My tec liquid chiller project

[Vinas]

I can confirm a similar problem with my e8400 and CoreTemp/Everest/RealTemp... The temperatures with these programs get stuck at 30C for CoreTemp and 20C for RealTemp. Smart Guardian reports idle temps at 242C which should translate to -13C. I had to calibrate and offset the BIOS temp using the RealTemp method... After that I undervolted and underclocked my CPU to 1.8GHz just to make certain, and the cores were still stuck on 20C/30C... The temps eventually go up as CPU load gets over 80% but never below 20C/30C.

[leuler]

I have some power consumption numbers.

Chiller ( PSUs, TECs, fans, and pumps ) - 925 watts at 30C ambient

This is just a little over the max wattage that I had planned for the
chiller. However, the ambient temperature at my house will be around 20C,
so the power consumption will be higher ( at peak loads ) when all is said
and done.

Here is how it breaks down :

Pumps - 104 watts
Fans - 54 watts ( approx)
PSUs and TECs - 767 watts

I forgot to try to see how many watts the PSUs use with no load. Anyway,
the calculator at http://www.ferrotec.com/products/the...tail.php?id=70
shows that ( assuming my Chinese TECs are similiar enough to the Ferrotec
TECs) the TECs are pulling about 4 amps at 35C hotside, for a total of
about 650 watts ( at 11.6 volts ).

Would a PSU consume 55 - 56 watts with no load?

[Uncle Jimbo]

I have some power consumption numbers.

Chiller ( PSUs, TECs, fans, and pumps ) - 925 watts at 30C ambient

This is just a little over the max wattage that I had planned for the
chiller. However, the ambient temperature at my house will be around 20C,
so the power consumption will be higher ( at peak loads ) when all is said
and done.

Here is how it breaks down :

Pumps - 104 watts
Fans - 54 watts ( approx)
PSUs and TECs - 767 watts

I forgot to try to see how many watts the PSUs use with no load. Anyway,
the calculator at http://www.hebeiltd.com.cn/peltier.d...TEC1-12706.pdf
shows that ( assuming my Chinese TECs are similiar enough to the Ferrotec
TECs) the TECs are pulling about 4 amps at 35C hotside, for a total of
about 650 watts ( at 11.6 volts ).

Would a PSU consume 55 - 56 watts with no load?

It is amazing how close your actual numbers came to the original design point calculations. If you tell me what gauge wir you are using between the PSU and the TEC, I can tell you how to measure current directly - well pretty close - from the voltage drop in the wire. Based on the performance you are getting, I suspect you are actuallly drawing a little less power than you think.

PSU efficiency varies quite a bit. 90% is pretty good, and if you are looking for the difference between calculated and measured power, then expect the PSU to use 70W or so to deliver your 675W to the TECs.

Just going to the 12710 TEC and using PWM to control the power, and making NO other changes, you get exactly the same cooling power for only 550W in.

The biggest 40mm that is easily available is the Kryotherm Drift-0.6, which is a 19915. Using those units, you will get the same cooling power for about 450W in.

Your power consumption will go down with a lower ambient, assuming the same cold side temps. You are also operating at a lower efficiency point but you also have a smaller dT requirement. If the hot side remains at 6C or so above ambient, and ambient drops to 20C, your current 12706 rig will drive the cold side to 10C lower than it is now. If you drop power to maintain your -1C load temp, you would need only 7.5V in, and total power in is only 260W.

This is really a killer design, and you have a lot of good options to either reduce the power in or get even lower temps.

[Uncle Jimbo]

noticed a problem with the circuit while testing a prototype. Pin 1 (the control) sets 100% duty cycle when the control is at .7 of Vdd. If the value becomes greater than .74 of Vdd, the MIC502 senses this as an overtemp fault, and shuts down the output. That will happen at 25C if the pot is set above 50%. To avoid this problem, put an 8.2V Zener Diode between pin 1 and ground. I will put up a revised schematic in a bit.

[leuler]

It is amazing how close your actual numbers came to the original design point calculations. If you tell me what gauge wir you are using between the PSU and the TEC, I can tell you how to measure current directly - well pretty close - from the voltage drop in the wire. Based on the performance you are getting, I suspect you are actuallly drawing a little less power than you think.

PSU efficiency varies quite a bit. 90% is pretty good, and if you are looking for the difference between calculated and measured power, then expect the PSU to use 70W or so to deliver your 675W to the TECs.

Just going to the 12710 TEC and using PWM to control the power, and making NO other changes, you get exactly the same cooling power for only 550W in.

The biggest 40mm that is easily available is the Kryotherm Drift-0.6, which is a 19915. Using those units, you will get the same cooling power for about 450W in.

Your power consumption will go down with a lower ambient, assuming the same cold side temps. You are also operating at a lower efficiency point but you also have a smaller dT requirement. If the hot side remains at 6C or so above ambient, and ambient drops to 20C, your current 12706 rig will drive the cold side to 10C lower than it is now. If you drop power to maintain your -1C load temp, you would need only 7.5V in, and total power in is only 260W.

This is really a killer design, and you have a lot of good options to either reduce the power in or get even lower temps.

Thanks to my math and physics background and the number of hours I spent
researching, I pretty much knew what to do to meet my design parameters.
However, constructing the chiller and it actually coming close to my
calculations surprises me; I thought it would take more than one try to get
this close, given that some of the construction involved doing things that
I have never done before.
--------------------------------------------------------------------------
The wiring is temporary; the final set up will decrease voltage drop ( which
is already pretty low ). But let me describe the PSUs and how everything
is wired :

Two 600w PSUs with four 12v rails each with double 18 awg leads for each rail.
12v rail1 = 18 amps
12v rail2 = 18 amps
12v rail3 = 15 amps
12v rail4 = 12 amps

rail1 + rail2 + rail3 rated at 45 amps total.

Both rail1s are each connected to three TECs. The 18 awg leads are approx
12" long. The 16 awg leads from the TECs are approx 3" long.

Ditto for the rail2s.

The rail3s are connected to one TEC and three heatsink fans each ( 0.3 amps
each ). Lead lengths are the same as above.

The rail4s are connected to three San Ace fans ( 0.48 amps each ). The
18 awg leads are approx 18" long.
--------------------------------------------------------------------------
I really believe that this design can be improved. PWM power control alone
will make this chiller much more versatile, and as you pointed out, allow
for more cooling power ( when needed ) and more efficiency ( for every day
use ) to be built in. Also, I think I can improve heat transfer efficiency.
Maybe not with this chiller ( though I've got some ideas ), but definitely
with my next one ( sometime down the road ).

[Uncle Jimbo]

noticed a problem with the circuit while testing a prototype. Pin 1 (the control) sets 100% duty cycle when the control is at .7 of Vdd. If the value becomes greater than .74 of Vdd, the MIC502 senses this as an overtemp fault, and shuts down the output. That will happen at 25C if the pot is set above 50%. To avoid this problem, put an 8.2V Zener Diode between pin 1 and ground. I will put up a revised schematic in a bit.

here's a revised schematic that fixes the problem

[leuler]

Thank you very much.

[Uncle Jimbo]

Thank you very much.

This is not actually my original suggestion - this is a better design because it also puts the thermistor at a constant voltage which gives tighter temp control. It also allows pretty much any supply voltage - I have tested it to 30V.

[boy_gameover]

Leuler ! where do you buy copper ? How much is that ?

[leuler]

Leuler ! where do you buy copper ? How much is that ?

I bought the copper bars on eBay. Counting shipping, I paid $85 to $90
for it.

There are at least three suppliers online in the US that sell small amounts
of copper for hobbyist. If I bought the copper from them, it would have cost
$150 to $180, counting shipping.

[leuler]

I can't seem to find a thermistor probe that will have a resistance of
10k ohms when the water temp is 0 C to 10 C. All that I can find are
at least 10k at 25 C ; I need one that has a resistance of 3.5k to 5k
at 25 C ( assuming Rt/Rt(25 C) is around 2 at 10 C ).

Jimbo, what could I expect from my chiller using your controller design
and a 10k (25 C) thermistor probe? Could I use two 10k thermistor probes
in parallel?

[leuler]

By the way, I've ordered all of the parts for the controller, minus the
thermistor probe. And I've designed my enclosure for the chiller, rads,
pumps, and PSUs. So, in the next couple of months I should have a
water chiller that looks and performs halfway decent.

[HDCHOPPER]

killer project luelur well done

[Uncle Jimbo]

I can't seem to find a thermistor probe that will have a resistance of
10k ohms when the water temp is 0 C to 10 C. All that I can find are
at least 10k at 25 C ; I need one that has a resistance of 3.5k to 5k
at 25 C ( assuming Rt/Rt(25 C) is around 2 at 10 C ).

Jimbo, what could I expect from my chiller using your controller design
and a 10k (25 C) thermistor probe? Could I use two 10k thermistor probes
in parallel?

You just need to change the other resistor values to match. A NTC probe that has 10K at 25C will have a much higher value at 0C - if you don't have the table, the easiest thing is just put it in an ice water bath and measure the resistance. Then adjust the other resistors in the bridge to match.

So let's say you measure 30K at 0C. Then use something close to that for VR1, and about 1.5 times the value for R2. They don't have to be exact - if you use 27K for VR1 and 47K for R2 that will work fine.

[leuler]

You just need to change the other resistor values to match. A NTC probe that has 10K at 25C will have a much higher value at 0C - if you don't have the table, the easiest thing is just put it in an ice water bath and measure the resistance. Then adjust the other resistors in the bridge to match.

So let's say you measure 30K at 0C. Then use something close to that for VR1, and about 1.5 times the value for R2. They don't have to be exact - if you use 27K for VR1 and 47K for R2 that will work fine.

Ah, so it is the relative resistances of the probe, VR1, and R2 that matter,
not the total resistance of that branch.

Okay, I'll get the sample thermistor probe from Ametherm and also get a
trimpot and resistor to match with it.

Thank you Jimbo

[leuler]

Well, here is my version of the controller, just lacking the thermistor probe.
My very first time constructing an electronic circuit, not counting working
on a breadboard in college.

I made the positive and negative rails by splitting in half a flat copper bus
from a meter base. I thought that it was a good idea until I tried soldering
to them. That's a lot of copper to heat up, and I'm inexperienced at soldering
anyway. So that explains the scorch marks.

Anyway, if I luck, the controller will work. If not, I will blame the assemblier,
not the designer.

[leuler]

I haven't finished the controller nor have built the case yet.

However, I have been tweaking the chiller and will be testing again this
weekend. I've exchanged the TEC1-12706 pelts for TEC1-12709s. I've
also improved the insulation, upgraded the wiring, and improved the
cooling for the hotside loop ( shrouds and push/pull fans on the rads ).
I'm also cooling the northbridge with a TEC ( direct die with a SanAce
Liquid Cooler to cool the hotside plate ).

I'm not sure if I will have a camera on Saturday to take pics, but I will
report the results.

[jens21]

I'm having some problems to get the right parts shipped to the netherlands. I just cant get them here!

You said it would only cost like $20 of parts.

Is there a SLIGHT possibilty that someone could sent me parts?

I would be very VERY gratefull!

I love to get my pelts started, with 100% on both pelts the water just freezes, I need to get a grip on these pelts.

The tricky part, is the PWM-module in the middle, its just out of my reach.

That in advance!

[zipdogso]

I'm having some problems to get the right parts shipped to the netherlands. I just cant get them here!

You said it would only cost like $20 of parts.

Is there a SLIGHT possibilty that someone could sent me parts?

I would be very VERY gratefull!

I love to get my pelts started, with 100% on both pelts the water just freezes, I need to get a grip on these pelts.

The tricky part, is the PWM-module in the middle, its just out of my reach.

That in advance!

See your pm's...