Upgrading to TECs

[-J-]

ok, i had my WC rig for 1 month, so i have to upgrade

ive started designing my new setup. im already using rotor blocks for my setup, and as far as i know those are the best for TECs.

is there any change i should make to the design of my blocks for using them with a TEC? appart from the clamping holes for the coldplate.

i mean, do i need a bigger base plate or something like that?

im actually using a 2mm baseplate, in a 12,5mm block.

that for the block, now about the rad, im using a 16x16cm heatcore, that would be enough for 2 170w TECs, or i should add another heatcore in parallel?

any tips r welcomed.

[8-Ball]

I would imagine, though I may be wrong here that you'd want the base plate as thin as possible.

As I understand it, the base plate only needs thickness when heat spreading is necessary. In the case of cooling a large TEC, there isn't going to be any block left for the heat to spread to, so I would say go thin, provided it maintains structural rigidity.

I would also make sure that the pin/channel region is at least as big as the pelt it self.

8-ball

[-J-]

thats a good one 8ball thx

anyone else?

about the heat, would that rad be enought?

[8-Ball]

Are you trying to design a block for two 170W pelts or is one on the cpu and one on the gpu?

8-ball

[killernoodle]

Make sure the cold plate isnt thin though. Get something with some mass to it.

[8-Ball]

The best be would be to have two cold plates and use one to establish the optimum thickness.

Start thick and mill a small thickness off layer after layer leaving recording the temperatures each tim.

8-ball

[Jabo]

With TECs 'cold plate' thicknesis pretty critical.
Thermoelectric Cooling module is build of large number (depending on size and cooling power) of thermocouples. Each thermocouple produces module's max dT but has only fraction of module's quoted Qmax. Block side should be as thin as possible (MCW50-T by Swiftech empluys direct module cooling). It does not always hold true since with smaller modules there's not enough surface area to dissipate thermal energy into your coolant at satisfactory rate.
Heat source side is a different matter. Due to size difference (thermal energy density per mm^2) between the module and heat source there's a need for a 'energy funnel' to channel/distribute (depends which way one is looking) thermal gradients. This is where 'cold plate' thickness comes into play.
To cut it short, the smaller the heat source the thicker the cold plate should be (A64 series of CPUs have quite large IHS, which is also of substantial thickness, which allows for thinner cold plate) and vice versa

[8-Ball]

Originally Posted by Jabo

With TECs 'cold plate' thicknesis pretty critical.
Thermoelectric Cooling module is build of large number (depending on size and cooling power) of thermocouples. Each thermocouple produces module's max dT but has only fraction of module's quoted Qmax. Block side should be as thin as possible (MCW50-T by Swiftech empluys direct module cooling). It does not always hold true since with smaller modules there's not enough surface area to dissipate thermal energy into your coolant at satisfactory rate.
Heat source side is a different matter. Due to size difference (thermal energy density per mm^2) between the module and heat source there's a need for a 'energy funnel' to channel/distribute (depends which way one is looking) thermal gradients. This is where 'cold plate' thickness comes into play.
To cut it short, the smaller the heat source the thicker the cold plate should be (A64 series of CPUs have quite large IHS, which is also of substantial thickness, which allows for thinner cold plate) and vice versa

I was under the impression that dTmax could only be achieved when Q=0, and as soon as Q>0 then dT<dTmax.

This is why you want as big a TEC as possible, as the smaller Q relative to Qmax, then the closer dT is to dTmax.

8-ball

[Jabo]

Originally Posted by 8-Ball

I was under the impression that dTmax could only be achieved when Q=0, and as soon as Q>0 then dT<dTmax.

This is why you want as big a TEC as possible, as the smaller Q relative to Qmax, then the closer dT is to dTmax.

8-ball

Yes, you are 100% right m8
I was talking about a single thermocouple and you are reffering to a module as a whole. With Q=zero dT of single thermocouple is the same as dT specified for the whole module. With dT=zero for single t/couple Q capacity of it is going to be 1/100th that of the module as a whole (I am using imagionary t/couples count of 100 in our imaginary module ). With thin 'cold plate' t/couples outside thermal energy propagation sphere will have little to no use therefore decreasing cooling power of thin 'cold plate' solution (only fraction of t/couples will be in use decreasing Qmax available but maintaining dTmax as specified for the whole module @ Q=zero).
Sb here linked to Slovakian site with quite nice thermal sim renderings for various thicknesses of cold plates and resulting temps.

[8-Ball]

I see where your coming from now.

Cheers for clearing that up.

8-ball

[Butcher]

Don't forget clamping pressure either - TECs need a lot of pressure.

[-J-]

8ball, the idea is to use 2 170w TECs on parallel on the same block yes.

each is 5x5cm big, so its hard to make that work.

i will take into account what you have mentioned here ppl, thx.

any other recomendation is welcomed.

[SysCrusher]

Originally Posted by -J-

8ball, the idea is to use 2 170w TECs on parallel on the same block yes.

each is 5x5cm big, so its hard to make that work.

i will take into account what you have mentioned here ppl, thx.

any other recomendation is welcomed.

Check this page out. This should help you get a good idea where to start on the cold plate issue.

http://thermal-management-testing.co...0materials.htm

[-J-]

thx

ill check it out

i would love to get a silver cold plate, but it exceeds my budget right now.

maybe that would be a nice future upgrade

[8-Ball]

Out of curiosity,

what do folks use to clamp the coldplate to the base of their blocks. On the one hand, we need to provide up to 300Psi over the TEC which would imply steel hardware, but on the other hand, we want to minimise the conduction of heat from the hot side to the cold side, something steel hardware would do quite nicely.

Can nylon bolts provide sufficient clamping pressure for a TEC, or would they lose their threads.

8-ball

[Jabo]

Originally Posted by 8-Ball

Out of curiosity,

what do folks use to clamp the coldplate to the base of their blocks. On the one hand, we need to provide up to 300Psi over the TEC which would imply steel hardware, but on the other hand, we want to minimise the conduction of heat from the hot side to the cold side, something steel hardware would do quite nicely.

Can nylon bolts provide sufficient clamping pressure for a TEC, or would they lose their threads.

8-ball

I'd go for the size here. Thermal spread in 3D is spherical in uniform materials (copper) so in theory if you place clamping pints at appropriate distance from the cooling element parasitic heat migration is going to be limited sufficiently to be disreaded. Comparing it to loses caused by improper mounting and non-perfect contact conditions... . Other solution (prick your ears manufacturers) is to use veru high tensile strenght material like high carbon steel (cheap) or many others in a form of wires therefore severly limiting cross sectional area of contact and increasing thermal resistance of the whole clamping assy. See construction industry for solutions.

[8-Ball]

I like the idea of using high tensile strength but thin bolts.

With the application I have in mind, there wouldn't really be room to have the clamping points sufficiently far away from the TEC to limit parasitic heat migration.

Will look into it.

8-ball

[Jabo]

Originally Posted by 8-Ball

I like the idea of using high tensile strength but thin bolts.

With the application I have in mind, there wouldn't really be room to have the clamping points sufficiently far away from the TEC to limit parasitic heat migration.

Will look into it.

8-ball

Yeah, MoBos can be a tad cramped at times The thing is that with appropriate wire you can achieve immense breaking point load values using extremely small cross sectional area! Fixig (tensing) such wires would be through normal, threaded type of bolt. Imagine regular threaded bolt of a size to your liking. Cut off two pieces of such a bolt (lenght dependant of your choice of nut) and drill a hole in the middle through the lenght of both pieces. Hole should be berely larger than your wire. I think you are catching my drift by now

[BillA]

to my limited knowledge, there is only one 'enthusiast' mfgr presently offering TEC head assemblies (assembled and insulated)
-> they ARE using multiple high-strength small fasteners having low thermal conductivity, as compared to copper

perhaps TEC offerings elsewhere are more limited

Jabo
it is sometimes worthwhile to distinguish between sellers of components, and manufacturers of devices
one is not accountable for performance, the other is

[Jabo]

Originally Posted by unregistered

to my limited knowledge, there is only one 'enthusiast' mfgr presently offering TEC head assemblies (assembled and insulated)
-> they ARE using multiple high-strength small fasteners having low thermal conductivity, as compared to copper

perhaps TEC offerings elsewhere are more limited

Jabo
it is sometimes worthwhile to distinguish between sellers of components, and manufacturers of devices
one is not accountable for performance, the other is

Sorry d00de but I don't have the faintest?
When and where did I otu the two of them in the same bag?
I know what Swiftech does and I am using Swiftech products exclusively atm.
Steel (or steel alloy) bolts have lower conductivity than copper indeed but as you perfectly know structural redundancy of a bolt is MASSIVE and wasteful compared to quite elegant tensile wire fastening solution not to mention expotential decrease in cross sectional area largely eliminating parasitic thermal gradient energy migration.

I does require more workman hours (marginally more if mass produced and appropriate assembly procedure is applied) but I think it is worjth it especially with the dawn of ~300watt TEC modules and Scotties

and I know you do not need to pric your ears snce you got a finger in almost every pie

EDIT: and I know that TECs bigger than 226 watters are older than most members of this forum

[BillA]

we are working with 360W TECs for present customers, I know of what you speak

in an earlier post you (I think ?) commented on the spherical heat propagation,
I have experimented with other, and insulated, fasteners but found no improvement
- not to say such cannot be done, just can't see it from the mfgrn/QC side
your wires are indeed most elegant, and as an engineer I say Bravo !
but the wire anchor, tensioning, and attachment leave me using threaded connectors

(small) economies of scale can be quite limiting, and TEC heads (sold to DIYers) are a really small part of the business

[Jabo]

Originally Posted by unregistered

we are working with 360W TECs for present customers, I know of what you speak

in an earlier post you (I think ?) commented on the spherical heat propagation,
I have experimented with other, and insulated, fasteners but found no improvement
- not to say such cannot be done, just can't see it from the mfgrn/QC side
your wires are indeed most elegant, and as an engineer I say Bravo !
but the wire anchor, tensioning, and attachment leave me using threaded connectors

(small) economies of scale can be quite limiting, and TEC heads (sold to DIYers) are a really small part of the business

Sometimes I really do hate monetary system!
Unfortunately it is alwyas different to speak from sidelines about feasibility of certain solution than having your 'rael world' new tach. fasibilty study igures in front of your very eyes ...
Well, it is nice to dream from time to time

[SysCrusher]

Parasitic heat migration is that much of a problem with clamping devices? I'm not so experienced in this area but thinking about trying it out with a block idea I'v been running through my mind.

[Jabo]

Originally Posted by SysCrusher

Parasitic heat migration is that much of a problem with clamping devices? I'm not so experienced in this area but thinking about trying it out with a block idea I'v been running through my mind.

According to BillA's experiences... who knows since he says in one sentence that in his testing it did not make a difference and later that his designes use thin bolts made off low conductivity material (compared to copper), sooo....

Being serious, it all depends on at what level of performance you are at/aiming for. To tier of extreme cooling madmen by elimination of number of these 'small' losses are able to performa 'miraculous' efficiency improvements

EDIT: You know that thermal transfer ratio depends proportional on dT. dT between hot and cold sides is arounbd 50C = substantial nuff to make a difference imho

[zer0signal667]

Nobody suggested this yet- why not use some nylon washers to provide another barrier between the colplate and bolts/nuts?