D.I.Y. heatpipes

The alternative to direct air cooling

Moderators: NeilBlanchard, Ralf Hutter, sthayashi, Lawrence Lee

Post Reply
whiic
Posts: 575
Joined: Wed Sep 06, 2006 11:48 pm
Location: Finland

D.I.Y. heatpipes

Post by whiic » Thu Sep 27, 2007 3:00 pm

I've been half-seriously thinking about trying to make them, without a wick, probably, for simplicity of manufacturing. You may call them thermosiphons or thermosyphons if you like.

Idea:
- buy a small diameter metallic tube(s) and cut them to proper length
- hit one end of the tube shut with a hammer and then solder it tight
- fill with proper amount of water
- keeping it upright, hammer the top end shut too
- heat the lower end of pipe until it boils and steam starts to escape from non-soldered end
- now solder that other end so that it's both water and airtight.

There should be only water vapor inside. Since it was at boiling point at ATM (100 deg C) there will me significantly lower than ambient pressure inside the tube when it cools down so proper sealing is extremely necessary (I wouldn't trust thermosiphons that transmit steam and condensed water via plastic tubes since the joints are just pressed shut - not soldered).

I think such tubes are available from hardware stores (typical plumbing sizes) and some hobbyist stores (even very small diameters). Copper, brass, aluminum... I wouldn't care what it is since it's single metal so no galvanic corrosion should occur. I just want it to be solderable with normal tin+lead or tin+silver soldering wire and a blowtorch (because there's no soldering iron capable of heating the pipe hot enough to make a realiable solder).

Anyone of you done such a thing? If so, what material was your pipe? If not, do some of you have experience on soldering (big) components that don't have tin-covering on them?

Here's a list of fillings for heatpipes copy-pasted from a D.I.Y. heatpipe thread from Finnish murobbs forum (it's not original research so they copy-pasted it from some place else, I don't know where):

Code: Select all

MEDIUM      MELTING PT. BOILING PT. AT ATM. PRESSURE. USEFUL RANGE  

Helium      -271        -261                          -271 to -269 
Nitrogen    -210        -196                          -203 to -160 
Ammonia     -78         -33                           -60 to 100 
Acetone     -95          57                            0 to 120 
Methanol    -98          64                            10 to 130 
Flutec PP2  -50          76                            10 to 160 
Ethanol     -112         78                            0 to 130 
Water        0           100                           30 to 200 
Toluene     -95          110                           50 to 200 
Mercury     -39          361                           250 to 650 
Sodium       98          892                           600 to 1200 
Lithium      179         1340                          1000 to 1800 
Silver       960         2212                          1800 to 2300
I have no idea the original creators of that table came up with "useful range" because that depends on amount of substance inside the tube. I'd choose plain water... maybe with small amount of normal antifreeze glycol since they also contain corrosion inhibitors to prevent car engines from rusting from the inside out. Ethanol wouldn't be bad either but I'd stay away from acetone since I'll be boiling it before sealing the pipe and don't want to get toxicated doing it in a poorly ventilated household.

I read many of the posts there but they were creating plastic heatpipes with flexible hoses... they also had a low-pressure pump to depressurize it and due to leaks in the system, had to be depressurized frequently.

I want a permanently sealed, all metal heatpipe (without a wick is OK too). If I could buy heatpipes from a store, I'd be happy with that (because they only cost around 2 eur each and would have a wick) but the problem is that in reality I'd have to order hundreds of them since I'd have to contact the manufacturer itself. DIY is probably the only reasonable way to obtain heatpipes in small quantities.

Apprentice_GM
Friend of SPCR
Posts: 28
Joined: Sun Oct 09, 2005 7:12 pm
Location: Central Coast, NSW, Australia
Contact:

Post by Apprentice_GM » Thu Sep 27, 2007 6:19 pm

I thought I'd read an article here, possibly by fmah, (in his fanless / passive cooled "reference" article? ) but although searches show a hit the page isn't available any more.

Maybe it's been archived or something . . . it had a fair bit of detail of someone building heatpipes from copper pipes and sealing as you described. There was a fair bit of experimentation involved in how much liquid to have in the pipe - it was quite finnicky - too much or too little dramatically changed the heat transfer capability of the heatpipe.

He used some kind of alcohol. Sorry I can't find the article, maybe MikeC or an editor with archive search access can dig it out?

whiic
Posts: 575
Joined: Wed Sep 06, 2006 11:48 pm
Location: Finland

Post by whiic » Fri Sep 28, 2007 2:57 am

Water expands to 1500 fold volume when vaporized, thus if we wanted to make a heatpipe that can handle temperatures up to 100 deg C and didn't want to go any further, filling 1/1500 (0.6%) of volume with liquid water before boiling and sealing would be adequate. Well, of course since we boil it, some steam would manage to escape before top end is successfully soldered tight. 1/1500 of volume of a small diameter heatpipe is ridiculously small amount of water.

Putting too much water wouldn't increase the pressure because as long as ozygen and nitrogen molecules are driven out and it doesn't leak and let new air enter later, vapor pressure at given temperature is constant (providing that liquid phase hasn't completely depleted). So, I could easily fill 1/150 of volume (10 times the needed amount) so that I can boil water properly and make sure that boiling the water "flushed" the pipe of any residual oxygen and nitrogen. 1/150 is still quite a small amount of water after all.

What you did say about the amount of water affecting performance, it's true. Too little, and liquid phase will be depleted within intended operating temperature range and when the heatpipe reaches certain temperature it's evaporation/condensation cycle halts, the cold end remains at constant temperature but hot end heats up rapidly.

Too much liquid phase only boosts upper end of temperature range. Lower end should be unaffected and only limited by amount of impurities inside the tube (mainly the pressure of residual oxygen and nitrogen). Too much water might cause boiling process itself to form too big bubbles. People who tried transparent plastic pipes at murobbs noticed that boiling water in small diameter pipes caused big air bubbles that were so big they pushed columns if liquid phase water up to the top-end of heatpipe. Obviously heatpipes without a wick rely on gravity pull it back but surface tension keeps droplets still big so too small diamater could make part of liquid phase stuck on the wrong end.

I'm not sure if that's a problem though. It merely transfers excess liquid phase to wrong end of heatpipe but after being stuck there, would it hurt cooling? Well, it's safer to assume it does. But I do think filling it with 1/150 of water prior to heating is within safe range. Of course by doing this I would create a heatpipe that under extreme thermal stress would not lose efficiency but increase to 10 bar internal pressure (9 bar above ambient). In worst case scenario: exploding heatpipes. It's impossible for them to explode at below 100 deg C, though, because pressure inside the pipe is lower than ambient.

A way to get around the problem might be to bend a pipe like "C" then bend it wither to "O" and solder it tight (not hammering anything). Drill a small hole to it, inject water into it with a needle, boil the water and solder the hole shut. Put the hot end on other other of vertical columns and condenser on the other. That way it should circulate without a water pump and condensed droplets don't have to flow down the one and only column down, against steam-flow which goes the opposite direction.

Other solution: increase heatpipe diameter so that droplets or bubbles don't fill the whole pipe and block it.

But keeping the amount of water in moderation 1/1500 to 1/150 might be the easiest way.
Last edited by whiic on Fri Sep 28, 2007 5:59 am, edited 1 time in total.

jhhoffma
Posts: 2131
Joined: Mon Apr 25, 2005 10:00 am
Location: Grand Rapids, MI

Post by jhhoffma » Fri Sep 28, 2007 5:15 am

The vapor would only stay "vaporized" if the heatpipe is heated to the boiling point of water inside the pipe (might be higher than 100C due to pressure increases). If the temperature decreases at all (turning the component off or lower power state) the vapor will condense on the coolest surface around (the top of the heatsink) leaving you with a very poor-cooling copper heatsink.

That's the reason most heatpipes use alcohol (methanol) which has a very low vapor (boiling) point, it's phase change point is right in the ball park for where we want our computers to be. Also, it doesn't expand as much as water when vaporized keeping you from exploding a pipe. I'm not sure what you mean when you say "vapor pressure at given temperature is constant". You may mean that water has a certain vapor pressure at a particular temperature in free air (no volume constriction), however, it most certainly does not if you restrict volume (keep it constant).

Ideal gas law says:

pV = nRT

where,
p=pressure
V=volume
n=moles of gas
R=gas law constant
T=temp (in Kelvins)

If you keep V constant and raise T, p must increase to account for it. You could try your idea, but I think one of two things will happen. 1) you will burn up whatever you're trying to cool, because of issue#1 or 2) you will blow up your heatpipe because of issue#2.

At least your thinking, though...

whiic
Posts: 575
Joined: Wed Sep 06, 2006 11:48 pm
Location: Finland

Post by whiic » Fri Sep 28, 2007 6:55 am

"The vapor would only stay "vaporized" if the heatpipe is heated to the boiling point of water inside the pipe (might be higher than 100C due to pressure increases)."

If I put more water in than 1/1500 of internal volume of heatpipe, yes, it can handle over 100 deg C. One correction though: the water inside the pipe is constantly at boiling point, even when system is shut down and both hot and cold end are at rooms temperature. It's in boiling point because there's near-vacuum inside the tube. In vacuum you can boil water at 0 deg C. I think 30 to 200 deg C is quite realistic operational range for water filled heatpipes. Some degree of cooling would be offered even at below 30 but cooling it that low is unnecessary at heatsink at cold end would be cooled by air in rooms temperature anyway

"If the temperature decreases at all (turning the component off or lower power state) the vapor will condense on the coolest surface around (the top of the heatsink) leaving you with a very poor-cooling copper heatsink."

I don't understand why condensing on cold end would be bad. It releases energy that was stored at the hot end when it was vaporized. That's how heatpipe works.

"That's the reason most heatpipes use alcohol (methanol) which has a very low vapor (boiling) point, it's phase change point is right in the ball park for where we want our computers to be."

True, but even water boils at 0 deg C when in vacuum. Of course there'll be some residual oxygen and nitrogen molecules to operating range starting from 30 deg C is more realistic. Also, even IF I could get rid of all the oxygen and nitrogen, the vapor pressure would be quite low at 0 deg C, meaning that number of heat carriers would be limited at such low temperature.

"Also, it doesn't expand as much as water when vaporized keeping you from exploding a pipe."

If you end up with a heatpipe hotter than 100 deg C, your CPU is toast already. I don't care about exploding heatpipes. Not going to happen.

"pV = nRT"

That formula handles expansion of gas vapor. No liquid phase there. That's how gasses expand but as they expand and increase pressure, boiling point increases too.

Still, if I seal the pipe with the water (both in liquid and vapor form) inside when heated to 100 deg C at pressure of 1 ATM, the pressure inside the tube will be 1 ATM even after it's sealed. If I continue to heat it, pressure will increase both due to water evaporating and transforming to vapor form, and due to gas expansion as temperature rises. But this doesn't happen if I don't heat it beyond 100 deg C. If I keep it at exactly 100 deg C (or if I heat it to 110 deg C and let it cool back to 100) the pressure will remain at exactly 1 ATM. There will be no mysterios pressure build-up because that increased from boiling and gas expansion will, given any constant temperature, reach an equilibrium state when evaporation=condension.

If I heat it beyond 100 deg C, pressure starts to exceed ambient and risk of explosion is created. Increase of pressure isn't due to expansion of ideal gasses because pressure created by gas expansion hinders evaporation in turn. It ends up canceled from the equation and it all comes down to vapor pressure.

http://en.wikipedia.org/wiki/Image:Wate ... _graph.jpg
(nore that at 100 deg C, vapor pressure is 760 Torr = 1 ATM)

You can use that graph to plot at which temperature water starts to boil, given some known pressure. Or alternatively you can determine the pressure of water vapour inside a heatpipe (with air sucked out) given a known temperature. Note that increasing temperature just 100->120 will double the pressure where as doubling the pressure of ideal gas require doubling temperature (in Kelvin scale) (so basically it means heating from 100 deg C to 750 deg C), so while ideal gas does expand, limiting the amount of coolant in practice does act as a safeguard against exploding.

Of course it's not much a safeguard to the cooled component at hot end...

whiic
Posts: 575
Joined: Wed Sep 06, 2006 11:48 pm
Location: Finland

Post by whiic » Sun Sep 30, 2007 5:43 am

Correction: even wickless heatpipes aren't thermosiphons. Thermosiphons rely on natural movement of warmed liquid but not on evaporation & condensation. Even if I were to build a O ring heatpipe, I'm not sure if could be called a thermosiphon. It's closer to principle of thermosiphon but it'd still have most of it's cooling potential due to evaporation & condensation (and ring shape is merely to ensure condensed water return with less resistance back to hot end).

I might consider ethanol after all, to lower the low end range of cooling. While 30 deg C is low enough, it's not like it's an ON/OFF operating range. Conduction is still far from perfect using water at 30 deg C. Not because it won't boil (because it will) but because vapor pressure is still quite low at boiling point with few carriers of heat. I'd rather not use acetone or methanol from health reasons but I could consider ethyl alcohol.

Pure (or at least 96%) alcohol would be preferrable but I'm not sure if Estonian booze (at 80%) would be too diluted to have significant effect of heating. It's the alcohol that vaporizes first so it's the alcohol that determines cooling efficiency. Water just stay inert and takes some space in the tube, doing absolutely nothing... unless I heat the pipe to a point when all alcohol remains in vapor phase and water starts to boil. Making any theoretical calculation on heatpipe with several substances as working fluid becomes more difficult.

If I even manage to produce a working heatpipe, I'll probably try making a CPU cooler that conducts all it's heat to computer chassis. I'd probably use my Prescott-based system as a lab rat since I don't really care if it overheats and dies as result of experiment. But I'd still make some testing using just a non-delicate heat source (such as hot water) and some heatsink in other end because intentionally killing a CPU doesn't sound too wise. I don't want to kill the only lab rat before I at least make sure the heatpipe itself is working well or at least decently. If CPU later dies because improper application of properly working heatpipe - that's a whole different thing and more difficult to eliminate by doing tests outside the system it's going to be used.

whiic
Posts: 575
Joined: Wed Sep 06, 2006 11:48 pm
Location: Finland

Post by whiic » Sun Sep 30, 2007 12:06 pm

Some reading for us interested in DIY CPU cooling solutions: http://www.silentmods.com/section2/item230/

They didn't make the heatpipes themselves but dissected a Thermaltake Typhoon. They also dissected Zalman heatpipe HDD cooler and I think a Zalman GPU cooler. HDD cooler's heatpipes were fragile (in addition to being short) and snapped while bending. Zalman GPU cooler had only 2 heatpipes which was too much price per heatpipes obtained.

It's a appear a bit non-tempting for me to dissect high-end coolers to make my own ghettocoolers. They did try making heatpipes themselves... even with a bendable joint. Again, I don't understand why most people who actually try to build heatpipes lack all common sense. No-one's DIY guide suggested to create vacuum by boiling working fluid... instead everyone suggested that you use some kind of vacuum pump. First: vacuum pumps cost money. Second: no vacuum pump can create perfect vacuum. It's also quite a hassle to fit some one-way-airflow valves for sucking air out, or trying to pump air out through some small hole and soldering it shut while there's pressure difference. No wonder they ended up failing in their experiment three times before they got it working. Heat the pipe up to boiling point of working fluid and pressure difference won't suck the molten solder into the heatpipe, for god's sake! Where's the common sense?

Also why on earth would I drill holes to copper blocks since it's so hard to drill into. Even then they ended up having the holes way too loose to work with any thermal paste and not transmitting heat until they soldered it. It's funny they originally didn't intend to solder the pipes ... even though they noticed that Typhoon they dissected had their heatpipes soldered. We're they in some hypnotic state at that point?

No, I'm not really the right person to bash their methods. I'm just jealous. I'm jealous about their silent build and I'm jealous that I don't have any of the equipment they have: drill bits for copper, pipe bender, 60W soldering iron (I have a crappy 15W one), etc.

If I were to try that, I would cut corners here and there to make it simpler to manufacture, both in worksmanship and in equipment required.

I would try to save a lot of trouble by directly soldering heatpipes onto the block. Some sort of clamping wouldn't hurt either as it would increase contact area. Flattening heatpipes should be OK since we've seen that on commercial heatpipe heatsinks(*). Didn't (at least) one manufacturer make tower heatsinks with heatpipes directly touching CPU IHS? If we tried to make something real bold, we could do the same thing and flip that block (with heatpipes soldered onto it) around: make direct contact with heatpipes to CPU. We could lap both the surfaces nice and smooth.

(*) Also, if I were to use DIY heatpipes, nothing forces me to fill the pipes with working fluid at this point. I wouldn't even need to worry about accidentally overheating them if they were just empty tubes at this point. Filling it with working fluid when block is done. As a side product I might have some nice bluish velding marks on my pipes. Looks kinda more hardcore that way. :p

ddrueding1
Posts: 419
Joined: Sun Sep 19, 2004 1:05 pm
Location: Palo Alto, CA

Post by ddrueding1 » Sun Sep 30, 2007 2:32 pm

I'm really appreciating all the thoughts and info here. What are the thoughts WRT the length of a heatpipe and it's efficiency?

whiic
Posts: 575
Joined: Wed Sep 06, 2006 11:48 pm
Location: Finland

Post by whiic » Sun Sep 30, 2007 4:11 pm

"I'm really appreciating all the thoughts and info here. What are the thoughts WRT the length of a heatpipe and it's efficiency?"

That's a tricky question since heatpipe doesn't have a constant resistance/distance (because it doesn't rely on conduction but instead evaporation+convection+condensation), nor does it have constant degrees/wattage (because vapor pressure changes with temperature). All I can say is that when properly manufactured with properly selected working fluid and right amount of it, and when operated within it's intended temperature range, heat transfer is phenomenal. Usually they say that outside the temperature range, heat conduction is only 1/80 of within it. Since the heatpipe is hollow, the conduction while in operating temperature range is probably similarish to about 10 time that of solid (non-hollow) copper pipe of same diameter... but that's just my guess. And that 1/80 was from wikipedia... and that too was just a generalization since the thickness of pipe wall and flow characteristics inside it varies from pipe to pipe. There's lots of different diameters even though computer industry mainly uses one or two of them.

I suspect even a single heatpipe could cool low-end CPUs enough. Having several of them is safer because not only will it add more theoretical heat transfer, it also allows to distribute the heat to a wider area (if they are for example soldered to HS fins). Also having many heatpipes adds redundancy in case one doesn't work or ceases to work (for example due to physical damage when transported).

To have enough fins and enough airflow at the cold end is important and in silent/quiet systems it may be the limiting factor. Also, how heatpipes are connected to fins and especially to CPU block plays a big role. If fins are replaced by attempt to try to conduct heat to computer chassis, that obviously is tricky as you can't just solder the heatpipe to stainless steel (nor aluminum) chassis. It may be best to make another block and then bolt that block to chassis, using thermal compound in between.

ddrueding1
Posts: 419
Joined: Sun Sep 19, 2004 1:05 pm
Location: Palo Alto, CA

Post by ddrueding1 » Sun Sep 30, 2007 8:31 pm

whiic wrote:It may be best to make another block and then bolt that block to chassis, using thermal compound in between.
That is my current plan for my next gaming system. Motherboard, SSD, GPU, DC-DC PSU, all surface-mounted to a huge aluminum plate with heatpipes as needed. Then wall-mount the plate on the wall and allow ambient air circulation to do the rest.

ddrueding1
Posts: 419
Joined: Sun Sep 19, 2004 1:05 pm
Location: Palo Alto, CA

Post by ddrueding1 » Sun Sep 30, 2007 11:35 pm

Interesting PowerPoint presentation can be found here.


ddrueding1
Posts: 419
Joined: Sun Sep 19, 2004 1:05 pm
Location: Palo Alto, CA

Post by ddrueding1 » Mon Oct 01, 2007 9:54 pm

This place also has an online store where you can order just about any heatpipes you want, from 3mmx15cm for $15 each to 8mmx30cm for $25.

That is cheap enough that I don't think I'll be DIYing it anytime soon.

whiic
Posts: 575
Joined: Wed Sep 06, 2006 11:48 pm
Location: Finland

Post by whiic » Fri Oct 05, 2007 5:22 pm

I have bought the material needed for making heatpipes. And I intend to make ones with a wick after all.

Supplies to be used in the project:
- brass pipe, 1 meter, 6mm inner, 7mm outer.
- regular SnPb soldering wire (which includes some flux in it)
- butane torch (1000+ watts)
- steel cable (=multistranded wire), diam 1.5mm
- Viru Valge Vägev 80% (or some other Estonian quality product)

Also tools like hammer, bench vise, pliers, eye protection, gloves, etc.

I will probably measure effectiveness by warming up another end with known temperature (for example boiling water at 100 deg C) and measuring the other end with a thermocouple. Of course with just the pipe, it'd heat up to ~99 even from cold end, so there has to be some method of cooling the other end. Ideas?

About the wick... well, that's what the steel wire is for. I'll unwind it a bit so that it'd become less dense, but not too much as it needs to be capillary. I do some capillary tests in bathroom sink before assembling the heatpipe.

Material cost (not counting butane torch since it's reusable): roughly 10 eur for 3...4 heatpipes. And there'll be quite a bit leftovers in everything but brass pipe. I could make 6...8 heatpipes with 16 eur.

Maybe I'll start actually making the heatpipes tomorrow. There's not much that I can plan further without trying. Except how much ethanol expands when it evaporates. It isn't exactly the same as water.

I'm not too good in chemistry so correct my if I'm wrong:
water: 1.00 g/cm³ divided by 18 g/mol = 0.055 mol/cm³
booze: 0.79 g/cm³ divided by 46 g/mol = 0.017 mol/cm³

Since pV=nRT applies to all gaseous molecules, I think it's proper to assume that ethanol expands 1500*17/55 =460 times the volume it takes when it's liquid.

peteamer
*Lifetime Patron*
Posts: 1740
Joined: Sun Dec 21, 2003 11:24 am
Location: 'Sunny' Cornwall U.K.

Post by peteamer » Fri Oct 05, 2007 11:43 pm

A link of possible usefullness.

Links to version 2 and 2.2 at bottom of page and in left side bar.

Good Luck. :D

Pete

drownmypc
Posts: 56
Joined: Mon Feb 11, 2008 2:14 pm
Location: chapel hill, nc

Post by drownmypc » Sun Feb 17, 2008 7:28 pm

whiic wrote:Maybe I'll start actually making the heatpipes tomorrow. There's not much that I can plan further without trying. Except how much ethanol expands when it evaporates. It isn't exactly the same as water.
Does anyone know if whiic made his heatpipes?

I tried to make some heatpipes today, using essentially the method he outlined. I took a copper pipe section, capped it (with solder), poured a little water in, heated it all to boiling, let it boil for a while hoping to push the air (mostly O2 and N2) out with water vapor, then capped it and soldered it shut.

Problem: sealing it shut. While trying to put on the final cap, still-boiling water was pushing water vapor through the molten solder (the solder sputtered while I was brazing the cap). I ended up with micro-holes. I couldn't see them, but I could hear the air rushing out when I reheated the pipe. And if air can get out, it can get in.

Problem: purity of low-pressure atmosphere. I have doubts about whether I managed to get only pure water vapor in the pipe when I capped it. After all, the air is mixing with the water vapor, and yes, mostly it is just water vapor (I guess), but I'm sure some air remains in the pipe. I don't know if a little air is a big problem.

Problem: controlling the amount of water in the pipe. I boil the water for a bit in order to force the air out. How much should I let it boil? I'm losing water while I do so, and I don't really know

Making heat pipes isn't easy. At least, not the first time I tried it. :-)

drownmypc
Posts: 56
Joined: Mon Feb 11, 2008 2:14 pm
Location: chapel hill, nc

Post by drownmypc » Wed Feb 20, 2008 4:33 pm

I tried something a little different today, to deal with this issue I had before ...
Problem: sealing it shut
I used a 1/4 inch copper pipe, 2 feet long, capped at one end and a ball valve attached to the other end.

Procedure: Clamp the tube vertically with valve on top and capped end on bottom, open the valve, pour in a little water, heat it up to boiling with a blow torth, when I see water splashing out (due to boilage), close the valve. Near instant seal.

The valve was threaded, so while it was shut I screwed on a pressure gauge (goes -30 to +30 psi) using plumber's tape (teflon) for an airtight seal.

Open the valve -- allowing the gauge to feel the pull of the tube's vacuum -- and the gauge instantly dropped to -18 psi.

My conclusion: the experiment mostly worked, and the pressure gauge lies (or is rather inaccurate).

Why: 1 atmosphere is 760 Torr = 14.7 psi. A perfect vacuum should register -14.7 psi, right? Am I missing something?

Anyway, the pipe didn't keep its seal very long. The negative pressure gradually moved to about -10 psi over the course of an hour. I think soldering job was okay (better than last time!). I think the teflon-taped joins were leaking.

But it was an interesting experiment, albeit rather sloppy.

-stefan

Post Reply