When does a MOSFET need a fan to cool?

[eem2am]

Hello,
I beg to know what is the maximum dissipation (Watts) of a TO220 or TO247 Mosfet attached to a heatsink of any size , where a fan is NOT needed?

In other words, what is the dissipation level in a mosfet, where any size increase of heatsink provides no further cooling improvement , and a fan becomes necessary?

I am speaking of a mosfet in a well ventilated enclosure with 30 degrees C ambient.

 

[DF96]

How long is a piece of string? Similar question, similar answer: it all depends on which piece of string and how much you stretch it.

Specify: ambient temperature, natural convection cooling rate, thermal resistances (chip-case, case-heatsink, heatsink-ambient) then do a heat flow calculation/simulation.

 

[boywonder]

Is there a rule-of-thumb for max temp? Clearly, more watts dissipated means larger heat sink, but if one where to measure the case temp, is there a reasonable target temp to stay below?

 

[eem2am]

For TO220 typical junction to case thermal resistance is 3 degrees per watt.

Usually, the golden rule (given to us by the chief engineer at a big smps company) was that the case should stay below 109C.

 

[KatieandDad]

For TO220 typical junction to case thermal resistance is 3 degrees per watt.

Usually, the golden rule (given to us by the chief engineer at a big smps company) was that the case should stay below 109C.

It depends on how reliable you want your project to be.

The cooler the better.

The maximum temperature of the silicon junction itself is about 150 Degrees C.

For good long reliable life I try to keep the case temperature below 80 degrees C, normally much lower, in the order of 50 degrees C.

You also have to consider the cooking effect on the other components - Electrolytics will rapidly dry out if they are overheated.

 

[AndrewT]

I would strive to keep junction temperature below 100degC. This Leaves 50Cdegrees of margin for peak temperatures etc.
Further, I would check for a variety of worst case operating conditions to determine that the circuit can survive these hopefully rare moments.

As for infinite heatsink. Just substitute Rth s-a = 0 in the equation to see what effect Rth j-c and Rth c-s have on the operating temperatures of the device/s being examined.

 

[eem2am]

thanks, i see,

Rth c-s is the difficult one to know.....because its through that heat paste.

Supposing there was an infinite heatsink, and Rth j-c was 3 degC per Watt.......

...then i presume that would mean the maximum dissipation before a fan was needed would be.......23 Watts? (taking 30C ampbient).

 

[KatieandDad]

The IRFP244 happilly runs dissipating 40 Watts on a large heatsink.

 

[KatieandDad]

It depends on how reliable you want your project to be.

The cooler the better.

The maximum temperature of the silicon junction itself is about 150 Degrees C.

For good long reliable life I try to keep the case temperature below 80 degrees C, normally much lower, in the order of 50 degrees C.

You also have to consider the cooking effect on the other components - Electrolytics will rapidly dry out if they are overheated.

I spent many years as a TV repair engineer. Most failures were down to Electrolytics being placed close to inadequate heatsinks. The caps would fail due to cooking and then they would take out the regulator ICs as they went down.

If you keep everything cool, this wont happen. If you must run semiconductors at their limits then keep heat sensitive components away from the heat.

 

[KatieandDad]

Be very sparing with the heatsink compound. Its only there to fill in the microscopic holes in the pads and the metal. Too much and you are defeating the object. You only need the meerest smear. In the ideal world you dont need it. The metal of the semiconductor should sit immediately next to the mica and the mica should sit immediately next to the heatsink. The world isnt perfect so the paste just fills in the microscopic gaps.

 

[eem2am]

Thanks, I usually get the heat paste all over my clothes, instead of the heatsink.

Anyway, are there any advances to 40W without a fan for a MOSFET?

(30 deg C ambient)

 

[toprepairman]

If you can electrically insulate the heatsink from everything else then you can lose the mica insulator which will improve the thermal conductivity.

 

[AndrewT]

A typical 175degC 200W FET mounted on a large passive heatsink can be operated at far above 40W.
The sink certainly does not need to be infinite, nor fan cooled.
A FET with Tc = 100degC (175degC - 25degC) is rated at 50% of Pmax, i.e. 100W.
A 0.2C/W sink operating at Ta=20degC will run with delta T ~ 100W * 0.2 * DF = 50*DF.
For 50Cdegrees DF ~1.1
Expect that sink to run ~ 20+60 = 80degC.
Tc = 100degC
Ts = 80degC
Delta Tc-s = 20Cdegrees.
The required Rth c-s <= 100W / 20Cdegrees <= 5C/W.
Any sensible 200W device will give <5C/W for Rth c-s

Now substitute your own numbers, to find what limiting value of Rth c-s is required to ensure the 200W device can operate within spec when dissipating well over 40W.

 

[!]

If you have a fet dissipating in the neighborhood of 40W or beyond without a fan, you'll tend to have ambient temperature around the heatsink in excess of 30C.

 

[AndrewT]

That is not the way to use, nor interpret, the datasheets and app notes that give guidance of the use and cooling of FETs.

 

[!]

Well you can't just take the ambient temperature outside of the entire chassis. Assuming this PSU is enclosed you have to account for a temperature rise to some extent and 30C ambient would be a fairly small rise. I'm actually thinking the more conservative figure would be closer to 50C ambient.

For example we can assume that if this were an open frame PSU sitting out on a desk, the fet is going to be cooler running than if it were in a metal box with a few vent slits here and there.

 

[eem2am]

The applications engineer from powerint.com thinks 5W is the maximum you can do with a heatsinked TO220 FET..............


When is a fan needed to cool a FET? | Power Integrations

i am thinking about heatsinks as the five on page 4 of this....

http://www.powerint.com/sites/default/files/PDFFiles/rdr203.pdf

...a FET on that kind of heatsink could only do 5W max?

 

[!]

^ Can't tell how tall those heatsinks on the PDF are, too lazy to see if there's a BOM in the document to look up heatsink part #'s, but yes that's a roughly 5W capable heatsink... with a poor design putting electrolytic capacitors so nearby, I would not want 5W dissipation that close as it will probably fail prematurely as a result unless... you add a fan.

 

[eem2am]

Thanks

I have picked the Aavid Thermalloy 530002B00000G heatsink for TO220 which is Rth_sa = 2.6C/W.

My FET is STF11NM80

http://docs-europe.electrocomponents.com/webdocs/0dbf/0900766b80dbfe7e.pdf

This has Rth_jc = 3.6C/W

So therefore, with 30degC ambient, if i want to keep the FET junction below 100degC then i can dissipate 11.3W in the FET.

Would you agree with this?

 

[!]

I agree with those numbers, but I would design in more margin, not wanting 100C die or assuming only 30C ambient at that power level... this is where chassis design and passive-active cooling come into play and even then, the finished product may end up in a non-ideal environment (room temperature, stacking components, dust may accumulate, etc.) hurting thermal dissipation.