Heat Management Issues

[The_Driver]

- A well designed FR4 board with filled vias is almost as good as most metal core boards.
- Metal core boards made of copper are only marginally better than aluminum core boards as losses through the thermal prepreg normally dominate due to the small cross-section of area for heat transfer
- Direct copper slug connections are the best

Have seen these tests of the new sinkpad PCBs? They clearly prove that a copper PCB, where the center (neutral) pad for the led has no isolation layer between it and the led (meaning that the center pad of the led is directly soldered to the PCB),makes all the difference in the world. The tetsing rig of those tests can be seen here. I know a German modder, who has tested if the material that the copper pcb is mounted to makes a big difference (i.e. of the heatsink is made up of copper or aluminium). His tests revealed that the brighness difference in percent was in the single digits. Unfortunately I don't have a link for this. He never made the results public.

The reason that these new copper boards are so much better is that normal aluminium PCBs always have an isolation layer between the LED and the aluminium. Removing this for the center pad is the important thing.

 

[SemiMan]

Have seen these tests of the new sinkpad PCBs? They clearly prove that a copper PCB, where the center (neutral) pad for the led has no isolation layer between it and the led (meaning that the center pad of the led is directly soldered to the PCB),makes all the difference in the world. The tetsing rig of those tests can be seen here. I know a German modder, who has tested if the material that the copper pcb is mounted to makes a big difference (i.e. of the heatsink is made up of copper or aluminium). His tests revealed that the brighness difference in percent was in the single digits. Unfortunately I don't have a link for this. He never made the results public.

The reason that these new copper boards are so much better is that normal aluminium PCBs always have an isolation layer between the LED and the aluminium. Removing this for the center pad is the important thing.

The SinkPAD is a unique product. Per my post above, it would be classified as a "direct" connection. It does not suffer the losses of a thermal prepreg. Where you have a thermal prepreg, using copper instead of aluminum for the base has limited overall impact.

Even for the SinkPAD, the difference between the copper and aluminum version may not be very large. Aluminum seems to be their standard material with copper by special order. I did a quick back of envelope calculation and figured the difference between the copper and aluminum versions may only be 0.3 C/W for an XML version.

Keep in mind that there is no "standard" metal core PCB. There are a wide range of dielectric materials and of varying thickness depending on the voltage isolation needed.

Semiman

 

[ATeadiesyngeag]

Heat Management I

Hello Thanks to your handy manual, I had a reasonably easy time replacing the Combo Drive in my brothers friends PB 1GHz 12 Aluminum DVI. Until, that is, I tried to reinstall the heat sink. No matter how much or little pressure or torque I applied, I couldnt get the 7.5mm screws with springs to tighten or even catch. I gave up and decided to post here only after one of the screws, propelled by its spring, shot across the shop, probably never to be found again. And you dont seem to sell them. In fact, looking at the remaining screw closely, I cant figure out how it could possibly work. It doesnt even appear to be threaded. It doesnt work if I try to screw it directly into the inserts in the logic board. Im not even sure the inserts in the logic board are threaded. The screw doesnt seem to be like the Southco 5T heat sink screws Ive dealt with in the past. Yikes Any advice would be greatly appreciated. Also, assuming Im not screwed and the screw problem can be solved, shouldnt I either replace the thermal pad for the processor or possibly add some thermal grease? Thanks. Regards, Mike .

 

[degarb]

"Copper's heat conductance is close to Silver, but reacts with copper"....http://www.engineeringtoolbox.com/thermal-conductivity-metals-d_858.html

Both Artic silver and Arctic Alumina are %100 electric insulators. So, it would follow that you could maybe make an "Arctic Copper" thermal epoxy that could bond with aluminum, without the corrosive effect of two dissimilar metals. Price should be around Artic Alumina, with heat conductance of Silver.


So, is it only the perceived menace of the galvanic corrosion of the copper and aluminum parts that would stop them from make this line of Thermal Epoxy? Or why not?

 

[DIWdiver]

Without electrical conductivity, there can be no galvanic corrosion.

Besides, the galvanic voltage between silver and aluminum (0.75V) is worse for corrosion than copper and aluminum (0.55V).

I suspect that the reason for using silver instead of copper has little to do with price of materials, and more to do with the price of creating the proper size and shape of microparticles required for high thermal conductivity of the finished product. Or perhaps neither of those is significant, and they use silver because it has a great reputation and high market value.

Any way you look a it, I wouldn't expect Arctic Copper to be notably cheaper than Arctic Silver.

 

[likevvii]

wow. I thought arctic silver was the stuff.... DIRRECT CONTACT IS BETTER!!! my whole life has been a lie! thank you so much for that tip.

 

[DIWdiver]

Beyond any doubt, that is true. However, achieving direct contact of the quality necessary to improve upon a very thin layer of Arctic Silver or other decent compound is rather difficult. Probably beyond the capability of most modders. So unless you want to dedicate the time and effort to develop the skills and tools necessary to achieve it, do like the vast majority of us and use compound.


One thing the OP didn't mention is that the degree to which a layer is a thermal barrier is proportional to the thickness of the layer. It's also inversely proportional to the thermal conductivity and the surface area. The thermal resistance of a layer is equal to t/(K*A) where 't' is the thickness, 'K' is the thermal conductivity, and 'A' is the surface area.


Let's compare the thermal resistance of several layers. I'm going to use metric, since thermal conductivities are most commonly available as W/m-K. That's Watts per meter-Kelvin. A Kelvin is the same as a degree Celcius. I'll also use the abbreviation um for a micro-meter.


Layer 1 - a 0.001" (25.4 um) layer of 63/37 tin/lead solder, K = 50
Layer 2 - a typical layer of anodizing; let's be generous and say 15 um (typical is less), K = 1 (typically 0.5 - 1.5)
Layer 3 - a 0.001" thick layer of Arctic Silver, K = 7 (from memory - please correct me if this is wrong)
Layer 4 - a 1/16" thick layer of aluminum (typical 'star'), K = 200
Layer 5 - a 0.001" layer of air, K = 0.03
Layer 6 - a 0.0001" layer of air, K = 0.03


Thermal conductivities are all in W/m-K, to keep everything fair.


I'll use the area of the thermal pad of an XM-L2, 2.782 x 4.7 mm, since this is a pretty common point of interest. This is 13 x 10-6 m^2.


Layer 1 - 0.039 K/W
Layer 2 - 1.2 K/W
Layer 3 - 0.28 K/W
Layer 4 - 8.1 K/W
Layer 5 - 65 K/W
Layer 6 - 6.5 K/W


These are interesting numbers, because you can see that a 0.001" air gap dominates the calculations completely. And if you reduce the air gap to 0.0001", which is hard to do, you still don't even come close to where the 0.001" Artic Silver numbers are. You have to get much better than 0.0001 to beat decent compound.


Note in particular that a single layer of anodizing is less important than the aluminum star! So you would be better off eliminating the star than eliminating the anodizing. Also, the solder represents a trivial thermal resistance compared to other things. So before you go looking for exotic solders, there are other things you can do with much better returns.


Also VERY important is to think about surface area (the area through which the heat has to pass). All these calculations are done with the surface area of the thermal pad on the XM-L2. Once you get through the star and into the head/body, the surface area is much larger, so the thermal resistances are much smaller.

 

[lucca brassi]

It's funny because all split hairs on a few micron layers with negligible thermal resistance, neglecting the only transport route and heat exchange with a cooling medium....

What helps block of copper, which can not to cool the super conductive layer of silver or copper paste and oversized or overloaded LED.....

 

[SemiMan]

Layer 4 - a 1/16" thick layer of aluminum (typical 'star'), K = 200

Layer 4 - 8.1 K/W

Note in particular that a single layer of anodizing is less important than the aluminum star! So you would be better off eliminating the star than eliminating the anodizing. Also, the solder represents a trivial thermal resistance compared to other things. So before you go looking for exotic solders, there are other things you can do with much better returns..

This part of your argument is significantly flawed, at least your calculations are for a typical star does not have 1 dimension of transmission hence the effective path is much wider than just the base of the star (unless you are using some sort of specially shaped slug).

Semiman

 

[DIWdiver]

This part of your argument is significantly flawed, at least your calculations are for a typical star does not have 1 dimension of transmission hence the effective path is much wider than just the base of the star (unless you are using some sort of specially shaped slug).

Semiman

Quite true. That was only for comparison purposes, and to begin a discussion of the truly important issues, based on actual numbers.

In reality, the star is probably more like 2 or 2.5 K/W. But it's beyond my ability to calculate with any kind of accuracy. I'm sure some of the other numbers will come under scrutiny as well. In fact, I hope they do. Maybe we can all learn something useful if we keep the conversation cordial and investigative. Like maybe someone knows how to calculate the thermal resistance of a star.

One could also point out that the area of an anodized surface, or the glue/compound layer, is probably at least an order of magnitude larger, making its thermal resistance proportionately smaller. We could talk about the layer of fiberglass/FR4 that's between the LED and the star on some of the cheaper ones. Someone could find the typical thickness and conductivity of a natural oxide layer on aluminum, so we could compare that. Maybe someone can discuss the average gap in a friction fit (that's probably enough for a separate thread all by itself).

As Lucca points out, the discussion can't be complete until we discuss the path all the way to the air (or water) and away from the light.

Maybe all of that doesn't belong here. I don't know. I really just wanted to get people thinking in terms of real numbers, and to show that something with low thermal conductivity isn't necessarily a bad thing.