Dynamic Peltier cooling.

[MadHacker]

I have an idea that I want to bounce off you guys.
Dynamic Peltier cooling.

In the figure below is a standard water cooling system with an extra heater core, pump,
and 2 water blocks with a peltier sandwiched between it. Either a 226watt or 2x 169watt I haven’t decided yet.

There is a temperature probe in the water reservoir and when the system is idle the water temperature
shouldn’t be much higher them ambient temperature. But when the CPU, GPU, North bridge under load,
the temperature in the reservoir would increase causing the Peltier and the Fans in the second heater
core to turn on. The “Device control” will control that. Exactly how I haven’t figured out yet but it can’t be all that hard

I have been reading about power usage of the peltier and having it on all the time would significantly increase the hydro bill.
This way the peltier would only be used when the system would run warm.



Think it would fly?

[Khledar]

I like the idea of the control, very smart, but I don't know enought about pelts to know if they're efficient in the dual block configuration.

[bigben2k]

If you look at it closely, how is it better than simply running with two rads?

If you look even closer, the pelt would only pull heat out of the coolant, then put it right back, and then some!

Now if you cooled the pelt on a seperate loop, you might be getting closer to something that's more functional, but you're back to the inefficient chiller.

[MadHacker]

Originally posted by bigben2k
If you look at it closely, how is it better than simply running with two rads?

I was thinking that running the peltier would cool the water enough to get it below ambient temperatures.

If you look even closer, the pelt would only pull heat out of the coolant, then put it right back, and then some!

the second heater core wouldn't be efficient enough to draw out the heat the peltier generated?

[/b]Now if you cooled the pelt on a seperate loop, you might be getting closer to something that's more functional, but you're back to the inefficient chiller. [/b]

I know it goes back to the standard inefficent chiller. I just thought the advantage was having it turn on only when the temperature went above a specified value. since my machine runs idle 80-90% of the time i thought it might be appropriate.

[Alchemy]

It won't work.

You need a hot loop and a cold loop if you want to use a TEC as a water cooler. Otherwise, you're just pumping heat into the loop and more or less banging your head against the Second Law.

At best, you'll get the same performance you would have if you added another heat source (like another CPU) instead of a peltier device.

Alchemy

[Khledar]

As to the comments re: adding more heat, that may be true, but if the Rad. After the hotside of the pelt and before the cpu etc is able to fully, or near fully dissipate that heat, I don't think that would be a concern. Obviously this would need to be a very good rad - but in theory - it is possible, unless I'm missing something. But I don't think my thermodynamics are that bad. Simple deltaQ + and deltaQ - no?


Edit: Also noticed you're from BC MadHacker, nice to see the representation

[bigben2k]

What you're not taking into account, in your assumptions, is the flow rate.

Under a low flow rate, what you say might be true, but in a PC cooling loop, the flow rate is such that the temp of the coolant varies very little, from one point to any other. In fact, most of us don't have the tools to be able to measure that difference!

Think about it.

[Alchemy]

Originally posted by Khledar
As to the comments re: adding more heat, that may be true, but if the Rad. After the hotside of the pelt and before the cpu etc is able to fully, or near fully dissipate that heat, I don't think that would be a concern.

Of course it can fully dissipate the heat - at least, the heat not dissipated by the other rad. But the higher heat load will result it a higher water temperature going through the rad, which will cause a higher water temperature through the waterblocks, which will cause a higher CPU temperature.

Since you can't get a very high temperature gradient going in the loop - that is, the difference in water temperature from one point to another will be fractions of a degree Celsius - the benefit of higher temperature through the radiator and lower temperature through the blocks will be totally overshadowed by the great amount of heat going into the system. The end result will be a WC system equilibrating with higher water temperatures everywhere.

The effect is something like placing an A/C unit in the middle of a room. It will just make the room hotter if the heat coming out the back isn't pulled away.

TECs require a heat sink thermodynamically separate from the cold side.

Alchemy

[jaydee]

Originally posted by Alchemy
It won't work.

You need a hot loop and a cold loop if you want to use a TEC as a water cooler. Otherwise, you're just pumping heat into the loop and more or less banging your head against the Second Law.

At best, you'll get the same performance you would have if you added another heat source (like another CPU) instead of a peltier device.

Alchemy

That pretty much sums it up.

Put the pelt on it's own loop if you want to do any good with it. Putting it in the same loop is just cooling one side and heating the other side which cancels each other out, except the pelt is inefeicent by design so in turn you will actually be ADDING more heat to the loop than removing.

[MadHacker]

I can quote everybody on how bad a design it is...
but i don't think everybody wants to reread the whole thread.
i just thought i'd bounce it off a few people.

I have lots of ideas.. most of them are pretty bad.
once in a while i have i good one...
thought i had one here...
but nope..
thanks for the input...

[jaydee]

Originally posted by MadHacker


I have lots of ideas.. most of them are pretty bad.
once in a while i have i good one...
thought i had one here...
but nope..

It is a collolation of bad ideas that will make you come up with one great idea usually. When you are told why your ideas are bad and can agree then you have made a great step forward. Wish some of the other people in this forum would catch onto the concept that criticism of a bad idea is needed for them to progress in a positive direction witch will create a good idea. Your on the right track though. This idea is not bad at all if you mod it as described. People have done it before with decent results.

[Khledar]

I see what you're saying now BigBen, but is this due to the radiator design being maximized for other applications?

Like, would it be more benficial if a "Unit B" in a loop could change the temperature of the water from "Unit A" to a lower temperature before "Unit C" (Where A -> B ->C) Is this possible, or are the flow rates too high - or is it the design of the radiators.

I'm just curious if this can somehow be improved, but maybe this mythical "Unit B" wouldn't actually help - that's why I'm asking - If it could help it might be worth investigating further.

[bigben2k]

The flow rate is indeed too high for that design to work the way it's intended to.

The hot side of the pelt really needs to have its own loop. Since your design already calls for two pumps, you already have all the elements to make it work.

If you try to reduce the flow rate to make it work, then all the elements (block and rad) will work with a much lower efficiency, and that will negate any benefit.

[Khledar]

What about reducing the flow rate in just the radiator - I'm not really talking about the above system speciaifcally but in general - would running two rads in parallel (which would ofcourse be how to recude flow only in a radiator increase their efficiency dramatically/at all?

Like is it unreasonable to expect some sort of temperature drop across a "unit"?

Like, what about an evaporative solution - wouldn't there be a significant temperature difference between the input water and the water that remains after evap.? Or is it similarly that the water is flowing too fast to "isolate" such a difference?

[bigben2k]

It'll be the same, any way you look at it.

Reducing the flow rate through the rad only might work, but then you're talking about two pumps, and if you're talking about two pumps, you might as well go with two loops. With two loops, you'll have a better performance.

[RoboTech]

Hey MadHacker,

I believe your idea could work if you made a few simple changes (as others have suggested). Like this...



Split the TEC into a cold-loop and hot-loop configuration. Get rid of the raddy on the cold-loop side (put it in parallel with the raddy on the hotside if you want to use it).

Coming up with a control circuit is your next challenge! That's a lot of power to control and regulate (and TECs don't like PWM).

Good luck...
(keep those posts and ideas coming)

[MadHacker]

Originally posted by RoboTech
Split the TEC into a cold-loop and hot-loop configuration. Get rid of the raddy on the cold-loop side (put it in parallel with the raddy on the hotside if you want to use it).

Coming up with a control circuit is your next challenge! That's a lot of power to control and regulate (and TECs don't like PWM).

I was thinking about that. but what i'm looking for is something dynamic, that will turn on the TEC only when the CPU is in full use.. To not use much power when the system is idle.
unfortunatly i have to pay for my power usage.. and did some calculations and discovered that a TEC or 2 running full time is EXPENSIVE.
There for the dynamic idea.

But i guess the only realy efficient way to cool the PC is to use Phase Change as in a water chiller.

[Broh'kehn]

There is a way to create a control system.
The first thing you needs is a pair of operational amplifiers, you can get these at an electronics store, also you need a pair of thermocoples or similar and a few variable resistors.
The op-amps can be tuned to give off a signal when the temperature is higher or lower than a given value. this is then fed into a logic circuit controlling a relay, this switches the peltier on.
Im currently trying to design a logic circuit to allow this to happen but the challenge is to make it switch on at a high value and stop it again at a low value.

[wymjym]

Originally posted by Broh'kehn
There is a way to create a control system.
The first thing you needs is a pair of operational amplifiers, you can get these at an electronics store, also you need a pair of thermocoples or similar and a few variable resistors.
The op-amps can be tuned to give off a signal when the temperature is higher or lower than a given value. this is then fed into a logic circuit controlling a relay, this switches the peltier on.
Im currently trying to design a logic circuit to allow this to happen but the challenge is to make it switch on at a high value and stop it again at a low value.

Pelts do not live a long life if they are swithced on/off a lot, as you have suggested with relays.
PWM is a better solution, contrary to what was stated earlier.
VPC makes the least expensive units I know of (and have used) to control the pelts based on temperature. TEDistributing has a line of controllers but they are more lab type pieces and expensive.
I have seen some folks that built their own, with great success. So this is a 'green' approach that works and makes sense.
wj

[bigben2k]

Pelts don't like to be switched, that's true, but a PWM solution would require to have its output smoothed out.

An adaptation of Yo_Duh's power sharing circuit might be easier, where the thermal probe controls the MOSFET, and lowers the curent supply.

[wymjym]

Since all of my pelt chiller parts are either hand/home-made or way out of warranty, I am planning on 'modifying' one of the vpc controllers I have. If it works out as planned I'll be able to 'control' a lot more than 200 watts. I measured the output of the vpc and found that in fact it is always putting out a slight amount of power and when the threshold is hit the power jumps as does the frequency. So I guess it (the vpc) is a combination of pwm and 'Yo_Duh's power sharing circuit '...don't really know but it has been totally reliable for some time now. I am going to add several high power mosfets, additional cooling area, and a higher AMP circuit breaker to the controller. I think that the logic is all low level and it simply triggers the onboard mosfets...if this is so then much higher amp loads can be handled with adequate cooling (large sheets of plate aluminum). The vpc circuit is potted so I cannot see what is inside and makes it tick but the mosfets and fuse(circuit breaker) are accessable which will allow me to add a couple of jumpers and see what I can do to up the power handling.
wj

[RoboTech]

Originally posted by wymjym
PWM is a better solution, contrary to what was stated earlier.

OK, I stand corrected...

The statement I made earlier "Pelts don't like PWM" is too general and as such not correct. PWM can be used successfully to regulate the TEC voltage as long as the voltage doesn't exceed Vmax and the frequency is kept above 2KHz to minimize thermal stresses. Power MOSFET's are a good way to go.