Post Rad Chiller Concept

[Scarlet Infidel]

attached is a shot of a cooler I helped my nephew build

Interesting stuff, but its obvious from the offset that's a really inefficient way of using Peltier elements, unless I've badly misunderstood. It would be nice to know a few more details about that project and how you have implement peltiers in the past.

[NaeKuh]

Interesting stuff, but its obvious from the offset that's a really inefficient way of using Peltier elements, unless I've badly misunderstood. It would be nice to know a few more details about that project and how you have implement peltiers in the past.

well i would actually have to agree with you on this statement.

Because water is a better medium for transfer on the cold side. Also this is what i felt on the hotside. Im trying to milk the entire unit for efficiency. By watercooling the hotside, the cold side gets cooler, and i can use less voltage to get desired results. no?

[Scarlet Infidel]

Yeah, that's correct and a very sensible way of getting the most out of your chiller.

I wouldn't recommend trying to air cool the hotside. The reasons I'm air cooling my hot side is:

Lower cost,
lower noise (that's not likely but still possible),
smaller size,
to be different,
to prove that it can be done,
because my heatload is spread over a large area,
and because I don't think it will be much worse than water cooling if done carefully.

[Uncle Jimbo]

Interesting stuff, but its obvious from the offset that's a really inefficient way of using Peltier elements, unless I've badly misunderstood. It would be nice to know a few more details about that project and how you have implement peltiers in the past.

I have designed pelt based control systems going back more than 20 years. Typically these have an instrument package in a sealed NEMA enclosure, and the pelt cools if load is too hot, heats if load is too cool. These are typically on 48VDC systems so we use 2 24V pelts in series with SCR chopper control. Those systems ran in ambient as low as -40C and as high as 50C.

The efficiency of the TEC improves with hot side temp, with best efficiency around 50C hot side, so air cooling the hot side is actually more efficient. For example, the curves on a CP1-12730, a big 62mm TEC, show 240 Qc W with 5C differential and the hot side at 50C - but only 190 Qc W at 25C for the same volts and amps. Since what you want is to take cold side below ambient, you want closer to 30C differential with the hot side at 50C. With a 16V supply and 30A, that TEC will move about 140W. It will use 480W getting there, so your heatsink needs to be able to dissipate 620W which is big.

the cooling effect you get is determined by the surface area of your heat exchanger (you have lots of that!) and the rate at which the water goes through. To change the temperature of 1 gram of water 1 degree C requires 4.18 Joule (watt-second). At 0C, a gallon of water is about 8.33 Lbs or about 3.77 kg - 3770 grams. For a 1GP/M flow, 1/60th of that goes by in a second, so 63 grams go by each second. Cooling that by 1C requires about 260W. Ideally, the TEC above would drop the water temperature by about .5C - assuming the heat exchanger transfers the energy. Higher flow rates will do less transfer.

In comparison, the guts from a standard wall mount water cooler will cool 1 GPM by .5C using about 200W. so it is more than twice as efficient, but not nearly as pretty!

In answer to your question, the setup I pictured used the same CP1-12730, but operated at 12V and 22A. This is a more efficient area of the operating curve, as I move 100W with only 260W input. If you plan on using 2 TECs, you might find this to be a good choice - it powers from a standard PSU. for about 500W input you move 200W, which would get you close to a 1C drop at 1 GPM. The heat dissipated would be about 350W for each of the 2 units. The big Melcor sink would give a 25C rise with that load.

The attached CP1-12730 curves can be used to tell pretty much what this TEC will do - there are similar curves for most commercial TEC units.

[Duh]

Yeah, that's correct and a very sensible way of getting the most out of your chiller.

I wouldn't recommend trying to air cool the hotside. The reasons I'm air cooling my hot side is:

Lower cost,
lower noise (that's not likely but still possible),
smaller size,
to be different,
to prove that it can be done,
because my heatload is spread over a large area,
and because I don't think it will be much worse than water cooling if done carefully.

very nice points.. in a distant future I will try to do the same ( no money ATM)

I will be using about 20 x 40w undervolted and air cooled..I am still looking for a big and effective HS for the hot side..would love to run it fanless if possible. the only thing that is going to be a real killer is a decent psu

@uncle jimbo: you could have said 1 calorie instead of 4.18 joule :P.. make it simple for us :P.. nice info

[Uncle Jimbo]

well i would actually have to agree with you on this statement.

Because water is a better medium for transfer on the cold side. Also this is what i felt on the hotside. Im trying to milk the entire unit for efficiency. By watercooling the hotside, the cold side gets cooler, and i can use less voltage to get desired results. no?

The cold side will not get much colder because it is effectively in a water bath. But you are right, the TEC will move more heat for the same voltage with a lower differential, and water cooling the hot side will reduce the differential. However, TECs have a negative coefficient of heat - so they draw more current when the hot side is cooler. Using the curve from my previous post, and assuming a 12V supply, we can see that lowering the differential from say 30C to 10C will give a 30% increase in Qc W (100W to 150W). But lowering the hot side from 50C to 30C will reduce the effective resistance of the TEC from about .55 Ohm to around .4 Ohm, so instead of supplying 22A, we now need 30A.

So with the hot side at 50C, you supply 264W to move 100W, or 2.64W in per watt moved. With the hot side at 30C, you supply 360W to move 150W, or 2.4W in per watt moved. That buys you a slight increase in efficiency and moves more heat. But doing 2 TECs with air and a 50C hot side moves 200W for a 520W cost, again at 2.6 W per w moved. For a 9% increase in efficiency, you need to create a whole additional cooling system for the hot side WC.

Or if, as you state, you reduce the voltage, you could go to 10V which puts the amps at 25, and you are moving 100W with 250W in. That's about where you started, except you added the complexity of WC on the hot side, and you need a 10V supply, which is not off the shelf. And now you are at 2.5W in per watt moved, so it is only 5% more efficient.

The other issue you would face is that with a small differential, you will swing the amps a lot depending on heat load. That puts a big demand on the PSU... with a 25C or 30C differential, the amp swings are much lower for a change in load.

I like to design these to use an off the shelf $40 PC supply and not a $200 lab supply. OC work is expensive enough without going too exotic. Getting a supply that can deliver 30A at 12V continuous is a lot harder and more expensive than one that will do 22-25.

[Uncle Jimbo]

very nice points.. in a distant future I will try to do the same ( no money ATM)

I will be using about 20 x 40w undervolted and air cooled..I am still looking for a big and effective HS for the hot side..would love to run it fanless if possible. the only thing that is going to be a real killer is a decent psu

@uncle jimbo: you could have said 1 calorie instead of 4.18 joule :P.. make it simple for us :P.. nice info

Not trying to be arcane - you have to put the 4.18 in somewhere to get to watts... too bad, it's handy to have 1 cal = 1 degree.

BTW you might do better to use bigger pelts even if you want to undervolt - they are really just more elements per package anyway. Per my previous post, the best efficiency is with the hot side hot - draws less amps to get to the same point on the curve. If you put 2 of the TECs I was discussing in series across 12V, you will probably get about 11A draw and move about 35W per unit. That gets to 1.9W in per watt moved, which is about as good as it gets for TECs. I have never tried running a unit that low and the curves don't go down there, so it is something you would need to test. BTW just work around 12V as a design center and you'll find cheap PSUs that can do the job. Also a smaller number of bigger TECs makes finding heat sinks easier. With only 100W or so per unit, you could go fanless with a standard CPU sink I think.

[Scarlet Infidel]

The attached CP1-12730 curves can be used to tell pretty much what this TEC will do - there are similar curves for most commercial TEC units.

Unfortunately I am using ten TEC1-12709 elements. I can easily find graphs for TEC1-12708 and 12710 but not for mine. This caused me to look in more depth at the equations that are used to make the graphs. I prefer this to using graphs because its gives me more control and more accurate numbers.

You seem to know a fair deal about thermodynamics and peltiers so it still concerns me that you don't think you can get a peltier to move more heat that it consumes. I have seen it happen with my own test setup.

The only thing that I can think of, is that either you are used to using peltiers from a different perspective or that I've messed up somewhere. With all your examples you always seem to be using the peltiers at quite high towards their maximum voltage.

I and others have found that about 1/3 of the maximum voltage and below is where they really become useful. Have you ever used peltiers in this kind of range? If so, how did you feel they performed?

[Scarlet Infidel]

If you put 2 of the TECs I was discussing in series across 12V, you will probably get about 11A draw and move about 35W per unit. That gets to 1.9W in per watt moved, which is about as good as it gets for TECs.

If I look at the curves for the TEC-12730 which seems to be similar to the element you mentioned.

For a 35w heat load with a dT of 15c it will draw 6A at 4v. So 24W power consumption to move 35W of heat. Why does this seem so far off what you said?

The answer, Im fairly certain, lies in the fact that in yours the dT will be much higher. Perhaps around 40C from that graph (excuse any mistakes, I really don't like these graphs)?

[THE JEW (RaVeN)]

Replace the bolded link with this
http://www.kryotherm.ru/dir2attz/soft.exe

On an aside, does anyone have a guide to using the Kyrotherm software? I think ls7corvette is pretty familiar with it. A link to a guide or "dummies" manual might help us all start calcing out some info.

Maybe I should just pick up one of my mech friend's thermal books.

[Uncle Jimbo]

If I look at the curves for the TEC-12730 which seems to be similar to the element you mentioned.

For a 35w heat load with a dT of 15c it will draw 6A at 4v. So 24W power consumption to move 35W of heat. Why does this seem so far off what you said?

The answer, Im fairly certain, lies in the fact that in yours the dT will be much higher. Perhaps around 40C from that graph (excuse any mistakes, I really don't like these graphs)?

I've never used pelts down in the bottom of their operating range, and I am usually looking for a larger delta T (30C or so), so I don't have any direct experience with the operating range you are discussing. You have definitely caught my attention though.

Many valuable innovations came from people who 'didn't know any better' than to try something that a more experienced person would dismiss as unworkable. Experience is a two edged sword - the things you know can filter your thinking and blind you to what you don't know. The whole point of doing this 'out of the mainstream' work is to break that view of how things work, and come at it from a fresh perspective.

Like many semiconductors, pelts develop internal EMF which affects the effective series resistance, and without published data, the only way to determine that is by experiment. Just to get an idea of the base operating levels at zero delta T and low voltage, I took a CP1-12730 ((which is supposed to have the same performance as the TEC 12730) and put it between two Melcor HX5-301 sinks ( http://www.melcor.com/bonded.html ) which have .07 C/W at 102 CFM and .13 in free air, so they should effectively keep the differential at zero for these low wattages. I then applied 3.3V and 5V with a series resistor of .1 Ohm to measure the current accurately.

I found that your estimates are close for amps draw - 3.3V drew 3.5A, and 5V drew 4.6A. The pelt actually drew a little less current than your estimate.

The next question is how much heat is moved. To do that I will put a gallon of water initially at 25C in an insulated container, and put the cold side in it, while maintaining the hot side at 25C in air. The drop in water temp over time will give us the heat transfer. But that will have to wait for another day...

[Scarlet Infidel]

Glad you can see some merit in what I'm saying and are even willing to give it a try.

I would almost be offended by your statement about people "who 'didn't know any better' " except I entirely agree. It can be a pain when 9/10 of those people are not suggesting anything sensible and you start to just dismiss them.

However, I would like to point out that I didn't stumble on the idea of running lower voltages out of chance, I did a whole load of reading and computer simulations first.

I'm looking forward to your results. I did a similar experiment with a water loop on the cold side to measure how quickly the water temp dropped. However mine was not as controlled as Id have liked and I had the added variable of the pumps heat load.

[Uncle Jimbo]

Glad you can see some merit in what I'm saying and are even willing to give it a try.

I would almost be offended by your statement about people "who 'didn't know any better' " except I entirely agree. It can be a pain when 9/10 of those people are not suggesting anything sensible and you start to just dismiss them.

However, I would like to point out that I didn't stumble on the idea of running lower voltages out of chance, I did a whole load of reading and computer simulations first.

I'm looking forward to your results. I did a similar experiment with a water loop on the cold side to measure how quickly the water temp dropped. However mine was not as controlled as Id have liked and I had the
added variable of the pumps heat load.

no offense meant at all - the 'experienced' people are often the first to dismiss a potentially good idea. I was only saying that many - maybe a majority - of innovative ideas come from people who are not 'experts'. For example almost all of the great discoveries of the 19th century were done by people who were not 'experts' but just had an interest in the problem, and the energy to go after a solution.

I will test your low voltage approach and give you some feedback, but can't do it for a few weeks as I am traveling...

[Scarlet Infidel]

No problem, enjoy yourself.

[desone]

Just one small suggestion. You have done an excellent job there. With so many peltiers you would have had no option other than to go for a heat exchange system using water on both sides. I am using two 68w peltiers with an air cooled system on the hot peltier side.

You will get some heat transfer between the hot and cold side with the clamping stainless steel bolts. My advice would be to go for slightly smaller diameter bolts and use transistor insulating mounts (plastic) to insulate the two sides. You could eliminate the heat transfer almost entirely.

These are the type of insulators I am referring to.

http://s263.photobucket.com/albums/i...e01/screws.jpg

You could also use nylon bolts and nuts.

[desone]

Just one more piece of advice. Don't run the peltiers without the pump running.

That waterblock will freeze very quickly, as there is only a small volume of water. Remember ice expands and can easily damage your cold waterblock.

You should keep the cold side water volume in the system to a minium. That way it will be more efficient.

If you keep good airflow inside your computer case, condensation will not be a problem.

You could run the water through say a 120 mm radiator and fan and use it to cool the air inside the whole case. That way everything is cooled down.

[Uncle Jimbo]

Just one small suggestion. You have done an excellent job there. With so mant peltiers you would have had no option other than to go for a heat exchange system using water on both sides. I am using two 68w peltiers with an air cooled system on the hot peltier side.

You will get some heat transfer between the hot and cold side with the clamping stainless steel bolts. My advice would be to go for slightly smaller diameter bolts and use transistor insulating mounts (plastic) to insulate the two sides. You could eliminate the heat transfer almost entirely.

These are the type of insulators I am referring to.

http://s263.photobucket.com/albums/i...e01/screws.jpg

You could also use nylon bolts and nuts.

Nylon and plastic are OK as long as everything stays cold, but not if the pelt ever gets hot - the nylon will not hold torque. Fiber washers and stainless steel bolts are a better choice - the SS does not transfer much heat. I use 3mm SS. Take care to apply torque evenly and not too much - the pelt can crack if one side is tightened before the others. A neoprene gasket will help to keep radiant transfer between hot and cold side down. But if the temp differential is small, loss from those factors is not really a factor.

[Uncle Jimbo]

Just one more piece of advice. Don't run the peltiers without the pump running.

That waterblock will freeze very quickly, as there is only a small volume of water. Remember ice expands and can easily damage your cold waterblock.

You should keep the cold side water volume in the system to a minium. That way it will be more efficient.

If you keep good airflow inside your computer case, condensation will not be a problem.

You could run the water through say a 120 mm radiator and fan and use it to cool the air inside the whole case. That way everything is cooled down.

Good point about freezing the block - if you are using a remotely controllable PS (like a converted ATX PS) you can switch the TEC supply on when the pump runs, and it is fairly easy to also shut it off if the cold side temp gets below a safe level.

[NaeKuh]

I like to design these to use an off the shelf $40 PC supply and not a $200 lab supply. OC work is expensive enough without going too exotic. Getting a supply that can deliver 30A at 12V continuous is a lot harder and more expensive than one that will do 22-25.

You see im the oposite.

When i got something in my head that i want, i go out and get whats needed.

The only thing im regretting right now is that i didnt grab the 600W meanwell like nol told me to. Now looks like i might need to grab another 320W meanwell.

Okey anyhow from my understanding and thank you jimbo, you've been an valuable resource to me in the past 2 threads, you think i should keep the hotside on air?

Let me rephrase this question after you know some background.

1. i have more excess h2o crap then air crap. So if i was to go out and get sinks, i would need to buy everything from ground up. The blocks are custom by Martin, so i would need to send out a block to a millist, and have him mill me a counter top where i can sink my screws in. Ummm... sounds like too much work.

2. The pump i designated for this system is still unknown. Most likely a DDC-3.2 for the cold side. The hotside i already picked up the DB-1. Since the loop was to be small to begin with, if your saying i want the hotside at a higher delta, i could always connect it to an MCR220. Kinda lost on this. But if your saying i want to keep a 50C on the hotside, then i figure MCR220 with a heat load of around 300-400W will probably make my coolants ~ 35-40C on a small radiator like that.

Blah, if i need to go out and get the sinks, man... No comment. :\

[n00b 0f l337]

Okay sorry for lack of pics but camera trouble. So far though, I'm not seeing any benefit at all from the Tec chiller when in parallel. (As in, liquid pumped through hot side then rad on one half of a loop then hopping into resevoir, and other side going pump through cold side to load tester and back). And I'm not even applying load to the load tester yet. Cool side gains 1-2C, hot side gains 3-5C, and I'm sure that will get worse with a load.

[NaeKuh]

Okay sorry for lack of pics but camera trouble. So far though, I'm not seeing any benefit at all from the Tec chiller when in parallel. (As in, liquid pumped through hot side then rad on one half of a loop then hopping into resevoir, and other side going pump through cold side to load tester and back). And I'm not even applying load to the load tester yet. Cool side gains 1-2C, hot side gains 3-5C, and I'm sure that will get worse with a load.

yeah i could of told you that.

I seriously give the MCR320 new respect after this project.

Nol, run it straight like it originally should of been designed as.

[n00b 0f l337]

Yep thats up next

[Uncle Jimbo]

You see im the oposite.

When i got something in my head that i want, i go out and get whats needed.

The only thing im regretting right now is that i didnt grab the 600W meanwell like nol told me to. Now looks like i might need to grab another 320W meanwell.

Okey anyhow from my understanding and thank you jimbo, you've been an valuable resource to me in the past 2 threads, you think i should keep the hotside on air?

Let me rephrase this question after you know some background.

1. i have more excess h2o crap then air crap. So if i was to go out and get sinks, i would need to buy everything from ground up. The blocks are custom by Martin, so i would need to send out a block to a millist, and have him mill me a counter top where i can sink my screws in. Ummm... sounds like too much work.

2. The pump i designated for this system is still unknown. Most likely a DDC-3.2 for the cold side. The hotside i already picked up the DB-1. Since the loop was to be small to begin with, if your saying i want the hotside at a higher delta, i could always connect it to an MCR220. Kinda lost on this. But if your saying i want to keep a 50C on the hotside, then i figure MCR220 with a heat load of around 300-400W will probably make my coolants ~ 35-40C on a small radiator like that.

Blah, if i need to go out and get the sinks, man... No comment. :\

I think you should use what you have around if you can. I have a 600W meanwell that I use for testing 24V stuff. But if you are planning on running 12V (i.e. 16V TEC), a regular ATX PSU does a fine job for a lot less money.

You can use WC on the hot side - you will lose a little efficiency but you will get more out of the TEC.

I'll look over the previous posts again and see if I can give you some advice on layout and design. Do you already have the TECs you want to use? That would help...

[NaeKuh]

I think you should use what you have around if you can. I have a 600W meanwell that I use for testing 24V stuff. But if you are planning on running 12V (i.e. 16V TEC), a regular ATX PSU does a fine job for a lot less money.

You can use WC on the hot side - you will lose a little efficiency but you will get more out of the TEC.

I'll look over the previous posts again and see if I can give you some advice on layout and design. Do you already have the TECs you want to use? That would help...

There 90W TEC's if my memory serves me. And theres 5 of them.

The block is uber free flowing. Placed right along with DD MPC Universal.

Nol is doing all the testing needed. :T Thats why he has it. After he figures out the best way to use it, thats how i will adopt it.

[n00b 0f l337]

Gotta switch setup, hopefully have time today but def will have testing done by this weekend I hope.