Learning heatsinking from computer overclockers

[akb1212]

Class A amps requires a lot off heatsinking. And there seems to be a constant stream of questions regarding this here in this forum.
There have been quite a lot of good info presented in this forum already. To collect all that and make a presentable info page could be a good idea, but I don’t feel quite up to something like that.

What I would like here is to share some of the things I have learned about it, and give some of you some ideas and food for thought. After all this is a DIY forum, and DIY’ers like to do as much as they can themselves.

There are three terms to consider. First is forced or convection cooling . Secondly the heat carrying medium, normally air or water (or other suitable fluids, like oil). Then there is passive or active cooling.

The main goal for heatsinking in amps is to enable the output devices to operate by in some way lead away the excess energy dissipated in the amplification. To most of you there is only one way of doing this, and that is to attach the power devices to conventional heatsinks that is capable of dissipating the heat generated at the temperatures required.
This is what’s known at passive convection cooling. The advantages of this method are many. First of all it is by far the easiest method to implement. All that is normally needed is to attach the heat-source mechanically to a suitable heatsink. Most amps use this method. It’s only when the heat dissipation is large that other methods is starting to become an alternative. And since the needs in the amps we are building here is in this category I thought I’d give you some ideas and facts on the other alternatives.

After “normal” heatsinks the most used method is forced cooling, and to most of you this is air, water or other fluids that is lead in good temperature contact with the heat-source. To get the needed heat transfer there normally still needs to be some kind of heatsinks. These heatsinks are normally made in other ways than normal heatsinks, and sometimes also in different materials.

About the heat-carrying media, this is normally air as this is the most convenient an accessible media. But a fluid is also possible to use as heat carrying media. Water is the easiest accessible, and at the same time one of the most efficient one (it is one of the fluids with the highest heat carrying capabilities).
It is possible to have both convection and forced fluid cooling. The most used one is forced. This is because it is much more efficient (as is forced air cooling compared to convection). It is often that since you have had to go to the step and use water-cooling in the first place it is most likely to be because you have such high cooling needs that active cooling is needed.

But for us amp builders passive water-cooling is a really attractive alternative. The way a normal heatsink is working is (partly, there is also radiation heat) by heating the surrounding air and make the air expand and rise, making new and cool air rush in and get heated in the same way. The exact same thing is possible in water-cooling. The thing about active cooling, be it air or water, is that it almost always makes noise! That is a very unwanted thing in a high-end amp (that is after all what we are trying to make here, right?). That is what makes passive water-cooling so attractive. It is possible to make it completely silent, and at the same time capable of handling large amounts of heat.

Now what can be done is to use water or oil to move the heat in a more efficient way out to large area heatsinks. By using oil or other media there is a possibility to take advantage of other attractive properties that is achievable, like higher expansion to heat, thus making it easier to make the medium flow.

There are several advantages of using fluids for cooling, and one of the most attractive one is the capability to take away large amounts of heat radiated in a small area. This is possible since it is possible to make the heat resistance lower. There are also lots of more advantages that can be made use of, but I’ll come back to that later.

The third way to cool things is the active way. There are several ways of achieving this, but the point is to take away the heat generated and moving it actively to other parts. The most common ways are known as phase-change and peltier. Common for these method is that they use energy to move more energy. So by using this method you are actually using energy on the cooling process itself in addition to the energy you want to get rid of. This is something to have in mind when using one these methods.

For amp cooling both of these methods can be used. In terms of ease the Peltier is by far the best. It is a device you putt between the device and the heatsink, and when applying power to it it takes energy from one side and transports it to the other side together with the energy used in the process to move the heat. One quite large downside is that you have to have even larger heatsinks, as they have to dissipate both the energy dissipated in the heat-source and the heat from the peltier used in moving the heat. This heat can often be a great part of the total as the efficiency of the peltier is quite low! So if your goal is to find a way to use smaller heatsinks peltier is NOT the way to go. If you goal on the other hand is to be able to dissipate large amounts of heat dissipated on small areas there can be an advantage in using peltiers. If you for example are trying to build amps using only a few output devices and want to run them at very large dissipation levels (like using only 4 power devices in a AlephX with 300 W dissipation, meaning 75 W idle on each device) peltiers could make this possible!

But the far most used active cooling is the phase-change. This is the way most refrigerators work, and is a perfect way to cool semiconductors. I don’t know of any amps that use this cooling method, but I have heard of people on the forum that have thought about it. And to be honest I think it is a perfect way of doing it. I’m even considering it myself. If interested in more knowledge go to www.phase-change.com .

Now lots of my knowledge on cooling comes from studying other amps and normal literature about the subject. But one great source of information and knowledge on the subject is from Over-Clockers. There have for a long time been a development in the computer industry to make more and more power-consuming CPU’s. And this has lead to an enormous development of cooling methods. The thing about computers is that at the present moment the CPU’s are dissipating so large amounts of heat that they are way past what we use in each output devices we are using in our power amps.

As an example I can mention that my trusty old AMD Athlon Thunderbird puts out something like 70W on 0.5 cm2 ! That is way more than any of us (except for maybe Nelsson Pass in this thread: http://www.diyaudio.com/forums/showt...016#post263016 ) dissipate in each power device.
I think we have a lot to learn from these guys as they have no alternative as we have to parallel devices. You can’t just parallel two CPU’s in a computer to give a lower heat resistance :-) . And when they even increase these figures by cranking up the voltage to mage the CPU’s go faster, and at the same time try to get them as cold as possible to make them go even faster than that…..

Then there is passive phase change, or heat pipes as they are normally called. These work by making a closed loop of evaporating and condensing a selected media. This media is picked to have a boiling point in a suitable range for us, and is put in thermal contact with the heat-source and evaporated taking energy in that process. The steam from this process is lead along a tube and condensed at the other end of the pipe that is normally finned to be able to give of the heat given of by the condensing fluid. The fluid is then brought back to the heat-source by a wick, and evaporated again closing the circle.

This is also an appealing way to do it in an amp because it’s efficiency and low thermal resistance. It is also noiseless if the heatsink in the dissipating end is made with large enough heat fins. It is probably not that easy to make a DIY-version of this, but it will probably be doable. This method has lots of appealing qualities, and is one I’m going to look deeper in to.

One method is kind of on the side, and is possible to combine with several of the above-mentioned methods. I for one have made serious thoughts about using it both in amp cooling and OC’ing. What I’m talking about is submersion. The whole idea is to submerge the power device itself in to the cooling liquid. The whole point is to take away one of the heat resistances.

Witch brings me to another thing to think about here. The heat resistance is built up of several resistances in series. The whole point is to make the total resistance as small as possible, but there are only some of the resistances that are possible to do something about.
In submersion one of these resistances that is normally always there is taken away. In applications with very high dissipation levels this is of outmost importance, and it is therefore a very effective tool in increasing dissipation capabilities. But it demands a lot more than normal cooling in mechanical design.

When it comes to calculating the different types of cooling devices mentioned here there are huge differences in how this is done. Especially when it comes to phase-change methods there is no longer talk about heat resistance, but energy moving capability. Remember that the heat is carried away by a medium, and the specific heat capacity together with the mass flow determines the energy movement. There are of course resistances, but there are what could be looked on as voltage or current sources in the heat path.

Due to measge length limitations to be continue

[akb1212]

I have plans to write up more detailed info on this and more on each of the above mentioned methods here if there is interest for it. And if others feel like contributing, or have specific questions this would be a great way to try to condense some knowledge that would be great to have when thinkering about Class-A amp building and cooling them by DIY-means. I for one think that there is lots of possible ways of doing this that deserves to be looked in to. And there are certainly lots of good sources out there to supply this knowledge. The problem is to find it…. And since this is such a great forum I think this should be a suitable place to find it.

Anders

[tpenguin]

Peltiers are most certainly NOT a viable method of cooling an amplifier. They work "acceptably" at cooling a computer's CPU or GPU because there are only one or two heat sources, and even then, massive (275+ Watts for a single CPU peltier) power supplies and actively-cooled radiators (for water-cooled pelts) are necessary. Now, take an amplifier with 75 watts of idle dissipation on each device. That means for a peltier to maintain the same temperature on both sides, you need a 75-watt peltier element (for each device). This would obvioulsly not be desirable since the peltier would dissipate 75 watts of heat itself, and thus would result in nothing more than a net doubling of the idle heat dissipation. In order to cool a device to ambient, you would need a minimum of a 200-watt peltier for each device, and a power supply capable of running all of them.

Then you must have heatsinks capable of dissipating 275+ watts per device, while keeping the "hot side" of the peltier no more than 70&degC above ambient. Obviously, this won't be possible with any "conventional heatsinking," and you have only managed to increase your idle dissipation for those same four devices to 1100 watts and also requiring a minimum of 800 watts DC to run the peltiers, all to lower the temperature of your output fets to ambient, and provide no real benefit. The mosfets won't last longer, because the system would be far more prone to failure if you were using TEC's. 226-watt peliters are used by overclockers in water-cooled (with powerful fans and pumps) systems to keep their CPU below ambient temperature. This creates its own problems in terms of preventing condensation, and if the peltier fails (as they can easily do), it becomes a layer of thermal insulation between the CPU and the water block, which is not good.

I also don't like the idea of "completely passive" water-cooling. While it is technically possible by using convetion currents, it would not be able to transfer sufficient heat without active liquid circulation. Even then, an even larger passive radiator (no, not a speaker) would be required than an ordinary heatsink. And, while you aren't adding any noise, it also would provide worse thermal performance and much greater cost than a normal aluminum heatsink.

The only truly effective way to silently improve cooling without going to extreme measures like air conditioning your amp chassis would be to use copper heatsinks. Copper has approximately twice the thermal conductivity of Aluminum, and can dissipate far more heat without raising the temperature. It is also much more expensive and much harder to find in large heatsinks suitable for amplifiers. Silver is more thermally conductive than copper, but only slightly so, and will give only about half a degree C of a performance improvement. Water cooling does work, and if you can put your pump and radiator in another room with the tubing as short as possible, you can have good, silent results, but a huge fan-cooled radiator, about the size of that needed for a car engine, will still be necessary to dissipate about 800W of heat (a couple of class-A amps), and still don't expect much of an improvement in temperature.

[Petter]

Originally posted by tpenguin
The only truly effective way to silently improve cooling without going to extreme measures like air conditioning your amp chassis would be to use copper heatsinks. Copper has approximately twice the thermal conductivity of Aluminum, and can dissipate far more heat without raising the temperature. It is also much more expensive and much harder to find in large heatsinks suitable for amplifiers. Silver is more thermally conductive than copper, but only slightly so, and will give only about half a degree C of a performance improvement. Water cooling does work, and if you can put your pump and radiator in another room with the tubing as short as possible, you can have good, silent results, but a huge fan-cooled radiator, about the size of that needed for a car engine, will still be necessary to dissipate about 800W of heat (a couple of class-A amps), and still don't expect much of an improvement in temperature.

The problem is not likely to be conductivity, and Aluminium has a higher emissivity - thus it is in fact wrong to say that Copper can dissipate more heat than Aluminium - it might but it would require major design changes. The advantages of Copper is that you can make finer, denser products but that means forced air cooling. Also, the weight of Copper is so high that you will likely not want to build a Copper heatsink. I am surprised that heatsink manufacturers don't make black anodized Aluminium over Copper heatsinks for their finest forced air CPU coolers.

Submersion is very cool (no pun intended).

The cheapest way to go is tap water cooled with exhaust being dumped. I had some friends with an amp like that - turned ugly when some fool shut the water off

Petter

[tpenguin]

Well, speaking as an overclocker who has tried various different cooling methods, including water, copper heatsinks provide significantly better performance than aluminum, whether using forced-air cooling or semi-passive (fans on case only). Copper-Aluminum hybrid heatsinks are also used quite successfully, but all-copper is still by far the best. Submersion in water (especially tap water) is a bad idea. Tap water contains lots of dissolved minerals, which will deposit on whatever it touches in addition to being highly electrially conductive, and distilled/deionized water is extremely corrosive and will destroy any metal it touches very quickly. Submersion cooling should be done with a completely inert liquid, which water most certainly is not. Pure water would work well for a while, though.

Originally posted by Petter


The problem is not likely to be conductivity, and Aluminium has a higher emissivity - thus it is in fact wrong to say that Copper can dissipate more heat than Aluminium - it might but it would require major design changes. The advantages of Copper is that you can make finer, denser products but that means forced air cooling. Also, the weight of Copper is so high that you will likely not want to build a Copper heatsink. I am surprised that heatsink manufacturers don't make black anodized Aluminium over Copper heatsinks for their finest forced air CPU coolers.

Submersion is very cool (no pun intended).

The cheapest way to go is tap water cooled with exhaust being dumped. I had some friends with an amp like that - turned ugly when some fool shut the water off

Petter

[tpenguin]

Here is Zalman's passive water-cooling radiator. This is intended to be used for a CPU, and maybe a video card, totalling less than 150watts. It is obviously quite large, and two would be needed to cool one chanel of an AlephX, along with a good pump. It still would be worse than direct heatsink mounting because of the extra thermal bariers between the waterblocks and the water and the water and the radiator.

[jackinnj]

Originally posted by Petter

The cheapest way to go is tap water cooled with exhaust being dumped. I had some friends with an amp like that - turned ugly when some fool shut the water off

Petter

It's not just dissolved "solids" -- if you aren't changing the water and using "agents" you get "biologicals" in the water!

30 years ago I used to follow a couple of companies in the water-treatment industry -- Betz, Nalco, Dearborn, Petrolite -- there are aspects of cooling water treatment which are fairly easy to deal with, but there's a reason that engineers from the aforementioned companies are on site at the Exxon or Dow Chemical refineries.

I would used distilled water or ethylene glycol.

[bigparsnip]

The 'best' quiet PC cooling system that I have seen so far is also from zalman, and as you can see from the pic below, they are basicaly follwing the same idea that you will find on a lot of the high power amps produced here. The only other way which I have seen to give good, quiet cooling, was to burry a large metal tank in the ground (quite deep down where it is always cold), fill it with water and use this with a pump to cool waterblocks (Have a look here).

You can also read a bit more about the zalman case here.

[5th element]

Althought the Zalman flower coolers are really no good for amplifier cooling they do work well in the PC environment. I have one (the ALCU varient) and needless to say the power supply produces more noise then the CPU cooler. Which is not a lot of noise, you have to be actively listening for the PC to be aware of it which is great!. You should hear the other two PC's in the house BUZZZZZZ HUMMM WHURRR!, nitemare I tell you. Oh and copper is better then AL for a cooler on a CPU. We recently got a Thermal take volcano 7+ for one of the PC's, on its low fan speed setting it is about 10 degrees cooler then an aluminium sink.

[Jarno]

Hello all,

Just thought I add my 2 cents.
I am collecting parts to build 4 channels of Aleph 5 and since I already have the heatsinks and they are a bit on the skimpy side, I will be using forced air cooling.
In my last job I worked at a firm that designs video projectors, and since I am a mechanical engineer, a lot of the design is influenced by thermal aspects. This is because a beamer (or LCD projector) with a 200watt bulb puts out a lot of heat. Getting rid of this heat is no picknick because the housing is getting smaller and smaller, so you can't just put a big fan in there. Furthermore, even if you can use a bigger fan, you can't just have the fan run at it's intended (or max) voltage because it then would produce to much noise. So all of the fans are controlled from software (temperature) and are as big as possible because a large fan running at a lower voltage produces less noise than a small fan running at a higher voltage (and both fans producing the same flow).
In a LCD projector two types of forced air cooling are used:
"Spot" cooling using radial fans
"General" cooling using axial fans

For instance LCD's and polarisers are cooled using a high velocity airstream but low cfm, to cool a specific spot.
The rest of the inside apparatus is cooled using a higher cfm and low velocity airstream.

In my Aleph 5's I will use the same technique, the fets will receive "spot" cooling with radial fans and airguides. While the fins on the heatsink will be cooled using axial fans

Regards,

Jarno.