# What tools do you use for thermal analysis?

I am using four 12 Ohm 0603 resistors in series to introduce thermal energy into a system (dissipating ~2W with 10V across the series string). The resistors are from the Vishay RCP a high power family, and are spec’d max 1.5W @ 25 deg (but dissipating 0.5W each).

My concern is that the resistors are getting hot enough in free air to melt the high temp solder I used (290 deg C melting point). The resistors will be potted into a thin metal tube using Stycast thermal encapsulant in the finished product, so there will be better dissipation - but how much better?

Given that I do NOT have a thermal background, what is the best / easiest way to understand the thermal conduction paths, expected temperatures at the nodes in the (thermal) network, what features of the system are important to pay attention to, and what can be ignored, etc.

Thanks,
Dale

Perhaps a little more information may help. Why are you introducing this extra heat? What are you hoping to understand, or measure?

In my experience, thermal simulation can be a tricky thing. The output is only as good as the inputs to the simulation, which includes all the geometry and thermal properties of all the materials involved, and accounts for conduction, convection, and radiation of heat.

A bunch of years ago I was trying to simulate the heat transfer from some high power LEDs through the filled vias of a PCB to the custom extruded aluminum mounting channel, to the overall outer stainless steel housing cooled by a flood of running water. Luckily it was very long in one axis, so could be simplified to a 2d open ended simulation. A steady state sim didn’t work, because convection of the air inside the housing turned out to be the largest contributor to heat transfer to the outside world. I had to do a time stepped simulation starting from t=0 to something like 40 minutes until the convection pattern stabilized before I felt I could trust the result. At the time I had to leave it to run overnight.

More recently I needed to know how much power I needed to dump inside a customers aluminum housing to keep the system from condensing moisture on the outside window. I put a bunch of 5w power resistors in the housing, and measured the temperature at different power settings with the system in my residential freezer. Was a lot easier and faster.

Gut instinct isn’t shocked that 0603s are melting their solder at 5/8W.

1 Like

Thanks @nickb for your input.

The resistors form the heating element in a “thermal dispersion mass flow switch”. This type of device uses the temperature difference between an un-heated temperature sensor and a heated temperature sensor both inserted into a fluid, typically in a pipe, to determine if the fluid is flowing or not by how much heat is carried away from the heated sensor by the fluid.

The original design (before my time) used two 25 Ohm 1206 resistors, which worked well for a number of years until the device was discontinued. The obsolete resistor was changed to a 1206 Vishay RCP (also before my time), but it’s not clear if the new resistor was ever actually used, or if production kept building with the old part from stock still on hand.

In attempt to miniaturize the design (before my time still), the design was changed to a single 0603 50 Ohm heater resistor, and then I became involved. The original designer had left, and the final product was found to have a lack of sensitivity using the 0603 heater resistor. I found the heat from the single 0603 resistor wasn’t getting carried away by the fluid, presumably due to having significantly less surface area, so I changed the design to four 12 Ohm 0603 resistors and the expected sensitivity was restored.

Then we started doing more testing, including full temperature range testing, which is when we realized we were melting the solder.

Dale

I worked on a rig witch did measure temperatures of up to 1100 Celsius. What they did was using a small micro spot welding machine I don’t know if that is possible if you don’t have legs on your part.

Maybe they could use a PT100. That is a totally different implementation but just putting it out there.

1 Like

One more thing for thermal analysis. Have you soked the board in IPA and run it and se the IPA dry? It’s a good way to see how the heat spreds if you don.t have a thermal camera.

2 Likes

Watch this, I think it holds up: https://www.youtube.com/watch?v=HikutbNOf0Y&t=7s
It was a talk I gave about thermal analysis.
For this problem, a pencil will show the problem.
Part power ratings don’t mean what you think they mean.
The ratings are still useful, but not in the simple sense that one number means that much power is okay.

2 Likes

As plenty of folks have mentioned this is a tricky question.

IC thermal dissipation is somewhat similar as 90-95% of the energy goes into the PCB. A great app note from TI (ignore the ad for their package) with some thermal vias is here: SLMA002 (the app note).

1 Like

Thanks @JohanHoltby. Intersting idea, but IPA? India Pale Ale? Isopropyl alcohol? ???

Haha… Isopropyl alcohol :). I saw it first on https://www.youtube.com/watch?v=t5fICjcaJ3E

It sounds like the circuit is a bolometer with a heat source. Thermal isolation of both the source and the sensor is essential. You don’t want a thermal path between elements other than thru the media being measured. I assume the mechanical design provides that.

You need to plot a curve of sensed temp change vs. flow rate for the media. Apparently this existed before and you need to match the heat and thermal conduction to the media or the curve will be different.

I’ve done something similar to measure low power rf. The rf heated a 50Ohm load whiich was thermaly close to a sense resistor, upsetting a balanced bridge.

Also, thermal ir cams with usb plugs into my android phone. Very cool. \$400 online.

Thermal analysis uses an analogy with electrical circuit analysis: Power is a current source, resistance is thermal resistance, temperature is voltage, specific heat is capacitance. Thermal resistance = rho * length / cross-sectional-area . Most references use G=1/(thermal resistance), because thermal analysis is taught in the mechanical engineering course “Heat Transfer”, and they like G not R.

A detailed problem like this could be approached with a program that uses a 3D meshed geometry of the system.

Another approach is to eyeball it, struggle through a few thermal resistance calculations, and then solve a few voltage divider equations. It might seem like black magic at first, but once you do it a few times, you can see the pattern. Then you realize that the simple approximations in books and app notes are not so bad, for example the great TI app note or the fun book “Hot Air Rises and Heat Sinks” https://www.amazon.com/Hot-Air-Rises-Heat-Sinks/dp/0791800741 .

1 Like