Peltier (TEC) + Liquid Cooling Project

[RnRollie]

Is this setup still assuming that the TEC polarity will be reversed when signaled to do so? If not, which is the "real cold side," and how will the TEC heat up the water to "body temp" if the waterblock is on the "cold" side of the TEC?

Found a Swiftech MCP35X lying around! (That saves me a 100 bucks heh.)
Added a water block to my Parts List

-Water block: http://www.performance-pcs.com/cata...oduct_info&cPath=59_971_294&products_id=30144
-Radiator: http://www.performance-pcs.com/cata...t_info&cPath=59_457_667_670&products_id=21112
-Pump: Swiftech MCP35X
-1/8" ID Tygon Tubing
-1/8'' ID Copper pipe: http://www.homedepot.com/p/Cerro-1-...943611&cm_mmc=CJ-_-3754110-_-11210757&cj=true
-1/8'' barb x 1/4'' Male iron pipe size (4 of these)

i think it can be done without reversing the polarity. Either way, the watrblock remains on the hot side. It's easier

Use Radiator cooling to drop the loop temps back down to room temp.. should be fast enough. Switch TEC ON for heating, with rad fans off/low, heating should be fast. With rad fans on high, should be able to keep it steady.. if we can match the TEC & RAD. OR... with a variable voltage PSU for the TEC. Then it just turning the knob to choose the temp
If the TEC is "small" enough then a Model Train Transformator might be an option.. But TECs tend to draw some serious amps, even at only 12V. Need the TEC-specialists for exact figures.


HOWEVER, most TECs are 40mmx40mm, you NEED a waterblock with a base of at least 40x40mm, the block you listed is 15x15mm , which is fine enough if a small TEC of that size with wattage is available. I dont know where exactly the size change comes in. Most 4-5 watt TEC are 15x15mm. Where are the TEC specialists when you need them?

 

[FNT-Puck]

Few random thoughts, sorry if its all out of order...

What response time do you need? Using TECs as chillers are more efficient, but there is a delay between heating/cooling and reaching potential temps depending on the heat load and volume of liquid being heated or cooled. Direct Die TEC setups are not as efficient, but can have measurable temperature differentials almost instantly. They are also simpler to make.

Needing heating AND cooling will normally be an issue, since you don't want a radiator on the cold side, but if your cold side goal is only 25c, that will not be as big of an issue since as long as your ambient is 25c or less it will work great. You can leave the fans on for the cold side to keep it from dropping below ambient, and slow them down to heat up to 36c.

As for controlling the TECs, PWM is the recommended way to go, especially if you have an arduino. Playing with voltage creates tons of waste heat, so it is best to spec the TEC to run efficiently in a feasible voltage range, then use PWM to adjust duty cycle. Optimum voltage for cooling potential:heat released is usually around half max rated voltage - for CPU use we usually use 24v TECs at 12-16v. No clue how you will implement PWM and also have the ability to switch polatiry...I'm sure its possible, but don't have enough experience with arduino's to say how.

Will you be able to connect this to a PC, or will it need to be completely portable? IF you can connect it to a PC, you can hook up the arduino via USB and use software to control the TECs with a probe on each plate of it for monitoring.

Your temp ranges are not extreme, but I still suggest using AS Ceramique 2 over AS5. It lasts longer at cold temps, and takes heat cycling better then most pastes.

 

[RnRollie]

You know.. there is one important question i forgot to ask...: Pocket depth ?


Did you get any feedback from the TEC specialists? I got some already from one of the great guys; i'll get into it later. It's w-e; we'll have to wait till monday/tues to see if more info comes in.. not everybody dwells on this board 24/7

Anyways, we'll have to power this contraption. Don't know what you have available/laying around. But an (old) PC PSU will do... i'm guessing something like 250 Watt will do, probably less. Needs to be at least 100W though, what with capacitor aging & such.

You know, i suggested some simple On/Off switching - basically something looking like this: http://www.frozencpu.com/products/7...ary_Switch_Baybus_-_Silver.html?tl=g47c17s817
But 4/5 simple on/off switches is easy enough to build oneself.

But, from the feedback on the tec, might be able to use a "simple" fancontroller. Which would be great, as it allows for much finer control in finding the "sweet spot".
Mind, not one of those cheap&nasty fan controller though.

As for tec.. it looks like the 15x15mm waterblock you're thinking off is not going to do it. It will probably going to need a 40x40mm TEC, so you'ld need a 40x40mm (or larger) waterblock for the hot side. But these are the dimensions of a "standard" WB nowadays anyways.
I think the EK Supreme LTX is the cheapest on the market.

For the "cold" side of the tec, to keep it from freezing, any CPU aircooler will do. Even the thing that came stock with the CPU As long as it can be mounted to the WB, it's ok. In order to avoid needless fiddling, if you've got an old AMD aircooler around, then its probably easier to get the WB with AMD mounting bracket.

Later

 

[FNT-Puck]

With any TEC setup, your first step should be calculating your heat load so you can spec a proper sized TEC.

From there we can determine your cooling needs and power requirements .

 

[RnRollie]

With any TEC setup, your first step should be calculating your heat load so you can spec a proper sized TEC.

From there we can determine your cooling needs and power requirements .

The load is zero ... or to be more precise ambient room...20-25 °C

The TEC hot side must raise loop temp to 30-35-max 40°C relatively fast

So, it probably a relative "small" TEC.. the question is.. which one

 

[myom]

What response time do you need?

The goal is maintaining 34C temps with heating, then dropping to an ambient 25C for maybe 30 seconds, and then back to 34C.

Direct Die TEC setups are not as efficient, but can have measurable temperature differentials almost instantly. They are also simpler to make.

"Direct Die setup is where the hot side of the TEC's is water cooled and the cold side touches the CPU directly via a cold plate." - Is that what a direct die setup is? I'm not sure I'll be able to heat up the water with this kind of setup?

Will you be able to connect this to a PC, or will it need to be completely portable? IF you can connect it to a PC, you can hook up the arduino via USB and use software to control the TECs with a probe on each plate of it for monitoring.

Currently testing with arduino + software control and a thermistor at the "site of interest" aka the fingertips.

You know.. there is one important question i forgot to ask...: Pocket depth ?

What do you mean by pocket depth? The depth of the fingers maybe?

Anyways, we'll have to power this contraption. Don't know what you have available/laying around. But an (old) PC PSU will do... i'm guessing something like 250 Watt will do, probably less. Needs to be at least 100W though, what with capacitor aging & such.

Right now, I'm not too concerned with getting the best response time/temp ranges. Just need a proof of concept that I can heat/cool the fingertips using water cooling and a TEC.

You know, i suggested some simple On/Off switching - basically something looking like this: http://www.frozencpu.com/products/7...ary_Switch_Baybus_-_Silver.html?tl=g47c17s817
But 4/5 simple on/off switches is easy enough to build oneself.

But, from the feedback on the tec, might be able to use a "simple" fancontroller. Which would be great, as it allows for much finer control in finding the "sweet spot".
Mind, not one of those cheap&nasty fan controller though.

Hoping that everything can be digitally controlled without any manual switches

As for tec.. it looks like the 15x15mm waterblock you're thinking off is not going to do it. It will probably going to need a 40x40mm TEC, so you'ld need a 40x40mm (or larger) waterblock for the hot side. But these are the dimensions of a "standard" WB nowadays anyways.
I think the EK Supreme LTX is the cheapest on the market.

For the "cold" side of the tec, to keep it from freezing, any CPU aircooler will do. Even the thing that came stock with the CPU As long as it can be mounted to the WB, it's ok. In order to avoid needless fiddling, if you've got an old AMD aircooler around, then its probably easier to get the WB with AMD mounting bracket.

I'll have to buy a fan (don't have anything lying around) Also, could you clarify where the fan/TEC/waterblocks are going? Is this fan separate from the radiator component? I was thinking the loop would be like:
Pump > Hand Loop > Rad(+fan) > TEC (waterblock on hot side, heat sink on cold side) > Pump

Did you mean that the fan would attach to the cold side of the TEC (and not a heatsink)?

With any TEC setup, your first step should be calculating your heat load so you can spec a proper sized TEC.

From there we can determine your cooling needs and power requirements .

Yup, as RnRollie said, there is no load. I'm not cooling a device (like a PC); I'm looking to heat/cool a thermochromic material. Right now I'm going for a proof of concept to simply heat/cool a water loop that can transfer heat/cold to a "site of interest" (the fingertips).

 

[myom]

I just realized how energy inefficient it would be to maintain ~36C using the TEC then turning it off temporarily to bring it to ambient.

CHANGE OF PLANS: I want to maintain ambient and turn the TEC on temporarily to cool the loop to ~17C (I was able to use this small TEC to cool a thermistor from ambient to ~17C). So now the TEC is only doing cooling. Does this change anything?

Loop: TEC(cold side) attached to WB > pump > copper hand loop > rad > TEC+WB

A copper heat sink will be attached to the hot side of the TEC.
Could I use the rad to dispel "cold energy" to bring the loop back to ambient?

 

[RnRollie]

I just realized how energy inefficient it would be to maintain ~36C using the TEC then turning it off temporarily to bring it to ambient.

CHANGE OF PLANS: I want to maintain ambient and turn the TEC on temporarily to cool the loop to ~17C (I was able to use this small TEC to cool a thermistor from ambient to ~17C). So now the TEC is only doing cooling. Does this change anything?

Loop: TEC(cold side) attached to WB > pump > copper hand loop > rad > TEC+WB

A copper heat sink will be attached to the hot side of the TEC.
Could I use the rad to dispel "cold energy" to bring the loop back to ambient?

I think the consensus is "not in this universe". The laws of physics kinda frown upon that idea creating "cold" is extracting heat ... creating heat is not "extracting cold"

You can use the rad to warm up the cold loop again.
This is where the fan + switch comes in. The fan(s) are attached to the radiator. When fan is OFF , the rad must use natural convection to add warmt to the loop (or dump heat to air in the previous setup); this is a relative slow process.

When switching on the fan(s) on the rad, it becomes forced convection and will be a lot faster to bring the loop back UP to ambient (or in previous setup to bring the loop back DOWN to ambient).

The previous setup was based on maintaining ambient and "heat up" on demand.
The "new" setup is based on on maintaining ambient and "cool down" on demand.
Either setup is of course less energy consuming as "keeping the loop at body temp" which i kinda ignored you had in mind then


The previous setup was "easier" and actually more energy efficient because there was "no load" on the tec. And the heat produced by the tec was used to warm up the water.
This can be done with a relative "small" tec

If you turn it around, the "cold produced" (actually heat extracted) by the tec is used to cool the water... this costs more energy as heating strangely enough.
But the kicker is, when you switch on the tec, you now have a load -ambient temp- in order to drop the loop below ambient, you'll need to "pump" more power/amps/watts through the tec, you'll probably need a "bigger" tec and you'll have to get rid of the ambient temp PLUS the heat produced by the tec itself on the hot side. You may be looking at 50-90°C on the hot side of the tec... a simple aircooler might struggle with that.

But... all needs to be calculated... hope to get some more TEC guys chiming in.



All being said, for control, it might be easier to get an "off-the shelf" TEC controller. Mind, you'll still need the loop & pump & heatsink(s) & psu & etc... but control will be more precise, which means less guessing on which tec. And it allows polarity change. Witha turn of the dial you set the temp (hot or cold) and with flip of the switch you turn on/off the fan(s) on the radiator.
Simple, "manual" controllers start below $100, the more advanced become a bit more spicy http://www.ovenind.com/bv/Products/5R7-350A-Temperature-Controller__5R7-350A.aspx


The "pockets deph" question is actually about budget...
anything to do with TECs tend to turn into bottomless money pits quickly.
So, without actually naming figures... you'll have to be prepared to sink more as $9.99 into this proof of concept

.

 

[myom]

The previous setup was based on maintaining ambient and "heat up" on demand.
The "new" setup is based on on maintaining ambient and "cool down" on demand.
Either setup is of course less energy consuming as "keeping the loop at body temp" which i kinda ignored you had in mind then

Hahahaha I'm glad you ignored that idea.

If you turn it around, the "cold produced" (actually heat extracted) by the tec is used to cool the water... this costs more energy as heating strangely enough.
But the kicker is, when you switch on the tec, you now have a load -ambient temp- in order to drop the loop below ambient, you'll need to "pump" more power/amps/watts through the tec, you'll probably need a "bigger" tec and you'll have to get rid of the ambient temp PLUS the heat produced by the tec itself on the hot side. You may be looking at 50-90°C on the hot side of the tec... a simple aircooler might struggle with that.

Unfortunately, the thermochromic materials I'm working with become "activated" (turn blue) at cooler temperatures, so I will have to keep the default state at ambient and activate cooling with the TECs. Or maybe I can ignore the TEC altogether and just use liquid-cooling to achieve that? And then use a radiator to warm the water back to ambient?

All being said, for control, it might be easier to get an "off-the shelf" TEC controller. Mind, you'll still need the loop & pump & heatsink(s) & psu & etc... but control will be more precise, which means less guessing on which tec. And it allows polarity change. Witha turn of the dial you set the temp (hot or cold) and with flip of the switch you turn on/off the fan(s) on the radiator.
Simple, "manual" controllers start below $100, the more advanced become a bit more spicy http://www.ovenind.com/bv/Products/5R7-350A-Temperature-Controller__5R7-350A.aspx

I've already got a TEC set up, PWM controllable via Arduino

The "pockets deph" question is actually about budget...
anything to do with TECs tend to turn into bottomless money pits quickly.
So, without actually naming figures... you'll have to be prepared to sink more as $9.99 into this proof of concept

Ah I see haha. I can shell out $100-200 even though it's mainly for proof of concept.

Did some math to calculate the minimum thermal conductivity to cool the hot-side of the TEC for my current setup: Need theta = 0.24333 W/cmK.

Now I'm no thermodynamics expert, but this seems achievable with a passive copper heat sink?

 

[RnRollie]

Did some math to calculate the minimum thermal conductivity to cool the hot-side of the TEC for my current setup: Need theta = 0.24333 W/cmK.

Now I'm no thermodynamics expert, but this seems achievable with a passive copper heat sink?

i guess so , since copper gets only beaten by silver & Ub120 as far as thermal conductivity in solids is concerned
Diamond & Graphene is not be included because of the need to get oriented "laticces"




Weeelll.. to abandon the TEC completely... a (largish) tube reservoir on top of the DDC pump ; open lid ; drop in icecube (or two) ; wait

Two issues with that
1. use "plastic" icecubes (like gelpacks) OR freeze distilled water in order to avoid loop gunking up
2. dropping temps by dropping icecubes in the res is not a "fast" cooling process

Although, with some experimenting, you can work out how much and how long in advance of the effect to drop.

 

[myom]

Weeelll.. to abandon the TEC completely... a (largish) tube reservoir on top of the DDC pump ; open lid ; drop in icecube (or two) ; wait

Two issues with that
1. use "plastic" icecubes (like gelpacks) OR freeze distilled water in order to avoid loop gunking up
2. dropping temps by dropping icecubes in the res is not a "fast" cooling process

Although, with some experimenting, you can work out how much and how long in advance of the effect to drop.

Ah I see.. yeah dropping ice cubes is way too inconvenient.. the hope is that once the setup is complete, everything can be controlled with Arduino I guess I'll be sticking with the TEC.

 

[FNT-Puck]

You can indeed use a rad to warm up a loop. Hot and cold are all just movement of energy...if you have a rad in your loop and the loop is colder then ambient, the rad will bring it to ambient. If the loop is higher then ambient, the rad will attempt to bring it down to ambient.

With TECs in the small size we are talking, copper heatsinks will be fine on the hot side - something like an array of ~4 small TECs with northbridge heatsink/fan combos may work if its a small amount of coolant to be chilled and no heat load besides the ambient - which is still a heatload, since it still takes XX watts to chill YY amount of ambient coolant, even without any heat load. The more coolant, the more watts needed to chill in the same time frame. A TEC setup will work. In some ways this is much easier then a PC TEC loop since it is all low power draw and very low cooling requirements, but in others more difficult since you have size constraints to worry about.

If it were my project, I would use a single ~40mm TEC at low voltage, with a heatsink on the hot side. The cold side would use a standard CPU waterblock(can be an old dirt cheap used one, as long as the base is 40mm it wont matter) run to a small pump and some 1/4" tubing arrangement, then have 5 small "waterblocks" to act as cold plates - one for each finger. Insulate the tubing with some thin walled neoprene tubing, and you're good to go. Simple and effective. The 5 cold plate blocks can be as simple as small plastic T fittings with a copper plug on the bottom - just something to transfer the chilled loop to a specific location.

Didn't have time to make a proper pic since I have to leave, but here is a quick paint to get the idea...the heatsink/TEC/waterblock can be kept in a small backpack or fannie pack on the hip, and the pump can be left inline if its small enough or relocated with the heatsink/tec combo if its too large.



With the TEC turned OFF or very low duty cycle, the plates will be ambient. Once you turn them on, the coolant will be chilled. You can add a single temp probe anywhere in the loop and a second to get the ambient to control the TEC.

 

[myom]

You can indeed use a rad to warm up a loop. Hot and cold are all just movement of energy...if you have a rad in your loop and the loop is colder then ambient, the rad will bring it to ambient. If the loop is higher then ambient, the rad will attempt to bring it down to ambient.

That clarifies rads. Thanks!

If it were my project, I would use a single ~40mm TEC at low voltage, with a heatsink on the hot side. The cold side would use a standard CPU waterblock(can be an old dirt cheap used one, as long as the base is 40mm it wont matter) run to a small pump and some 1/4" tubing arrangement, then have 5 small "waterblocks" to act as cold plates - one for each finger. Insulate the tubing with some thin walled neoprene tubing, and you're good to go. Simple and effective. The 5 cold plate blocks can be as simple as small plastic T fittings with a copper plug on the bottom - just something to transfer the chilled loop to a specific location.

Does the copper wire loop not seem plausible? Or would these T-fittings with copper plugs be more effective you think?



Wouldn't I need a radiator to heat up the water after it's been cooled? Thanks a bunch for your input!.

 

[FNT-Puck]

You will not be able to get sub ambient with a radiator in the loop and those small TECs. Once you turn off the power to the TEC though, the heatsink+fan that cools the hot side will start warming the cold side since there will be no more heat output.

If you need faster switching from cooling to heating, you can just reverse the polarity on the TEC and it will heat up FAST.

Keep in mind a small loop like this with no res and low volume will heat and cool very quickly.

 

[RnRollie]

i suggested the radiator as a "mechanical" safeguard against overshooting.
When the fan(s) on the Radiator are OFF, then the loop can only slowly return to ambient (from cold/hot) because it has to be done through natural convection/exchange.
When the fan(s) on the Radiator are ON, then the loop can return much faster to ambient (from cold/hot) because it is be done through forced convection/exchange.

OF course, if arduino programming can set high/low safeguards and provide for the TEC polarity switch, then one can leave out a Radiator.
I would leave it in until i'm 100% sure of my programming & safeguards

The coldplate fingertips is not a bad idea... but it would likely involve "extra plumbing" (possibly under a magnifying glass) and thus introduce extra fail points.

Of course, maybe you have some specific blocks/Ts/things/gizmos/... in mind that will work a lot better. Please elaborate on that if you will, thanks.

 

[myom]

You will not be able to get sub ambient with a radiator in the loop and those small TECs. Once you turn off the power to the TEC though, the heatsink+fan that cools the hot side will start warming the cold side since there will be no more heat output.

If you need faster switching from cooling to heating, you can just reverse the polarity on the TEC and it will heat up FAST.

Keep in mind a small loop like this with no res and low volume will heat and cool very quickly.

^That's good news! The less parts, the cheaper and the more space-efficient

i suggested the radiator as a "mechanical" safeguard against overshooting.
When the fan(s) on the Radiator are OFF, then the loop can only slowly return to ambient (from cold/hot) because it has to be done through natural convection/exchange.
When the fan(s) on the Radiator are ON, then the loop can return much faster to ambient (from cold/hot) because it is be done through forced convection/exchange.

OF course, if arduino programming can set high/low safeguards and provide for the TEC polarity switch, then one can leave out a Radiator.
I would leave it in until i'm 100% sure of my programming & safeguards

The coldplate fingertips is not a bad idea... but it would likely involve "extra plumbing" (possibly under a magnifying glass) and thus introduce extra fail points.

Of course, maybe you have some specific blocks/Ts/things/gizmos/... in mind that will work a lot better. Please elaborate on that if you will, thanks.

I'm going to trust that I can set good safeguards for overshooting temps
And yeah, I'm wondering where I can find such T-fittings with copper plugs. Could a T fitting that is entirely made of copper work too?

EDIT: I will be modifying an existing baby-sized mannequin leg that's a little longer than 1ft. (switched from fingers to toes!), and there is no need to cool/heat each toe individually. They are small enough that I can cool/heat all of them with a rectangular water block that spans the width of the foot.

 

[myom]

Question about fittings:
http://www.performance-pcs.com/cata...oduct_info&cPath=59_971_294&products_id=30145

^Is that a built-in fitting sticking out of the block? Or do I need to buy a barb fitting? And the specs say Tubing size: 4mm ID, so does that mean only 4mm ID tubes will fit? I've got 1/8'' ID tubes.

Also, is the below connection complete? No clamp necessary?

 

[RnRollie]

thats indeed a build in fitting

depending on the quality of the hose... it can be stretched over a 1/4"... need to put hose end in near boiling water before slipping/forcing over the barb



The DCC does have some head, with this thin hose who knows how much pressure is created..
if you want to be sure, some jubilee clips are advised

 

[myom]

thats indeed a build in fitting

depending on the quality of the hose... it can be stretched over a 1/4"... need to put hose end in near boiling water before slipping/forcing over the barb

Luckily this fitting is only 0.03248'' bigger than a 1/8''

The DCC does have some head, with this thin hose who knows how much pressure is created..
if you want to be sure, some jubilee clips are advised

What's DCC? Does that refer to the liquid? Also, can't seem to find 1/4" OD clamps..

Edit: Found this clamp
Product description: These reusable clamps are designed for 1/4" ID, 3/8" OD tubing. Ideal for Swiftech PVC tubing and various 1/4" OD tubing systems. Black color.
^Is that a typo that meant to say 1/4" ID or 3/8" OD?

 

[RnRollie]

sorry DDC also known as MCP35x

Jubilee clips, in any HW store Home Depot et all

The plastic re-usables have mixed feelings, some people never had an issue, some people hate them
Jubilee clips are a proven technology for almost a century