TO−247 stray capacitance to heatsink

[luky315]

Is there a way to estimate the stray capacitance from a (TO−247) MOSFET to the heatsink? My application is a very fast switching half bridge. For mechanical reasons the heatsink is connected to PE and the heatsink is shared between a lot of components (multiple output stages). The arrangement is not my choice, but unfortunately I cant't do anything about it.
Is it ok to simply estimate the drain to heatsink coupling capacity as a parallel-plate capacitor with the size of the electrically active tab and the thermal pad as dielectric? How much would the effective capacitance be bigger? Does anyone have some informations on this topic?

[Vovk_Z]

Is it ok to simply estimate the drain to heatsink coupling capacity as a parallel-plate capacitor with the size of the electrically active tab and the thermal pad as dielectric?

I guess yes. I don't see anything else which may change or add to that. Don't forget about the dielectric properties of ceramic insulation. It has some dielectric constant. I guess you may just straight measure this capacitance if you have appropriate capacitance meter.

[Siwastaja]

Is it ok to simply estimate the drain to heatsink coupling capacity as a parallel-plate capacitor with the size of the electrically active tab and the thermal pad as dielectric?

I don't think you can do much better than that without practical prototype measurement; I mean, that's what it physically is; the "electrically active tab" as you say, or drain, would effectively shield away any transistor internals, so it's really just a two-plate capacitor. That should do for first-order approximation.

Remember that the silpad squeezes when compressed so you would need to estimate real, compressed thickness. Probably easier to measure after being torqued to whatever you specify as mounting torque.

[Kleinstein]

If really deparate for low capacitance and still good heat transfer one would need a good thermal conductor and electric isolator that is a little thicker. Old time classic was some BeO ceramic, but this is problamtic with dust. An alternative would be a sythetic saphire disc - sounds expensive but surprisingly not as bad as one might think.

[Roehrenonkel]

Hi luky315,
 
consider using shielded thermal-pads.
It's a copper-foil between two layers of kapton(?)
with a pin/drain-wire.
I have no photo at hand right now, sorry.

Good luck

[mawyatt]

If you can ground the heatsink that will help with various components attached to the heatsink from "talking" to each other thru coupling to heatsink capacitance.

Any TO-247 should have a similar footprint so one could use an available TO-247 on the expected insulator of choice and measure capacitance to any large metallic surface, this should give a good measured estimate of actual capacitance in use.

Of course you can calculate the capacitance based upon simple parallel plate, if you know the Dielectric Constant of the insulator, and dimensions of components. Use a method which estimates fringe capacitance for a more accurate representation.

Best,

[luky315]

The grounded heatsink will create EMV problems, the question is only how severe they are. I know that it would be beneficial to connect the heatsink to some other potential and than with an impedance to PE, but this is not allowed here because the heatsink can be touched (it is part of the case)
The fringe effect is interesting, but I never saw it taken into account when calculating coupling capacitances.
Shielded thermal-pads could be interesting, at the moment we are using a very high thermal performance 0.2mm polyurethane
film for thermal and insulation reasons

[wraper]

You can "ground" heatsink to negative side of rectified mains voltage. Also you can connect it to earth too, many "industrial" SMPS do so, as aluminum enclosure is the main heatsink. The question is about adding enough of insulation in form of thermal interface.

The grounded heatsink will create EMV problems

Do you mean EMC? It's exactly opposite, leaving unconnected heatsink can be the issue in this regard.

[luky315]

Ungrounded is bad, but grounded can also be bad - the PE connection is never perfect

[wraper]

Ungrounded is bad, but grounded can also be bad - the PE connection is never perfect

Connecting something to PE does not help with EMC at all. What helps is containing interference within enclosure and Y capacitors that do not allow large conductors bounce at high frequency and radiate interference.

[Hydron]

If really deparate for low capacitance and still good heat transfer one would need a good thermal conductor and electric isolator that is a little thicker. Old time classic was some BeO ceramic, but this is problamtic with dust. An alternative would be a sythetic saphire disc - sounds expensive but surprisingly not as bad as one might think.

You can get aluminium nitride insulators from aliexpress for reasonable money - in theory they can be as good as BeO (i.e. in the ballpark of aluminium metal) without being made of death dust, though the aliexpress product may not be the finest. A decent thickness should give a significant reduction in capacitance, though do take into account the dielectric constant vs alternative pad materials.

[jbb]

An often (always?) impractical suggestion: I assume your power stage is hard switching (no shame there, a whole lot of products work just fine with it!); could you redesign it to use a soft switching topology? These tend to have lower dV/dt on internal nodes which can help.

[Vovk_Z]

You can get aluminium nitride insulators from aliexpress for reasonable money

They are sold not only on Aliexpres, thankfully. TS may find insulators of different thickness. I know up to 3 mm wide are sold widely. 3 mm can be quite ok.

[trobbins]

Assess where you want parasitic capacitance current to loop, to best avoid that current from looping where you don't want.  The shielded silpad option has been around for decades, so you may be able to dig through old conference papers that determined that stray capacitance.  I will check what I have, as I did some pspice Sims back then of emi spectrum and how to mitigate using shielded silpad.

PS.
I didn't measure the cap, but any dmm that measures capacitance should be able to give you a ballpark value.
https://dalmura.com.au/static/Int89a.pdf