Overclock.net banner

1 - 20 of 71 Posts

·
Registered
Joined
·
744 Posts
Discussion Starter #1 (Edited)
Introduction

So after weeks and weeks of deliberation I've made up my mind. This coming week I can begin drinking legally and I've decided to celebrate by making my first peltier build. But I've set restrictions for myself and I'm approaching this from a way I don't often see. My goal with this is to use peltier(s) in such a way that it is cost effective, pleasing to the eye and works well. The side goal is to make a thread that covers in detail all steps required so that an inspired individual could follow along and build a system around the same concept as mine.


The Proposition

So I mentioned I'm going off a principle that I don't often see used. That is, almost nobody keeps their computer at full load for even several minutes let alone hours. So instead of increasing cooling capacity to account for these spikes, I'm going to lower cooling capacity to basically bare minimum and instead increase the buffer space I have. In otherwords, I'm going to get a bigger reservoir. Let's talk about why this will work. Essentially we're giving ourselves a big enough buffer that we're actually working on the average heat being given off by the processor instead of the spikes.

To understand this, we'll have to understand a few concepts. Watts and BTUs. Watts is a measurement of energy (joules) over a period of time (1 second). BTU/hr's is ALSO a measurement of energy over a period of time. A BTU is the amount of energy required to raise 1lb of water by 1F. We're going to convert these units into units that make sense.


1 BTU = 1Lb of Water raised 1f
= 0.45L of Water raised .56C
= 1L of water raised .25C

In addition, you can convert watts to BTU/hr's by multiplying by 3.41. So 300w would be 1023BTU/hrs. So if we have a computer with a processor that is producing 300w of heat overclocked at full load then it would be producing 1023BTUs per hour. That is, it would:
  1. raise 1L of water by 255C in an hour
  2. raise 1023L of water by .25C in an hour

So let's play with the math. Suppose we had a reservoir that is 4L. Our 300w would raise 4L by 64c over the course of an hour. 8L? 32c. 16L? 16c. For reference, that's about 4 gallons. Now let's suppose we have a peltier cooling the water that's providing 220w of cooling power or about 750BTUs of cooling. You can subtract that from the heat being added by the processor to find that while the peltier is running and the processor is running at full load, we are adding 272BTU/hrs to the water instead of 1023BTU/hrs. So now if we crunch the numbers again: 272BTU/hrs would raise 4L by 17c over the course of an hour. 272BTU/hrs would raise 8L by only 9C over an hour and only 4C for 16L


The Specifics

The system will use the following components:

Computer hardware:
Mainboard: Asus P6T - Received
CPU: Xeon X5670 - Received
GPU: 2 x GTX 690 - Received
Memory: 16GB EVGA SuperSC - Received
Case: ThermalTake The Tower 900 - Received

Cooling Hardware:
Waterblocks:
  • Peltier block - Ultrasonic2 Chiller Blocks Received
  • Raystorm CPU block - Received
2 x 360mm Radiators: - Received
2 x Reservoir: Phobya LT 50 - Received
2 x D5 Pumps: Received
Fittings: Monsoon Matte Black 3/8" 1/2" - Received
Tubing: - Received

Controller/Peltier Hardware:
Peltier: CTE 400 (http://customthermoelectric.com/tecs/pdf/19911-5M31-28CZ_spec_sht.pdf) - Received
Arduino UNO R3 - Received
DHT22 Temperature/Humidity Sensor - Received]
DS18B20 Waterproof Temperature Probe - Received
Female USB to USB Header Adapter - Received
IRFZ34N NPN 30A 55V Mosfet - Received
100k Ohm Resistor - Received
12V Switching Power Supply - Received


Controller Theory

The controller will be built to be capable of maintaining above dew point temperature while also being capable of being set to any desired temperature. It's operation will be very simple. The Arduino will have the capability of monitoring three things: humidity, coolant temperature and air temperature. Air temperature and humidity will be used to calculate dew point. This combined with coolant temperature will determine if the peltier should be running or not.

The controller will be connected to the computer via USB and use a serial connection to report runtime data in JSON. This includes humidity, coolant temperature, air temperature, peltier status, and dewpoint.The arduino will also be capable of receiving simple commands such as setting a target coolant temperature.

All code will be open source and free to use.


Controller Links, Source, Downloads

Windows x86 Controller Interface: Github
Fritzing Wiring/PCB Source Files: Google Drive
Arduino Program:: *WILL UPLOAD TO GITHUB WHEN I GET A CHANCE - THANKS*


End

That's all for now! I will update this post as I begin purchasing parts and have pictures to show off

Just noticed I forgot to give this thread a name lol. Oh well.
 

·
TEC Veteran
Joined
·
2,007 Posts
Awesome!

What are the clocks of the Xeon going to be? Keep in mind the CTE at 12v cant really handle over ~150w before it breaks even and runs out of delta.

Only issue i see is that the negative of a large buffer is that if you ever turn off your rig, there will be a much longer time before coolant stabilizes.

Can't wait to see the progress!
 

·
Registered
Joined
·
744 Posts
Discussion Starter #3
Quote:
Originally Posted by Puck View Post

Awesome!

What are the clocks of the Xeon going to be? Keep in mind the CTE at 12v cant really handle over ~150w before it breaks even and runs out of delta.

Only issue i see is that the negative of a large buffer is that if you ever turn off your rig, there will be a much longer time before coolant stabilizes.

Can't wait to see the progress!
The clocks are going to be as high as the cooling will allow
biggrin.gif
Not sure what that number will be yet. I'm sort of hoping somewhere from 4.2 to 4.5GHz.
Thanks for the info regarding the CTE tec. What I was thinking is if the CPU is putting off more than 130w of heat at idle, I'm sort of dead in the water. I may end up adding a second TEC or increasing the voltage. The first time putting it together and testing should be enough to give direction going forward.

In regards to the coolant reaching temp, I was thinking about adding a bypass switch which would run the peltier with the system off (basically just bypass the transistor). But then there goes my safety measures for temperature control. It seems I either sacrifice the convenience of everything being in one contained package (no external box or reservoir) or I just suck it up and wait for the temps to drop on boot up.

I should have everything needed to begin working on the controller on Saturday. So I'll setup a github repo then and begin working on the controller. I'm very excited to see what could be done with the data and temperature control. I'm also interested to see what sort of voltage control I can get by controlling the transistor with the arduino. Now it's just time to sit back and collect parts and be productive.
 

·
Banned
Joined
·
3,726 Posts
ok, so I am bad with communication, in truth I have no idea how this sounds to others. so let me be clear I am not trying to shoot you down or trying to change your mind in any way about this project.

I wanted to use one of the four arduino's I own to control my own TEC's, I also have three different large transistors. the first thing I noticed is that the arduino PWM single was not strong enough to open and close the transistor with a 15Amp current wanting to flow through it. this was solved easy enough with the ardunio single amplifier about a $50 upgrade. now that I could drive my transistors, I begain my quest for control, the only single style transistor that I did not burn up was the glass tube style transistor, that was a $70 transistor. my final conclusion was that I needed to build a transistor bank of 9 or more transistors wired in a parallel connection with each other for this to work without costing a lot. that is where I stopped and decided to try a PWM DC motor controller. I really don't want to start developing electronics for my TEC project.
 

·
Registered
Joined
·
744 Posts
Discussion Starter #5
Quote:
Originally Posted by toolmaker03 View Post

I wanted to use one of the four arduino's I own to control my own TEC's, I also have three different large transistors. the first thing I noticed is that the arduino PWM single was not strong enough to open and close the transistor with a 15Amp current wanting to flow through it. this was solved easy enough with the ardunio single amplifier about a $50 upgrade. now that I could drive my transistors, I begain my quest for control, the only single style transistor that I did not burn up was the glass tube style transistor, that was a $70 transistor. my final conclusion was that I needed to build a transistor bank of 9 or more transistors wired in a parallel connection with each other for this to work without costing a lot. that is where I stopped and decided to try a PWM DC motor controller. I really don't want to start developing electronics for my TEC project.
Gotcha. I hope I got the right transistor for the job; but I'll go ahead and wire a test bench up to see if it'll work. Thanks for the info!
 

·
Registered
Joined
·
744 Posts
Discussion Starter #6

Threw together a little test. I was able to turn the mosfet on; but I realized PWM probably won't be a possibility if I'm driving the gate from the arduino. In addition I have realized my drain resistor is not nearly big enough, so I've ordered some 100k ohm resistors.
 

·
Banned
Joined
·
3,726 Posts
Quote:
Originally Posted by Skyl3r View Post

Threw together a little test. I was able to turn the mosfet on; but I realized PWM probably won't be a possibility if I'm driving the gate from the arduino. In addition I have realized my drain resistor is not nearly big enough, so I've ordered some 100k ohm resistors.
ok, so your going to put a resistor between the transistor and the TEC?
have you considered how many amps will be traveling through that resistor?
 

·
Registered
Joined
·
498 Posts
In order to drive a decent sized mosfet from an MCU you'll need to make sure it's logic level (VGS 3.3 - 5v). If the threshold voltage is higher than 5 volts it won't open fully so you'll generate a lot of heat in the mosfet.
 

·
Registered
Joined
·
744 Posts
Discussion Starter #9
Quote:
Originally Posted by toolmaker03 View Post

ok, so your going to put a resistor between the transistor and the TEC?
have you considered how many amps will be traveling through that resistor?
No no, the resistor goes from gate to source.

First let me clear up some confusion I started
tongue.gif
I kept calling my mosfets transistors. They're mosfets, not transistors.

On a mosfet the gate has capacitance and builds up charge which will allow a connection between source and drain which are the other two pins. The gate needs a way to drain or it will be stuck closed (always on) so you tie it to source. The problem with this is that when your sending a signal to the gate from the arduino, the electricity will take the path of least resistance, which is not the mosfet. So to fix this, you throw a transistor in the way so it's forced to charge the mosfet but still drains.

Quote:
Originally Posted by flak-spammer View Post

In order to drive a decent sized mosfet from an MCU you'll need to make sure it's logic level (VGS 3.3 - 5v). If the threshold voltage is higher than 5 volts it won't open fully so you'll generate a lot of heat in the mosfet.
I checked the data sheet and it said 2-4v was the range for the gate.

I have metered it and it does fully open and close; so I think it's working correctly.
 

·
Banned
Joined
·
3,726 Posts
Quote:
Originally Posted by Skyl3r View Post

No no, the resistor goes from gate to source.

First let me clear up some confusion I started
tongue.gif
I kept calling my mosfets transistors. They're mosfets, not transistors.

On a mosfet the gate has capacitance and builds up charge which will allow a connection between source and drain which are the other two pins. This pin needs a way to drain or it will be stuck closed (always on) so you tie it to source. The problem with this is that when your sending a signal to the gate from the arduino, the electricity will take the path of least resistance, which is not the mosfet. So to fix this, you throw a transistor in the way so it's forced to charge the mosfet but still drains.
I checked the data sheet and it said 2-4v was the range for the gate.

I have metered it and it does fully open and close; so I think it's working correctly.
ok, got it a source, gate, drain. I am totally following you at this point.

I only saw the one mosfet so I made some assumptions. pleases continue, this build I will watch with lots of anticipation.
thumb.gif


http://www.ni.com/en-us/shop/labview/download.html

this could help with the design process if your interested.
 

·
Registered
Joined
·
744 Posts
Discussion Starter #11
Quote:
Originally Posted by toolmaker03 View Post

ok, got it a source, gate, drain. I am totally following you at this point.

I only saw the one mosfet so I made some assumptions. pleases continue, this build I will watch with lots of anticipation.
thumb.gif


http://www.ni.com/en-us/shop/labview/download.html

this could help with the design process if your interested.
Thanks!
biggrin.gif

I'm excited to join the TEC cooling club.

I'll check the software out tomorrow. If the extent of my wiring is 2 sensors, a mosfet and a resistor, I might not need it; but it's hard to say right now.
 

·
Registered
Joined
·
744 Posts
Discussion Starter #12
Well I went ahead and placed an order for TEC chiller blocks from ultrasonic2's site. I also am working on getting another 360mm radiator. That'll be either today or tomorrow. Left on my list...
Case, Reservoirs and tubing!
There's probably a bit of cable management things I'll need.. velcro, cable ties, a way to mount the TEC chiller blocks. We'll see once I determine a case.

The cases I'm looking at right now are:

Phanteks Enthoo Primo - $230

Corsair 900D - $340

Corsair AIR 740 (Won't hold 2 360mm rads
frown.gif
) - $170

I'm open to suggestions, only restrictions are no Thermaltake or Coolermaster cases.
I've got to squeeze in at least 2 reservoirs, 2 pumps, 2 360mm rads and the chiller block assembly (including arduino and sensors). So there needs to be space. I was liking the idea of the AIR 740, because I could hide the insulated reservoir in the back and then only the hot loop would be visible from the front which would give it the appearance of a standard watercooled system. But losing out on a 360mm rad feels like a big hit. A thick 240mm could still go a long way, but I'll survey my options before making a decision.
 

·
TEC Veteran
Joined
·
2,007 Posts
Quote:
Originally Posted by Skyl3r View Post

Well I went ahead and placed an order for TEC chiller blocks from ultrasonic2's site. I also am working on getting another 360mm radiator. That'll be either today or tomorrow. Left on my list...
Case, Reservoirs and tubing!
There's probably a bit of cable management things I'll need.. velcro, cable ties, a way to mount the TEC chiller blocks. We'll see once I determine a case.

The cases I'm looking at right now are:

Phanteks Enthoo Primo - $230

Corsair 900D - $340

Corsair AIR 740 (Won't hold 2 360mm rads
frown.gif
) - $170

I'm open to suggestions, only restrictions are no Thermaltake or Coolermaster cases.
I've got to squeeze in at least 2 reservoirs, 2 pumps, 2 360mm rads and the chiller block assembly (including arduino and sensors). So there needs to be space. I was liking the idea of the AIR 740, because I could hide the insulated reservoir in the back and then only the hot loop would be visible from the front which would give it the appearance of a standard watercooled system. But losing out on a 360mm rad feels like a big hit. A thick 240mm could still go a long way, but I'll survey my options before making a decision.
Is it the triple 62 block? That with three CTE400s would be a nice setup!

For the case, get whatever can hold the most rads. I like horizontal cases since they are roomy and you have less risk of ruining GPUs or SSDs in case of any leaks. You can eliminate that 740 right off the bat, since once the case is filled with hardware and chiller stuff you will only be able to fit 2x280mm push only rads max, which is NOT enough. Hot side coolant would ruin any delta the TEC gives you and not make it worth while. I would aim for at least 2x360s...but 3x would be even better
smile.gif
.
 

·
Registered
Joined
·
744 Posts
Discussion Starter #14
Quote:
Originally Posted by Puck View Post

Is it the triple 62 block? That with three CTE400s would be a nice setup!
Yep! I am getting it with the intentions of eventually being able to run 2 or 3 CTE 400's
smile.gif


My first goal though is to prove a couple things to myself and hopefully also put some would-be peltier users' minds at easy regarding the following problems:
  1. A chilled build could be made in a reasonable sized case and look good
  2. It doesn't have to be insanely expensive over watercooling
  3. A simplistic controller could be relatively easy to make with an arduino and a couple sensors
Once I'm content with how these conditions are satisfied, then the board is getting insulated and I'm adding chilling power
biggrin.gif


Quote:
Originally Posted by Puck View Post

For the case, get whatever can hold the most rads. I like horizontal cases since they are roomy and you have less risk of ruining GPUs or SSDs in case of any leaks. You can eliminate that 740 right off the bat, since once the case is filled with hardware and chiller stuff you will only be able to fit 2x280mm push only rads max, which is NOT enough. Hot side coolant would ruin any delta the TEC gives you and not make it worth while. I would aim for at least 2x360s...but 3x would be even better
smile.gif
.
Yeah... That's depressing, but you're right. I really wanted to get the 740 because I'm running my main system in an AIR 540
tongue.gif
I'll keep looking around.
 

·
TEC Veteran
Joined
·
2,007 Posts
Quote:
Originally Posted by Skyl3r View Post

Yep! I am getting it with the intentions of eventually being able to run 2 or 3 CTE 400's
smile.gif


My first goal though is to prove a couple things to myself and hopefully also put some would-be peltier users' minds at easy regarding the following problems:
  1. A chilled build could be made in a reasonable sized case and look good
  2. It doesn't have to be insanely expensive over watercooling
  3. A simplistic controller could be relatively easy to make with an arduino and a couple sensors
Once I'm content with how these conditions are satisfied, then the board is getting insulated and I'm adding chilling power
biggrin.gif

Yeah... That's depressing, but you're right. I really wanted to get the 740 because I'm running my main system in an AIR 540
tongue.gif
I'll keep looking around.
Sweet, two CTE 400s were able to chill a CTE400 direct die block on top of an OC'd ~4.8ghz 3770k down to 8c coolant temp under load, so two should be able to get the coolant going to your Xeon down to at least 10c, which is around the dew point for most climate controlled ambients so would reach your goal without needing the third one
smile.gif
.

Im estimating the Xeon heat output @ 4.3-4.5 is probably fairly similar to the 4.8ghz 3770k+150w of TEC on top so should be a decent rough comparison? Hopefully less...never OC'd one of those chips before so not sure exactly just how hot they are.
 

·
Registered
Joined
·
744 Posts
Discussion Starter #16
Quote:
Originally Posted by Puck View Post

Sweet, two CTE 400s were able to chill a CTE400 direct die block on top of an OC'd ~4.8ghz 3770k down to 8c coolant temp under load, so two should be able to get the coolant going to your Xeon down to at least 10c, which is around the dew point for most climate controlled ambients so would reach your goal without needing the third one
smile.gif
.

Im estimating the Xeon heat output @ 4.3-4.5 is probably fairly similar to the 4.8ghz 3770k+150w of TEC on top so should be a decent rough comparison? Hopefully less...never OC'd one of those chips before so not sure exactly just how hot they are.
I'd guess the i7 + TEC would be a little higher output. I found this: https://www.tomrei.com/2015/06/intel-x5650-overclocking-resulting-a-4-37-ghz-hexacore/ which gives some power consumption figures. He doesn't specify if that's full system or how he's getting the numbers. So I'm not sure what the outlook will be yet. I think I'll really just have to get the CPU and test.

BTW, anyone know of a way to search for cases by radiator support? I can't find a search that will allow filtering on radiator support
tongue.gif

Newegg only filters by fans and PCPartPicker doesn't even have that.
 

·
Registered
Joined
·
744 Posts
Discussion Starter #18
Quote:
Originally Posted by Elloquin View Post

Not a huge fan of the manufacturer but they are inexpensive and you can fit almost anything in there... http://www.thermaltake.com/products-model.aspx?id=C_00003042
Yeah, I don't really like Thermaltake but that case looks pretty ideal for what I'm trying to do. I googled a little bit and it looks like the x9 is the same thing, just a bit bigger for about the same price. I'll have to keep checking my options.

If I'm interpreting their images right you can fit 4 360mm rads or 3 480mm rads. That's pretty good radiator support. And modular is cool. But I've always had issues with Thermaltake build quality. We'll see.

The pictures make it look deceptively small. It's 19.8"x15"x25.2"
 

·
TEC Veteran
Joined
·
2,007 Posts
Quote:
Originally Posted by Skyl3r View Post

Yeah, I don't really like Thermaltake but that case looks pretty ideal for what I'm trying to do. I googled a little bit and it looks like the x9 is the same thing, just a bit bigger for about the same price. I'll have to keep checking my options.

If I'm interpreting their images right you can fit 4 360mm rads or 3 480mm rads. That's pretty good radiator support. And modular is cool. But I've always had issues with Thermaltake build quality. We'll see.

The pictures make it look deceptively small. It's 19.8"x15"x25.2"
While still not the quality of a top tier case, Thermaltake quality has increased a LOT in the last few years. I would try to find one displayed if you have any PC shops nearby...you may be surprised.
 

·
Banned
Joined
·
3,726 Posts
well if you want a nice case, that will last for 20 years, do yourself a favor, and get a case labs PC case. they are the best built cases I have ever owned.

http://www.caselabs-store.com/the-magnum-double-wide-case-line/

my first case was a lan li $300 all aluminum case.

the next two cases I got where $400 server cases.

than I finally broke down and got a $1000 TX10 case by case labs, it is the best case I have ever owned.
 
1 - 20 of 71 Posts
Top