Overclock.net banner
1 - 9 of 9 Posts

· Registered
Joined
·
50 Posts
Discussion Starter · #1 ·
So my design team at uni is building an autonomous underwater vehicle and the autonomous part means our code is pretty taxing on our processor. Computer vision can cause the cpu to go to nearly 100% load. Needless to say, stock cooler just won't cut it (especially in a case with no airflow to exterior). There are two designs floating around, one with a large aluminum heatsink that we would manufacture. The other (my design
smile.gif
), would have heatpipes going straight from our cpu to the ends of our cylindrical case (pressure vessel) which are aluminum and touching the water. Also, pleeeeease don't answer watercooling with the outside water, OCN is where I learnt that that would be a terrible terrible idea.

I was wondering if you guys would know where to find heatpipes, how to work with them (just use a little pipe bender?), how to attach them to a cpu block (which we would probably make out of copper), suggestions for tim pads (heatpipe to endcap can't be tim paste it will be too messy when we take it apart) or any other options or suggestions that you guys would have.
 

· Premium Member
Joined
·
67,312 Posts
Quote:
Originally Posted by RandomPerson View Post

So my design team at uni is building an autonomous underwater vehicle and the autonomous part means our code is pretty taxing on our processor. Computer vision can cause the cpu to go to nearly 100% load. Needless to say, stock cooler just won't cut it (especially in a case with no airflow to exterior). There are two designs floating around, one with a large aluminum heatsink that we would manufacture. The other (my design
smile.gif
), would have heatpipes going straight from our cpu to the ends of our cylindrical case (pressure vessel) which are aluminum and touching the water. Also, pleeeeease don't answer watercooling with the outside water, OCN is where I learnt that that would be a terrible terrible idea.

I was wondering if you guys would know where to find heatpipes, how to work with them (just use a little pipe bender?), how to attach them to a cpu block (which we would probably make out of copper), suggestions for tim pads (heatpipe to endcap can't be tim paste it will be too messy when we take it apart) or any other options or suggestions that you guys would have.
Sounds like an awesome project!

You can get various heatpipes from manufacturers but they are geared toward large industrial orders. However, I know of one that sells their spare/sample heatpipes... I'll go look for the company.

Another thing you might want to consider is sending out emails to heatpipe companies explaining your a student and are interested in their technology. You might be able to get samples or sponsorship.

Here's a $600 kit of MANY parts: http://www.aavid.com/products/standard/heat-pipe-exploration-kit
Here's a individual heatpipe site: http://www.enertron-inc.com/heatpipe.asp
 

· Registered
Joined
·
725 Posts
In the case of the heatsink, where would you mount it? Pictures help, even if crude.

In the case of using the vessel itself, you would still need fins on the surface of the vessel to increase heat dissipation. Basically turning it into a rad.

Either of the aforementioned cases would either require exceptionally large surface areas or for the vessel to be constantly propelled so as to maintain flow over the surface.

To the point of watercooling using outside water, using either of the above solutions use the surrounding water. You're merely carrying the heat to a different point first before transfer. If it were me, I'd pump water in from outside, straight through the blocks, and then back outside. Autonomous to me means a large environment, like a pool or lake, where the heat transfer from system will be negligible relative to the volume of water contained therein. Using something small, like a fish tank, simply won't work for extended periods.

You'll need to put water blocks on EVERY component as well. Fans will not work in a closed environment like that unless you plan on conditioning the air contained in the vessel as well.

You're best bet may be mineral oil submersion, and then cooling the oil with one of the rad options with the outside water. This would affect buoyancy, however, so you'll need a double walled vessel to allow for the system in the centre vessel and air/water mix to adjust buoyancy in the outer vessel.

Hurray, I just over engineered it.
 

· Premium Member
Joined
·
67,312 Posts
Quote:
Originally Posted by Firehawk View Post

In the case of the heatsink, where would you mount it? Pictures help, even if crude.

In the case of using the vessel itself, you would still need fins on the surface of the vessel to increase heat dissipation. Basically turning it into a rad.

Either of the aforementioned cases would either require exceptionally large surface areas or for the vessel to be constantly propelled so as to maintain flow over the surface.

To the point of watercooling using outside water, using either of the above solutions use the surrounding water. You're merely carrying the heat to a different point first before transfer. If it were me, I'd pump water in from outside, straight through the blocks, and then back outside. Autonomous to me means a large environment, like a pool or lake, where the heat transfer from system will be negligible relative to the volume of water contained therein. Using something small, like a fish tank, simply won't work for extended periods.

You'll need to put water blocks on EVERY component as well. Fans will not work in a closed environment like that unless you plan on conditioning the air contained in the vessel as well.

You're best bet may be mineral oil submersion, and then cooling the oil with one of the rad options with the outside water. This would affect buoyancy, however, so you'll need a double walled vessel to allow for the system in the centre vessel and air/water mix to adjust buoyancy in the outer vessel.

Hurray, I just over engineered it.
I would avoid pumps or internal watercooling since:
1) pumps mean more weight/power and add another point of failure.
2) the required internal tubing would drive up complexity
3) the water is not purified so there might be crude or salt residue
 

· Registered
Joined
·
725 Posts
Quote:
Originally Posted by DuckieHo View Post

I would avoid pumps or internal watercooling since:
1) pumps mean more weight/power and add another point of failure.
2) the required internal tubing would drive up complexity
3) the water is not purified so there might be crude or salt residue
Fair points. So you're suggesting they use the heat pipes to transfer the heat directly outside the vessel? That would have to be an impressive heat pipe to move 200W for the system.

1-Can't get around that except to go to a completely passive solution
2-mineral oil would require little in the way of tubing since its total submersion, it would also cool all components without the added complexity of putting heat pipes on everything, including the power supply, and an extra fun fact, its less dense than water.
3-everything you submerse in water will be subject to that. Water drawn in from the outside can be filtered, though obviously not purified. The heat sinks or rads will be subject to degradation too, unless you use rarer metals.

@OP what will you be using as a power supply?
 

· Registered
Joined
·
50 Posts
Discussion Starter · #6 ·
Duckie, thanks for the links, will definitely look into that

Firehawk, our case looks like

Below the colourful psu cable is the computer. It is not an ideal position and its cramped, which ain't so great, but its al the space we have. I agree that the endcap needs cooling fins, would welding them on work? Or is there a better way to attach these things for good conductivity?

As far as mineral oil, it would be a hassle to service it when testing at a pool (draining it will be an issue if we have to fix something)

I'm hoping I could fit 6 heatpipes from the cpu and bring them straight to the endcap. The other components may use heatpipes, but I was hoping we could circulate some air and put fins on certain parts on the inside of the endcap. I'm a little worried about this cause we'll have vrms an integrated gfx card motor controllers etc so we may need to route even more heatpipes. Even of all parts have heatpipes, I still would like some air circulation just to be on the safe side.

As far as psu, we're powering it with 2 lithium polymers 22.2 batteries, but I forgot what the actual psu is.
 

· Premium Member
Joined
·
67,312 Posts
Quote:
Originally Posted by Firehawk View Post

Fair points. So you're suggesting they use the heat pipes to transfer the heat directly outside the vessel? That would have to be an impressive heat pipe to move 200W for the system.

1-Can't get around that except to go to a completely passive solution
2-mineral oil would require little in the way of tubing since its total submersion, it would also cool all components without the added complexity of putting heat pipes on everything, including the power supply, and an extra fun fact, its less dense than water.
3-everything you submerse in water will be subject to that. Water drawn in from the outside can be filtered, though obviously not purified. The heat sinks or rads will be subject to degradation too, unless you use rarer metals.

@OP what will you be using as a power supply?
200w actually is nothing for heatpipes. You just have to size them right.

Mineral oil would not work here.
1) It would not make the system serviceable. Try opening the robot for a repair or tweak on a rocking boat...
2) It would drive up cost since it's not the cheap in volume.
3) There are other mechanical parts in the robot. Sealing off the oil will increase complexity again.

Quote:
Originally Posted by RandomPerson View Post

I'm hoping I could fit 6 heatpipes from the cpu and bring them straight to the endcap. The other components may use heatpipes, but I was hoping we could circulate some air and put fins on certain parts on the inside of the endcap. I'm a little worried about this cause we'll have vrms an integrated gfx card motor controllers etc so we may need to route even more heatpipes. Even of all parts have heatpipes, I still would like some air circulation just to be on the safe side.
Air circulation won't help. It looks like the container is some type of plastic? The heat will most get trapped inside so eventually you will be just circulating very hot air.

Even if you place a heatsink outside the container, where would that heat go? Into the cavity of the robot or directly to it's hull?

Again, I would draft a nice email and hit up a few heatpipe/heatsink companies. You might find an engineer who is willing to help your team out for fun or corporate sponsership.

BTW, you need some Velcro ties and work on cable management!
smile.gif


Depending on how long the robot runs for, I think you might end up with overheating issues....
 

· Registered
Joined
·
725 Posts
Most heatsink companies solder the fins onto the heatpipes. Keep in mind that soldering is not a strong mechanical bond, but as long as the fins aren't moved too much you should be fine. You'd be hard pressed to find someone who can weld fins in place. The difference in thickness between the plate you're welding them too and the fins themselves makes it particularly challenging. You could try buying some premade heatsinks and attaching them with thermal cement (not sure about the name).

From a serviceability standpoint, heatpipes definitely take the win. Every component that generates heat is going to need at least one pipe, or you'll create a hot box. CPU, chipset, VRMs, etc.
Air circulation would mean nothing. Actually, you'd be generating more heat because the fan motors create heat. They would also affect your battery life.

Any chance you could switch to a fully metal cylinder? It would allow you to use the entire length for more heatsinks.
 

· Banned
Joined
·
3,704 Posts
Quote:
Originally Posted by DuckieHo View Post

Sounds like an awesome project!

You can get various heatpipes from manufacturers but they are geared toward large industrial orders. However, I know of one that sells their spare/sample heatpipes... I'll go look for the company.

Another thing you might want to consider is sending out emails to heatpipe companies explaining your a student and are interested in their technology. You might be able to get samples or sponsorship.

Here's a $600 kit of MANY parts: http://www.aavid.com/products/standard/heat-pipe-exploration-kit
Here's a individual heatpipe site: http://www.enertron-inc.com/heatpipe.asp
Dude that is really neat, thanks for the info! How much is the explore kit? I haven't seen a price....
biggrin.gif
 
1 - 9 of 9 Posts
This is an older thread, you may not receive a response, and could be reviving an old thread. Please consider creating a new thread.
Top