[highbias]

Hello, new to the forum! I have more of a thermal question, but couldn't find a proper place to put it, I apologize if this is the wrong place to pose this question.

I am building a DIY Class A amplifier, that will be water-cooled. The Aluminum block is 3"x"3 @ 12" length. I am not an engineer, so go easy on me. I've attached a pic to describe what I'm trying to achieve here. The block will be run to a radiator out of a 190Hp gas vehicle with two large fans providing the cooling, and of course a water pump. I am trying to figure out the °C/W of the block at various flow rates ie 1, 1.5, 2, etc in Gallons Per Minute. My hope is to dissipate up to 1300-1400 watts, and keep the Mosfets at roughly 75-80C junction temps. There will be 12 Mosfets attached to the block, 6 on top, 6 on the bottom. The materials used are whats on hand to build this cheaply; advice and design change if necessary would be helpful.

regards-
John

[ckelloug]

The first thing that I can say looking at your plan is that you most likely want to use a much thinner block so that the water channel runs as close as possible to the transistor attachment in the block. Secondly, using thermal compound like heatsink grease, self sticking thermal tape etc., where the transistors connect to the block is also essential.

Experimentation is the only way to be sure of the exact temperatures achieved in the system.

1400 W means 1400 J/s of energy. 4.18 joules is enough energy to heat 1 g of water 1 degree C. Dividing it out, this means that the amplifier output can heat 334g of water 1 degree C per second.

If we assume the flow rate through the block leaves the water there for 10 seconds and has about 0.5 meters of flow length, and that the diameter is .023m then the flow rate for the pipe is 1/3 of a gallon per minute or 21g/second.

Given these reasonable but arbitrary assumptions, the temperature rise for 1/3 gal/minute flow rate will be about 16 degrees coming out of the block. An automobile radiator is designed to dissipate thousands of watts albeit at a higher temperature so with a fan on it, it's pretty sure to keep up with the load.

While I haven't given you numbers for system performance answering your exact question, I am fairly certain that my back of the enveloping says that this design will work under any vaguely plausible conditions and that it is likely to be safe designing for .5 gpm flow rate. It may also not be necessary to use a thinner aluminum block but it definitely would be better as the thick block increase the heat transfer path length to the water which is bad.

Regards,

Cameron

[123CNC]

In addition to smaller cross section between source and transfer medium, you may want to consider a common header/manifold at top and bottom of heat exchanger rather than s/zig-zag single pass. If I have interpreted your drawing correctly. Another consideration is more and smaller holes to increase your surface area of transfer.

Are you planning on a fixed continuos flow rate or temperature control loop? Added complexity and cost, depends greatly though on materials on hand.

[martinw]

Dear highbias,

The two last posts give excellent advice, but I have a slightly different slant.

You have twelve MOSFETS, and you want to get rid of 1400 watts. That averages out at about 116 watts per device. I do not know which devices you are using, but the TO247 package typically has a junction to case resistance of about 0.4 degrees C per watt. Using a good heatsink paste,between the case of the TO 247 and the aluminium block will add (say) another 0.3 degrees C per watt. That makes a total junction temperature rise of

(0.4 +0.3) x 116= 81.2 degrees C.

That 81.2 degrees C figure, is above ambient which might be at least 40C on a warm day with the amplifier in an enclosure.

Now, this takes no account of what is going on in the aluminium block, or the heat transfer mechanism to the water loop.

By best guess is that what you are limited by is the ability of a TO 247 package to get the heat out onto the heatsink.

Incidentally, my other guesses are

(1) that the distance from the device to the water channel is of no real importance if the heat has to pass through aluminium. That stuff is a really good conductor of heat.

(2) if you go the water-cooled route, you want to make the water flow as turbulent as possible in order to increase the heat transfer coefficient at the aluminium/water boundary. Manifolds will probably lessen the HTC.

BTW, I could be entirely wrong.

Best wishes,

Martin

[highbias]

Thank you for the input. The fets will be mounted with kapton or mica insulators, with thermal compound. The Al block is unfinished, and is a 6061 alloy.

As for the amplifier, I was incorrect with the original dissipation figures. Each block will contain 12 fets for one channel. The fet used is FQA19N20C. Each block will be asked to dissipate 860 watts (the plans change by the second here, sry bout that). Each fet dissipates roughly 72 watts which is whithin reason being that were water-cooling.

The Amplifier is a Nelson Pass Aleph design. You can add as many output fets to it as you'd like withing reason. His Aleph 3 has 4 outputs per channel, and his Aleph 1.2 has 24. You can vary the voltages, bias per fet etc. Its a robust design. With 12 fets mounted, you can run it as an Aleph 2, which would be one channel. With 12 fets, you could also run a stereo Aleph 5 at various bias settings. I initially stated 1400watts, because in a particular config, thats how much i'd be dissipating. If I can do that within reason, going to lower dissipation figures wouldn't be an issue. Regardless of the amp config, I don't want to exceed 80 watts per fet. This particular fet has a Power Dissipation (TC = 25°C) of 180watts, Derate above 25°C 1.45 W/°C, and a max junciton temp of 150C. If I can dissipate above 80 and keep junction temps @ max 100C, I think it will survive. This amp has a nasty habit if fets are blown, it sends full rail voltage to the speakers, so DC protection is a must.

The main figure I'm after is Rth of the block itself. But to get that, I need to figure hole sizes/length etc. I particularly like the manifold idea, as the fets would be at a constant temp, versus the differentiation between fets with the s-zig-zag flow(fets are cooler at water input than the fets at water output). Lowering HTC is not good though. It seems that the consensus is more/smaller holes (3/8" or 1/2" or smaller?)would be much better, and closer to the fets themselves. Disrupting the laminar flow is a good idea, but all I have is a drill press to make the holes, I gotta figure that one out. The flow will be continuous, and will not be temp regulated. Of course, a manual resetting thermal switch on the block will be used if the water pump fails or in the event of a leak.

The are commercial cooling plates available, that tout 0.0033Rth figures, I'm trying to come as close as possible to that or better.
http://www.d6industries.com/HeatSinks.htm
Scroll down to Hydroblok_A16P, the model I was looking at is part#6P-12, but they want too much $$.

If i've missed any details, they will be provided.

see pic

Thanks for the help!!

regards
-john

[martinw]

Dear highbias,

There are a whole load of "consumer electronics" people who make stuff for cooling "over-clocked" PCs. They are probably pretty good too. You could easily modify their components.

Just a thought.

Best wishes,

Martin