasible, at least thermally speaking. – Tim Williams Mar 29 '23 at 19:21
  • If you only need functional isolation to an internal heatsink, do not worry. If you for example couple the transistors via _reinforced isolation_ to chassis, you do need to pay close attention to your isolation to be compliant. – winny Mar 30 '23 at 12:38
  • Why your MOSFET has 4 terminals? – fraxinus Mar 30 '23 at 13:52
  • 1
    @fraxinus there is an extra leg for source so the currents don't affect gate driver – TQQQ Mar 30 '23 at 17:37
  • @winny yes! let's say, it might at any time became a matter of reinforced isolation. Any practical advice? – TQQQ Mar 30 '23 at 17:37
  • related (if not duplicate) [question about TO-220 mounting kits](https://electronics.stackexchange.com/a/227854/7036) – Nick Alexeev Mar 30 '23 at 23:22

4 Answers4

8

You can use "squishable" silicone gap pads like this one. They provide high insulation (e.g. 10 kV per mm) as well as good thermal conductivity.

Choose one with high thermal conductivity (given in W per metre-Kelvin), generally 2 W/mK or higher is good. You need to apply pressure to "squish" them, and as they get thinner the thermal conductivity increases.

The problem is: You need to prepare it yourself for your application i.e. there's no ready-made pads for TO-247 or something else.

Rohat Kılıç
  • 26,954
  • 3
  • 25
  • 67
  • You refer to a 2mm pad, it seems that 0.5mm is better for 50W, what do you think about 0.5mm? – TQQQ Mar 29 '23 at 15:34
  • You have two requirements: Heat transfer and insulation. Check if 0.5mm one still provides the same insulation (in your case you should think about at least 2 kV) and good heat transfer. As I indicated in my answer, even if you get 2mm you can "squish" it down to 0.5mm by applying external pressure. It'll give you 5 kV insulation and higher conductivity than initial (2 or 3 W/mK). – Rohat Kılıç Mar 29 '23 at 15:41
  • Do they still insulate 10kV per unsquished-mm when squished? – user253751 Mar 30 '23 at 11:17
  • @user253751 Depends on the specific product but generally yes, although here's a physical limit for squishing. Here in the factory we use 2- and 3-mm thick ones in our medical-certified power supply designs, and we squish them down to 1~1.2mm. Some of them are under the pressure of component leads (e.g. PFC output cap) and still provide the maximum insulation. – Rohat Kılıç Mar 30 '23 at 11:29
5

One way is by the thermal pad, but since I am using relatively high voltage (400V) it scares me that a thin pad may be punctured.

Don't be scared. The data sheets for the pad will give you the safe working voltage and, given that plastics are generally reliable for voltages up to about 10 kV per mm you should have no problem finding a suitable thermal pad. I've used thermal pads for 900 volts but, do check the data sheets to see what it says.

Andy aka
  • 434,556
  • 28
  • 351
  • 777
5

Silicone pads are probably fine for this application. You can buy larger sheet with glue surface and apply it over the whole heatsink area, and then punch through holes only for the screws. This reduces the chances of a small pad coming loose, and is easier than cutting out custom sized pads for TO-247 if you can't find them pre-cut.

(Image from this question)

If you want something with better heat conduction and higher breakdown voltage, try mica. It requires the use of a heat transfer grease on both sides of the pad, so not as convenient to install as the silicone pads. The TO-247 size is not commonly available, but TO 218 pads are slightly larger so should work fine for the purpose.

since I am using relatively high voltage (400V) it scares me that a thin pad may be punctured.

Look into solutions that avoid hazards if there is a short-circuit involving the high voltage. For example grounding the heat sink can make the difference between electric shock and just blowing the fuse.

jpa
  • 6,804
  • 17
  • 32
  • Thank you, grounding advice goes straight to manual. Looking at mica now. Big pad - great idea, but i am probably going to have headaches accurately locating the mosfets. The thing is i need to place 18 of them and then put on the PCB. pretty hard without having a grove for each MOSFET – TQQQ Mar 30 '23 at 17:41
  • Thk: 0.05mm wow. And yet 0.4K/W... At 50W it's 20 degrees – TQQQ Mar 30 '23 at 17:42
4

Well, if expense is no barrier you could use very thin diamond sheet. You won't get better thermal conductivity than that.

Dirk Bruere
  • 13,425
  • 9
  • 53
  • 111
  • 3
    Uhhmmm... thank you for the idea, will not work for this project :) The whole BOM is less than this wafer :D – TQQQ Mar 29 '23 at 15:24
  • @Dirk, have you used thin diamond sheet like that? What methods can cut it into a desired shape? – Nick Alexeev Mar 30 '23 at 23:25
  • @NickAlexeev I have not used such sheet but given that they would be very thin for this purpose you could probably snap them quite easily. Diamond is brittle – Dirk Bruere Mar 31 '23 at 16:05