PCB Thermal Analysis

I have a very small board that I would like to model for thermal analysis. I am pretty confident that I can get by with a first order analysis since I only want to know if I have a prayer of a chance of dissipating the heat from this board, rather than a detailed analysis of the range of temperatures across the board. I was thinking about drawing a simple spice circuit with current to represent heat, resistors to represent thermal resistance and voltage to represent temperature.

Sounds pretty simple, right? But even after doing some web searching, I have no idea where to find details on the thermal conductivity of PCB material and the values to use for transmitting heat from the board to the air. I am surprised that this is not commonly used, but it appears that most IC data just *assumes* that your board is large enough that it is a good heatsink. I did not find a single reference with info on the PCB.

Am I not looking hard enough? Can anyone point me to the raw data on PCB thermal characteristics I need for this analysis?

Reply to
rickman
Loading thread data ...

Hello,

I think you are going about this the wrong way - my experience of thermal modelling is that you need fancy finite element software, good accurate models and good material info. The first is expensive and the second two very expensive.

If this is your tiny amplifier problem I suggest you design a test board that you can measure temperatures on and get it made. It will cost you about three days. Another trick that I have used is to measure a similar board, putting heat in by wiring directly to resistors on any handy footprints.

Our thermal models were never very close to the real thing but the measurements on mockups were.

Good Luck

Michael Kellett

formatting link

Reply to
MK

Yes, this is the small line driver board. I am surprised that you say you found thermal modeling to be inaccurate. I understand that finite analysis software can be expensive... I don't know of any CAD software that is cheap. But the model seems rather simple to me if you know the values to use. I have been searching some more and I think I have a paper that describes the thermal conductivity of a multilayer PCB pretty well. I don't see anything in it to describe thermal vias, but I do have some data on calculating the thermal conductivity using material parameters, so I guess I could attempt to model the vias myself.

I have considered building a trial board. I could even do that with board stock I have around. But I don't have a good way to model the thermal resistance of the chip to board connection. Of course, even this would be better than *no* information. I may give that a try.

Reply to
rickman

1 oz copper foil has a thermal resistance of about 70 K/W per square. The electrical resistance is about 500 uohms/square.

A 30-mil via, 1 oz plated, has a theta of about 50 K/W from top to bottom of an 0.062 thick board. Less if it's filled with solder.

The thermal resistance through a piece of 0.062 inch FR-4, from top to bottom, is roughly 50 K/W per square cm. How's that for mixed units?

The thermal dissipation of a surface to air is hard to nail down, since it depends on orientation and convection. VERY roughly 150 K/W per exposed square inch, half that for both sides.

The thermal resistance of a part soldered to a surface layer of copper foil depends on, among other things, the size of the part footprint. The smaller the part, the higher theta, without limit.

You can Spice thermal systems using the equivalents...

ELECTRICAL == THERMAL

1 ohm == 1 K/W

1 amp == 1 watt

1 volt == 1 degree K

1 farad == 1 gram of aluminum

1 second == 1 second

which works to 5% or so.

The nasty part about thermal modeling is that everything is distributed, unlike electrical circuits where everything is usually nicely lumped.

Somebody should write a practical book about this stuff.

Really, the easiest way to do thermal design is to build a mockup and test it.

John

Reply to
John Larkin

I agree that a mockup can be best if you want detailed results, but even a "mockup" can have sigificant errors based on the details of the chips and the chip attachments. A 20 pin TSSOP with a thermal pad will be very different from just plopping a resistor onto a PCB, both dissipating the same power. Also the details of the PCB can make a big difference. The thermal vias under the IC thermal pad make a significant difference too.

I just want a first order approximation to see if I am even in the ballpark. At first I thought the value of 150 K/W for removing heat from the board was a bit scary. My board is less than 4 sq in, but then I realized that your figure was in sq cm which is much better! Is the 150 K/W per sq cm value for a horizontal board or vertical? If horizontal, I would expect the bottom side to have a much poorer value rather than just dividing the number in two to get the total. If vertical, which my board is (I think) then the 75 K/W per sq cm sounds even better!

Thanks for the info. I think this is good enough for a rough estimate. I might even try working up that spice model with a 1 mm mesh, or maybe 1 cm to start.

Reply to
rickman

I *said* it was messy. Personally, I'd still do a mockup. Hack a piece of copperclad with an x-acto knife, add resistors. Use wire vias to approximate power-pad ic's. Put it in the actual box you plan to use, fire it up, measure. You said you wanted a "first order analysis."

Or buy a good FEA thermal analysis program for $30K and spend a couple of months learing how to use it. And then wonder if you can trust it.

I said square inches, and I meant it.

A resistor mesh works. There was one commercial product, Sauna, that did thermal analysis with an explicit resistor mesh model. But you could x-acto the thing in an hour and get much more realistic data.

We often mock up thermal systems. Metal, wood, cardboard, duct tape, actual fans/power supplies/heatsinks. This sort of stuff can be seriously counter-intuitive.

John

Reply to
John Larkin

The case the whole thing goes into can make a huge difference, as the air FLOW is what does the cooling. You can get some area figures from NXP SOT223 data sheets for a PCB first order, but a mockup that _includes_ a case, is a good idea. If that case might go into air-pocket, or pre-heated locations itself, include that too.

Generally, the PCB design is done to spread the heat, and avoids hot-spots, and then you have the whole PCB area to radiate.

One design we did used thick copper PCB, and thermal VIAS (paste filled), then expanded to an Al heat spreader on both sides, and that Al plate used the whole of one side of the plastic case to radiate to the outside air. Basically the poorer the thermal conduction, the larger the area you have to budget.

-jg

Reply to
Jim Granville

Opps, thanks for the correction. I was mixing up the 50 K/W per square cm of resistance *through* the board with the 150 K/W per exposed square inch between the board to the air. So now I am concerned again. Heck with just 4 sq inches of total board, that is nearly 20 degrees if both sides are used or nearly 40 degrees if just one is used. I may have to tell the customer that forced airflow will be required for a worst case condition (shorted outputs).

However, I did find a quad opamp in a TSSOP20 package that should do the job well. With a power pad which lowers the thermal resistance between chip and board *and* a thermal cutoff, this design should allow normal operation along with output short protection without forced airflow.

I would still like to know the basis of the 150 K/W number. Do you have a reference? You also didn't say if this was for horizontal or vertical orientation. So I don't know if I need to adjust up or down for the other case.

Yes, I am sure, but I am not designing a system, I have to live in one. I have not gotten detailed info on the case, but I do have some info. The bottom line is that this is looking like I have a pretty optimal design for the application and the thermal limitation will be in the system cooling.

Thanks again for the info.

Reply to
rickman

ElectronDepot website is not affiliated with any of the manufacturers or service providers discussed here. All logos and trade names are the property of their respective owners.