PDA

View Full Version : Heat conduction, how is it measured?

10-25-2001, 11:08 PM
I am doing a project in which I need to make my own heatsink. I wanted to know how the heatconduction of metals (and other materials) is measured. What metal can conduct heat the best? I see that all of the "best" CPU sinks are made from copper.

scotth
10-25-2001, 11:32 PM
Well here is a link to heat conductivity of common materials.

http://www.glacierbay.com/Heatprop.htm

It is fairly easy to set up an experiment to compare to materials relative heat conduction. Get/make two identically shaped/sized rods of two different materials. Apply the same amount of heat (or cold) to one end of each rod. Add a thermometer to the other end of each rod. Then make a graph (temp vs time) for each. The one with the steepest graph has the higher thermal conductivity.

Copper is very good. There are probably better substances out there, but I am not aware of them. It is probably the best common (read fairly cheap) substance.

cynic
10-25-2001, 11:36 PM
Silver has the highest thermal (and electrical) conductivity of any metal. Pure silver at 70 deg. F has a conductivity of 430 W/m-K, compared with 400 W/m-K for copper.

LoverBoy
10-25-2001, 11:37 PM
A physics expert will probably come along and add more to my very rudimentary knowledge of the topic, but:

In general, materials that conduct electricity well conduct heat well, because electron movement is responsible for the conduction of both electricity and heat. This is why you see copper in a lot of heatsinks (it's an excellent conductor.)

From a bit of research, it appears thermal conductivity can be measured in Watts*cm*ºC (I don't know why it's not Watts*cm^3*ºC, but that's what the site says.) A table that I found of a few thermal conductivities is available here: http://www.reade.com/Particle_Briefings/thermal_con_metals.html. Higher values are better (more watts transferred).

Another table of conductivities I found here: http://www.amm.com/ref/conduct.HTM
This table indicates that silver is a better conductor than copper, but I'm guessing it would be cost-prohibitive for your project.

Would be interested to hear what the heatsink is for.

- Rob

bouv
10-26-2001, 01:35 AM
If you actually want to find, say, the temperature at one point on a surface while it is being heated at a certain temp at a certain point after so much time, you use a big, long, and utterly ass-fucking differential equation.
Ugh...It involves fourier series, boundry value problems, other stuff too, I'm sure. I failed that course first time around and am taking it again, maybe I'll fair better this time...

Crafter_Man
10-26-2001, 09:51 AM
I’ll attempt to answer your first question…

Measuring thermal conductivity is conceptually simple, but a real bitch to actually pull-off.

I’m glossing over a lot of details here, but it basically goes like this: you fabricate two plates made from the “unknown” material. Typical dimensions would be 4" x 4" x ¼" for each plate. You sandwich a thin, flexible, Kapton heater between the two plates. On the outside of each plate is another plate that is kept at a constant temperature. (This is done by pumping fluid from a constant-temperature bath through holes in the plate.) The heater is hooked up to a DC power supply, and thermocouples are attached in a differential configuration. You take two measurements: 1) Heater power, 2) temperature difference. You combine this with the plate thickness (1/4”), and you calculate the thermal conductivity.

Again, I have severely oversimplified the operation. I didn’t even bring up the issues of edge losses, control of the constant temperature bath, heater power measurements, differential temperature measurement(s), stability, etc.

For more info, get the actual specification from the American Society of Testing and Materials (ASTM).

Now if you want to write a report on the subject, go get the ASTM spec and knock yourself out. However, if you want to measure the thermal conductivity of a material, you would probably not have the time, money, and resources to actually build this instrument. (Am I correct in assuming this?) So there are other ways of doing it. (Albeit they do not conform to an ASTM standard, but they’re a lot cheaper to do.) One way is to use a “Thin Film Heat Flux” sensor, which integrates a thermopile and discrete thermocouple in one convenient package. They cost about \$130. Probably the cheapest way to do it is simply with an ice bath and temperature sensor: Stick one end of the material in an ice bath, and measure the temperature of the other end. There are lots of obvious problems with this approach (changes in ambient temperature, etc.), but it will give you a 1st order approximation.

douglips
10-26-2001, 10:18 AM
Originally posted by LoverBoy
In general, materials that conduct electricity well conduct heat well, because electron movement is responsible for the conduction of both electricity and heat. This is why you see copper in a lot of heatsinks (it's an excellent conductor.)

This is true of metals - the conduction electrons are in general responsible for the conduction of heat. These electrons move very quickly, around 1% of the speed of light. The hotter or more impure the metal, the more the electrons are deflected in their journey, and the slower heat is conducted. So, pure metals are much better conductors than alloys.

In an electrically non-conductive material, heat is entirely conducted by sound waves, so the heat is conducted at the much lower speed of sound in that material. Nevertheless, an electrically non-conductive material with a very strong crystal structure can be very efficient at transferring heat, as all the vibration that goes in one end comes out the other. For example, pure diamond is the most thermally conductive solid at common temperatures.

10-26-2001, 01:24 PM
Make certain that you account for surface effects. A small layer of oxide, a bit of dirt or grease, or a small bit of foreign material can really change the reading in your thermal flux measurement.

10-26-2001, 07:56 PM
Does anyone know where I can find a more complete list? I silver really the best?

Manlob
10-26-2001, 10:18 PM
At room temperature silver is the best conducting metal. Diamond and highly oriented graphite are better thermal conductors than any metal (at room temperature), but not very good electrical conductors.

Electrical conductivity is easier to measure than thermal conductivity. There is a simple correlation between electrical and thermal conductivity of metals, called the Wiedemann-Franz-Lorenz law. So you can measure electrical conductivity and use this law as an approximation. However this is not a direct measurement of thermal conductivity, and does not apply to non-metals like diamond.

To measure thermal conductivity directly you need to know a heat flux and a temperature gradient. One common way to do this is the Kolrausch method, where a rod of the material is thermally insulated except at the ends where it is cooled. The material is heated by an electrical current (of known power) and allowed to reach a steady state. Temperature is monitored with thermocouples at three locations. Knowing the power and three temperatures the thermal conductivity can be found.

The method described by scotth is actually a measure of thermal diffusivity, not conductivity, although they are related. To measure conductivity this way requires knowledge of the material's specific heat and density. Laser flash diffusivity is a common method of this type.

Copper has a specific heat quite a bit larger than silver, and depending on the situation copper could make a better heat sink. Although for something at steady state, like a CPU heat sink, conductivity would matter more than specific heat. For an air-cooled CPU heat sink it is possible that the transfer of heat from the sink to the air is more of a factor than the conductivity.