Dissipated power vs. temperature delta

[Eelcos]

We are working on development of a new electronic product which is mounted in a control panel. We have run some thermal simulations to predict the thermal behavior of the product in the panel. The panel is a fully closed cabinet, so there is no airflow and only natural convection occurs.
We have simulated the product by using the 3D model of the product enclosure and have added the PCB's as 'power dissipating plates' in it. So the total power dissipation at a PCB is averaged over the whole PCB area. In the simulation model some barriers (like wiring ducts) are added around the product. A fixed ambient temperature of 50 degree Celsius within the panel is used.

When simulating this setup I see that the PCB area at the top of the product has the maximum temperature (of course) and is 82 degree Celsius (delta T = 32 oC).
I now have reduced the PCB power consumption to 50% and I expected to get a maximum temperature of ~66 oC (delta T = 16 oC). But my simulation result was a maximum temperature of 72 oC (delta T = 22 oC) at the same location as the first simulation. This means that a 50% power dissipation reduction leads to only 33% of delta T reduction.

Because I am not very experienced in thermal studies, I hope someone can explain why the delta T is not proportional to the power dissipation when leaving all other parameters equal.

 

[IRstuff]

It may have something to do with you're fixing of the ambient temperature of the interior, which is not normally done.

Additionally, the natural convection coefficient changes as a function of temperature, since the coefficient is reflecting the buoyancy of the air, which is temperature dependent

TTFN
faq731-376

Need help writing a question or understanding a reply? forum1529

Of course I can. I can do anything. I can do absolutely anything. I'm an expert!

 

[chicopee]

A funny thing about heat transfer, my heat transfer professor said, over 40 years ago, that if your results were within 50% of the predicted value, you did "good", so a 6dC difference is pretty good.

 

[racookpe1978]

You said it was an enclosed cabinet, no air flow from inside to outside.

How are you simulating the heat loss of the total electronics heat out of the package?

Looking at the heat load and heat balance (final temperature) of just a single small part inside, when the air around that part is also heating up at the same time, MANDATES that you include in the "part" heat transfer model the loss of energy from the whole enclose.

 

[MintJulep]

Using an arbitrary surrounding temperature doesn't seem right, and it's likely not ok to neglect radiation.

 

[Eelcos]

Thank you for all your useful responses.
One complicating aspect is that we sell only the product itself and that our customers build it (together with other components and installation materials) into a control panel (cabinet).
We specify our product at a maximum ambient temperature of 50 degrees Celsius, as a direct ambient temperature (within the control panel). For that reason we have run our simulations without the control panel enclosure and with an environment with a fixed temperature of 50 degrees Celsius. In that setup I see the behavior described above.
The question is whether this simulation setup is representative for the real situation, but that's a separate question.