How many of you are familiar with air heatsinks?

[MrDeaf]

I wonder what the limitations are to designing unconventional designs?

 

First, let's look at some of the traditional designs.

 

Well established standards

• Solid Copper: Good transmitter of heat from metal to metal, but heavy and not as efficient at transmitting heat to air. Typically used as the base plate contact point for chips, mosfets and other high heat parts.
• Solid Aluminum: Lighter than copper and better than copper at transmitting heat to air, but not as good at metal to metal heat transfer. Typically used as fins attached to copper plates or pipes in various designs.
• Copper Heatpipe: Even better than solid copper and light weight. Some cheaper designs have limited ideal orientations, but better ones have wicks to fix this limitation.
• Axial fan: This is your typical case fan. It displaces large volumes of air, but tends to have poor static pressure and has a dead spot behind the hub. Some axial fans have designs that improve static pressure.
• Radial fan: (AKA blower fan) This type of fan displaces an average volume of air with high static pressure, but tend to be more cumbersome and heavier than Axial fans.

 

Some examples of Typical heatsink + Fan designs

• Al core + Al pin: This type of design has been mostly discontinued for high power chips, but can still be found on lower power chips with passive cooling. (examples: Southbridge heatsink, GPU RAM heatsink)
• Al core + Al straight fin: This type can be found on many medium power chips and mosfets. Can be found with passive cooling, as well as with active cooling. (examples: low end GPUs)
• Al core + Al flared fin: (AKA Flower) This is your typical stock intel heatsink for lower power chips. (examples: Intel stock heatsink for Pentium and i3)
• Cu core + Al flared fin: (AKA Flower, Orb) This is your stock Intel heatsink for medium power chips. (examples: Intel stock heatsink for i5)
• Cu heatpipe + Al stacked fins: (AKA Tower, low profile, top-flow) This is your typical aftermarket heatsink that boasts extremely good efficiency and can even be found, stock, on high end graphics cards. This is more of a brute force approach to cooling hot parts with maximal surface area, but does not consider things like fan deadspots. (examples: Noctua U14S)

 

 

---

Now, here are two designs, that I have wondered if they are possible.

 

Fin design

• Aluminum sandwiched Copper fin. Instead of your typical heavy Cu fin or less efficient Al fin, I have wondered if it was possible to have a fin that encompasses the better parts of both, without adding too much weight. Enter the Al sandwiched Cu fin. The Cu cored, Al sandwiched fin should, theoretically, transmit heat better than pure Al fins, while maintaining Al's good metal to air heat transfer capability. On top of that, due to the Cu core, it should have even better thermal conductivity between the Cu heatpipe/plate and fin.
• Possible Caveats. Chemically, Al and Cu don't mix well. There is such a thing as copper clad aluminum cables, but they are not bonded well. It is typically not practical to try and chemically bond copper with Al due to a bunch of chemical issues. The typical connection methods are mechanical, with tight tolerances or use solder/welding as an intermediate mechanical bond between copper and aluminum.

 

Heatsink design

• XO design. In this heatsink, there is a heatpipe that is shaped like an upside-down Ω. This starts life off as a typical tower type heatsink, but instead of stacked fins in a tower, the fins are laid out radially. As I have mentioned above, axial fans have a dead spot behind their hub, and have poor static pressure. On top of that, the amount of static pressure and air volume moved differs between the tips and hub of the fan. If the fins are laid out radially, such issues can be negated and the performance of the heatsink could be improved with optimal fin layout.
• Possible Caveats. The heatsink would have to be optimized for the selected fan and may not work well with a user selected fan. I am not entirely sure how much surface area this design offers. Manufacturing this would be more complicated than a typical "U" heatpipe tower, as each fin is different and requires 3 to 4 pieces on each layer, instead of a single, interlocking, fin for the full tower.

 

older

 

 

Normally, I would post such a thing onto a more PC oriented forum, but I only lurk on hardocp and the sorts and don't feel like signing up, just for a single post.

So, what do you think?

[The_GoDDfather]

Mind=Blown

 

Though i don't see how one can go wrong with just pushing the heat out with a huge fan....

[MrDeaf]

Mind=Blown

 

Though i don't see how one can go wrong with just pushing the heat out with a huge fan....

 

That's how things are done these days, but there are a multitude of limitations to that, however.

• Weight. There is only so much weight you can place on the motherboard. To counteract that, these days, water cooling with twin, triple and quadruple sized radiators is becoming the norm. A triple 140mm radiator with 420mm x 140mm x 30mm dimension is not going to fit on the motherboard, but it can be fitted to the case.
• Size. Some of these air heatsinks are pushing 150mm fans, but they run into space constraints around the motherboard and need to be designed around it. Even then, it is typically limited to 2x 150mm fans, or a mismatched pair of fans.
• Air flow. With larger sizes, axial fans improve in airflow and sound frequency, but degrade in static pressure. This means that usage of larger fans can give marginal improvements and diminishing returns.

 

A radial fan would work better for CPU heatsinks, but those are typically very cumbersome and a majority of home PC cases are not designed with those in mind. On servers, laptops and graphics cards, however, smaller radial fans can be commonly found, as they have more leeway with how their internal bits are oriented.

It would be somewhat interesting to see how large of a blower fan would be needed for a 120mm² heatsink would be

[madgiecool]

[NegativeWavesMoriarty]

One thing to consider is the amount of physical space available. Many heat sink designs are to maximize heat transfer while minimizing space. If you want to go crazy with a design there really shouldn't be much in the way of technical/physics limitations (well technically if the heat sink was very large the heat might not transfer all the way to the edges where it can dissipate fast enough) but for mass market heat sinks, space is a consideration.  

[CapnCappy]

While heat sinks are passive, optimal performance requires a fan. I often see two fans.

 

And still liquid cooling wins by a landslide. It's quieter and it does a better job. And now it's just as affordable.

[MrDeaf]

One thing to consider is the amount of physical space available. Many heat sink designs are to maximize heat transfer while minimizing space. If you want to go crazy with a design there really shouldn't be much in the way of technical/physics limitations (well technically if the heat sink was very large the heat might not transfer all the way to the edges where it can dissipate fast enough) but for mass market heat sinks, space is a consideration.  

 

technically, the XO design could be the same height as 140mm and 150mm fan towers, as there is no heatpipe protruding atop.

The practical limitation of using 150mm sized out diameter rings, is how thick/deep the heatsink is. 150mm O.D sized ring would obviously be paired with a 150mm fan, which is fairly tall already. For instance, dual tower heatsinks have interference issues with RAM and tall mosfet heatsinks and tend to use a 140mm + 120mm to improve clearance.

 

While heat sinks are passive, optimal performance requires a fan. I often see two fans.

 

And still liquid cooling wins by a landslide. It's quieter and it does a better job. And now it's just as affordable.

 

my design is for use with a fan. sorry if that wasn't clear.

But yes, AIW watercooling has come a long way and is very practical.

[AVR_Project]

Heat dissipates from metal to air if it is painted an oil-black color.

http://www.offroaders.com/tech/paint-it-black.htm

[m373x]

 I see a couple of design flaws with the XO design for a tower heatsink.

 

• First one is that it is circular. A square profile heatsink will have more area than a circular profile given that the side of the square is equal to the diameter of the circle. A square will allow you to fit more fins, resulting in bigger surface area and ultimately better heat transfer. You will need circular fins which will be harder and more expensive to produce. You will also need a round frame fan.
• Second one comes from the three copper pipes that are behind each other and cross in the center. This will create a lot of turbulence as the air passes around each one of them, impairing airflow and increasing noise.

[MrDeaf]

 I see a couple of design flaws with the XO design for a tower heatsink.

 

• First one is that it is circular. A square profile heatsink will have more area than a circular profile given that the side of the square is equal to the diameter of the circle. A square will allow you to fit more fins, resulting in bigger surface area and ultimately better heat transfer. You will need circular fins which will be harder and more expensive to produce. You will also need a round frame fan.
• Second one comes from the three copper pipes that are behind each other and cross in the center. This will create a lot of turbulence as the air passes around each one of them, impairing airflow and increasing noise.

 

theoretically, there is almost no air flow in the middle, because that part is hidden behind the hub, which is a dead spot, as far as airflow is concerned. The only thing the middle section should be doing is balancing out heat dissipation between the left and right sides of the same heatpipe.

 

As for the fin density, it does indeed have less surface area, but the layout is also more optimal for an axial fan, which should increase airflow over the fins. Higher airflow, but less SA could match lower air flow, but high SA. I know manufacturers like Thermalright, Noctua, be quiet! and Swiftech try and add extra features to their design, to try and improve axial fan efficiency, but they are all limited to things like shrouds, uneven fins and bending fins.

• Radiator shrouds reduce deadspots in axial fans have, but also add a lot of height
• FPI density in radiators is matched with fan static pressure performance
• e.g. Noctua split offset fins, Thermalright Ultra concave bent fins, be quiet! wave contour fins

I really wish someone I knew had a thermal conductivity program to try this design out, virtually.

 

and yeah, the complexity of the design really adds to the cost, while probably not adding much performance.

[m373x]

I didn't present it well enough. I meant that turbulence will be created around the pipes along their length in the spaces between each fin not just in the central section. It would be indeed very interesting to test this and adjust the design as tests are carried out but I don't know anyone with thermal conductivity program either.

[MrDeaf]

I didn't present it well enough. I meant that turbulence will be created around the pipes along their length in the spaces between each fin not just in the central section. It would be indeed very interesting to test this and adjust the design as tests are carried out but I don't know anyone with thermal conductivity program either.

 

oh, I see what you mean.

Isn't that, technically, a problem with all heatpipe heatsinks?

 

The heatpipes in the picture are overly large. Smaller heatpipes or a fat and flat heatpipes to reduce turbulence should work too.

 

[m373x]

Isn't that, technically, a problem with all heatpipe heatsinks?

 

It is but to a lesser degree as the pipes are uniformly spaced from the edges of the heatsink meeting a similar in speed stream of air. It is further negated if the heatsink is open which is the common case. The XO design forms a short length pipe in pipe pattern that is crossed from wall to wall with a pipe that meets airflow with various speeds.

 

The more I talk about this the more I get confused tbh. This really needs a simulation or an expert. 

[Carrier_Lexington]

Fin design

• Aluminum sandwiched Copper fin. Instead of your typical heavy Cu fin or less efficient Al fin, I have wondered if it was possible to have a fin that encompasses the better parts of both, without adding too much weight. Enter the Al sandwiched Cu fin. The Cu cored, Al sandwiched fin should, theoretically, transmit heat better than pure Al fins, while maintaining Al's good metal to air heat transfer capability. On top of that, due to the Cu core, it should have even better thermal conductivity between the Cu heatpipe/plate and fin.
• Possible Caveats. Chemically, Al and Cu don't mix well. There is such a thing as copper clad aluminum cables, but they are not bonded well. It is typically not practical to try and chemically bond copper with Al due to a bunch of chemical issues. The typical connection methods are mechanical, with tight tolerances or use solder/welding as an intermediate mechanical bond between copper and aluminum.

Just the curious, engineering-side of me: Why not use other materials, then?

Substitutes for Copper:

Silver has a better thermal conductivity than Copper, at 438 W/mK vs 401 W/mK at 273.2K.

Diamonds have a huge thermal conductivity, at 1000 W/mK for an impure diamond. Natural and isotopically-enriched diamonds can far exceed this, with the latter having values in the 3,000-4,000 W/mK range. (all at 273-293 Kelvin)

Graphene exceeds diamonds, with 5300 W/mK at 293 Kelvin.

TPRC Pure Aluminum, if supercooled, has figures going into the 20,000 W/mK range (of course, this requires the effort of supercooling, at which point, your battery would die just like Mark Whatney's did in The Martian)

 

Aluminum Alternatives:

Thermoplastics are more spatially efficient, although less thermally conductive, and would be more stable in this kind of bond.

 

[MrDeaf]

 

It is but to a lesser degree as the pipes are uniformly spaced from the edges of the heatsink meeting a similar in speed stream of air. It is further negated if the heatsink is open which is the common case. The XO design forms a short length pipe in pipe pattern that is crossed from wall to wall with a pipe that meets airflow with various speeds.

 

The more I talk about this the more I get confused tbh. This really needs a simulation or an expert. 

 

feel free to disseminate to any other forums.

 

 

 

Just the curious, engineering-side of me: Why not use other materials, then?

Substitutes for Copper:

Silver has a better thermal conductivity than Copper, at 438 W/mK vs 401 W/mK at 273.2K.

Diamonds have a huge thermal conductivity, at 1000 W/mK for an impure diamond. Natural and isotopically-enriched diamonds can far exceed this, with the latter having values in the 3,000-4,000 W/mK range. (all at 273-293 Kelvin)

Graphene exceeds diamonds, with 5300 W/mK at 293 Kelvin.

TPRC Pure Aluminum, if supercooled, has figures going into the 20,000 W/mK range (of course, this requires the effort of supercooling, at which point, your battery would die just like Mark Whatney's did in The Martian)

 

Aluminum Alternatives:

Thermoplastics are more spatially efficient, although less thermally conductive, and would be more stable in this kind of bond.

 

 

Good question...

 

Silver is very expensive compared to Copper, although it is an excellent thermal conductor. I still foresee there being bonding issues between silver and aluminum, since it's the aluminum that's hard to work with, while silver is not so different from copper, chemically speaking.

Diamond is difficult to work with and I've seen major heatsink manufacturers trying to work with it, but I have yet to see a working product.

Graphene is even more experimental than diamond, last I heard

Thermoplastics sound interesting, but I have zero knowledge of them so I can't make any form of judgement on them.