Ultrasonic2's TEC liquid chilled pc

[Ultrasonic2]

lol just found pics of my quad tec direct die pc from may years ago

[Duh]

Do I see a DFI lan party ultra B? :O

[Ultrasonic2]

Update time

IMG_0961.JPG

[flak-spammer]

Looks awesome, I'm really starting to think I should go liquid cooling. I just can't stomach the cost right now. Not considering a 4 1/2 year old system. But it's looking great. Looking forward to see the end result.

[Ultrasonic2]

done burger

IMG_0964.JPG
IMG_0965.JPG

Well almost i dont have a OS at the moment

[thePoison.]

Hello Ultrasonic2, I messaged you via your website to see if you could fabricate one of those nifty PWM Boosters, or if you could give me the schematics to assemble one. I currently am on that very costly adventure into water chilling, and it has been a pain in the neck. Here is my current setup that I am using. Yes it is an air cooled water chiller, I have been running these four Qmax 89.2 TEC's at 5v get's, 12 gets mighty hot, and have been considering changing it to 9 of them. As of the loop I run 2 580s, and an i7 2600k. The radiators are in place have their fans disabled, because well radiators kinda defeat the purpose of chilling the liquid. My Temps have been anything but desired. If you could please get back to me about fabrication of one of those amazing boosters @ thepoisonlord@gmail.com I would be highly appreciative. I have followed up on a lot of your work and it is quite impressive and has been the source of my inspiration since I have started my own. The copper block is 6x6x1 with plenty of surface area. The block is suspended so it does not come into contact with the aluminum heat sink. Most is insulated but the exposed side. What are the temperatures you achieve with your current dual water chiller setup? And what type of TEC's would you recommend for this setup? I have been looking at getting 9 50mm techs but Have not came up with a solid decision. The temps shown are best case scenario with the system being on for about 30 minutes on light load. With the radiator fans being off. Ambient was around 68 F, so a lot of that low level cooling is credit to the low ambient temperature :P

[Ultrasonic2]

hi there i didn't get any emails from you which site were you referring too?

I wouldn't bother with a PWM booster like i used to make it's a billion times easier to use a DC-DC SSR like this one http://jaycar.co.nz/productView.asp?...y&form=KEYWORD

you say you have 4 90 Qmax tecs to cool your 2 gpu's and your CPU ? Well that is NO where near enough Qmax for just the CPU let alone 2 gpus as well !!

you have a total of 360Q where you'll need at least 700 for the cpu alone. however i would never recommend 700 i would recommend about 1000Qmax for a CPU ( obviously undervolted )

Mine is 972Qmax and works. At 1/2 that 486 it's unable to achieve anything.(mind you my cpu is pulling more wattage i would think)

Now keep in mind my TEC's are water cooled which is WAY better than your air cooled system so to compensate for less efficient hot side cooling you'll want an even greater Qmax that is undervolted even more. if it was me i would be starting with about 2000Qmax

Why are aren't you water cooling the hot side of your TEC's as you obviously have a radiator to do so ?


As for my results i'll post that when the pc is booting again. but i will achieve my goal of 20 cold side water and 32-35c hot side water (ambient of 25c) at full load overclocked With heaps of head room.

[Duh]

a masterpiece . I love your stuff man.

Keep it coming. they are each time more impressive !!

[thePoison.]

2000Qmax after half leaving 1000 Qmax remaining or the other way? To be honest the design to water cool the hot side wasn't in the plans, but I am starting to lean towards that. Fund restrictions at the time you know how that can be :P The aluminum enclosure only cost 40 dollars to construct. In your opinion what and how many peltiers and what Qmax would you recommend to obtain a 20 C cold side water temperature? Is the DC-DC SSR what you are using right now? My knowledge of circuitry is limited to basic things, I can construct a schematic if given to me but putting together one in logic is a whole different story, I am a novice in that department. I severely appreciate your input Ultra. The physical dimension I have to work with is 152 mm square.

[Ultrasonic2]

First off give up Trying to air cool the TEC's, then buy TEC's that total about 1000Qmax then water cool both sides and it will work

If you want to cool the gpus as well then apply 2000Qmax

All at about 6 volts should be fine

The SSR and the PWM boost are dumb they require a PWM controller to tell them what frequency to run at, in my case that is provided by a T-balancer

[thePoison.]

I have already begin designs for a water block for the hot side as we speak kind sir, so for cpu + gpu + gpu, I need a total of 2000 QMAX. K got it. Doesn't the QMAX exponentially drop when you lower the voltage? I have a T-Balancer in hand, actually the exact same one you have. It is a nifty little device, and when I saw you made a mechanism for it to control virtually any size device I was stunned. It was a very interesting way of connecting via PWM.

[Ultrasonic2]

The 2000 is just an estimate no ones tried that before

No the Qc does not exponentially drop. Actually it’s the other way around. As you lower the input voltage the more efficient (greater the COP ) will become

So the T-balancer controls the SSR so you can control any voltage and any current. The ssr is what I use now and is much simpler

[thePoison.]

Would you happen to have a schematic of how you would have the T-Balancer communicate with the pelts?

[Ultrasonic2]

Would you happen to have a schematic of how you would have the T-Balancer communicate with the pelts?

I confirmed this morning that while the T-balancer shows a change in duty cycle the result appears to be off below 10 and on above that which is FAR from ideal and explains why everything is much louder than what it should be currently the load on the TEC’s is sort of exponential.

Tec’s on above 10% moves the heat to the hot side then below 10% the TEC’s are off and the heat moved to the hot side is conducted back to the cold side. So the TEC’s come back on and the TEC’s now have to move the cpu load plus the load conducted back. This results in an ever worsening condition.

To be honest I thought this might happen because of the Caps on the drain of the T-balancer so I’ll try applying a load across the T-Balancer

[thePoison.]

Let me know how it works for you Ultra! You are the man. I already got my CNC guy up and running blue prints for another 6x6 block.

[Ultrasonic2]

right so no amount of resisters or zennor diodes will help. it's back to plane -A . which is to remove the cap from the T-balancer that is causing the problem in the first place.

I have removed it and now it works much better now but not ideal

It now has 3 speeds instead of 2. I must admit I find it odd that it’s got only 3 speeds now. In a million years I might make a my own controller

[chaos_being]

Wow, those chillers and that cpu block you made are works of art...very impressive!

I was really facinated with TECs years ago when they seemed to be more in use, but I never had the cojones to try building a setup that used them.

[bds71]

/off topic somewhat

i have something in the works that some on this forum might be interested in. a 40A/480W "4" channel PWM controller (2x 10A channels with a single header output each and 2x 10A channels with 3 headers each). runs off 2x PCIe PSU plugs and could run a TEC on each 10A ch and pumps/fans on the other 2x3 channels with combined 10A/120W. all controlled by a manual temperature setting (via thermistor) with temp sensors feeding back info via loopback until you reach the "set" point. ideally adjustable from -70 to 70 (some details are still being worked - NOT finalized). it's a work in progress and i hope to hear more soon. i can't give any more information/details out of respect to the company in question, but as soon as i get the OK from the company i can maybe answer any questions. the information i've given so far is simply what i have **requested**. whether or not this will be possible is still open for discussion. more to follow.

question: does this sound like a product that the LC/TEC/OC community would be interested in?


/thread on

[Ultrasonic2]

so if i get that right your saying 240 watts for tec's ? if so that's not enough for current TEC blocks or chillers.

[thePoison.]

Ultra could you throw me an e-mail of what your current SSR looks like on your bread board to your t-balancer, throw me an email at thepoisonlord@gmail.com if you could, i would really appreciate it. I also need to find a good place to get a hold of some higher powered tecs, ebay is not ideal for this :P

[bds71]

so if i get that right your saying 240 watts for tec's ? if so that's not enough for current TEC blocks or chillers.

a module has (would have had) a 10A channel for "a" TEC and another 10A channel with three outputs for fans and pumps. but, the deal fell through. the vender wanted me to front the cash to develop the product - which i don't have, and would not have given even if i did have it. it was meant to be usable by a regular 1000-1500W PSU rather than an external and would have come with two 20A modules powered by two 6-pin PCIe connectors (the same ones that power modern GPUs - each being 20-30A depending on PSU rating) - all PWM controlled from 5-12V.

i'm still learning about "what's enough" and what isn't. from what i understand, it is a common practice to get multiple (4 seems to be the popular number) overlarge TEC's and under volt them to run near their 25% range for maximum COP, and then run them in series to increase the dT. doing this will make the TEC's more "efficient" (ie: more cooling power per input volts/amperage) and would consume less current per TEC as well. but i think 10A per TEC should still be well within the perameters of this kind of setup. a 360W 24V TEC (for example) would pull 15A and give approx 80C dT. you could run two of these units in series at 6V and it would still pull less than 10A. your dT would probably only be about 40C dT but another pair (and another 10A - increasing to a total of 20A@24V) and would give you about 80C dT?? obviously the numbers aren't exact (or even correct necessarily) - as i said, i am still trying to figure everything out. essentially, by operating at a reduced voltage you are able to better utilize the TEC's cooling capacity (much higher COP=better cooling per Amp) - and, can obtain MUCH greater dT's with the same or lower amperage when using multiple TECs. does that sound about right, or am i still missing something???

[Ultrasonic2]

First off their is no doubt that you can run a chiller off a normal PSU if you want especially via a GPU Plug thing.

2 yes reducing input voltage will increase COP but it does decrease Qmax and dTmax. Still your better off with the TEC undervolted

3 if you mean physically placing them in series ( Stacking ) over electrically connecting them in series then this will increase the delta. However this significantly reduces the Qmax meaning you'll need MANY MANY more TEC's so it's very expensive over just increasing the input voltage to the original TEC's

The reality is you'll never achieve a practical chiller that could hold a 300watt load to 80c that would require a huge array of TEC's with multiple layers stacked.

Keep up the investigating though

[bds71]

if you mean physically placing them in series ( Stacking ) over electrically connecting them in series then this will increase the delta. However this significantly reduces the Qmax meaning you'll need MANY MANY more TEC's

no, i meant electronically - not stacked (i know that has a set of problems all by itself). if you connect TECs in series (electrically) will you still get the desired dT increase? or do they have to be stacked to get that effect? ie: if they were side-by-side would you have a greater dT with half the voltage each vs a single at full voltage? what effect on QMax does this have? (still working exactly what QMax does for a TEC)

so far i understand UMax (max voltage), IMax (max current), QMax (max heat transfer), dT (temp differential between hot and cold sides at IMax), and COP (heat transference curve relative to current (amperage) at specific hotside temp) and how they all interrelate (mostly). i understand why more TECs is better than a single with similar total UMax/IMax. i assume the massive QMax you guys recommend is due to the undervolting of the TECs, whereas the actual power draw from those undervolted TECs is considerably less than total UMax/IMax for all the TECs combined?? still trying to figure out the relationship between UMax/Imax and QMax - i thought QMax was simply the product of volts and amps, but then i saw high QMax TEC's that were well under what i thought they would be. (35V/20A@403QMax) note: CoP will always be highest as dT approaches zero. but, since our chiller loops depend on a greater dT, how do we compensate for the lack of efficiency (ie: higher power requirement/current draw) of an ever increasing dT - at what point are we simply wasting power for diminishing returns?........anyway - still reading/learning/trying to figure this all out. thanks for all the input and help - and thanks for putting up with my MANY questions!!

idea: i was thinking about a barely sub-ambient (no condensation) water loop. since TEC's are most efficient at dT=0 then (ideally) you would want your hot side close to your low side. now, since it's impossible to get your hotside below ambient.......why not put a radiator in teh cold loop? (hear me out)...by doing so you would increase the cold side of the TEC to ambient - and if you can keep your hotside to within a few degrees of ambient you would then have a super efficient cooling method that would ensure ambient even under high load. you would have more cooling **potential** than you would at higher a dT and would (should) be able to remove more heat from your cold side loop (note: and with your radiator taking out some of the heat - as well as the TEC keeping the temps at room temp) you would then have a C/W of ZERO!! right? granted, you would not have a negative C/W but the cooling potential would be the same or greater than a subambient loop.

[thePoison.]

no, i meant electronically - not stacked (i know that has a set of problems all by itself). if you connect TECs in series (electrically) will you still get the desired dT increase? or do they have to be stacked to get that effect? ie: if they were side-by-side would you have a greater dT with half the voltage each vs a single at full voltage? what effect on QMax does this have? (still working exactly what QMax does for a TEC)

so far i understand UMax (max voltage), IMax (max current), QMax (max heat transfer), dT (temp differential between hot and cold sides at IMax), and COP (heat transference curve relative to current (amperage) at specific hotside temp) and how they all interrelate (mostly). i understand why more TECs is better than a single with similar total UMax/IMax. i assume the massive QMax you guys recommend is due to the undervolting of the TECs, whereas the actual power draw from those undervolted TECs is considerably less than total UMax/IMax for all the TECs combined?? still trying to figure out the relationship between UMax/Imax and QMax - i thought QMax was simply the product of volts and amps, but then i saw high QMax TEC's that were well under what i thought they would be. (35V/20A@403QMax) note: CoP will always be highest as dT approaches zero. but, since our chiller loops depend on a greater dT, how do we compensate for the lack of efficiency (ie: higher power requirement/current draw) of an ever increasing dT - at what point are we simply wasting power for diminishing returns?........anyway - still reading/learning/trying to figure this all out. thanks for all the input and help - and thanks for putting up with my MANY questions!!

idea: i was thinking about a barely sub-ambient (no condensation) water loop. since TEC's are most efficient at dT=0 then (ideally) you would want your hot side close to your low side. now, since it's impossible to get your hotside below ambient.......why not put a radiator in teh cold loop? (hear me out)...by doing so you would increase the cold side of the TEC to ambient - and if you can keep your hotside to within a few degrees of ambient you would then have a super efficient cooling method that would ensure ambient even under high load. you would have more cooling **potential** than you would at higher a dT and would (should) be able to remove more heat from your cold side loop (note: and with your radiator taking out some of the heat - as well as the TEC keeping the temps at room temp) you would then have a C/W of ZERO!! right? granted, you would not have a negative C/W but the cooling potential would be the same or greater than a subambient loop.

TEC's are heat pumps, you are effectively taking the "coldness" generated from the water chiller into a radiator that is then WARMED to ambient, completely defeating the purpose. I figured this out myself. Having it in the cold loop is a no no, that is like making a pseudo air chiller involuntarily.

[bds71]

but that was the whole point - to warm the cold side (to get closer to dT=0). i know - i have seen it said here many times not to do this because it defeats the purpose. and if you are trying to go subambient than yes, it certainly does, but if you only want to maintain ambient (better than normal LC, and without the increase in temps under load) then why wouldnt this make an efficient heat pump. it is still doing what it is intended for - it is removing heat generated by your heat sources.