Water Cooled Amp?

This old topic is closed. If you want to reopen this topic, contact a moderator using the "Report Post" button.
Has anyone tried to implement a computer water cooling system to a chip amp? I know CPUs can put out some serious heat, and people have said for very high end computer systems, to run water cooling for maximum cooling. I see when you buy a WC kit, it comes with copper pads for the CPU, could someone retro-fit these onto their amp ICs and potentially run cooler?
You're late to the party, but welcome nonetheless.
Remember that silly "Did you search?" button when you started this thread? You might have tried searching.
I started the first (as far as I know) water-cooled amplifier thread here years ago and there have been a fair number of threads since--roughly one every six months or so. Going with water wasn't my first choice, but I live in a nearly Stone Age part of the country and heat sinks simply don't exist. As the expression goes, "Necessity is a mother..."
I had monoblock class A amps that needed to dissipate 300W of heat per channel and later built a second pair. That's 1200W of heat, or about the same amount put out by a hair dryer. Yikes!
The good news is that the water cooling system could not only swallow that 1200W whole, but that in real terms its cooling capacity for home audio use is essentially infinite. The gooder news (yes, I know that ain't gud English) is that it was far cheaper than an equivalent number of passive heatsinks.
The bad news is that it's not portable, but I don't lose a lot of sleep over that.
My system isn't based on CPU coolers. I don't even know that there were any commercially available at the time. I just used things that I had on hand. Since then some pretty fancy looking CPU do-dads have come on the market and I have no doubt that they can do the job.

Well i know one thing everyone on here is concerned with these chip amps is heat. I'm working with the LM3886T chip, and those little suckers can really handle some power, it's just they over heat quickly. So they say 65W output, which is limited because of the heat needed to sink is just crazy. So I was figuring using those CPU coolers, (single CPU cooling pad per chip) and running water cooling. That way (hopefully) you can push them a little bit harder, just providing you can get the heat out. I guess my question is, does water cooling provide more heat loss then a standard finned heat sink? Even with a fan? I think that would mad cool to have an external heatsink with a single fan cooling my amp. :bigeyes: But mostly I want to see what peoples reactions are. If its better performance, or even worth it.
We use liquid cooling of various sorts where I work, both straight water and water and glycol mix. The cooling limitation is really only a function of your flow rate and the temperature of the water before it enters the heat exchanger that your device is mounted to. The last company I worked for used a large liquid cooling system to cool a high volume semiconductor test system. This machine used a 480vac 3 phase supply at 100 amps per phase, so you can guess that the various devices could produce some heat. These machines were used to test memory chips and microprocessors to verify that they meet spec.
The only thing I worry about with using a water based coolant is its proximity to electricity. If you can spring the money, using a nonconductive liquid like flourinert is a safer thing, but it's expensive. Alternatively, if you can keep it from getting contaminated, distilled water is not conductive, its only when it has impurities that it conducts electricity.
Water cooling would probably be a good solution, but there are several problems. The first is safety, but if you have your mains side in a separate enclosure, good grounding, and good fuses this shouldn't be a problem (still, be careful). The other problem is noise - water cooling requires a pump to run the entire time. You would probably get equally good results spending the same cash on a bigger heatsink and some big, slow (120mm or bigger) fans.
A few good places to get computer-based watercooling supplies that you can easily use for amp-based projects are:


Also, they have non-condictive fluids available in multiple colors, and even some de-algae solutions for if you use water.

They also have peltier+water cooling solutions, if you want to go overboard


quick, guess what one of my side jobs is?
Heat Pipe cpu coolers look way easier to apply, finding hard # for degree C/Watt, max Watt isn't easy though

these just move the heat to a fan blown fin array so you can achieve very high power density at the copper chill plate but are not "magic" when it comes to air flow, surface area and temp rise for ultimately dumping the heat into the air

also remember you can't "beat" the junction-case thermal resistance heat dissapation limit without acutally chilling the case below the 25 C "infinite" heat sink rating

and SOA current limiting inside the chipamps will limit available power with likely only modest gains for "infinite" heatsinking
jcx said:
Heat Pipe cpu coolers look way easier to apply, finding hard # for degree C/Watt, max Watt isn't easy though

You can find it experimentally -- I use a screw mounting Dale 50 watt resistor with silicone grease -- put enough current through it to draw 10 watts -- I attach a thermometer to the heat sink, measure ambient temperature and the temperature about 10 minutes later. Air flow should be unobstructed and still. Just measure the temperature gradiant and le voila.

Ive mentioned this article before -- the "Stanley Steamer" was a ham radio 2 kW amplifier described in QST in the 1960's.
jcx said:
Heat Pipe cpu coolers look way easier to apply...

Yeah I was thinking that. However I'm going stereo in one chassis, and I'm going to have 4 chips per channel (maybe more later). For some reason I see computer cooling devices as ALOT more efficient source of cooling than standard audio heat sinks. I'm going to be putting alot of power in a small case, I really dont have the room nor budget to put 8+ CPU heat sinks in the case. So I was thinking of using CPU water cooling pads, one for each IC, and making a custom radiator (possibly as the lid of the chassis).

danielwritesbac said:
...Also, LM3875T, no detectable difference in decibels, but less heat output. ;)

What do you mean by that? Whats the difference between the LM3875T and the LM3886T? They both are a single mono amps non insulated right? They look as if to have the same specs.
simply using more chips, paralleled (w/small current sharing Rs) on a conventional heat sink is likely to be cheaper and less hassle

paralleled outputs give bigger current headroom for loads with impedance dips, and the multiple packages spread out the heat over the heatsink, using it more effectively and puts the unavoidable package thermal resistances in parallel as well

as I mentioned you can't beat the surface area and air flow requirements in the end

just measuring thermal resistnace is fine, but heat pipes have a max transfer rate - when you've boiled all of the fluid off of the hot end the incremental thermal resistnce skyrockets, but I believe some of the cpu coolers must be over 200 Watt capacity to do the job, this number should be on the datasheet - requires lots of power to measure yourself
Well looking at the numbers I see that CPU finned sinks are VERY efficient. I dont know the stock specs on the average audio heat sink (according to the spread sheet above, for basically maximum output on the chip will need a sink that is 1.5 C'/W or better. Reading some specs on CPU coolers, they have ratings of 0.16 C'/W or better. And with what people say, that a low end water coolers is equal to a high end finned sink, leads me to believe that water cooling may not be a bad choice. Now fans and pumps is a worry, maybe I incorporate a small heat sink with a water cooling as a thermal backup. When you really start pushing the amp (high volumes) the pump then kicks in and adds a bit extra. I know this sounds uber complicated, and it is. But one part of DIY (atleast for me) isnt really making something that you want to have, but more the learning process. Adding bells and whistles. Building something no one else has. If I wanted the easy out, I would just go buy an amp.

But I'm still unsure exactly the route I'm going to take here, just doing some research looking for possibilities to try something different.
A good PC heatsink can dissipate in the region of 60W with a good fan speed...

Is it quiet? No

Is it hell to keep the noise generated by the fan out of the circuit ? yes

Generaly I'd do anything to avoid haveing a fan running in the background...this would include watercooling....

With watercooling the only limit is what size radiator you can connect and how much air you can push through it... There is also a tower based watercooler which drops droplets back into reservoir, cooling them in th process.... but over $200.

As far as I'm concerned it is not a matter of is it possible... is more... is it practical...?
MotoMan_Yz400 said:

and making a custom radiator (possibly as the lid of the chassis).

Not a good idea. All you would end up with is a ridiculously expensive hot plate.

There's a reason why passive heatsinks are vertical and not horizontal: hot air rises. You want the hot air to flow along the surface of whatever you are using to dissipate the heat. A flat horizontal metal plate is not going to dissipate heat very efficiently.

There are some very nice computer water cooling radiators you can get. Some of them even fit double or triple 120mm fans. Put some low-speed fans on there and your amp will be both reasonably quiet and cool enough.

If quiet fans are not quiet enough, you could add a temperature control for the fans, so they don't kick in until needed. Maybe even incorporate a passive cooling component into the loop so the radiator fans are not needed until you really crank that thing!
Hot plate? Actually that is a type of passive cooler, like this:

The aluminum top plate is up on spacers, at least 3/16" not touching the sides--there's an air gap.
And there are ordinary heatsinks hanging down from the aluminum top plate (+thermal paste between join).
And there are vents underneath the ordinary heatsinks.
The heatsinks don't touch the bottom of the enclosure.

Although the contact points between heatsinks and plate are critical, if airflow path is successful, hotplate style is a mighty effective passive cooler. Moving air surface = outside area of the plate.

For the "T" chips, this big aluminum plate can be lacquered (on top and sides, not bottom) and then the chips don't require micas.
I put together a water cooling system for my PC several years ago, all from dangerden parts. Of course it looked cool, and I was able to overclock my 2.8ghz p4 to 3.5ghz with barely any increase in cpu running temperature. In it's final configuration I was using a 15" by 5" by 1" radiator mounted externally to the back of the case.

I grew out of my water cooling phase in the end. It was great at first, but after 3 years of running 24/7 the maintenance only starts to increase. When 2 nights before a major final exam in college a plastic seam in my reservoir started to separate and leak, I decided the gains weren't worth it anymore and just switched to one of the big zalman heatsinks. Costs several times less than my water cooling system, I haven't noticed any loss in performance, and it's been running for 2 years now without being touched once.

Basically I'm just saying I don't think the long term maintenance in water cooling is worth it. It's likely that after several years when parts start becoming brittle, and you start getting the ugly mineral deposits everywhere, you'll have wished you just went with a good old heatsink from the start.

I also just want to add some info. As others have said, there are some non-water solutions or additives you can use, but none of these conduct heat as well as water. If you use water, you should use distilled as it won't have minerals as others said. However, the water will still leach minerals off of the insides of the metal in your cooling path anyway though, so it will still just be a matter of time before everything gets a coating of mineral deposits.

This old topic is closed. If you want to reopen this topic, contact a moderator using the "Report Post" button.