Thermal Fuse Effectiveness 1

[TurboXS]

I have been having a debate with another engineer over the effectiveness of a thermal fuse being used to monitor parts on a PCB.

Imagine the following;

A PCB say 100 x 100mm fitted with parts such as ICs, resistors and MOSFETS.

The assembled PCB is encapsulated in a potting material such as Sylgard and a thermal fuse rated at 150 deg C is potted in the centre of the potting but say 5mm away from the PCB. The box that the potted assembly is in is say 100 x 100 x 50mm.

Lets say there is a resistor 5mm directly below the thermal fuse.

The discussion we are having here is;

How hot would the resistor get before it causes the thermal fuse to fail?

The other engineer tells me that Sylgard is 12 times more conductive than air. However given that air is a poor conductor this doesn't seem like a good indicator of its effectiveness.

I've tried searching but it looks like I might have to pull out my old Heat Transfer books out.

thanks
Pete

 

[Skogsgurra]

Either do an accurate thermal modelling in a program made for this or do a full scale test. You have to put a thermocouple close to the hottest point and then start misusing the board so it gets hot. Record temperature and the thermal fuse. When fuse trips, you have your answer.

To get more results out of the test, use several different fuses and see when they trip. Then choose the one that best fits your needs.

Remember that there is a thermal lag. Heating slowly (low overvoltage/overcurrent) will produce a more accurate answer. Also remember to use a "thin" thermocouple so that the cooling effect of the thermocouple (the "stem" effect) doesn't give you false readings.

Gunnar Englund

http://www.gke.org

 

[itsmoked]

I agree with skogs.. Break out the fine wire T/Cs.

Keith Cress
Flamin Systems, Inc.-

http://www.flaminsystems.com

 

[OperaHouse]

What kind of thermal fuse are you talking about, A one time use or solid state resetable PTTC type? A PTTC would act on a combination of current and temperature. Again, specifications are like nailing jelly to a tree. Get out those thermocouples!

 

[TurboXS]

The thermal fuses are one time types.

I don't have access to a heat transfer program.

Yes I agree that a test would be the best way however I always like to back it up with some analysis so that we have a "feel" for the situation and at least have some understanding of the effects.

 

[IRstuff]

I don't see that you've given enough information to determine a solution. What's the heat load from rest of the board? What's the ambient temperature? Is it convectively cooled on the outside? Is the board equally potted on both sides?


TTFN

 

[ko99]

TurboXS,

With 5mm of epoxy separating them, the thermal fuse will be a very poor indicator of the resistor temperature unless you are very lucky.

The temperature difference due to conduction is:
power x thickness / conductivity x area

Air conductivity is 0.026 W/mK
Thermal epoxy is about 2-3 W/mK
(this is 100X air but still poor, aluminum is 180 W/mK)

So, for example, if ALL the heat went from the resistor to the fuse, a 1 watt resistor with a .1 x .4" area would be 100C hotter than the fuse! In reality, most of the heat from the resistor will search for better paths (through the circuit board, etc). In fact, the path from the resistor to the fuse is so poor that it's very difficult to predict how much heat will travel to the fuse. Hence all the suggestions to carefully measure it.

The poor tharmal path also means there's a high likelihood that the thermal lag mentioned by scogsgurra will make your fuse useless.

Before you build and test anything, I suggest you come up with a design where the fuse is better connected, thermally, to the device you want to track.



ko (

http://www.ecooling.biz)

 

[TurboXS]

Thanks Kevin.

A few "thought" experiments we have been playing with here also gave us similiar numbers.

regards
Peter