Sticking a thermocouple junction to a FET case so as to measure its temperature

[treez]

About air currents: they are present with and without the thermocouple bead touching the tab. The disturbance caused by the thermocouple wire may be small.

Thanks, yes we appreciate that. The thing is , if the tip of the thermocouple is actually touching the drain tab, and is glued to it, then static glue immediately surrounds the thermocouple end, and therefore any air currents are not absolutely local to the thermocouple end, and therefore there is less chance of getting a falsely too_cool measurement result. Do you agree?

 

[c_mitra]

Thanks, yes we appreciate that. The thing is , if the tip of the thermocouple is actually touching the drain tab, and is glued to it, then static glue immediately surrounds the thermocouple end, and therefore any air currents are not absolutely local to the thermocouple end, and therefore there is less chance of getting a falsely too_cool measurement result. Do you agree?

Air currents around the device will reduce the temperature (device temp >> ambient temp) and if you put a small dab of glue, it will act as an insulating layer and the temp of the tab below will be higher than (without the glue) the bare case.

When the temp diff is high, radiation is the major cause of loss of heat (most heat is lost by radiation) but when the temp diff is rather low (say ~100 or 200C) it is the conduction /convection that is the primary way the device loses heat.

If you just put a tape and cover the tab, the temp will rise (because heat loss will be reduced).

 

[treez]

Thanks, i am speaking about the attachment of the thermocouple to the metal drain tab. Is it ok to just rest it against the tab, or does it need a small spot of glue?

If it doesnt have a small spot of glue round it, then doesnt the local flow of air over the thermocouples surface cool the thermocouple down somewhat and result in the temperature measurement of the metal drain tab being cooler than it really is?

(As you know, The tab's major heat conduction path is into the copper of the PCB ,and then down the thermal vias to the heatsink)

 

[treez]

Hello,
We are going to stick our thermocouple end to the metal drain tab of our D2PAK FET with this glue…Electrolube TBS20S…
**broken link removed**
Its datasheet says its only for -40degC to 120degC, but I assume it wont all melt away if the FET drain tab goes up to 150degC?

TBS20S Glue Datasheet
https://www.farnell.com/datasheets/...48.380353493.1499278297-1679746183.1489787856

 

[c_mitra]

Solder melts around 200C; max temp you can use during soldering is around 300C. Max junction temp is (common Si devices) is around 150C.

Your glue must stay firm till 200C. They are plastic and they do not have a melting point and they will become soft around 150.

Thermoset glues (epoxy class) do not melt or get soft at higher temp. If your thermocouple wires do not have physical support, they will fall off around 150C.

 

[treez]

Thermoset glues (epoxy class) do not melt or get soft at higher temp.

Thanks, do you know of such a glue (Manufac P/N)? Every glue on farnell thats for 200degC plus and cures within 6 hours costs >£100.

- - - Updated - - -

If you place the thermo couple on the drain tab, you usually need high voltage isolation. Or a temperature meter that can accept superimposed switching voltage.

Thanks, will a normal battery thermocouple reader be OK?
Also, previously we've always put a bit of kapton tape over the metal of the drain tab and then fixed the thermocouple to that for isolation purposes.
Do you think the superimposed switching voltage (0 to 330V at many kHz) will make the thermocouple reading inaccurate?

 

[KlausST]

Hi,

Thanks, will a normal battery thermocouple reader be OK?

It will work, maybe the shown values are not that stable.

But my concern is safety. Those battery devices are far away from safety standards for 330V.

Klaus

 

[betwixt]

Treez, bear in mind that the origin point of the voltage you are measuring is tiny, even in comparison with an SMD transistor. It's small mass also means it very quickly adapts to the temperature around it. Unless you need high accuracy (< ~0.25C) all you have to do is ensure there is no 'escape route' for the heat that could cause a thermal gradient at the measurement point.

I'm not sure whether you are logging temperature over a long period or just taking spot measurements. If all you need to to is occasional measurements manually, use a poor heat conductor as a handle (a 4mm wood stick or plastic tube, something like a ball point pen body), glue the thermocouple to it's end with epoxy (not cyanoacrylate - it decomposes at high temperature and releases cyanide!) then glue a small heat conductive pad over the end of the junction. You can make the pad by cutting up a transistor heat pad, they are also designed for high voltage isolation.

You get an almost zero cost, fast reacting, accurate thermometer with an end that flexes slightly to ensure good thermal contact. As an added bonus, if you 'acquire' the bosses favorite pen to make it, you get the satisfaction they congratulate you for ingenuity while going nuts searching for it :laugh:

Brian.

 

[c_mitra]

Thanks, do you know of such a glue (Manufac P/N)? Every glue on farnell thats for 200degC plus and cures within 6 hours costs >£100.

Most common two component glues are epoxy based; they are readily available at the neighborhood shopping complex. If they take long time to set, just use a hot air gun (but hair dryers are cheaper and equally good).

Most semiconductor packaging is epoxy.

Also, previously we've always put a bit of kapton tape over the metal of the drain tab and then fixed the thermocouple to that for isolation purposes.

Anything that prevents or retards heat flow will cause heat buildup and increase in temperature. Usually you will reach a steady state when the heat produced equals to the heat lost to the environment. With a kapton tape, the temp reading will certainly be higher (than without it).

If you see too much noise on the thermocouple, please twist the two wires neatly and hopefully the problem will be solved.

 

[treez]

(not cyanoacrylate - it decomposes at high temperature and releases cyanide!)

Thanks, i was wondering at what temperature it decomposes for Loctite 382?
It only gives the "glass transistion temperature" in the datasheet...
Loctite 382..cyanoacrylate glue...
https://www.farnell.com/datasheets/...07.977596787.1499370874-1679746183.1489787856

Most common two component glues are epoxy based; they are readily available at the neighborhood shopping complex.

Thanks, though we need it to be thermally conductive.

As well as thermocouples, we would liek to stick a thermistor to the fet drain tab and see how hot that gets.....any thermal glue recomendations for this much appreciated.

 

[mtwieg]

I think someone already mentioned JBweld. I've used it for gluing heatsinks since it has decent thermal conductivity, though I don't think I've taken it past 100C.

 

[treez]

I think someone already mentioned JBweld. I've used it for gluing heatsinks since it has decent thermal conductivity, though I don't think I've taken it past 100C.

Thanks JB Weld doesnt say what is the thermal conductivity...
**broken link removed**

..Our FET is in the same (non ventilated) enclosure as 150W worth of LEDs, and so will likely go above 100degC. Our D2PAK FET is on FR4 PCB and has no thermal vias in its pad, so as you can imagine, it will be getting toasting hot. The FR4 PCB sits on a heatsink...but its the same heatsink as the LEDs, so its well hot.

- - - Updated - - -

Unless you need high accuracy (< ~0.25C) all you have to do is ensure there is no 'escape route' for the heat that could cause a thermal gradient at the measurement point.

Thanks, and I assume that to achieve this we should perhaps smother heat paste round the thermocouple end as it lies against the metal drain tab?…then we wont have this situation of the end of the thermocouple having different temperatures.

- - - Updated - - -

---------------------------------------------------------------------------------------------
Years ago there was a cheap thermal glue from Dow corning that was in a purple tube, and was a slight purple color, and it was hard up to 240degC, and set rock hard in 4 hours at 40degC.
It doesnt seem to be available now. Anybody heard of it?

 

[c_mitra]

Thanks, though we need it to be thermally conductive.

As well as thermocouples, we would liek to stick a thermistor to the fet drain tab and see how hot that gets.....any thermal glue recomendations for this much appreciated.

I wonder if any common glue is thermally conductive but it may be mixed with some metal oxides (like ZnO, or Al2O3, TiO2 etc) to increase the thermal conductivity. You can make out by looking- they appear white (not clear or transparent). Some glues use carbon power as fillers (they become thermally and electrically conducting) they look dark black. Some use metal powders (Silver or Aluminum) they look shiney (metal like).

A thin layer of glue will drop about 0.1 to 0.5C around 100C (so do not use a thick layer -more than 0.2-0.3mm- of glue).

Most common thermistors have thin glass insulation and they stick well with regular glue.

 

[betwixt]

Thanks, and I assume that to achieve this we should perhaps smother heat paste round the thermocouple end as it lies against the metal drain tab?…then we wont have this situation of the end of the thermocouple having different temperatures.

What I meant was to thermally insulate the back of the thermocouple to prevent heat loss but to fit a small (~4mm) pad of silicone heat pad on the front of it. The pads I mean are the ones used under T0220 and similar packages instead of the old mica washers. The pad material is designed for good heat conduction AND good electrical insulation, the two properties you are needing. The bonus is it is slightly compressible so you can apply pressure to it and it will mold itself to eliminate air and give good thermal contact. The settling time from first touch to stable temperature reading should be no more than about a second or so. Remember it is only the spot weld between wires that generates the voltage so it's mass is very, very tiny.

I think you need to be more specific about the accuracy you need though, it seems you are going to extreme measures to take a simple measurement. Also take note of the previous warnings about isolation if you bond directly to the tab of the transistor. You will be measuring microvolts of DC against potentially several hundreds of volts of background DC or AC so even if you trusted a sliver of glue as an insulator, bear in mind the whole thermometer circuit could be very 'live'.

Brian.

 

[mtwieg]

I wonder if any common glue is thermally conductive but it may be mixed with some metal oxides (like ZnO, or Al2O3, TiO2 etc) to increase the thermal conductivity.

JBweld does this, with steel particles.

 

[treez]

Also take note of the previous warnings about isolation if you bond directly to the tab of the transistor. You will be measuring microvolts of DC against potentially several hundreds of volts of background DC or AC so even if you trusted a sliver of glue as an insulator, bear in mind the whole thermometer circuit could be very 'live'.

Thanks yes, this is why we've also put a thermocouple on the plastic case of the D2PAK FET, and will also do a thermistor reading for comparison, as having microvolt measurements riding atop the voltage thats fluctuating over 100's of volts always sounds like a recipe for inaccuracies to creep in to the temperature measurement.

 

[treez]

I accidentally glued the thermocouple end directly on to the metal drain for one of the FETs, and the tecpel 316 thermocouple reader would not read it due to the noise...(the unit has to be powered off in order to read the thermocouple temperature or else the reading comes out ridiculously low).
Tecpel 316 Thermocouple reader
http://www.tecpel.com/Product-20021018174531.html

 

[c_mitra]

Please do not forget to twist the two wires nicely- 5-6 turns/inch would be ideal. At least you will reduce inductive pickups.

 

[c_mitra]

I accidentally glued the thermocouple end directly on to the metal drain for one of the FETs, and the tecpel 316 thermocouple reader would not read it due to the noise...(the unit has to be powered off in order to read the thermocouple temperature or else the reading comes out ridiculously low).

If the potential is +/-300V, and you are trying to measure a few uVs, you will need a CMRR of about 1e9 - not that easy.

 

[FvM]

the tecpel 316 thermocouple reader would not read it due to the noise

Not too surprising.

As previously mentioned:

Most people will prefer an isolated thermo couple mount.

- - - Updated - - -

I have a similar 4-channel thermocouple reader, the input channels are not isolated and not even differential, you can't use multiple non-isolated thermo couples with it simultaneously. The thermo couple might be even tied to instrument COM with one terminal, CMRR of the thermo couple input is not necessarily the problem (it can be though). But the instrument COM is tied to high frequent AC, injecting noise to the unshielded PCB through stray capacitance.