Use a peltier to chill air ??

[Cacker]

Hi,

I was thinking today, what if you sandwiched a peltier between two heatsink/fan units ?

One side would produce cold air, the other hot. With a little air flow control, via ducting and housing you could blow ambient air over the chilled heatsink to produce chilled air which is vented into your box whilst ambient air is blown over the hot heatsink and vented outside your box.

You'd need thermal insulation around the peltier and between the fan bases to stop them being in contact anywhere but the peltier.

The peltier wouldn't have to struggle that much to freeze the cold heat sink since it isn't 'actively' producing any heat (as opposed to normal peltier solutions where you put the cold side onto the cpu core or a water block to chill heated water etc ec).

Workable do you think ?

Chairs

Cacker

[Aedan]

There's a reason that aircon doesn't use peltier devices. I worked through an example to see how workable it was.

I made a few assumptions here. I've not modelled the air realistically. Things like humidity make a *huge* difference to the figures, as do changes in the air temperature and all sorts of other things. Generally, this means that more cooling will be required, not less cooling!

Anyhow, I assumed we'd bought a 120mm YS-TECH FD1238 fan, which puts out 125CFM worth of air. Now, these units are not great for this work, so I'll convert them.

125CFM is about the same as 0.059m3/sec (That's meters cubed per second).

Air has a heat capacity of around 1.08kJ/kgK (That's kilo joules per kilogram kelvin). If you want to drop the temperature of 1kg of air by 1C, then it'll take 1.08kJ!

Now, air weighs approximately 1.25kg/m3 (kilogram per meter cubed). Our fan blows 0.059m3/sec, so it moves 0.07375kg of air per second (Density of air times throughput). Not much admittedly, but it is only a 120mm fan!

I've also assumed we want to drop the air temperature by 10C. We know that we have to cool 0.07375kg of air a second, as that's what the fan blows. Cooling a whole kg of air by 1C would take 1.08kJ. Cooling the same amount by 10C would take 10.8kJ.

So we multiply 10.8 by 0.07375 and get 0.7965. As we were dealing with kg/sec, we end up with 0.7965kJ/sec. 1J/sec is 1W. Hence, we need 0.7965KW to cool that air. Converting that into watts, gives up 796.5W of cooling required.

That's just for a 10C drop on a single 120mm fan. Your peltier will need to pump 796.5W worth of heat, which would leave your hotside heatsink trying to dispose of the best part of 1300W of heat!

That's why peltier devices are not used to cool air.

[Aedan]

Having done all that work, I've made a small mistake. The fan actually blows 135CFM, not 125CFM, so it would require more cooling than I calculated!

To fix that, pretend the fan is a YS-Tech FD1212387B, rather than a FD1212389B

[Sideburns]

man, you know your physics

[cloasters]

Welcome to AOA Forums, Cacker!

It's a very interesting question! Those dang engineers are always dashing the hopes of theoreticians upon the rocks of hard science. Of course, theoreticians are usually engineers too.

[Aedan]

Originally Posted by PornoMonkey

man, you know your physics

Nah, the answer's only an approximation erring on the low side. On the other hand, it gives a good guide to feasability, which is all that is needed at this point in time!

I will admit, I did collar Kaitian to get some figures from him.

[Aedan]

Originally Posted by cloasters

Those dang engineers are always dashing the hopes of theoreticians upon the rocks of hard science.

Yes, I had entertained doing something similar to Cacker once upon a time, then thought the better of it!

[cloasters]

Consulted Kaitain for some figures? Holy Maloney, shocked I am! Not much flummoxes the dynamic duo of Kaitain and Aedan. Boy howdy!

[Soul99]

also you need drip trays for the cold heatsink >

This has been tried before, with hot heatsink poking out the top and the cold one inside,
as aedan points out its not very effective at the power levels its capable of doing.

TEC's are really only any good when you can cool the hotside with watercooling and have a suitable way of removing the heat. They just produce so much heat tho its a pain in the ass till you get the set up just right.

Its fun playing with them tho, that first time when you turn it on and watch it ice up its impressive (if you can get the hot side cold enough), i measured -36c once with chilled water.

[Kaitain]

Originally Posted by Soul99

TEC's are really only any good when you can cool the hotside with watercooling and have a suitable way of removing the heat. They just produce so much heat tho its a pain in the ass till you get the set up just right.

In fact, TECs are only suitable for small applications where the amount of heat to be moved is only about 100W.

[Soul99]

Just about what an over-volted over-clocked cpu puts out >

I got 1 226w pelt and it does its job wonderfully, i even snapped the wire off and soldiered it back on and it was ok.
Going back to the chilled air idea, i think the pelt will prolly die pritty quick due to thermal death as the hot side will be v hot even with a large heatsink and loud fan and the cold side may well be warmer than the air in the case!

anyone had any experiance with the max temp a pelt can take?

[Soul99]

just had a thought, a 226watt air cooled pelt would die quick but what about a really low power one? it might be able to get decent sub ambient temps if the hot side aint putting out too muich heat and you can cool it well enough.
Anyone sandwiched a low power pelt between 2 heatsink/fans before?

even if it did work its not efficient use of the pelt and the actually effect it will have on overall cpu and case temps will prolly be small.

[Kaitain]

Originally Posted by Soul99

just had a thought, a 226watt air cooled pelt would die quick but what about a really low power one? it might be able to get decent sub ambient temps if the hot side aint putting out too muich heat and you can cool it well enough.
Anyone sandwiched a low power pelt between 2 heatsink/fans before?

even if it did work its not efficient use of the pelt and the actually effect it will have on overall cpu and case temps will prolly be small.

If you have a smaller TEC then you pump less energy. If you pump less energy then you don't cool the air so much. If you don't cool the air so much, then what's the point? The amount of extra o/c you can get out of a drop in temperature of only a ºC is not a lot. Whichever way you look at it, peltiers are just not appropriate for this application, however large or small they are.

[drow_elf]

what if you used this in a sealed case? and not for staying much lower than room temp but for "maintenance"

it's still not very feasible (you would need to move tons of heat from the CPU and PSU)

[Kaitain]

Originally Posted by drow_elf

what if you used this in a sealed case? and not for staying much lower than room temp but for "maintenance"

it's still not very feasible (you would need to move tons of heat from the CPU and PSU)

Even with a sealed case, you'd be better off using more traditional phase-change systems. I love this idea: an air-conditioned case!! Anyway, phase change (think: refrigerator) is much more efficient for large heat loads and for cooling fluids such as water or air than TECs. TECs have their advantage in that they're compact and don't really require any expertise to set up and use.

I quite like the air-conditioned case idea. Use the cold side loop to chill the CPU, GPU and northbridge directly in a vapochill kind of way, but typically if you use a fridge compressor it'll end up being so vastly overrated that you'll have enough spare capacity for an air inlet chiller/filter/dehumidifier. Much healthier for components

Anyone feel like building one? I don't have either the time, nor the ability to lay my hands on freon in a hurry...

[drow_elf]

i've been wondering about using an in window air conditioner unit (the R134 coolant or whatever) dunking the "cool" radiator into a resivor and chilling some water down.

not sure if it would work though: even though you have a very efficient heat transfer the capacity of the A/C unit may be too small to be effective on such a mass of water (that also has a very high specific heat compared to air) basically the compressor would have to run on such a long cycle that its life would be too short or if you ran it for shorter cycles the temp decrease wouldn't really justify the extra cost.

[VIPER_HWBG]

And what you think about this:
http://www.swiftnets.com/products/mcx462plusT.asp

It has a 226W pelt and it is very good performer, but it can be used with CPU's producing no more than 100W. And for example mine CPU @ 2500 or 2600mhz produces more than 100W. The condensation is the other problem.

[Soul99]

I think that you'll have a heatsink trying to dissapate about 300w of heat inside your case!
Thats gonna be one warm case. (and room after a while)

[Kaitain]

Originally Posted by VIPER_HWBG

And what you think about this:
http://www.swiftnets.com/products/mcx462plusT.asp

It has a 226W pelt and it is very good performer, but it can be used with CPU's producing no more than 100W. And for example mine CPU @ 2500 or 2600mhz produces more than 100W. The condensation is the other problem.

EDIT: There I go sounding like an arrogant git again. It's early on Monday morning and I haven't had any coffee yet... forgive me. Swiftec claim it's only designed to dissipate 70W continuous with the stock fan. They have a rather nice characteristic curve on there which shows that the increase in core temperature with thermal load: http://www.swiftnets.com/assets/imag...ange_chart.jpg

Optimal cooling capacity: will dissipate 70 Watts continuous thermal load to ambient temperature with 80CFM fan.

Obviously the 100W figure I gave is not fixed. It's a rough guide for where the top end of the "efficient" operating range lies. Depending on the TEC this could be anywhere between 50 and 150W...

The cooling curve Swiftec gave looks about right for a TEC. It's so unusual for a manufacturer actually to test their product properly and report realistic (even conservative) results.

I must say, I liked their "crash test", too

Crash tests were conducted by repeatedly (3 times) dropping a case equipped with an MC462 heatsink from the roof of a second story building. The heatsink remained securely attached to the processor, and the processor didn't suffer any damage.

[VIPER_HWBG]

Originally Posted by Kaitain

EDIT: There I go sounding like an arrogant git again. It's early on Monday morning and I haven't had any coffee yet... forgive me. Swiftec claim it's only designed to dissipate 70W continuous with the stock fan. They have a rather nice characteristic curve on there which shows that the increase in core temperature with thermal load: http://www.swiftnets.com/assets/imag...ange_chart.jpg



Obviously the 100W figure I gave is not fixed. It's a rough guide for where the top end of the "efficient" operating range lies. Depending on the TEC this could be anywhere between 50 and 150W...

The cooling curve Swiftec gave looks about right for a TEC. It's so unusual for a manufacturer actually to test their product properly and report realistic (even conservative) results.

I must say, I liked their "crash test", too

Yes, you are right man, but I think that this cooler is good for not so insane clockers, who want to achieve better temps, or for people who live in warm places. As for the "hot case" - Righ now I'm building a case that will have more than 600CFM airflow, so I don't bother about the "hot case" and room :lol:. My room temp at this point is 35C - yeah its hot here .