Swiftech
Page 1 of 2 12 LastLast
Results 1 to 25 of 47

Thread: Someone help me see if I have this right...

  1. #1
    Xtreme Addict Kayin's Avatar
    Join Date
    Apr 2005
    Posts
    2,333

    Someone help me see if I have this right...

    So I've been reading quite a bit on the concepts of TECs and how to raise efficiency in loops, and most people seem to point to two things-air-cooled hot sides and using them as chillers made up of multiple smaller pelts rather than one large one on the CPU. First off, if I'm wrong there, please stop me and tell me what I'm doing wrong.

    So reading around a bit, I've been kicking about making a pelt chiller and using in place of a rad in my system (I have nothing but time here to mess with something. It may as well be something fun.) System consists of i7-2600K (currently stock cooled and underclocked, it's hot as three hells here in Alabama) and would also cover mosfet block and RAM. Currently GPU is air cooled, if I replace it I may or may not liquid cool it, but for now it's ok as is. RAID card has its own cooling system (I told you, I have a lot of time to play) though it may not stay in the system.

    From what I gathered with multiple perusals of multiple chiller threads, my setup should consist of a copper block with low pressure drop on one side for my liquid to flow through for chilling, multiple TECs laid out in a grid with high clamping force, and a heatsink and fan either for each TEC or an extrusion and large fans. The latter looks more like what would happen here, so I could use the heatsink extrusion to apply more even pressure to the TECs.

    The first question is on what TECs should I choose? Should I do an array of 10 or so small TECs, or should I do something like 3 or 4 of the large 62mm pelts? I'm using 12v max for this, mostly so I can use more easily sourced and modded ATX PSUs (modding a PS for 12v only is rather easy for me, did it for my Arima already) or even one of the 12v only VGA PSUs. With many of the larger TECs being 15V or more, this would already be underdriving them.

    The second is what kind of general temps should I expect? For the sake of conversation, let's say that I'm staying stock clocked on CPU. At that point, is there a way to guess at what temps I would be likely to achieve? If so, would I be able to take that number and determine what it would be if I OCed? I know I'm full of questions. I'd never learn if I didn't ask.

    I'm already well cognizant of things like having to insulate possibly, using TIM on pelts, using more power and that phase might be easier, but it certainly won't fit in my case like this could. This is just as much about "because I need something to do" as it is about the actual temps. If I was simply able to cool to the equivalent of a standard 3x120 in a bit smaller space (as that's my biggest issue in the case) I could call that a success. Anything past that is just plain great.

  2. #2
    Xtreme Addict Hell Hound's Avatar
    Join Date
    Oct 2007
    Location
    Chicago,Illinois
    Posts
    1,088
    Use 15v tec @ 12v,550w@62mm x6 should handle 4-600w load.
    Last edited by Hell Hound; 05-01-2012 at 04:57 PM.


  3. #3
    Registered User
    Join Date
    Sep 2009
    Posts
    85
    Quote Originally Posted by Kayin View Post
    The second is what kind of general temps should I expect?
    Short answer: it depends. How much cooling capacity are you going to provide and how much electricity are you willing to use?

    Here is a idealized calculator you can play with (written by ultrasonic2): http://www.tecpeltier.com/viewtopic.php?f=4&t=19

    Search for "Where can i get a TEC calculator app to do all the math for me ?" on the page and download the calculator.

  4. #4
    Hackintosh Lover davidkozat's Avatar
    Join Date
    Aug 2009
    Location
    Between London & Eastbourne, United Kingdom
    Posts
    559
    Quote Originally Posted by Hell Hound View Post
    Use 15v tec @ 12v,550w@62mm x6 should handle 4-600w load.
    are you saying 6 peltier take 600w load? you are gonna need another PSU!
    Why Not Try a Hackintosh for A Change?? Steps HERE!!
    Want to installing Mountain/Lion from a DVD or USB on you Apple PC?? Steps Here!


    Super Stable aKa 24/7




    Winners of EURO 08 | World Cup 10 | EURO 12
    Quote: Killin' Tube Kinks One Coil At A Time
    Hackintosh Lover
    "Dumb people" shouldn't use "Smart Phones"


    New Personal Wallpapers Selection

  5. #5
    Xtreme Addict Hell Hound's Avatar
    Join Date
    Oct 2007
    Location
    Chicago,Illinois
    Posts
    1,088
    It will be around 1.8kw to cool 600w load (depending on temp goal) because no rad will be used.
    I have seen 2 trues used for a tec setup but no high overclocks period w/ a setup like this.
    Quote Originally Posted by Kayin View Post
    So reading around a bit, I've been kicking about making a pelt chiller and using in place of a rad in my system (I have nothing but time here to mess with something. It may as well be something fun.) System consists of i7-2600K (currently stock cooled and underclocked, it's hot as three hells here in Alabama) and would also cover mosfet block and RAM. Currently GPU is air cooled, if I replace it I may or may not liquid cool it, but for now it's ok as is. RAID card has its own cooling system (I told you, I have a lot of time to play) though it may not stay in the system.

    From what I gathered with multiple perusals of multiple chiller threads, my setup should consist of a copper block with low pressure drop on one side for my liquid to flow through for chilling, multiple TECs laid out in a grid with high clamping force, and a heatsink and fan either for each TEC or an extrusion and large fans. The latter looks more like what would happen here, so I could use the heatsink extrusion to apply more even pressure to the TECs.

    I'm using 12v max for this, mostly so I can use more easily sourced and modded ATX PSUs (modding a PS for 12v only is rather easy for me, did it for my Arima already) or even one of the 12v only VGA PSUs. With many of the larger TECs being 15V or more, this would already be underdriving them.

    The second is what kind of general temps should I expect? For the sake of conversation, let's say that I'm staying stock clocked on CPU. At that point, is there a way to guess at what temps I would be likely to achieve? If so, would I be able to take that number and determine what it would be if I OCed? I know I'm full of questions. I'd never learn if I didn't ask.

    I'm already well cognizant of things like having to insulate possibly, using TIM on pelts, using more power and that phase might be easier, but it certainly won't fit in my case like this could. This is just as much about "because I need something to do" as it is about the actual temps. If I was simply able to cool to the equivalent of a standard 3x120 in a bit smaller space (as that's my biggest issue in the case) I could call that a success. Anything past that is just plain great.


  6. #6
    Xtreme Enthusiast Ultrasonic2's Avatar
    Join Date
    Apr 2008
    Location
    Auckland New Zealand
    Posts
    568
    From my point of view you haven't said enough to give any kind of estimations.

    Air cooling the hot side will always reduce it's efficiency as water cooling has a FAR greater capacity to move heat

    If you want to air cool the hot side you'll need about twice as many TEC's than you would need if you water cooled the hot side.

    The reality is you'll need about 1000 watts (4x 50mm) of raw TEC QMax too cool a modern PC and if your gonna air cool the hot side you'd need 1500 to 2000Qmax. Now of course you'll be massively undervolting the TEC's to increase their COP (efficiency) so electricity usage will be very low.


    To increase cooling efficency you want more TEC's undervolted more

  7. #7
    Xtreme Addict Kayin's Avatar
    Join Date
    Apr 2005
    Posts
    2,333
    Sounds to me like this is gonna be more trouble than it's worth. Certainly more expense. For what it was costing to look into this, I saw a Vapochill on the FS here for $200. I was hoping to help my temps, but it looks like I'd honestly be hurting them. At least with what's in my budget.

    I could stretch to the TEC block and controller that had a thread around here, but I couldn't go past that. And it seems that if I wanted to do this right, I would have to. I don't see a point in halfway doing something. That's really not productive.

    Thanks to all who replied, I'll keep this in the back of my mind for something to play with if I get some spare money, but as I rely on this PC, I'm gonna have to go with something more immediately functional.

  8. #8
    Xtreme Enthusiast Ultrasonic2's Avatar
    Join Date
    Apr 2008
    Location
    Auckland New Zealand
    Posts
    568
    I wouldn't even bother with the TEC Water blocks they dont have enough Qmax to satisfy a heavily Over clocked CPU.

    A 4 TEC water to water chiller is what you want anything less will disappoint.

  9. #9
    Xtreme Member bds71's Avatar
    Join Date
    Jul 2011
    Location
    Kaiserslautern, GE
    Posts
    325
    i had a thought (i know - never a good thing : 16 TECs in a 4x4 par/ser config. if i remember ohms law correctly this would effectively cause each TEC to operate at 25% voltage and current - and would provide a 2-4 CoP (coefficient of performance). here is my question: sinec the total power draw - in watts - would essentially be that of a single TEC, what would the Qmax be with a overall CoP of 2-4? i'm fairly certain dTmax would remain unchanged from single (for example 70C).

    theory of operation (as i understand it):
    assume 240W Qmax, 24A Imax, and 16V Vmax for a single TEC - hypothetical:
    each TEC would "see" 4V/6A and would pull 24W - total power draw: 384W - equal to a single TEC max power consumption
    each "leg" (series of 4 TECs) would see 16V (Vmax)/6A (1/4 Imax) (voltage adds in series while current remains constant)
    reasoning: assuming all TECs are same model - they would have same "resistance" so voltage drops in series would be "equal" and would add up to Vmax. same thing would apply to current, only applying to circuits in parallel rather than series (ie: each "leg" of 4 serially connected TECs would present itself as a single TEC in parallel with 3 others, thereby showing 1/4 Imax (assuming Vmax on each leg) on each "leg."

    normally, a single TEC running at 100% would have a CoP of .5 to .7 - this would give us (using the above example) 240W Qmax (power out) and would consume 384W (power in) for .625 CoP. does this mean that we could potentially see 768-1536W Qmax (actually useable) for this set-up? (assuming 2-4 CoP @ 384W power draw) if so, this would make an even better chiller than most compressors (which operate at less than 3 CoP)!!
    i7 3930@4.5GHz (EK Supreme HF), GTX690@1.2GHz (Koolance NX-690), 128G 4M + 2x128G 4M raid 0, Silverstone TJ07, Custom Enclosure w/MoRa, 18x GT AP-31, 401X2 dual PMP-400


  10. #10
    Xtreme Enthusiast Ultrasonic2's Avatar
    Join Date
    Apr 2008
    Location
    Auckland New Zealand
    Posts
    568
    im now sure how pedantic i want to be about the maths so i wont, but say it's not correct. However the point your trying to make is valid and is why we under volt heaps of TEC's

    You biggest problem is this

    i'm fairly certain dTmax would remain unchanged from single (for example 70C).

    dTmax is also relative to the in put voltage as you lower the input voltage it will be decreased. To continue the overly simplified maths you could say the dTmax would be 1/4 of 70 or 17.5c ( Though this is not actually correct maths )

    Also yes with highly undervolted TEC's you can have a MASSIVE COP of like 24 (yes 24 in theory)

  11. #11
    Xtreme Enthusiast Ultrasonic2's Avatar
    Join Date
    Apr 2008
    Location
    Auckland New Zealand
    Posts
    568
    err also when you decrease the input voltage you also decrease the Qmax of each TEC So based on your original incorrect maths there would be no difference in having 4 TEC's at 100% than 16tec all running at 1/4 their input voltage. As the totals would be the Same Pmax, Qmax AND dTmax of 1536,960,70

    HOWEVER in reality there is a big difference
    Last edited by Ultrasonic2; 06-01-2012 at 03:02 AM.

  12. #12
    Xtreme Member bds71's Avatar
    Join Date
    Jul 2011
    Location
    Kaiserslautern, GE
    Posts
    325
    dTmax is also relative to the in put voltage as you lower the input voltage it will be decreased
    that's what i dont understand: the total input voltage is Vmax. same thing for the Imax. your input voltage/current would still be Imax/Vmax - that's why i was saying the total dT would still be the original - because nothing changed (relative to a single TEC). the only thing that really changed is CoP which reflects in the total Qmax; but Imax, Vmax and dT *should* all still be the same the way it's all wired? (of course i have been known to be wrong - don't take that as a statement of fact, but as a question ) either way - my overarching point/question wasn't the actual math, but rather what would the total Qmax of such a circuit look like? would the theoretical value be close to 2-4 (CoP) times the input wattage (Vmax * Imax)? the reason i ask is because that's the way it usually comes out (more or less) - since CoP at 100% is usually around .6 you get 60% of the input wattage (that's why single TECs trying to cool a CPU never work - because your only getting a fraction of the applied power due to heat loss and inefficiency. right? but because you can get such a larger CoP with larger numbers of TECs you start getting CoP's above 1 (more cooling power than what is input). in the case above you *should* still have about 1150W Qmax at CoP of 3 with 384W input. the Imax, Vmax, dT, etc was admittedly, guesses based on the way i believe electronics to work throw that out if it's wrong - but tell me if my relationship between CoP and input Wattage (assuming Imax * Vmax) is correct?

    as always - thanks for your information Ultrasonic. i have truly learned a LOT from our previous discussions. think of this as continuing to build on that knowledge sorry, i know it must be painful for you - but, thanks again for your understanding and patience!!

    sudden thought: stupid question: where is the best CoP? 10%? 5%? is there diminishing returns the lower you go? (i know it's slightly different for each TEC - just looking for ballpark figure/best guess) most charts i've seen tend to go off the chart below 20-30% or so, but it looks there would be a peak/downturn at some point between 5-15% can you shed any light on that?
    i7 3930@4.5GHz (EK Supreme HF), GTX690@1.2GHz (Koolance NX-690), 128G 4M + 2x128G 4M raid 0, Silverstone TJ07, Custom Enclosure w/MoRa, 18x GT AP-31, 401X2 dual PMP-400


  13. #13
    Xtreme Enthusiast Ultrasonic2's Avatar
    Join Date
    Apr 2008
    Location
    Auckland New Zealand
    Posts
    568
    that's what i dont understand: the total input voltage is Vmax. same thing for the Imax. your input voltage/current would still be Imax/Vmax - that's why i was saying the total dT would still be the original - because nothing changed (relative to a single TEC).

    lets simplify it further and compare one resister to two of the same resister across the same input voltage. Assuming your applying 10V to the the first single TEC then when you mesure the voltage drop across it at the TEC's + & - you will get a voltage drop of 10 volts so you are applying 10 volts to this TEC.

    When you wire 2 resisters of the same resistance in series or +R1 to input P+ then R1- to R2 + and R2- to the input p- then measuring the voltage applied to the whole circuit from +R1 to -R2 will be = to the input voltage of 10V However this is NOT the applied voltage across each Resister (or input voltage per Resister). if you were to take a Volt meter and measure the voltage between +R1 and -R1 you would get 5volts or 1/2 the input voltage or the input voltage divided by the number of resisters

    http://www.youtube.com/watch?v=8RyRci2ntWc
    At about 3:50 minutes

    but rather what would the total Qmax of such a circuit look like?
    sticking with simplified maths the Qmax would not change 1 TEC at 100% Vmax = 100% of Qmax . 2 TEC's. TEC 1 50% of Vmax = 50% of Qmax * 2 (multiplied by 2 TEC's) = 100% of Qmax for a single TEC. So there is no point ( with simplified maths how ever with the real maths you do get a gain )


    6 you get 60% of the input wattage (that's why single TECs trying to cool a CPU never work - because your only getting a fraction of the applied power due to heat loss and inefficiency. right?
    Um no a single TEC does work. It however nolonger works because the CPU load it to great for the TEC ( keeping with simplified maths )

    COP has no relevance on whether or not a TEC can cool a load

    COP is simply an equation based on HEAT MOVED V the APPLIED POWER TO DO SO

    Up to this point you have talked about potential COP not Actual COP as you've never mentioned the load your moving so the COP calculation isn't actually real.

    back to your 240Qmax and 384watts this has a COP potential of 0.625 ( overly simplified maths ) However this assumes your applying a heat load of 240watts. before i move on you have to realise that for this COP you aren't achieving ANY delta so the hot and cold side will be the same temp ( assuming perfect thermal resistance of the hot and cold side water blocks).

    If your CPU is not putting out a load of 240 but 120 then the COP would be 0.3125 ( assuming the used watts stays the same which it doesn't) HOWEVER you'll be achieving a large delta between the 2 sides now

    COP is normally rated to a delta of 0 but a delta of 0 is quite pointless to us

    because your only getting a fraction of the applied power due to heat loss and inefficiency

    If it was possible to have a hot side water block with a thermal resistance of 0 then you could cool a 239watt cpu with a 240watt TEC

    in the case above you *should* still have about 1150W Qmax
    Not a chance as explained at the top

    just looking for ballpark figure/best guess) most charts i've seen tend to go off the chart below 20-30% or so, but it looks there would be a peak/downturn at some point between 5-15% can you shed any light on that?
    YES
    COP doesn't continue to increase to 0% input voltages because the TEC has a thermal resistance of it's self that you have to apply a certain % to over come.

    Now if you turn the TEC off it will still transfer heat from the cold to the hot side , since all a TEC is is two dissimilar metals stuck together. However this conduction is quite poor. If we apply no power to our TEC but apply a load to the cold side. because of the TEC's thermal resistance you will end put with the cold sides water being hotter than the the TEC's hot side OR a negative delta. SO the first X % of power you apply to a TEC is to overcome the thermal resistance of the TEC it's self and get the Delta back to positive

  14. #14
    Xtreme Member Liam_G's Avatar
    Join Date
    Dec 2009
    Location
    Brisbane, Australia
    Posts
    146
    I think if you used this TEC http://www.customthermoelectric.com/...Q_spec_sht.pdf to do your 4 parallel sets of 4 TECs in Series, so 16 TEC's on a 12v PSU you could achieve some decent numbers. Aim for 20*dT, each TEC get's 3v and about 5.5amps, and uses 16.5watts and has a Qmax of about 20-22 at 20*dT, so with all 16 TECs your using about 264 watts to give you a Qmax of about 320w. So keeping it simplified like that your doing pretty well, of course a lot of other stuff will come into play, but that is a basic approximation of what can be done with lots of these lil suckers.

  15. #15
    Xtreme Enthusiast Ultrasonic2's Avatar
    Join Date
    Apr 2008
    Location
    Auckland New Zealand
    Posts
    568
    Quote Originally Posted by Liam_G View Post
    I think if you used this TEC http://www.customthermoelectric.com/...Q_spec_sht.pdf to do your 4 parallel sets of 4 TECs in Series, so 16 TEC's on a 12v PSU you could achieve some decent numbers. Aim for 20*dT, each TEC get's 3v and about 5.5amps, and uses 16.5watts and has a Qmax of about 20-22 at 20*dT, so with all 16 TECs your using about 264 watts to give you a Qmax of about 320w. So keeping it simplified like that your doing pretty well, of course a lot of other stuff will come into play, but that is a basic approximation of what can be done with lots of these lil suckers.
    yeah so up until this point we have been pretending that a TEC worked like a simple resister and that a change in input voltage resulted in a change in everything else with a straight line. This meant (because we over simplified everything ) that there was no benefit in having TEC's undervolted. However there is an advantage to undervolting because the resistance isn't in affective constant and a change in input voltage does NOT = a straight line change in everything else. So undervolting does give you increased Q while decreasing P used . Like Liam has just posted

  16. #16
    Xtreme Member bds71's Avatar
    Join Date
    Jul 2011
    Location
    Kaiserslautern, GE
    Posts
    325
    back to your 240Qmax and 384watts this has a COP potential of 0.625 ( overly simplified maths ) However this assumes your applying a heat load of 240watts. before i move on you have to realise that for this COP you aren't achieving ANY delta so the hot and cold side will be the same temp ( assuming perfect thermal resistance of the hot and cold side water blocks).

    If your CPU is not putting out a load of 240 but 120 then the COP would be 0.3125 ( assuming the used watts stays the same which it doesn't) HOWEVER you'll be achieving a large delta between the 2 sides now
    i thought at Qmax (and Imax/Vmax/Pmax) - basically, with everything maxed out - is the only way to achieve the full dT? to get to the potential dT you have to max out the TEC's voltage and current (ie: operating at 100%). dT=0 is the point that the TEC is most efficient (which should be at the lower operating ranges between 10-20% - which is also where you have high CoP). in other words CoP is inversley proportional to dT - one goes up, the other goes down. the way you stated that sounds like your saying dT=0 occurs at max power (Qmax/Pmax) - maybe i'm misunderstanding you? i understand what you say about heat load changing the variables. i know a heat load pretty much changes all the rules - i'm still working that part out.

    yeah so up until this point we have been pretending that a TEC worked like a simple resister and that a change in input voltage resulted in a change in everything else with a straight line. This meant (because we over simplified everything ) that there was no benefit in having TEC's undervolted.
    i was never saying the resistance was linear - i know TECs operate on a curve. i was however assuming a fixed input voltage of Vmax...which would have a fixed current at that voltage of Imax. with a set specific input voltage, the TECs would have an easily calculated voltage drop and related current. (the chart proves this because it *does* show a linear voltage vs amps) and, since the current and voltage are proportional they both can easily be calculated using the methods used both by me and Liam_G. i'm not trying to debate these methods...whether they are right or wrong....or if they correct or not - it's all hypothetical and really doesn't need to be any more complicated. what is harder to calculate (and i think this is what you are referring to Ultrasonic2) is Qc at any given voltage/current and since dT is dependant on the heat load (and i wouldn't even begin to know how to calculate that!!) there are some things that we can't calculate, but simple voltage drops and their related amperage can be easily determined with a specified input.

    sticking with simplified maths the Qmax would not change 1 TEC at 100% Vmax = 100% of Qmax . 2 TEC's. TEC 1 50% of Vmax = 50% of Qmax * 2 (multiplied by 2 TEC's) = 100% of Qmax for a single TEC. So there is no point ( with simplified maths how ever with the real maths you do get a gain)
    i think i understand this: because at 50%, CoP increases so you actually get a higher Qc - something less than Qmax, but more than 50%Qmax (maybe around 55-60%). Q and CoP are not linear in relation to each other. so in the circuit above, you *can* have an overall/total Qmax higher than that of an individual TEC.

    back to your 240Qmax and 384watts this has a COP potential of 0.625 ( overly simplified maths ) However this assumes your applying a heat load of 240watts. before i move on you have to realise that for this COP you aren't achieving ANY delta so the hot and cold side will be the same temp

    If your CPU is not putting out a load of 240 but 120 then the COP would be 0.3125 ( assuming the used watts stays the same which it doesn't) HOWEVER you'll be achieving a large delta between the 2 sides now

    COP is normally rated to a delta of 0 but a delta of 0 is quite pointless to us
    i think i see what your saying here (though not completely about dT): in my example 240Wmax is the cooling "potential." so if your at "max" cooling capacity you should technically be at dT=0? however, if your cooling potential is greater than your load then you end up with a temp differential between the hot and cold sides. the greater the difference, the closer you get to dTmax at load. so in my example: at CoP of 3 you would have a "potential" to cool ~1150 watts at max dT with no load. if you apply this to a modern CPU that dissipates 200W (?) your dT will go down, as will your CoP (?) but you should still be well within your power envelope and should still get pretty high dT, right? (like, maybe 50/70).

    COP doesn't continue to increase to 0% input voltages because the TEC has a thermal resistance of it's self that you have to apply a certain % to over come.
    so the downturn is due to the thermal resistance of the TEC itself - cool, got it.

    ------------------------------------------------------------------------

    since i think we are getting hung up on the numbers let me try this. tell me if my "relationships" are correct or not:

    dT is inversley proportional to CoP (one goes up, the other goes down)
    dT is directly proportional to Q (one goes up, the other goes up also)
    Q is directly proportional to CoP - but not linear (one goes up, the other goes up, but not by the same amount)
    i7 3930@4.5GHz (EK Supreme HF), GTX690@1.2GHz (Koolance NX-690), 128G 4M + 2x128G 4M raid 0, Silverstone TJ07, Custom Enclosure w/MoRa, 18x GT AP-31, 401X2 dual PMP-400


  17. #17
    Xtreme Member bds71's Avatar
    Join Date
    Jul 2011
    Location
    Kaiserslautern, GE
    Posts
    325
    http://www.customthermoelectric.com/...Z_spec_sht.pdf

    tell me if these numbers are pretty close for 4 of these in series at dT 20C ($200):

    4 in series with 12V input would pull 13Amps (each TEC would get 3V/13A)
    162 Qc per TEC (20dT on the line graph at 13Amps) (400 Qmax)
    648 Qmax Total (20dT)

    adding a second leg with 4 more TECs would drop Qc from 162@13A down to 55@6.5A and would lower (?!?!) Qmax to 440 and would cost twice as much ($400) note: at zero dT the Qmax/CoP increase, but at 20dT it actually goes down?? so this means 4 (under load) is better than 8, even though 8 has more cooling power without a load? (damn, now i confused myself again)
    i7 3930@4.5GHz (EK Supreme HF), GTX690@1.2GHz (Koolance NX-690), 128G 4M + 2x128G 4M raid 0, Silverstone TJ07, Custom Enclosure w/MoRa, 18x GT AP-31, 401X2 dual PMP-400


  18. #18
    Xtreme Member bds71's Avatar
    Join Date
    Jul 2011
    Location
    Kaiserslautern, GE
    Posts
    325
    OK, i get it now (i think):

    with the 400W TEC linked above i can keep a 395W heat source at room temperature (0dT), or i can cool a 130W source to 50C below room temp (50dT) (or anywhere in between). with 4 in series i can cool 1120W@0dT or 370@50dT ("very" ballpark - but hopefully, you get the point). is that about right? (not the actual math - but the relationships)

    of course i understand that you'll never be able to run a TEC at 100% and get 0dT (room temp with a 400W source) - the CoP at 100% would probably be about .6 - so you could maybe cool a 240W heat source and keep 0dT...maybe, i don't know...
    i7 3930@4.5GHz (EK Supreme HF), GTX690@1.2GHz (Koolance NX-690), 128G 4M + 2x128G 4M raid 0, Silverstone TJ07, Custom Enclosure w/MoRa, 18x GT AP-31, 401X2 dual PMP-400


  19. #19
    Xtreme Member Liam_G's Avatar
    Join Date
    Dec 2009
    Location
    Brisbane, Australia
    Posts
    146
    Ok, so you've just posted a bunch of different stuff up there, i can't be bothered trying to interpret or explain it via your messages, so i will try to explain to you how i understand the "relationships" between all the variables and how i choose a tec and how i know it will cool the load i require.

    So,

    Find out the load you want to cool, so lets say i want to cool 300+ watts from a 3930k at 5ghz, so this lets me know i need enough Qmax to cool 300+ watts,

    but i need it to cool that at a dT of greater than 20, otherwise i will end up right back where watercooling had me before, basically ambient and no lower. Now, it is quite easy to find a TEC that will cool 300watts, but it is not so easy to find one or a combination that will give you 300watts Qmax at greater than 20*dT and not consume a ridiculous amount of power. So thats why we go with multiple tecs and undervolt them. I would also ignore the CoP for now, optimise that later, if you try to choose your TEC based on optimal CoP you will probably never be able to cool your heat load, i would say the best you will ever get with CoP is about 1.5 with tecs, and only then with a small load.

    So from this i now know what to look for when i am choosing a TEC, i need a combination of TEC's that will give me 300wQmax, i would aim for a 30*dT, to give you room to lose dT because of thermal resistance of waterblocks and radiators etc. and you want to do it with a decent amount of power, however for such a large heat load i would say you will not achieve a CoP any better than about .75 without using a rediculous amount of tecs, and being reasonable anything more than 8 large tecs is prob not ideal for space and cost, even if it makes sense for power usage.

    So at this point your power supply voltage comes into play, lets stick with 12v for now, even though 24v or greater actually makes more sense. so 12v limits you in that you can only run 1 to 4 TECs in series realistically, i would even say no more than 2 in series is realistic, i personally think 16 tecs is too many and any less than 6v will not net you enough qmax for each individual tec to be able to use 8 tecs or less.

    so lets see where we are at, so 300w load, target of 30*dT, 12v PSU.

    So at this point i would say, browse through a bunch of tec performance graphs and apply either 12v or 6v to them and see what amperage that uses at 30*dT. So at 30*dT you need an amperage that will give you 75wQmax per tec if using 12v per tec and 4 tecs, or 37.5wQmax per tec if using 6v per tec and 8 TECs. (by the way, applying 3v to that 400w tec will not give you 13amps if you use four tecs in series, it will actually use less than 4amps and give you no cooling whatsoever, as far as i understand these things, when you combine tecs in series, the voltage doubles and the amps stay the same, so if you apply 12v to one tec you get 12 to 13amps, so to get 12-13 amps with 2 tecs in series you would have to apply 24volts, if you apply 12v to 2 tecs each tec will get 6v and you will have a current of 4 amps through the 2, once again, no cooling power there at all. if you want to keep the amps the same, double the volts.

    so when playing with 12v you can either chose a high Vmax and high Imax tec if you want to run them singly, or a low Vmax high Imax tec if you want to run them in series.

    this is all getting a bit confusing but hopefully you can understand me enough to help you. one key thing to remember is the only thing you can control with a tec is the voltage you give it, everything else is a consequence of that, the amps, dT and Qmax are all dependent on choosing a tec that suits the voltage you will apply to it. also remember when running tecs in series you need to double the voltage applied to keep the same cooling capacity from each tec.

    so to try and sum up my initial point, find out your heat load (you can do this by finding how much power your cpu uses, that is a good approximation of how much heat your cpu puts out, same for gpu's. so if your cpu uses 300watts at full load then that is your target for qmax. so, find out your heat load, decide on your PSU voltage, take that voltage and combined with your target dT you can see the amps you will use and from that you will find out the qmax of each tec at that volts and amps. if you happen to find a combo that gives you good CoP then great, you also now have a ball park figure to work to to try and achieve a better CoP from a different combo of tecs, but really tyhre only way to achive great CoP is to have low heat load, say about 200, or lots and lots of tecs like that group of 16 i described and even then it was only about a CoP of 1.2 (which is great in my books)

    i hope this is understandable at all, bit all over the shop but i'm too tired to make more sense of myself :P

    so here is a quick example of my system at work:

    - 300w+ heat load for tec to transfer
    - 12v PSU
    - 30*dT target
    - so, 12v > 30dT gives me 12 amps for that 400w tec
    - 12 amps gives me about 90wQmax per tec at 30*dT
    - so i need 4 tecs at 12v each to give me greater than 300w Qmax at 30*dT (4 x 90 = 360wQmax) note that they are running at full 12v so this means 4 tecs in parallel, not series.
    - so 12v x 12a = 144watts power usage/tec, so 576w power usage all up
    - so 360w Qmax and 576w usage gives a CoP of .625, not bad considering only using 4 tecs and cooling up to 360watts, but not the greater than CoP of 1 you are looking for, you will only achieve that with a larger number of tecs if you want to cool a large load.
    - now don't forget, that for your tec to be able to transfer 360w from the cold side to the hot side you need to be able to cool your total heat load on your hot side of tec, this heat load is the sum of your transfered heat (360w) and your power usage, so 360+576 = 936 watts that you need to be able to cool to even achieve these numbers, so your looking at 2x360 or 480 radiators to cool that load, and after all thermal resistance is taken into account you will end up about 20* below ambient.

    there is so much more to this than i know and also more than i even thought possible, ultrasonic is your guy for all the nitty gritty, but that is the level of understanding that i am at and it is enough to know how to do it for now.

  20. #20
    Xtreme Member Liam_G's Avatar
    Join Date
    Dec 2009
    Location
    Brisbane, Australia
    Posts
    146
    If you want better CoP with a smaller number of TEC's you will have to move to a 24v PSU, i would say 8v is a sweet spot for some tecs, so you could parallel 2 sets of 3 tecs in series. I would think a 48v PSU would be pretty sweet to play with as well, say 5 tecs in series at 9.6v, that would be fun although the more tecs you put in series the more risk that one bad tec will take out your whole cooling system, 1 bad tec in a series connection will take them all out, whereas parallel they run independently regardless if one fails, unless you have paralleled them in series, in which case it will just take out that arm of the parallel.

    a 24v PSU allows you to do 24, 12, 8 or 6v with 1, 2, 3 or 4 tecs in series, whereas 12v only allows you to do 12, 6, 4 or 3. so that gap between 12v and 6v with one or two tecs is why 12v PSU isn't as good for efficiency. some tecs really shine in that 8-10v range.

    in saying all this, if cost and space is not an issue, and you just want the best efficiency and CoP, then by all means, go for 16+ TEC's, the possibilities are virtually endless with TEC's, but after a while you will be spending so much and the cooling system will take up so much space that you will probably want to go elsewhere, ie phase change.

    my reasons for using tecs are first and foremost that it gives me the opportunity to cool below ambient and also that i can still do it quietly and i also can fit it inside my case, i am prepared to sacrifice some efficiency for those things. if noise and space are not an issue then phase might be a better choice, i have no knowledge of them so i'm not sure of there average CoP. i think the cost of more than 16 tecs puts you well and truly in the realm of good phase change, Single Stage if not Cascade, but they are noisy, huge and a whole different beast really.
    Last edited by Liam_G; 06-02-2012 at 06:08 PM.

  21. #21
    Xtreme Enthusiast Ultrasonic2's Avatar
    Join Date
    Apr 2008
    Location
    Auckland New Zealand
    Posts
    568
    WOW there has been to many posts now and it's going to get out of control.

    i Will REPLY to EACH post Separately. Because this will take me many hours to reply to them all

    i thought at Qmax (and Imax/Vmax/Pmax) - basically, with everything maxed out - is the only way to achieve the full dT?
    YES

    to get to the potential dT you have to max out the TEC's voltage and current (ie: operating at 100%)
    YES

    . dT=0 is the point that the TEC has the greatest COP (which should be at the lower operating ranges between 10-20% - which is also where you have high CoP). in other words CoP is inversley proportional to dT - one goes up, the other goes down
    YES

    The way you stated that sounds like your saying dT=0 occurs at max power (Qmax/Pmax) -

    The point i was trying to make was that the MAX COP you can achieve for any given input voltage ocures when you apply the relevant QC to the cold side so the delta is 0

    ie 240 watts moved consumes 384watts = a COP of 0.625
    However 120 watts moved and assuming it's still consuming 384watts ( which it wouldn't ) = a COP of 0.3125


    . i was however assuming a fixed input voltage of Vmax...which would have a fixed current at that voltage of Imax. with a set specific input voltage, the TECs would have an easily calculated voltage drop and related current. (the chart proves this because it *does* show a linear voltage vs amps) and, since the current and voltage are proportional they both can easily be calculated using the methods used both by me and Liam_G. i'm not trying to debate these methods...whether they are right or wrong....or if they correct or not - it's all hypothetical and really doesn't need to be any more complicated. what is harder to calculate (and i think this is what you are referring to Ultrasonic2) is Qc at any given voltage/current and since dT is dependant on the heat load (and i wouldn't even begin to know how to calculate that!!) there are some things that we can't calculate, but simple voltage drops and their related amperage can be easily determined with a specified input.

    i dont know what your looking at but that chart shows that the resistance of the TEC is NOT constant. It varies based on the load applied and therefore the achieved delta

    With 2 volts applied while achieving a delta of 60 it uses 1.5 amps = a resistance of 1.33333
    However with 2 volts applied while achieving a delta of 0 it uses 4.2amps = a resistance of 0.47619

    i think i understand this: because at 50%, CoP increases so you actually get a higher Qc - something less than Qmax, but more than 50%Qmax (maybe around 55-60%). Q and CoP are not linear in relation to each other. so in the circuit above, you *can* have an overall/total Qmax higher than that of an individual TEC.
    YES

    i think i see what your saying here (though not completely about dT): in my example 240Wmax is the cooling "potential." so if your at "max" cooling capacity you should technically be at dT=0?
    YES

    however, if your cooling potential is greater than your load then you end up with a temp differential between the hot and cold sides. the greater the difference, the closer you get to dTmax at load.
    As the delta increases the COP decreases from it's max theoretical ( when appling Qmax )


    so in my example: at CoP of 3 you would have a "potential" to cool ~1150 watts at max dT with no load.
    If your not cooling any load then the COP is 0 . The COP is a Potential of 3

    if you apply this to a modern CPU that dissipates 200W (?) your dT will go down, as will your CoP (?)
    YES your COP will be down


    since i think we are getting hung up on the numbers let me try this. tell me if my "relationships" are correct or not:

    dT is inversley proportional to CoP (one goes up, the other goes down)
    YES

    dT is directly proportional to Q (one goes up, the other goes up also)
    YES is if we are talking about applying more voltage.

    Q is directly proportional to CoP - but not linear (one goes up, the other goes up, but not by the same amount)
    NO As Q increases for a given TEC COP decreases.

    for the above answers id have to check if the were directly proportional or not.

  22. #22
    Xtreme Enthusiast Ultrasonic2's Avatar
    Join Date
    Apr 2008
    Location
    Auckland New Zealand
    Posts
    568
    4 in series with 12V input would pull 13Amps (each TEC would get 3V/13A)


    NOPE you still haven't grasped the original ohms law point that i made.

    No point answering the rest since the primes is incorrect

  23. #23
    Xtreme Enthusiast Ultrasonic2's Avatar
    Join Date
    Apr 2008
    Location
    Auckland New Zealand
    Posts
    568
    Quote Originally Posted by bds71 View Post
    OK, i get it now (i think):

    with the 400W TEC linked above i can keep a 395W heat source at room temperature (0dT), or i can cool a 130W source to 50C below room temp (50dT) (or anywhere in between). with 4 in series i can cool 1120W@0dT or 370@50dT ("very" ballpark - but hopefully, you get the point). is that about right? (not the actual math - but the relationships)

    of course i understand that you'll never be able to run a TEC at 100% and get 0dT (room temp with a 400W source) - the CoP at 100% would probably be about .6 - so you could maybe cool a 240W heat source and keep 0dT...maybe, i don't know...
    with the 400W TEC linked above i can keep a 395W heat source at room temperature (0dT),
    YES ish

    or i can cool a 130W source to 50C below room temp (50dT) (or anywhere in between)
    NO the graph shows it can only move 70ish to a delta of 50

    with 4 in series i can cool 1120W@0dT or 370@50dT ("very" ballpark - but hopefully, you get the point). is that about right? (not the actual math - but the relationships)
    NO
    they could only move 700ish to a delta of 0. 50dt is not possible at 6volts (1/4 Vmax)

  24. #24
    Xtreme Enthusiast Ultrasonic2's Avatar
    Join Date
    Apr 2008
    Location
    Auckland New Zealand
    Posts
    568
    Quote Originally Posted by Liam_G View Post
    Ok, so you've just posted a bunch of different stuff up there, i can't be bothered trying to interpret or explain it via your messages, so i will try to explain to you how i understand the "relationships" between all the variables and how i choose a tec and how i know it will cool the load i require.

    So,

    Find out the load you want to cool, so lets say i want to cool 300+ watts from a 3930k at 5ghz, so this lets me know i need enough Qmax to cool 300+ watts,

    but i need it to cool that at a dT of greater than 20, otherwise i will end up right back where watercooling had me before, basically ambient and no lower. Now, it is quite easy to find a TEC that will cool 300watts, but it is not so easy to find one or a combination that will give you 300watts Qmax at greater than 20*dT and not consume a ridiculous amount of power. So thats why we go with multiple tecs and undervolt them. I would also ignore the CoP for now, optimise that later, if you try to choose your TEC based on optimal CoP you will probably never be able to cool your heat load, i would say the best you will ever get with CoP is about 1.5 with tecs, and only then with a small load.

    So from this i now know what to look for when i am choosing a TEC, i need a combination of TEC's that will give me 300wQmax, i would aim for a 30*dT, to give you room to lose dT because of thermal resistance of waterblocks and radiators etc. and you want to do it with a decent amount of power, however for such a large heat load i would say you will not achieve a CoP any better than about .75 without using a rediculous amount of tecs, and being reasonable anything more than 8 large tecs is prob not ideal for space and cost, even if it makes sense for power usage.

    So at this point your power supply voltage comes into play, lets stick with 12v for now, even though 24v or greater actually makes more sense. so 12v limits you in that you can only run 1 to 4 TECs in series realistically, i would even say no more than 2 in series is realistic, i personally think 16 tecs is too many and any less than 6v will not net you enough qmax for each individual tec to be able to use 8 tecs or less.

    so lets see where we are at, so 300w load, target of 30*dT, 12v PSU.

    So at this point i would say, browse through a bunch of tec performance graphs and apply either 12v or 6v to them and see what amperage that uses at 30*dT. So at 30*dT you need an amperage that will give you 75wQmax per tec if using 12v per tec and 4 tecs, or 37.5wQmax per tec if using 6v per tec and 8 TECs. (by the way, applying 3v to that 400w tec will not give you 13amps if you use four tecs in series, it will actually use less than 4amps and give you no cooling whatsoever, as far as i understand these things, when you combine tecs in series, the voltage doubles and the amps stay the same, so if you apply 12v to one tec you get 12 to 13amps, so to get 12-13 amps with 2 tecs in series you would have to apply 24volts, if you apply 12v to 2 tecs each tec will get 6v and you will have a current of 4 amps through the 2, once again, no cooling power there at all. if you want to keep the amps the same, double the volts.

    so when playing with 12v you can either chose a high Vmax and high Imax tec if you want to run them singly, or a low Vmax high Imax tec if you want to run them in series.

    this is all getting a bit confusing but hopefully you can understand me enough to help you. one key thing to remember is the only thing you can control with a tec is the voltage you give it, everything else is a consequence of that, the amps, dT and Qmax are all dependent on choosing a tec that suits the voltage you will apply to it. also remember when running tecs in series you need to double the voltage applied to keep the same cooling capacity from each tec.

    so to try and sum up my initial point, find out your heat load (you can do this by finding how much power your cpu uses, that is a good approximation of how much heat your cpu puts out, same for gpu's. so if your cpu uses 300watts at full load then that is your target for qmax. so, find out your heat load, decide on your PSU voltage, take that voltage and combined with your target dT you can see the amps you will use and from that you will find out the qmax of each tec at that volts and amps. if you happen to find a combo that gives you good CoP then great, you also now have a ball park figure to work to to try and achieve a better CoP from a different combo of tecs, but really tyhre only way to achive great CoP is to have low heat load, say about 200, or lots and lots of tecs like that group of 16 i described and even then it was only about a CoP of 1.2 (which is great in my books)

    i hope this is understandable at all, bit all over the shop but i'm too tired to make more sense of myself :P

    so here is a quick example of my system at work:

    - 300w+ heat load for tec to transfer
    - 12v PSU
    - 30*dT target
    - so, 12v > 30dT gives me 12 amps for that 400w tec
    - 12 amps gives me about 90wQmax per tec at 30*dT
    - so i need 4 tecs at 12v each to give me greater than 300w Qmax at 30*dT (4 x 90 = 360wQmax) note that they are running at full 12v so this means 4 tecs in parallel, not series.
    - so 12v x 12a = 144watts power usage/tec, so 576w power usage all up
    - so 360w Qmax and 576w usage gives a CoP of .625, not bad considering only using 4 tecs and cooling up to 360watts, but not the greater than CoP of 1 you are looking for, you will only achieve that with a larger number of tecs if you want to cool a large load.
    - now don't forget, that for your tec to be able to transfer 360w from the cold side to the hot side you need to be able to cool your total heat load on your hot side of tec, this heat load is the sum of your transfered heat (360w) and your power usage, so 360+576 = 936 watts that you need to be able to cool to even achieve these numbers, so your looking at 2x360 or 480 radiators to cool that load, and after all thermal resistance is taken into account you will end up about 20* below ambient.

    there is so much more to this than i know and also more than i even thought possible, ultrasonic is your guy for all the nitty gritty, but that is the level of understanding that i am at and it is enough to know how to do it for now.
    PERFECT POST (pretty much)

  25. #25
    Xtreme Enthusiast Ultrasonic2's Avatar
    Join Date
    Apr 2008
    Location
    Auckland New Zealand
    Posts
    568
    Quote Originally Posted by liam_g View Post
    if you want better cop with a smaller number of tec's you will have to move to a 24v psu, i would say 8v is a sweet spot for some tecs, so you could parallel 2 sets of 3 tecs in series. I would think a 48v psu would be pretty sweet to play with as well, say 5 tecs in series at 9.6v, that would be fun :d although the more tecs you put in series the more risk that one bad tec will take out your whole cooling system, 1 bad tec in a series connection will take them all out, whereas parallel they run independently regardless if one fails, unless you have paralleled them in series, in which case it will just take out that arm of the parallel.

    A 24v psu allows you to do 24, 12, 8 or 6v with 1, 2, 3 or 4 tecs in series, whereas 12v only allows you to do 12, 6, 4 or 3. So that gap between 12v and 6v with one or two tecs is why 12v psu isn't as good for efficiency. Some tecs really shine in that 8-10v range.

    In saying all this, if cost and space is not an issue, and you just want the best efficiency and cop, then by all means, go for 16+ tec's, the possibilities are virtually endless with tec's, but after a while you will be spending so much and the cooling system will take up so much space that you will probably want to go elsewhere, ie phase change.

    My reasons for using tecs are first and foremost that it gives me the opportunity to cool below ambient and also that i can still do it quietly and i also can fit it inside my case, i am prepared to sacrifice some efficiency for those things. If noise and space are not an issue then phase might be a better choice, i have no knowledge of them so i'm not sure of there average cop. I think the cost of more than 16 tecs puts you well and truly in the realm of good phase change, single stage if not cascade, but they are noisy, huge and a whole different beast really.
    marry me

Page 1 of 2 12 LastLast

Posting Permissions

  • You may not post new threads
  • You may not post replies
  • You may not post attachments
  • You may not edit your posts
  •