Open source DIY TEC and Fan controller

[Liam_G]

Seeing as i am insulating anyway and i have 2 480 rads, is a direct die TEC waterblock a better option? Just trying to fit it all in one case and if i can eliminate the chiller blocks and extra pump that would help, plus might be easier to go sub zero?? How many TEC's can you fit under one direct die waterblock? I saw digital storm did something like this, direct die TEC block and i think CPU idle was below zero.

EDIT, sorry that was a chiller setup in digital storm case, not direct die.

Double EDIT digital storm had a revised version that is direct die.
http://www.techfresh.net/digital-sto...ooling-system/
http://www.digitalstormonline.com/cryo-tec.asp

[Ultrasonic2]

This is just my block that i used to sell with a different top.

Dual 50mm TEC's

The reality is it's not enough cooling power

[Liam_G]

mhm, interesting, they are certainly cashing in on it...... I see mindchill has made a quad TEC Direct Die block for ivy bridge http://www.overclock.net/t/1254258/q...d-2500k-tested , seems to work ok, hopefully he is doing something for socket 2011 too so i can see results...

what do you think would be the coldest cold side liquid temps with a 6 TEC chiller setup? I've been doing some research on OC'ing the 3930k and seems i wouldn't get much benefit from the TEC chiller setup in terms of Overclock unless i could get close to Single Stage temps, so in the realm of -30*.... I don't want SS because of noise, if TEC's can't get me there then your sub ambient but above dew point controller/chiller setup might be the best bet, giving me 5ghz at decent temps for a safe 24/7 OC.

[Ultrasonic2]

i dont want to comment on his block so i wont. It looks nice from the out side that's for sure.

He's using high couple count high voltage TEC's i'd certainly NEVER DO THAT AGAIN as i found them to be very un reliable in the long run

Tec's can't compete with SS for pure cooling number's but can as you say be Much quite and use less power

[Liam_G]

ok, so back to TEC chillers for a second. With the chiller block is it best to have the water remain in the block as long as possible on the hot and cold side?

Would it be better to have the water remain in the block for a long time on the cold side to give it time to cool the water, but be more like a micro-fin setup on the hot side so the water doesn't remain in the block very long and get too hot?

[MadHacker]

ok, so back to TEC chillers for a second. With the chiller block is it best to have the water remain in the block as long as possible on the hot and cold side?

Would it be better to have the water remain in the block for a long time on the cold side to give it time to cool the water, but be more like a micro-fin setup on the hot side so the water doesn't remain in the block very long and get too hot?

so if the water stays in the block for a second and the water in the process increases by 1 deg, would it not be better to have the water stay in the block 1/2 a second and only raising the temperature of the water by 1/2 deg?
this way the block would constantly get fresh cooler water.

however it doesn't work that way.
no matter what the flow rate the water will remain in the block for the same amount of time in average.
whether the water flows through the block once every 10 seconds and remains in the block for 1 second or if the water flows through the block once every 5 seconds for 1/2 second.
over a 10 second interval the water will still stay in the block for a second.

the key factor is turbulence, getting the water to impact the insides of the block to get the most contact to transfer the heat is key.

perhaps someone can explain it better...

[Liam_G]

ok, so basically the water doesn't heat up that fast so are you saying that if i run the water through a 6 TEC chiller block that the water at input and output won't change a great deal and that the loop will just reach a balanced temp at the time?

I just don't want to end up with a chiller block where the first TEC gets cooler water than the last. And considering the huge heat load of 6 TECs is it still the same principal?

Also i think it will be hard to create an impingement type block to cool 6 TEC's, so creating turbulence would be nearly impossible and reverting to the older maze style might be the only option, which is why i am asking about this water temp change across the block? I see ultrasonic uses a maze style block on hot and cold side and it works so i guess i'll stick with that, but if i could gain more efficiency in the hot side by creating an impingement type scenario then i will pursue that, resistance would be huge though across 6 CPU style waterblocks, and thats not including the 3 GPU blocks in the hot side loop...

[Hell Hound]

Dual iwaki plus the 2 480 rads should be external,what about a DD external 360 w/fan shroud or a water box plus.Should be able to house chiller and rads/pumps.Maybe a third 480 is needed for ambient load 24/7.

[MadHacker]

Dual iwaki plus the 2 480 rads should be external,what about a DD external 360 w/fan shroud or a water box plus.Should be able to house chiller and rads/pumps.Maybe a third 480 is needed for ambient load 24/7.

i supose one thing to consider is how your CPU load is used.
Myself i use my system for work where i only need ocasional boosts in speed when i compile code, longest is 15 min or so.
however if your using it for video rendering or video encoding or something along the lines where your CPU is pused to max for extended periods of time then then you will need more rad.

[n00b 0f l337]

When using multiple TEC's you would be looking to provide "fresh" water to each peltier unit. This will help maximize efficiency. I would suggest multiple pin matrix and using a manifold block on top of this to allow flow to all blocks in parallel.

[Ultrasonic2]

This thread reminds me why i dont make any blocks any more and dont bother with forums. Forums are just full of stuff that people choose to believe and spout as facts as opposed to any actual facts.

Facts have become what the majority of people believe as apposed to reality

Forums have become a place where the world is flat and always will be.

[MadHacker]

This thread reminds me why i dont make any blocks any more and dont bother with forums. Forums are just full of stuff that people choose to believe and spout as facts as opposed to any actual facts.

Facts have become what the majority of people believe as apposed to reality

Forums have become a place where the world is flat and always will be.

Can you expand on this?
if i posted something in error then i would like to be corrected....

[n00b 0f l337]

He's simply trolling. Forums are full of people looking for answers or theories. Not a hubbub of flat earthers.
It's also full of trolls. At least you know where you stand.

[Ultrasonic2]

Well the problem becomes why should anyone believe me as apposed to what they already believe to be true. So there is no point in me saying how it is or what I believe to be true, I have to prove it somehow. So in the simplest form I have to build a complex Solid works model and come up with some simulations to prove it. This can take weeks of my time and then what.

What I have realised after 16 years of TEC’ing is that even if I do the above people simply choose to believe what they want anyway

So in the end I waste my life

Rather amusingly to me I got bad from Overclock.net because the majority of people got pissed off with me correcting them all the time. I could sign up again, but I loose nothing by not commenting on their site it’s in fact their loss not mine. I wish it didn’t have to be like this but it would seem it has to be


Here is an example

“When using multiple TEC's you would be looking to provide "fresh" water to each peltier unit. This will help maximize efficiency. I would suggest multiple pin matrix and using a manifold block on top of this to allow flow to all blocks in parallel.”

Where did this come from it’s stated as a fact but there is nothing to back it up .

If it was expressed as an opinion that would be fine but it’s not


Im not Trolling this was my thread about how to make a TEC controller THIS THREAD EXIST BECAUSE IM TRYING TO HELP OTHERS but yet again i wonder why i bother.

Others have tolled it and taken it WELL off topic.

I value others opinion but please keep to the topic for my sanity

[n00b 0f l337]

I can extrapolate in theory if you'd like.
Fresh water would be provided at a lower temperature than preheated water as seen in serial peltier blocks. The reduced water temperature results in greater heat transfer due to an increase in delta T. Thus to increase efficiency you could either attempt to create counter flow (which does not appear to apply well to heat pumps), or you can provide fresh water to each peltier unit as yours and most designs seem to encompass. As having only 7 years in TEC'ing might seem a bit meh, I've also created multiple blocks and tested this. Sadly I Teampuss and Procooling just aren't what they once were.

[MadHacker]

I can extrapolate in theory if you'd like.
Fresh water would be provided at a lower temperature than preheated water as seen in serial peltier blocks. The reduced water temperature results in greater heat transfer due to an increase in delta T. Thus to increase efficiency you could either attempt to create counter flow (which does not appear to apply well to heat pumps), or you can provide fresh water to each peltier unit as yours and most designs seem to encompass. As having only 7 years in TEC'ing might seem a bit meh, I've also created multiple blocks and tested this. Sadly I Teampuss and Procooling just aren't what they once were.

so what is the difference in temperature of the water from when it goes into the block with a TEC compared to the temperature of the water leaving the block...
difference of 1/2 a degree max? I never measured the temp difference so I can't say...

back to subject...

so this should be the final numbers
Across the 2 resisters there is a 1.6721v drop while dissipating 20.735watts of heat total.
Across the TEC there is a 24.768v drop while dissipating 306watts of heat total.
Across the Mosfets there is a 0.0285v drop while dissipating 0.353watts of heat total.

just out of curiosity... is there any way to not dump 20.735watts in the 2 resistors?
dumping power there seems like a waste.
if it can be done via mosfets, wouldn't it be more efficient?
I know this is extreme but but for a machine that is using high CPU loads for extended periods of time,
would it not be better to minimize the wasted watts?

[Liam_G]

Hey Ultrasonic, just to clarify I really respect your knowledge of TEC's and appreciate the time and effort you go to to respond to me as well as all the research and coding you are doing to get this open source controller up. I'm sorry for hijacking your thread, I never thought everybody else would jump in and have a crack, I asked here because I know you have experience in this and can tell me one way or the other what works, I should have just PM'd you.

The question about hot side waterblock style was legitimate though and i would appreciate your opinion/knowledge because i do believe there could be room for improvement there, i know the maze style block works as evidenced by your chiller blocks, but is there more to be gained, have you tested a pin matrix block or a microfin block? With such a large number of TEC's it seems like there could be a need for a change...

To simplify i may just run the loop like this with your style of chiller blocks:

RES>PUMP>480 RAD>3 TEC CHILLER>480 RAD>3 TEC CHILLER> All other components.

PS, no need to waste your time modeling anything and running simulations in solidworks, i can do that myself if i can't get the answer i need out of you Ultrasonic.

Thanks for taking the time to educate me!

Also, to any moderators, feel free to snip all this out from my first post in thread onwards and create a new thread.

[Ultrasonic2]

I can extrapolate in theory if you'd like.
Fresh water would be provided at a lower temperature than preheated water as seen in serial peltier blocks. The reduced water temperature results in greater heat transfer due to an increase in delta T. Thus to increase efficiency you could either attempt to create counter flow (which does not appear to apply well to heat pumps), or you can provide fresh water to each peltier unit as yours and most designs seem to encompass. As having only 7 years in TEC'ing might seem a bit meh, I've also created multiple blocks and tested this. Sadly I Teampuss and Procooling just aren't what they once were.

And thus a fight that will never end will begin. But i wont cos it really doesn't matter to me

[n00b 0f l337]

The ad hominem isn't getting you very far here either.
Explain your counter point or ignore away.

[Ultrasonic2]

just out of curiosity... is there any way to not dump 20.735watts in the 2 resistors?
dumping power there seems like a waste.
if it can be done via mosfets, wouldn't it be more efficient?
I know this is extreme but but for a machine that is using high CPU loads for extended periods of time,
would it not be better to minimize the wasted watts?

Well im no expert on electronics and as you'll know the Low pass filter wasn't my idea but it will work. So to some degree i can't answer your question cos im not all knowing.

You could further reduce the resister losses by reducing their resistance but that requires more capacitance for the same ripple. In an ideal word i would beable to increase the frequency of the mosfets thus requiring less capacitance and therefore you could reduce the resistance.

However it would seem the arduino is screwing up at high frequencies.

in reality the limiting factor is size if you reduce the resistance you'll need more capacitance and therefore you'll need a lot of space for them.

By the way the 20watts is only when the TEC are at 100% and obviously scales down as the PWM frequency is reduced.

To some what fudge the numbers you could have an additional mosfet in parallel to the resister that came on at 100% duty and therefore bypassed the resisters and therefore bypassing their losses (but only at 100%)

[Ultrasonic2]

Hey Ultrasonic, just to clarify I really respect your knowledge of TEC's and appreciate the time and effort you go to to respond to me as well as all the research and coding you are doing to get this open source controller up. I'm sorry for hijacking your thread, I never thought everybody else would jump in and have a crack, I asked here because I know you have experience in this and can tell me one way or the other what works, I should have just PM'd you.

The question about hot side waterblock style was legitimate though and i would appreciate your opinion/knowledge because i do believe there could be room for improvement there, i know the maze style block works as evidenced by your chiller blocks, but is there more to be gained, have you tested a pin matrix block or a microfin block? With such a large number of TEC's it seems like there could be a need for a change...

To simplify i may just run the loop like this with your style of chiller blocks:

RES>PUMP>480 RAD>3 TEC CHILLER>480 RAD>3 TEC CHILLER> All other components.

PS, no need to waste your time modeling anything and running simulations in solidworks, i can do that myself if i can't get the answer i need out of you Ultrasonic.

Thanks for taking the time to educate me!

Also, to any moderators, feel free to snip all this out from my first post in thread onwards and create a new thread.

To be honest there are some many different variables that there is no perfect global answer to your question. So while I may have the perfect answer for my set of variables if you change the variables the answer can be quite different.

Lets look at CPU water blocks for a second the manufactures can never seem to make up their mind what is the best approach for cooling a cpu

IE swiftech have their HD block which uses a matrix system while EK have the XTL which uses the channel method and both are with in a degree of each other. It would seem the Biggest factor to which is better, actually has nothing to do with the design at all but more about improved manufacturing processes allowing for finer fins and channels



In the water block I made I achieved the best performance with a pin matrix, whereas with my TEC blocks a matrix was a failure (WITH MY VARIABLES). That doesn’t mean that fins are best.

The fundamental differences between a Matrix V Strait fins are normally (assuming the water is entering from the left and exiting the right .)are ( in my opinion )



When milling the channels into pins you increase turbulence, which reduces laminar flow which is good for cooling. However you have now effectively made the block more restrictive and reduced the flow which is bad. While milling more metal out of the block could result in the same mathematical result the reality is you’ve milled out conductive metal that would have wicked heat up to be cooled

Now a HUGE benefit to pin matrix’s approach is you can have a variable restrictive model . If you have the inlet in the centre then as the water flows further from the centre of the block it has more spaces to go through and therefore water velocity per gap decreases as does it’s restrictiveness. But this is a good thing because you have the coldest and fastest moving water against the centre of the cpu where all the heat is coming from.



Now the above doesn’t apply to a TEC where each square CM of the TEC can only move X watts so it’s normally not beneficial to apply the majority of your cooling to one Spot.

So what about a number of channels



Well you haven’t milled out the extra copper so there is less turbulence which is a bad thing, But because there is less turbulence the block is less restrictive which leads to higher water velocities, which is good. To have the same total block restrictiveness you could have more channels and therefore more surface area which is good.

So which fundamental design concepts works the best can be more about implementation OR milling technology than anything else.

SO that leaves the maze system.

Now this is something I doubt you’ll ever see in CPU water blocks every again but it has some advantages im my situation. As I mentioned before since every CM of the TEC can move only X watts and no more, it’s HIGHLY beneficial to try and cool the entire TEC the same. Unlike pretty much all CPU blocks. The maze enables me to have guaranteed the same flow rate across the block and therefore the same cooling(in theory).

Now the most important factor with the design of my block is I wanted to use a 3mm end mill, had I gone for a pin matrix or a whole bunch of channels. That would have meant the velocity per slot was FAR lower than me jamming it all down 3 slots. This reduction in velocity per slot when having 16 (say) slots results in far less performance. Of course the block with more slots will have more water flowing though the block (through the barbs) because it’s less restrictive but it will actually perform worse.

Now it’s important to look at my next variable (first was 3mm endmill ) and that is me milling my block from ½ copper. Now this is SOOOOO much more copper than normal blocks . Because the copper is so thick I can have very deep channels so the block isn’t resistance at all when pushing all the water down its 3 slots.







Is this ^ the Best design I could have done ? Nope I could have implement what was on my 50mm x2 blocks but didn’t





As a side note pushing water down 2 4mm slots is not the same as pushing it down 4 2mm slots even though the area is the same.

So while some of you may or may not agree with this I certainly hope I’ve made you realise that if some says they have the magical best performance answer it’s not the truth with all possible variables. If this was the case then all CPU water blocks would be designed the same.

Now I have simulations for all of this and most of that is still on OC.net

There is a HUGE trap that people fall into when using solid works simulations, and that is how they apply the water to their block. There are 2 options out of the box and that is a volume or pressure. Both will massively screw reality.

If you apply a volume the solidworks will apply as much force as necessary to get your specify volume though it. This means your best performing block will be an extremely restrictive one !

You have the same problem if you apply a pressure to it but in reverse.

Now in the real world a pump has both flow and pressure restrictions as you increase flow the pressure drops and vis versa.
So for my testing in solid works I applied a proper 655 ( cos that’s what I have ) pump curve to all my blocks.


SOOO to sum up there is no perfect answer to what is the best set-up for all variables. Im not giving you an answer but hopefully giving you something to ponder.

[Liam_G]

For sure you have given me a lot to ponder, will take me a while to get my lil old head around that and this whole TEC thing in general..

Just to clarify though, I was only raising the question about the hot side of TEC Chiller block, I actually really like your CPU block and i can see that the extra mass works.... I also didn't realize your tech chiller block was 3 channels, I thought it was only 1, checked it now and see that its 3, that makes more sense to me now, I just didn't pay enough attention last time. I thought it was one little channel snaking all the way through the block which would take a while and thats why I raised the question. Now i understand it is very low restriction and quite a high flow rate through the block it allays any fears i had. So that is the only answer i need.

Got lots of ideas flying around in my head, i'll retreat back to my little man cave and get back to you when i have something to show for all your help.

Cheers

[Ultrasonic2]

could you remind me what your actual question was ? Sorry

I more that willing to share my idea's and thoughts with you one block design

There is something i suggest you read before designing a block

http://www.tecpeltier.com/viewtopic.php?f=4&t=33

and here is some thermal tests i ran

http://www.tecpeltier.com/viewtopic.php?f=4&t=38

[Liam_G]

ok, so back to TEC chillers for a second. With the chiller block is it best to have the water remain in the block as long as possible on the hot and cold side?

Would it be better to have the water remain in the block for a long time on the cold side to give it time to cool the water, but be more like a micro-fin setup on the hot side so the water doesn't remain in the block very long and get too hot?

That was my original question, but somewhere in the chaos of this thread it has been answered, the rest i will just continue to read, research and learn myself.

Thanks

[Ultrasonic2]

That was my original question, but somewhere in the chaos of this thread it has been answered, the rest i will just continue to read, research and learn myself.

Thanks

yeah it was .

Now i need some help guys i'm trying to stop the off charge usb changing on my gigabyte 890fxa-ud7 but can't ?