Post Rad Chiller Concept

[NaeKuh]

Okey guys, you know i bought some time off martin to run his mill. He has so far designed this for me for my post rad chiller concept.

Im thinking it will work, but i could use some advice from people who are more experienced with this stuff.

This unit will have its own loop. Cold side will be connected to main loop after radiator and before the cpu. Hot side will be on its own loop with a dedicated PA160.

I was thinking intially 3x62W [1 meanwell S320 should handle this] TEC's. However martain can stretch the area wide enough to take 437W tec's if i ask him to.

Heres our wonderful martin's creation:







[/QUOTE][/QUOTE]



So lets hear it. How would i make this better without causing him too much work or is it ready to go?

[Cupcake]

I just think you need more powerful TECs

Im trynig something similar but with two 226w TECs cooled by apogee GTs and I havent decided what the cold side will be using. maybe fuzions since I have a few extra

[Martinm210]

I'm pretty rough on the watts and thermal side of things.

But if I figured this out right, it takes about 395watts of heat energy to change a 1.5 GPM flow rate by only 1 degree celcius.

I could be wrong, but this is what I used:

1.5gallons per minute.

1.5 gallons = 5,678 milliliters = 5,678 grams

Divide by 60 for one second

= 94.6grams water / second

q=(specific heat of H20)X(grams of H20) x dT
=(4.18 J/g-K)x(94.6 grams H20)x(1K)
=395 Joules

A Joule is a watt-sec and we were measuring for one second, so that 395 watts.

And that's assuming 100% efficiency.

Probably need to go with the bigger TECs?

That seems like alot of TEC to change the water temperature though...Maybe I figured something wrong..?

Yeah I think that's wrong, I'll go digging some more...

[NickS]

I'm pretty rough on the watts and thermal side of things.

But if I figured this out right, it takes about 395watts of heat energy to change a 1.5 GPM flow rate by only 1 degree celcius.

I could be wrong, but this is what I used:

1.5gallons per minute.

1.5 gallons = 5,678 milliliters = 5,678 grams

Divide by 60 for one second

= 94.6grams water / second

q=(specific heat of H20)X(grams of H20) x dT
=(4.18 J/g-K)x(94.6 grams H20)x(1K)
=395 Joules

A Joule is a watt-sec and we were measuring for one second, so that 395 watts.

And that's assuming 100% efficiency.

Probably need to go with the bigger TECs?

That seems like alot of TEC to change the water temperature though...Maybe I figured something wrong..?

Oh god, SHUT UP!

I just took a test on q=MCdT in chem yesterday. It's 2AM saturday morning, I'm not supposed to be reading this!!!

[Martinm210]

Oh god, SHUT UP!

I just took a test on q=MCdT in chem yesterday. It's 2AM saturday morning, I'm not supposed to be reading this!!!

Well you're all fresh on this, I've been out of school for WAY too long..getting old and crusty...

I did find what you just said though on this french radiator review article:

Q = Mx Cp x dT (Equation 1)

Q = power dissipated by the radiator in W (calculate)
M = mass flow of water kg / s (measured)
Cp = heat capacity of water J / kg K, it is 4186 J / kg K here (known)
DT = temperature difference between the entrance and exit of the radiator for Water in K (1 K = 1 ° C) (measured)

So at 1.5GPM, you need 395 watts to change the temperature 1 degree...that's ALOT!

But I'm sure once you've exceed the heat dissipated by the parts getting cooled it would continue to cool down until it reached the ambient then it would start stabilizing by the cold loop radiator?

[NickS]

Yeah but 395 joules doesn't equal 395w. 395 joules is 0.109722222 watt-hour. Whatever that means, I'm not sure.

[Martinm210]

Yeah but 395 joules doesn't equal 395w. 395 joules is 0.109722222 watt-hour. Whatever that means, I'm not sure.

I might be messing that up.

I assumed that since I figured out how much water flowed per second that seconds would cancel out. Maybe I'm missing something

.10972222 watt-hour = 395 watt-seconds

Would it mean 395 watts?

It's getting late...

[NickS]

Hm, maybe..

[Fujimitsu]

I like the idea, and the design.

But I'd go with beefier tecs. I'd rather have something capable of going much colder and back it down a bit, than have something that's too weak.

For what you're spending though, you could have a pretty sweet phase chiller... could have martin design a nice heat exchanger or something.

This is by far the most thought out attempt at a TEC chiller I've seen, and although I'm hesitant to attempt it myself because of the cost/complexity/performance.. I'm eager to see some knowledgable people give it a go.

[Martinm210]

It's sad, but I woke up this morning thinking about this and realized I'm wrong in how I was thinking about this.

While it does take 395 watts to change a 1.5 gpm water delta 1 degree, that's assuming there is no radiator at work. As if you had nothing in the system pumping water at 1.5gpm that was automatically getting reheated the exact same amount it's getting cooled, that's not what we have here.

To figure out the result, you'd have to figure out the heat being dumped in and the radiator in use.

If the heat going in from blocks and pumps is say 200 watts, and without the chiller the water to ambient temperature was 8 degrees. Even only removing 100 watts of that with a chiller should be good for almost a 4 degree improvement.

So if you only had 200 watts dumping in and used 395 watts for removal, you'd be below ambient with your coolant temperature and over 8 degrees cooler, not 1 degree.

A chiller with a radiator like this should be an easy way to get coolant temperatures down to ambient and a little lower. Since it's going to be hard to get below that because of the radiator, it would be an easy and safe way to avoid any of the typical TEC problems with condensation The 395 watts per degree doesn't really apply unless you matched that heat going in...eventually below ambient, the radiator would match that when it reached an equillibrium. The radiator inline on the coolant side will slowly start working against the effort as you dip below ambient. You could also figure that out just like radiator cooling. If you had an excess of 200watts going into the chiller than being dumped. a PA120.2 would dump that amount back in once you got down to about 10C below ambient.

You should see improvements in temperature even with smaller TECs where the radiator is still helping.
That's why my numbers weren't clicking for me, I wasn't thinking about the radiator.

But one example would be like this
250 watts in
PA 120.2 with nexus fans at 12V
This is probably off a little because of the curves used(based on 10C delta), but the normal water delta would be 12C without the chiller

If you put 3 62watt TEC in line, assuming 80% efficiency, you'd remove about 150 watts of heat.
In the end your net heat in is only about 100 watts, which would give you coolant temps that are only about 5C higher than ambient (7 degree improvement)

And that's all under full load, on idle you'd probably have less than the 150 watts being removed by the TECs, so you'd probably have coolant temps near or lower than ambient.

I think it's a neat idea, and a low risk/maintenance one too. No condensation to worry about because you'll never get block temps below the dew point, yet it will do a really nice job of getting coolant temps near or to ambient. You'd basically have all the same conveniences and easy installation of a regular water cooling setup, but along with an extra bump in TEC performance.

I'm changing my mind, I'd probably start with smaller TEC's, if you go too big, you'll probably get way down below ambient and may get into dew point/condensation territory.

[systemviper]

I love he idea, i need to think about it for a few but it's going in the right direction, i have been building a fantastic tec TORTURE rack but i have been trying to figure out how to use more tec;s to cool a resi or something similar to this. keep up the good work, i think with the right temp reduction vs energy used is the formula that makes it turn into a real system.

I have been looking and they have been making much beefier TEC's so the combo should have a sweet spot somewhere.

I have just about all the parts for my rig except for the DD Torture Rack, which should get here in 2 weeks, i hope this is farther along, i could i find some space for it,

Once again I have to say, Martin, your presence here and with the other minds throwing good idea;s around. this is Nirvana.. I got no desire to use LN2 so my only Extreme Chance is a Mother of a TEC rig.... and it's designed, 90% parts in and should be operational in about 3 weeks.

I can't get there fast enough, I also have to read more on TEC usage to make sure my vision can expand and Rock the Chip word....... lovin it......

Keep up the great work people!


Man i would love to have some of the tools you have, but then i would have to rush retirement and that just isn't an option

[n00b 0f l337]

Been wanting to get something milled though, got autocad's exactly like that.
What your going to want instead is 8 or so 80w tec's. That way you have a possibly of 640w or so and can tune that down. Wire each pair in series, and get a 18V 600W.

[NaeKuh]

Been wanting to get something milled though, got autocad's exactly like that.
What your going to want instead is 8 or so 80w tec's. That way you have a possibly of 640w or so and can tune that down. Wire each pair in series, and get a 18V 600W.

WOW...

thanks for all the replys and i thank martin for his effort as well.

I really want this to be a fun project for us. Martin is getting a shipment from me next week with a PA160, DDC-3.2, Meanwell S12-320 [hence why my TEC's are limited] EK-Xres.

Thats the cooling setup on the loop. I really liked this idea because if my TEC's ever fell, the Radiator i have would be my safety net. That loop can always be ignored and pretended to be a gpuless block.

Man.... but nol if you think i need 8 TEC's total on this unit... its going to be massive. im not looking at replacing the radiator like the coolit product. i just want a bump in performance without frying my breaker. I think serialkiller was correct in having more tecs downvolted.

Also if the water froze, that would present a very big problem to me. [Martin, dont worry about budget to get this done correctly. if needbe, i can dish another payment of = amount to get it done correctly.]

[have a feeling at this point nol will shake his head going you should of phased] But i think this will be more fun!


Man.... looks like i might get a dualy and side mount the meanwells on the other mobo try. Hopefully i can get this setup all done by my next birthday in May.

[Martinm210]

[Martin, dont worry about budget to get this done correctly. if needbe, i can dish another payment of = amount to get it done correctly.]

No problem, this little project of yours has my 100% attention, I might have to try something very similar.

In fact, I just bought a 158 watt pelt off ebay to play around with, I like this chiller idea

Keep me posted on when you're ready with a design to start milling. I already have the copper for 4" base, I just might need to order another wide piece of Delrin if you want a wider top. We could also go with a brass top if you wanted.

[Martinm210]

Also, I was curious,

What TECs are bigger than 62mm? Seems like most are 50mm, and 62 is the biggest.

The above concept was with three 62mm square TECs.

The biggest TEC I could find was this one:
http://cgi.ebay.com/Gigantic-62mm-54...QQcmdZViewItem
?

[Duh]

@Nick: your shut up comment really made me laugh...

@martin:
seems like a very nice idea although i cant tell if more powerful Tecs will be better or not..gues I had to think a bit and I am way too lazy

have not seen any bigger than 62 also.. IIRC those 545 watter tec had something that did not make em attractive ( cant say what it was :S but was said here )

[littleowl]

@ martin thanks for helping out on this project! The 62mm tec is the biggest size you will find that can be of any good to us. There are much bigger ones but they are also lower wattage. IMO I would make the hole thing to fit and cool 62mm tecs even if Naekuh only wants to use 50mm ones.

@NaeKuh it is great to finaly see you guys starting this project. like I told you before I think if you guys can make the fins inside a little higher to get better heat to water transfere.

I am not too sure I agree with nol on that statement. I think the 62w tecs will work ok but I would consider going with 80w tecs. or even bigger and under volt them.

[n00b 0f l337]

Sure the 62's will work okay, thats very true, tuned down in wattage and used again in large quantities will be even better then the 80's.

[NaeKuh]

Sure the 62's will work okay, thats very true, tuned down in wattage and used again in large quantities will be even better then the 80's.

so nol, what combo should i use then? 80x3 = 240W draw, which means i should be safe on a 320W.

the next higher tec i can think of is 120W ? having 3 of those would overshoot my meamwell no?

[interDOS]

Hi Guys, I’m very new to this forum and several other similar forums. I joined to get info on a project I’m planning to start designing in a few weeks time, and have read tons of comments on different ways to cool systems. This is an area I need to pin down because it would have a direct impact on the design of my project.

Just about everything I’ve read about Tec’s, state that they are a bad idea for 24/7 cooling, however I wasn't ready to give them up so set about trying to design a Tec system that didn’t double your electric bill, create condensation or killed the Tec by use of on/off power switching to control temperature.

Your design is one of many that I have sketched as a possible method, and will probably sketch many more, and for that reason I’ve been looking at how to accurately control the temperature of Tec’s.

The most efficient method I have found is to use PWM (Pulse Width Modulation) temperature controllers. I have viewed several PWM controllers, but found none that had all the circuitry I required, due to the designs being more suited to industrial use for lasers, ovens etc.

PWM is also an efficient way to control the speed of cooling fans.

You can see/learn more about Tec’s (faq's) and PWM (#9 on site page) of the following link
http://www.tetech.com/techinfo/

[THE JEW (RaVeN)]

Also, I was curious,

What TECs are bigger than 62mm? Seems like most are 50mm, and 62 is the biggest.

The above concept was with three 62mm square TECs.

The biggest TEC I could find was this one:
http://cgi.ebay.com/Gigantic-62mm-54...QQcmdZViewItem
?

That one and this one are about the largest easily available to computer enthusiasts:

http://www.frozencpu.com/products/24...tl=g30c105s187

There's been a big discussion on Thermal Enterprises TEC's and how they're wattage is based on draw and not cooling power. Though, the one you linked runs 0-16 volts which makes it easier to tune for some (ie. me I have a 12 volt 600 watt Meanwell so this find really intrigues me.......thanks for this )

The only larger ones I know of are in lab equipment (ie. laser coolers). They're hard to find. If you find a place that has them readily available and relatively cheap, you'd be the TEC forum's new best friend

[Pete]

Take it your having 2 of these made to sandwich between the tecs?

Make sure you get it done up very right along each part of the tec, maybe 4 mounts per tech.

Making the trasfer from the copper block to the path of the water is what will help, Mart knows this well so ya need to sort that i think.

437w tecs x 3 will be the best thing you can do i think. It'd cool more.

However you'd need more than a PA160 too cool the hot side!

It's going to eat lots of power too sadly

[THE JEW (RaVeN)]

For reference, you might want to look at Swiftech's older TEC chiller, the MCW-Chill 452:

http://www.procooling.com/index.php?...s&disp=84&pg=1

It was, pretty much, just two MCW5002's sandwiched over a couple of MCW5000 plates. However, it was fairly successful for the time. I think if the TEC's were tuned, it might be worth it.

It's an interesting concept either way

[Polizei]

Been wanting to get something milled though, got autocad's exactly like that.
What your going to want instead is 8 or so 80w tec's. That way you have a possibly of 640w or so and can tune that down. Wire each pair in series, and get a 18V 600W.

Why in series? There is going to be a voltage drop across them so its going to be hard to keep them at the same voltage. I vote parallel.

[NaeKuh]

Take it your having 2 of these made to sandwich between the tecs?

Make sure you get it done up very right along each part of the tec, maybe 4 mounts per tech.

Making the trasfer from the copper block to the path of the water is what will help, Mart knows this well so ya need to sort that i think.

437w tecs x 3 will be the best thing you can do i think. It'd cool more.

However you'd need more than a PA160 too cool the hot side!

It's going to eat lots of power too sadly

rofl pete!

@ that voltage and those tec's my case would be an AC! thanks to the radiator.