Rough Estimate for temperature rise in an electronics box 2

[Rvanpelt]

I am looking to find a way to estimate what the temperature would be in an electronics box without getting into the complicated heat transfer equations. Hoffman provides a simplistic method here:

http://www.oksolar.com/pdf/components/enclosures_heat_dissipation-ChAp36_37.pdf

This seems to apply mostly to their product and might be overly simplistic. I was looking for some method similar to this but one I feel comfortable applying to something beyond sizing a Hoffman box.

 

[nbucska]

simplest:
Calculate or measure the power dissipated by the circuit.
Put a power resistor into the box and with a variable
power supply dissipate the identical amount of power.
After a few hours measure the temperature.

Plesae read FAQ240-1032
My WEB: <

http://geocities.com/nbucska/>

 

[Rvanpelt]

If I knew what the power dissipation from the circuitry was and the dimensions of the box, could I estimate what the temperature differential would be between inside the box and outside assuming it has been generating heat over a long time, i.e. steady state. The link I gave does this but I am not sure if I am comfortable with simly taking that method as accurate.

That method takes the power dissipation and divides it by the available surface area of the box and then uses a curve to determine the associated temperature rise over the outide ambient temperature. The curve being used most likely takes into account the material of the box (they specify it applies to painted steel), free convection and radiation from the outside surface, and any internal effects such as convection and radiation to the internal surface.

In order to generate a curve such as that, there is either collected data or a function driving it so I was wondering if their is a function or set of functions available to make estimates like that.

 

[IRstuff]

The graph corresponds to the "complicated heat transfer equations" for a heat transfer coefficient of 4.5 W/m²*ºC, which is the typical value used for still-air convection. You can verify this by dividing the power density by the temperature rise given in the graph.

A forced convection condition would have a lower temperature rise. A more restricted installation may have a higher temperature rise.

Note that this tells you the temperature ONLY at the surface of the box. The interior will be hotter than the values given.

TTFN

 

[itsmoked]

Sorry but nbucska has the best idea.

The boxes surface area has just about ZERO to do with the heat transfer. The heat generated in the box goes straight to the top of the box so starting down a few inches from the top of the box you can consider the rest of the box to be wrapped with an infinite layer of insulation. The hotter the top strata in the box the lower down the sides the rest of the box participates in heat exchange. Unfortunately this goes hand in hand with greater temperature at the top of the box.

We have gone to great expensive and some hassle to get more of the box to participate.

The least expensive is to place weather stripping along the rear sides where the inside panel edges are. This is to create a plenum using the panel. You then punch a hole in the bottom of the panel in the center. You mount a fan in this hole. The fan then forces the bottom box air into the plenum which sets up air flow that will now include all the box surfaces. This increases the the power capacity of the box by about 6X. Downside is you need to monitor the temp in the box or the fan rotation to catch a failed fan.

The passive method we use is to cast a massive vertical finned heatsink the size of the back face of the box. We then cut the back out of the box and replace it with the heat sink... Passive. Effective. Heavy. Not cheap.

Keith Cress
Flamin Systems, Inc.-

http://www.flaminsystems.com

 

[Warpspeed]

Keith is spot on here (as usual).

The reality of a fully sealed box is all the internal heat usually just rises to the top and stagnates. A small internal fan to just stir the air gently may assist, or it could just as easily be a complete waste of time.

If possible, any major heat dissipating items should have a direct thermal path to the internal wall of the box, particularly if the box is metal. This can be improved even further by fitting an external heatsink as previously described.

I doubt very much if the whole thing can be worked out theoretically from first principles, but the simple test proposed by nbucska is an excellent way to get an initial feel for the problem.

 

[BrianG]

Beware modelling internal temperature rise with resistors! Some years ago I designed a visual display terminal which had internal fins on a cast aluminium case plus a small "stirrer" fan. For convenience we "modelled" heat input to the case with some under-run filament lamps and a variac.

The result was completly wrong due to the fact that the lamps transferred lots of radiated heat to the case, resulting in a much lower temperature than when the real electronics were subsequently fitted.
We had to cut holes in the case to let hot air out at the top!

 

[LionelHutz]

How conservative do you want to be? We've used 0.6W/sq-ft = 1 degree C rise fairly reliably as a conservative calculating factor. This is for enclosures with air movement caused by at least 2 - 4" internal fans.

If you want to size the enclosure very tightly then you really do need to do some testing.

 

[Rvanpelt]

Yes, testing and using that data to come up with my own curves or functions seems to be the easiest way to arrive at a somewhat accurate prediction tool.

The value of 0.6 W/sq.ft. equates to 1 degree C rise would have been something you have developed from testing that particular set-up, right?

 

[ko99]

There are many ways to estimate the air temperature rise in a sealed box, but it's not clear how you plan to make design decisions with this number. It doesn't mean very much by itself. I would think what you really need to know is the temperature of the electronics (die or case temperatures).

Most components specify "j-a" in natural convection, but that assumes a single component and no obstructions to natural convection, radiation, and pcb conduction. Don't assume this j-a applies to your situation!

As was said, without a fan, hot air will stagnate at the top and only a fraction of the enclosure will be effective for heat transfer. Perhaps you can estimate this temperature but this doesn't tell you the air temperature near your critical components and just as importantly, it doesn't tell you what the natural convection airflow is near your components.

A resistor mock-up that approximates your component layout will give you a pretty good idea of what's going on, tho it's difficult to mimic radiation and pcb conduction with resistors.

I know you want a rough estimate but without real parts the only other practical way I know of is to make a CFD model (full disclosure: i do this for a living). A CFD model can mimic the actual components, pcb layers, enclosure, fans, vents, etc. You can run various scenarios to see the effects of fans, heat sinks, etc, on the die temperatures.







ko (

http://www.ecooling.biz)

 

[itsmoked]

Unless directly ducted it is kind of appalling how poorly internal fans actually "stir the air" in an enclosure. I don't exactly know why but it doesn't work very well at all. The fans energy input seems to pretty much cancel the stirred heat transfer effectiveness. Now mounting the fan at the bottom of a tube that spans one side of the box thereby forcing air to the top of one side of the box promotes pretty good circulation as it must then return to the bottom low pressure of the fan suction. This method still pretty much discards the back of the box behind the panel which was why we favored that method. But if you use a method to include the back of the box then make sure you don't flush-mount the box against a surface!

BrianG; Interesting bit about the radiant/oops study of yours. We often used the cheap trusty lamp resistor method to test designs. We were still amazed that the top of a 20 X 24 Hoffman would get hot with a 100Ws at the bottom with the sides appearing to be room temperature. When we turned our "panelback" fan on, in about 2 minutes, the entire box became uniformly a few degrees above ambient.

Keith Cress
Flamin Systems, Inc.-

http://www.flaminsystems.com