High Power Cooling / General Science and Electronics / Forums

Forums

4hv.org :: Forums :: General Science and Electronics

« Previous topic | Next topic »

High Power Cooling

1 2 next

Move Thread

LAN_403

...

Tue Jan 23 2007, 05:08AM

Registered Member #56 Joined: Thu Feb 09 2006, 05:02AM
Location: Southern Califorina, USA
Posts: 2445

Well guys, I am in need of some major cooling.
In short, I need to keep a 300w load at 25c, self contained, for as cheep as possible.

In a little more detail...
The load is actually 3 tiny pieces of silicon (with a footprint of 1mm by 1cm--50w laser diode bar) which is soldered down to a copper block 1" square by 1/4" thick. These copper blocks are then bolted down to a larger copper plate (~2" square). There is a .020" gap between the block and the copper plate, which is supposed to be filled with a piece of an electrical insulator (some type of ceramic?) and something between the copper/ceramic (probably indium foil).

The lasers need to be at 25c.

The thinnest ceramic I could find at McMaster was p/n 8462K21, which is .025" (I would need to somehow sand down the blocks .01" or something to make that work) but if I recall alumina has a a thermal conductivity of like 20W/mK, so for that .025" by 1" square it will drop ~5C.

So I am tempted to just run all of the lasers in a common anode configuration (I am custom building the power supply anyway, so I don't think it that big of a problem) and just fill the gap with indium. Indium is a pretty solid 80W/mK, which would give me a 1degree drop across the .020" of metal. Then the 2 metal-metal interfaces which Indium corp. says the pair should drop about 1C (depending on how tight I can get it, 1C is 30lbs/cm^2.). So that is a 2c drop there.

I am assuming a 2C drop across the copper in the block (educated guess, since there are some holes and screws and stuff in it), and another 1c drop across the mount. All of the thermal transfers from there are huge by comparison (like 30cm^2) so only a few C there.

So, to keep the lasers are 25c, I need the bottom of the mount at 20c.

That is where the active cooling comes in. The biggest peltier I could find were some 2" 300w monsters. I couldn't find any data on them (nor test them since I don't have my waterblock) but for a 80w peltier I am supposed to drop 35c at 40w with the hot side at 35c (I plan to have 35c water, which would mean hotside temps of ~60c). So, if I use a pair of the 300w peltiers, 7.5x the cooling power, which comes out at a hair over 300w of cooling power.

So, I think that I should be able to use a pair of my bigass peltiers to get the block cool enough. My plan would be to use a 3/8" copper block that is 2" by 4" with the mount bolted to it, and to have the peltiers each bolted down to its own swiftech Apogee waterblock.

Each of those blocks is .05c/w, so with the pair I should be at .025c/w, which is still dropping a painfully high 25C. I am not sure where to go from there, really the waterblock is dropping the most painful amount of temp.

From there I need a radiator that will keep my water below 35c with a 1kw load, which I don't think is all that hard. Just lots of fans ;) I figure with 25c air going in and 35c going out I need enough radiators to give me 100w/c... I am thinking that thermodynamically to raise 1kw worth of air 10C is 70cfm.

Overall it will be very tight, and every degree will be painfully earned, but I am hoping that the 100w/laser figure is a little high. I plan to put 100w into each diode of electrical power, but there should me 30-50w of optical power leaving the mess so that could help me get an extra 10%. And in the end, I am not even sure if this bar can handle 150w of pump power (similar looking mounts are only rated for 15w 1064nm, which would be like 60w of pump. So if I only run the diodes are 25w, or better yet put in 25w diodes (still have 3 of them around) which would require considerably less power I would really be home free and trying to get away with a single peltier.

Any comments/suggestions? I already know that I am insane

Dr. Slack

Tue Jan 23 2007, 08:39AM

Registered Member #72 Joined: Thu Feb 09 2006, 08:29AM
Location: UK St. Albans
Posts: 1659

If you are planning watercooling the hot side of the peltiers, and you need only get the hot side down to 20C, then have you considered leaving the peltiers (and their tens/hundreds of watts of additional drive power) out completely, and simply watercooling the hot side?

The trick is to have a resevoir containing ice and water so the heatsink feed water is near 0 degrees C. The enthalpy of fusion of ice is about 334kJ/kg, which means that melting 1kg per hour takes about 92 watts. If you have a limited run time, then you could simply load a cool-box with a few kg of ice and a little water, and circulate it till the ice has gone. That's self-contained in operation, and about as cheap as it gets (assuming access to a supply of free ice). Add salt and you can get down below 0 C, if your pump can handle the corrosive solution. For more than 1 hour or so run, you would need to consider replenishing the ice and draining off some water.

As you are now only cooling the prime load and not the heat from peltier operation, you can get away with much lower water flow rate, perhaps low enough to dispense with a pump, and drive the water by manually scooping a jug full from the lower resevoir to the upper resevoir. With the specific heat capacity of water being 4.2kJ/kG/C, that means 1 litre of water per 400seconds would cool 100 watts with a 10C temperature rise. Very cheap if you've not got the pump already, and can remember to scoop every 5 minutes. Seperate the laser and the water cooling on different tables, and let the tubes drop into a U between them, to avoid drips running down into your expensive electrical loads.

However, if the need to supply or replenish ice does not meet your 'self-contained' requirement, you might consider an evaporative 'bong cooler' (Google) on the Peltiers, which has the potential to get the cooling temperature down to the dew point of the atmosphere, rather than its temperature. Various fillers can be used, from none, through plastic packing chips to pebbles. Even this is not strictly self-contained as the water needs to be replenished, but that's easier than carting ice around.

Steve Conner

Tue Jan 23 2007, 11:09AM

Registered Member #30 Joined: Fri Feb 03 2006, 10:52AM
Location: Glasgow, Scotland
Posts: 6706

Peltiers are incredibly inefficient and it's usually best to avoid them as far as possible. By introducing peltiers you suddenly have to get rid of 900W of heat instead of 300, and the extra Delta-T caused by this extra heat flow can negate the cooling effect of the peltiers altogether.

In your case, with .025'C/W waterblocks and 35'C water, the lasers will run at about 45'C with the peltiers deleted, assuming the water is still at 35 degrees. So there is still a net benefit to them in this case. Having said that, wth no peltiers I'd expect the water to cool down nearer 28 and the lasers to run around 40. Because of the horrible economics of active cooling, any halfway decent pump diode bars should be designed to run at 40 degrees anyway. Nobody would buy a 5kW diode laser welding machine if it needed 15kW of TEC cooling.

Vapour cycle cooling is far more efficient and sometimes even more cost-effective. Can you steal the guts out of a refrigerator, or get hold of an old draught beer chiller, or something?

A beer chiller would be great, since it refrigerates a tank of water that you could pump round your waterblocks. And you can chill a 6 pack in it for when you get old enough to drink

Of course, if you already have a pair of bigass TECs and a power supply and you're looking for something to do with them, feel free to ignore this advice

Neil's bong cooler is probably the best value for money solution for cooling the hot sides of your TECs: it's basically a mini version of an industrial cooling tower.

Sulaiman

Tue Jan 23 2007, 12:29PM

Registered Member #162 Joined: Mon Feb 13 2006, 10:25AM
Location: United Kingdom
Posts: 3140

Would something like this be useful?

Marko

Tue Jan 23 2007, 12:52PM

Registered Member #89 Joined: Thu Feb 09 2006, 02:40PM
Location: Zadar, Croatia
Posts: 3145

I don't think that forced air or just water cooling alone will keep temperature that low.

I vote for Steve's idea, combination of fridge guts and water cooling. Using freon directly without water cycle would probably be more efficient but impractical unless you are a metalworking god.

With refrigerator you also get heatsink (condenser) and all stuff you need.

Peltiers are indeed very inefficient and practically useless with significant heat loads.

thedatastream

Tue Jan 23 2007, 04:13PM

Registered Member #505 Joined: Sun Nov 19 2006, 06:42PM
Location: Yorkshire!
Posts: 329

Difference in temperature is the biggest factor in any kind of heat transfer. The higher the difference the more heat will be transferred.

Q = KA [ (T1 - T2 ) / H ]

where:
Q = conductive heat transfer in Watts
K = thermal conductivity in W/mK
A = area of conductor
T1 = temperature of oneface
T2 = temperature of the other face
H = thickness of conductor

So you can see that the higher the temperature difference, the higher the heat transfer. Assuming your cooling medium is at an average room temperature of 22ÂºC, then you only have 3ÂºC delta - this will make for poor cooling.

Using some kind of refrigerated coolant, like iced water, will enable you to remove much more heat. The guts from a fridge or freezer sounds ideal and you can put a fan on the heat exchanger to increase efficiency.

As for the interface material, Bergquist

make some decent insulative material with high thermal conductivity - give them a look. They have been amenable to giving me sample material in the past. Mica washers with lots of thermal grease are also pretty good (if a little messy)

Pressure between the heat source and the heatsink will help to decrease the thermal resistance of the interfaces.

HTH
James

...

Tue Jan 23 2007, 11:34PM

Registered Member #56 Joined: Thu Feb 09 2006, 05:02AM
Location: Southern Califorina, USA
Posts: 2445

Ok, as to the temp of the bars. I need to keep them at the correct temp so that they put out exactly 807.5mn of light (best for yag absorption). However, I don't have the ability to measure the wavelength of the bars for a given temp

I am basing my 25c guesstimate on a small dpss laser I repaired using one of the emitters from one of the bars, which worked best at 25c, and by the time it was up to 35c there was almost no lasing action. I believe this is the case with almost all laser bars, as most any dpss laser has some type of active cooling, and the lifetime of a laser diode is cut in 1/2 for every 10c cooler you run it. So while they could design the lasers to run at 40c, they will only last 1/3 as long... Generally the huge lasers have the bars mounted directly on a waterblock, and an external waterchiller that will give 10-15c water at some ungodly rate But if I can get away with a $200 worth of pielters/waterblocks that would fit in a tabletop sized laser I would greatly prefer that over an external phase change watercooler. Have you ever lifted a refrigerator?

I can't use ice as the coolant, I have a 300w load, which would burn through a 1kg of ice in 15 minutes. You people don't seem to realise that this needs to be a system that you can plug in and run for a few hours. Just for alignment I will need probably 10's of hours of runtime at modest power. And then I need to be able to use it...

And I know how to calc the amount of heat drop over a conductor, that is how I got all of the numbers I have so far. Also, if the diodes really do need to be kept at 25c, the cold side of the cooler will need to be at 20c (due to losses in the block). There is no way you will get water at 20c without any active cooling.

Really, peltiers were the only thing I could think of that would work inside the laser. Remember that if I have the laser on a waterblock, I will have 10c drop there, so I need 3gpm of 10c water. If you think that it will be cheaper/easier to build a waterchiller that can produce 3gpm of 10c water with a 300w load, that will is easily transportable...

BTW, we are homebrewing the SMPSs to power all of this, so as long as I can keep my power budget <1.58kw (max us outlet can handle) I don't really care about how much power I am burning.

Dr. Slack

Wed Jan 24 2007, 08:35AM

Registered Member #72 Joined: Thu Feb 09 2006, 08:29AM
Location: UK St. Albans
Posts: 1659

Whether to fix peltiers to the load, and water cool their hot side, or to fix peltiers to a heat sink, and use them to chill water to send to cool the load?

There are several advantages to the second option.
1) You have more space per peltier, you can bolt each to a bigass air-exchange heat sink, the water circuit can "collect the cold" from each. Water cooling is a very dense and a good match for getting in close to where a lot of heat is being generated in a small space.
2) You can switch the inefficient peltiers for another type of cooler later on without altering the laser bars, say a refridgerator or even (if you want to run from batteries for a short demo in an out-of-the-way place) a truckload of ice.
3) The cold water in the cooling circuit gives you a buffer. Should your active cooling fail, you may have minutes rather than seconds to notice and switch off before damage occurs.

Peltiers are at their least inefficient when running with a low delta T and lower than design current, so use the best ambient cooling for their hot side that you can.

You people don't seem to realise that this needs to be a system that you can plug in and run for a few hours.

A little more tolerance and politeness to your respondants please. If you check your first post you'll find you mentioned cheap and self-contained as specifications, and a whole load of detail on the components, but nothing about run time. I don't expect thanks for my very cheap but off the wall suggestion, but I spelt out the run time issues so I don't expect to be dissed.

Steve Conner

Wed Jan 24 2007, 11:41AM

Registered Member #30 Joined: Fri Feb 03 2006, 10:52AM
Location: Glasgow, Scotland
Posts: 6706

Yes, you asked for advice, and you got it, so no complaining!

I see why the temperature needs controlled accurately, if you need to use temperature tuning of the diodes. That would shift the balance in favour of peltiers, since they're so easy to control. But you never specified that. Also, do you already own the big peltiers? That would obviously be another big argument in favour of them, since 600W of TEC cooling is liable to cost a few hundred dollars. If you do, I'd say stick waterblocks and a bong cooler on them. If not, I'd be looking for a beer chiller.

The last product I designed was a tunable laser source using a temperature tuned DFB. That's a lot easier since it comes in a package with a TEC, but to meet the wavelength accuracy spec, I needed to make a TEC controller accurate to 0.05'C.

...

Wed Jan 24 2007, 03:30PM

Registered Member #56 Joined: Thu Feb 09 2006, 05:02AM
Location: Southern Califorina, USA
Posts: 2445

I apologise for any comments that may have offended anyone. I was hoping that self contained would keep suggestions of hooking it up to a bucket of ice away, but I guess we have different ideas of what 'self contained' mean.

I also forgot that the rest of 4hv isn't well educated in laser diodes, and didn't feel the need to explain when I said I needed to keep them at 25C. I will try to be clearer next time.

But, in any case, I was thinking of putting the waterblock on the diodes, and then putting another pair of waterblocks on the peltiers, and then somehow cooling the other side, but I think that deltaT wise I am better off with the peltiers on the mount directly.

My reasoning is that, I need the mount at 20c, and it has 300w going out. So if I put a pair of blocks on it, that will drop a little over 8c, I need the water at 12c. So then I have another pair of peltiers on the blocks, dropping another 8c, I need the cold side of the peltiers at 14c. So that means I need to keep the 1kw side of the peltiers at 50c. Which means that between the peltier and the monster heat sink, I have 25c to burn 1kw, or 40w/c... I doubt that you can get a normal heat sink that good, or the overclockers wouldn't use watercooling... And I can't even think of using another set of waterblocks there as I would need to have water at 15c to cool the peltiers.

Finally, peltiers are cheep. A 300w module is buy-it-now on ebay for $25. The waterblocks I am using are $55 a piece. The pump is $70. Radiators are like $30 a piece, and I already have one. The power supply will cost <$20 is parts (since I already have most of what we will need). Overall, the cooling should cost about $300 with all of the extra little bits.

Also, at thedatastream, do you have any particular item from Bergquist that beats out alumina? I could only find thermal pads with like 5w/c (eek!) numbers.

1 2 next

Moderator(s): Chris Russell, Noelle, Alex, Tesladownunder, Dave Marshall, Dave Billington, Bjørn, Steve Conner, Wolfram, Kizmo, Mads Barnkob