Overclock.net banner

i7 8700K TEC chiller build planning

7K views 43 replies 9 participants last post by  Master Chicken 
#1 ·
It started with a simple question. What's the most single-threaded performance I can get in a practical 24/7 stable setup? It was clear that the answer was taking a binned + delidded i7 8700K to the highest overclock possible.

But what is the highest overclock possible for 24/7? That depends on cooling, which is when I started looking around this forum for ideas. In particular the builds done by OCAddict and foxrena stood out as excellent examples of what I want to do. I decided to go with the water chiller approach because it doesn't require making a custom CPU block.

The next thing I needed to figure out was how cold I would be able to run the loop. Without really good insulation the limit would be just above the dew point. In this case I'm fortunate that I live in Colorado where the dew point is below freezing for much of the year. The plan is to only insulate the parts of the loop that are not visible. The only visible parts will be a short tubing run to the CPU block and back. The end result is that I'll be getting coolant temps between 1-9 C, which is 20-25 C better than high end watercooling. That should make enough of a difference to be worth it.

At this point it was time to build a prototype. I decided to start with 2x CTE 19911-5P31-15CQ (40mm, 24v, 225W) sandwiched between some EK RAM blocks. I will be running them at 12V from 6-pin PCI-E power connectors. I also decided to use an Arduino to control everything. This has the advantage of being capable of easy software temperature monitoring and control. For the initial tests I just used some electromechanical relays stuck to the back of the control board.

This is the test setup





The CPU is an i7 3930K 4GHz @ 1.19v.

Running prime95 AVX hits about 35C initially, but slowly creeps up because the test setup is unable to maintain the cold loop temperature at full load.

Based on this testing, I decided to use 4x CTE 19911-5M31-28CZ, which is a 24.8V, 62x62mm, 400W TEC.

I will have the TECs sandwiched between a pair of waterblocks similar to this:
http://www.shop.customthermoelectric.com/Water-Block-575-x-575-x-115-WBA-575-115-AL-01.htm

I was able to custom order a pair of the above blocks with G1/4" threading, and holes to mount 2 of them together. The material was also changed from aluminum to copper for compatibility with PC watercooling parts. The blocks ended up being $595 each, which isn't cheap but I will be able to re-use them in future builds for a long time.

If I'm reading the datasheets correctly, this setup should be more than enough to maintain 5C in the cold loop at full load.

Now, on to the hard part: controlling the TECs.

The current test setup is not ideal. I don't want to deal with clicking relays all the time. I plan to use a PWM signal from the Arduino to drive MOSFETs. Generating the PWM signal is simple enough, but I don't have very much experience with designing circuits. As far as I know, the following setup should work:


Datasheet for that MOSFET:
https://www.infineon.com/dgdl/irlb3034pbf.pdf?fileId=5546d462533600a40153566027b22585

The plan is to replicate that circuit 4 times, once for each TEC. I know that using PWM isn't as efficient as voltage control, but I'm not too concerned with that since I'm already running the TECs at around half their max voltage. I don't see any way to make this compact without getting a custom PCB fabricated. Solderless breadboard probably can't handle the ~12.5A current that each TEC is going to pull. I don't mind the cost of getting a custom PCB, but I have no idea how to design one.

So my questions are:

1. Is my choice of TECs/waterblocks enough to cool a i7 8700K at around 1.45-1.475v?

2. Will that PWM/MOSFET circuit work?

3. What should I do to make the PWM/MOSFET setup compact/clean?

4. Is there a better option for the controller? It needs to run off the PC's PSU.
 
See less See more
5
#2 ·
#3 ·
Quote:
Originally Posted by toolmaker03 View Post

for that cost, on water blocks, talk to foxrena you will get a much better water block. at $300 each.

http://www.overclock.net/t/1406591/tec-block-on-4770k

I wanted to use his blocks for my build but the cost was too much for me to handle.
I didn't realize that he was selling them. I'll send a PM to see if that would be an option.
 
#4 ·
yes, he does it out of his house, so if you wanted like 4 of them it might take him a couple of months to build them.
thumb.gif



this is what the new ones look like
 
#5 ·
That's $300/block, you need 2 blocks / 62mm TEC, that's 8 blocks, these are direct die blocks and are too complicated to be economical for chiller blocks. I'd ask foxrena to make you a special 2x62mm chiller block, If you were happy to pay $1200 for 2 of those CTE blocks I would imagine you can get foxrena to make you a superior custom chiller block for less than that that will still allow you to use 4x62mm TEC's

Everything else seems good. You'll need more hotside rad's for 4x62mm TEC's though.
 
#6 ·
Quote:
Originally Posted by LiamG6 View Post

Everything else seems good. You'll need more hotside rad's for 4x62mm TEC's though.
I'm reasonably sure that the single 560mm radiator would work.

If I'm reading the datasheet for the TECs correctly this setup would be able to maintain a 30C delta with 400W of heat load on the cold loop. I don't see a i7 8700K going anywhere near that even at full load. I could probably get away with 3 TECs, but if I end up going with the CTE block I might as well have 4. The result is that I won't need to run the TECs at full power except to initially chill the cold loop.

Given that PWM control does not increase COP at lower duty cycle, I'm going to assume it's roughly linear. (eg: 50% duty cycle reduces Qc by 50% at the same delta T)

Doing the math for a 250W load. (I figure this is a nice worst case estimate)

250/400 = 62.5% duty cycle.

Full power is 12V/12.5A per TEC for a total of 600W. Running at 62.5% duty cycle brings that down to 375W for the TEC power.

Total heat load would be 375W + 250W = 625W.

The coldest I would want to ever go is 1 C, to avoid freezing/slush.

This is for a 30C delta, so to get 1 C cold loop I need the hot loop to remain at or below 31 C. My typical room temperature is 21-22 C so I need to keep a hot loop delta of 9-10 C at most.

Based on this review, it should be fine:

http://thermalbench.com/2015/03/28/hardwarelabs-black-ice-nemesis-560gtx/5/

This test is using a system that has a heat load somewhere between 500-550W. (650W at the wall, CPU + GPU watercooled)

I am using EK Vardar 2000 RPM fans. According to that graph I would be looking at ~1400-1500 RPM for this example. Is there something I'm doing wrong with the math here?

I'm putting this setup into a CaseLabs Magnum STH10. I have room to add another 240/280mm radiator in the lower chamber if the single 560mm isn't enough.
 
#7 ·
Quote:
Originally Posted by jedi95 View Post

I'm reasonably sure that the single 560mm radiator would work.
not even close to enough.
redface.gif


the reality is that you will need about twice he amount in radiator wattage, as the amount of TEC wattage being used. to give a example from your own test setup. the two 224watt 24volt 15amp TEC's you used at 12volts are producing about 200watts of TEC cooling. your 560mm radiator is capable of removing 600watts of heat the TEC's where only producing about 400watts of heat, so this is why that setup seemed fine.

now the proposed setup is 4 X 400watt TEC's running at 12volts or 200watts each, that is a total of 800watts in TEC cooling. now to cool that down, you will need at least 1600watts of radiator. so I would say a minimum of 3 X 560mm radiators will get you what your after.

on my own system I am using 6 X 360mm radiators to cool 3 X TEC's running at 300watts each, and 4 X 480mm radiators to cool 3 X more TEC's running at 300 watts each.
biggrin.gif
all of that radiator space, will just barley keep my hot side water loop at a 10C delta T, when the PC system is at load.

the 360mm radiators cool 350watts each X 6 that's 2100watts.
the 480mm radiators cool 535watts each X 4 that's 2140watts.
the 3 X TEC's at 900watts total, are producing about 1800watts of heat.
 
#8 ·
Quote:
Originally Posted by jedi95 View Post

I'm reasonably sure that the single 560mm radiator would work.

If I'm reading the datasheet for the TECs correctly this setup would be able to maintain a 30C delta with 400W of heat load on the cold loop. I don't see a i7 8700K going anywhere near that even at full load. I could probably get away with 3 TECs, but if I end up going with the CTE block I might as well have 4. The result is that I won't need to run the TECs at full power except to initially chill the cold loop.

Given that PWM control does not increase COP at lower duty cycle, I'm going to assume it's roughly linear. (eg: 50% duty cycle reduces Qc by 50% at the same delta T)

Doing the math for a 250W load. (I figure this is a nice worst case estimate)

250/400 = 62.5% duty cycle.

Full power is 12V/12.5A per TEC for a total of 600W. Running at 62.5% duty cycle brings that down to 375W for the TEC power.

Total heat load would be 375W + 250W = 625W.

The coldest I would want to ever go is 1 C, to avoid freezing/slush.

This is for a 30C delta, so to get 1 C cold loop I need the hot loop to remain at or below 31 C. My typical room temperature is 21-22 C so I need to keep a hot loop delta of 9-10 C at most.

Based on this review, it should be fine:

http://thermalbench.com/2015/03/28/hardwarelabs-black-ice-nemesis-560gtx/5/

This test is using a system that has a heat load somewhere between 500-550W. (650W at the wall, CPU + GPU watercooled)

I am using EK Vardar 2000 RPM fans. According to that graph I would be looking at ~1400-1500 RPM for this example. Is there something I'm doing wrong with the math here?

I'm putting this setup into a CaseLabs Magnum STH10. I have room to add another 240/280mm radiator in the lower chamber if the single 560mm isn't enough.
If you are only chilling the 8700k I would try your luck with 2x 62mm 400w TEC's at 12v and keep the rad you have, possibly add in the 280 if you can. I don't think 4 x 62mm 400w TEC's is required. If you have way too much Qc for your load your duty cycle will be low, this will punish the mosfets as they will be dissipating lots of heat, better to find the balance and use only what you need. I suggest 2 x 62mm 400w CTE TEC's and look into buying some Threadripper waterblocks as they have a big enough base to handle a 62mm TEC, try the XSPC NEO TR4 or the Phanteks C399
 
#9 ·
so you are wanting to cool a CPU and GPU right?
if so than 4 X 400watt TEC's at 12volts would be about right to keep the water below ambient but above the condensation point.
if it is just the CPU being cooled, than liamG6 has the right idea, two of the 400watt TEC's should be enough, at most 3.
 
#11 ·
Quote:
Originally Posted by toolmaker03 View Post

not even close to enough.
redface.gif


the reality is that you will need about twice he amount in radiator wattage, as the amount of TEC wattage being used. to give a example from your own test setup. the two 224watt 24volt 15amp TEC's you used at 12volts are producing about 200watts of TEC cooling. your 560mm radiator is capable of removing 600watts of heat the TEC's where only producing about 400watts of heat, so this is why that setup seemed fine.

now the proposed setup is 4 X 400watt TEC's running at 12volts or 200watts each, that is a total of 800watts in TEC cooling. now to cool that down, you will need at least 1600watts of radiator. so I would say a minimum of 3 X 560mm radiators will get you what your after.

on my own system I am using 6 X 360mm radiators to cool 3 X TEC's running at 300watts each, and 4 X 480mm radiators to cool 3 X more TEC's running at 300 watts each.
biggrin.gif
all of that radiator space, will just barley keep my hot side water loop at a 10C delta T, when the PC system is at load.

the 360mm radiators cool 350watts each X 6 that's 2100watts.
the 480mm radiators cool 535watts each X 4 that's 2140watts.
the 3 X TEC's at 900watts total, are producing about 1800watts of heat.
I don't follow the math here. At 12V each TEC will pull 12.5A of current for a total power consumption of 150W per TEC. That's 600W for 4 of them. If we figure 250W from the CPU that's a worst case of 850W with the TECs at full power. It's not possible for more heat to be generated than power is being consumed, that's simple conservation of energy.

Yes, the TECs are rated for Qmax of 400W, but that's at the Vmax of 24.8V. With 12V it should be closer to this:


Quote:
Originally Posted by LiamG6 View Post

If you are only chilling the 8700k I would try your luck with 2x 62mm 400w TEC's at 12v and keep the rad you have, possibly add in the 280 if you can. I don't think 4 x 62mm 400w TEC's is required. If you have way too much Qc for your load your duty cycle will be low, this will punish the mosfets as they will be dissipating lots of heat, better to find the balance and use only what you need. I suggest 2 x 62mm 400w CTE TEC's and look into buying some Threadripper waterblocks as they have a big enough base to handle a 62mm TEC, try the XSPC NEO TR4 or the Phanteks C399
Using 2 TECs gives Qc of 200W @ 30C delta. That 200W also includes losses to the environment, which will be significant since I'm not perfectly insulating everything. That's why I think 3 would be the right size.

Why would running a lower duty cycle be punishing on the MOSFETs? From what I understand the heat generated by a MOSFET is R * I²
The MOSFET I plan to use has a Rds(on) maximum of 2.0 mΩ with 4.5V gate drive. I'm using 1 per TEC so that gives 2.0 * (12.5)² = 312 mW = 0.312W
Am I missing something here?

Do those waterblocks have flat contact surfaces, or would I need to lap them?
 
#12 ·
Quote:
Originally Posted by toolmaker03 View Post

so you are wanting to cool a CPU and GPU right?
if so than 4 X 400watt TEC's at 12volts would be about right to keep the water below ambient but above the condensation point.
if it is just the CPU being cooled, than liamG6 has the right idea, two of the 400watt TEC's should be enough, at most 3.
CPU only. I don't see any point to running a GPU on a chiller because I can already get 35-37 C load temps with a good water loop.
 
#13 ·
Quote:
Originally Posted by jedi95 View Post

CPU only. I don't see any point to running a GPU on a chiller because I can already get 35-37 C load temps with a good water loop.
for the two TEC's that your considering, I would really suggest getting a nice adjustable DC power supply like this one.

https://www.amazon.com/TekPower-TP50SW-Analog-Supply-Cigarette/dp/B00L5QU616/ref=sr_1_95?ie=UTF8&qid=1511208039&sr=8-95&keywords=adjustable+dc+power+supply

a PC power supply may function fine, but there are a lot of safeties built into a PC power supply, so it may not be a great choice for powering TEC's.

I don't follow the math here. At 12V each TEC will pull 12.5A of current for a total power consumption of 150W per TEC. That's 600W for 4 of them. If we figure 250W from the CPU that's a worst case of 850W with the TECs at full power. It's not possible for more heat to be generated than power is being consumed, that's simple conservation of energy.

I agree I don't really get it either???
I think it has to do with surface area, the actual CPU die that is getting hot is really small, but capable of producing 200watts of heat. a 62mm TEC is at least 4 times the size of a CPU die, and also produces 200watts of heat. I think that because the TEC is 4 times larger it actually dumps more of that 200watts of heat into the water loop. as a result we see that as the water loop needing more radiators to dissipate what should be the same 200watts of heat.
 
#14 ·
Quote:
Originally Posted by jedi95 View Post

I don't follow the math here. At 12V each TEC will pull 12.5A of current for a total power consumption of 150W per TEC. That's 600W for 4 of them. If we figure 250W from the CPU that's a worst case of 850W with the TECs at full power. It's not possible for more heat to be generated than power is being consumed, that's simple conservation of energy.

Yes, the TECs are rated for Qmax of 400W, but that's at the Vmax of 24.8V. With 12V it should be closer to this:

Using 2 TECs gives Qc of 200W @ 30C delta. That 200W also includes losses to the environment, which will be significant since I'm not perfectly insulating everything. That's why I think 3 would be the right size.

Why would running a lower duty cycle be punishing on the MOSFETs? From what I understand the heat generated by a MOSFET is R * I²
The MOSFET I plan to use has a Rds(on) maximum of 2.0 mΩ with 4.5V gate drive. I'm using 1 per TEC so that gives 2.0 * (12.5)² = 312 mW = 0.312W
Am I missing something here?

Do those waterblocks have flat contact surfaces, or would I need to lap them?
yes 3 TEC's should be the sweet spot but 2 might be able to handle it, with all the extra losses of a chiller setup prob need 3, I'm used to direct die setups.

If you choose your mosfets well with low rds(on) you should be fine. I'm using an off the shelf controller and I don't know the specs of the mosfet and it gets quite hot when my CPU is idle and my TN669w TEC is at low duty cycle, I have a TO-220 heatsink and a blower fan on it to keep it cool.

The waterblocks are flat in as much as the threadripper IHS is flat, meaning it's pretty flat but lapping will improve things. The threadripper IHS is much flatter than intel's offering but the blocks could still be improved by lapping, but it is probably fine as is. I would recommend the XSPC block, it performs up to 10*c better than the phanteks on a 1950x and the fin coverage is better, about 55x50mm fins with a massive base plate that will handle a 62mm TEC. The fins are very thin so it will have vastly superior heat transfer over those crazy expensive CTE blocks. With one small mod to the XSPC it will be well suited to TEC cooling, basically just swap the thin metal jetplate out for a larger one to force the coolant to flow through the channels for the majority of their length.

Quote:
Originally Posted by toolmaker03 View Post

for the two TEC's that your considering, I would really suggest getting a nice adjustable DC power supply like this one.

https://www.amazon.com/TekPower-TP50SW-Analog-Supply-Cigarette/dp/B00L5QU616/ref=sr_1_95?ie=UTF8&qid=1511208039&sr=8-95&keywords=adjustable+dc+power+supply

a PC power supply may function fine, but there are a lot of safeties built into a PC power supply, so it may not be a great choice for powering TEC's.

I agree I don't really get it either???
I think it has to do with surface area, the actual CPU die that is getting hot is really small, but capable of producing 200watts of heat. a 62mm TEC is at least 4 times the size of a CPU die, and also produces 200watts of heat. I think that because the TEC is 4 times larger it actually dumps more of that 200watts of heat into the water loop. as a result we see that as the water loop needing more radiators to dissipate what should be the same 200watts of heat.
PC PSU's are perfect for powering TEC's, they are much more efficient and quieter than the majority of other shirty PSU's out there, and 12v is plenty.

Tools, your issues with heat stem from your waterblocks not being up to the task and overall losses in your loops, not from your lack of rad space, you have more than enough rads. If he uses 3 TEC's I would recommend adding the extra 280mm rad but those 2000rpm EK vardars move some serious air, I know because I use the same fans, so he may be ok with the 560 rad.

FYI OP, I use a single TEC at 12v 18a direct with foxrena block on a 4670k, a GTX1080 block, GENE VII mosfet block, Aqauero 5 LT block and I have a 420x80 rad, a 360x45 rad and a 280x30 rad, all with EK Vardar 2k 140mm or 3k 120mm fans on them and it keeps my loop at less than 5*c delta above ambient. That's probably a 600w total heat load.
 
#15 ·
Quote:
Originally Posted by LiamG6 View Post

yes 3 TEC's should be the sweet spot but 2 might be able to handle it, with all the extra losses of a chiller setup prob need 3, I'm used to direct die setups.

If you choose your mosfets well with low rds(on) you should be fine. I'm using an off the shelf controller and I don't know the specs of the mosfet and it gets quite hot when my CPU is idle and my TN669w TEC is at low duty cycle, I have a TO-220 heatsink and a blower fan on it to keep it cool.

The waterblocks are flat in as much as the threadripper IHS is flat, meaning it's pretty flat but lapping will improve things. The threadripper IHS is much flatter than intel's offering but the blocks could still be improved by lapping, but it is probably fine as is. I would recommend the XSPC block, it performs up to 10*c better than the phanteks on a 1950x and the fin coverage is better, about 55x50mm fins with a massive base plate that will handle a 62mm TEC. The fins are very thin so it will have vastly superior heat transfer over those crazy expensive CTE blocks. With one small mod to the XSPC it will be well suited to TEC cooling, basically just swap the thin metal jetplate out for a larger one to force the coolant to flow through the channels for the majority of their length.
PC PSU's are perfect for powering TEC's, they are much more efficient and quieter than the majority of other shirty PSU's out there, and 12v is plenty.

Tools, your issues with heat stem from your waterblocks not being up to the task and overall losses in your loops, not from your lack of rad space, you have more than enough rads. If he uses 3 TEC's I would recommend adding the extra 280mm rad but those 2000rpm EK vardars move some serious air, I know because I use the same fans, so he may be ok with the 560 rad.

FYI OP, I use a single TEC at 12v 18a direct with foxrena block on a 4670k, a GTX1080 block, GENE VII mosfet block, Aqauero 5 LT block and I have a 420x80 rad, a 360x45 rad and a 280x30 rad, all with EK Vardar 2k 140mm or 3k 120mm fans on them and it keeps my loop at less than 5*c delta above ambient. That's probably a 600w total heat load.
I'll go ahead and add the additional 280mm rad for the hot loop. It should help efficiency a bit by reducing the delta that the TECs need to maintain.

Going back to the controller, I added some capacitors to suppress ripple created by switching the TECs with PWM. I also have a rough idea of how I want to lay out the PCB.



The control board will be in the top chamber of the case, so it will be in the airflow path from the 480mm GPU loop radiator.
 
#16 ·
Looking good.

The pic of the Archimedes block was from my build - it is great for a direct die block, but cost prohibitive for a chiller. The waterblocks you are using are not perfect, but they will work - and as you have noticed, its very difficult to get a good chiller block sourced without paying a ton for a custom one off part.

Ultrasonic used similar ones with success.
 
#17 ·
I made some progress with the control board. It has a few more components added, and more powerful MOSFETs.
wheee.gif




I also built it to handle 4 TECs if needed. Small run PCBs were not going to be cheap either way, so it was worth the small size increase.



Never hurts to have a few spares



PCB up and running in the test loop with control software
thumb.gif




The test loop can't actually reach that 4C target, so I set 12C to verify that the controller can maintain a target temperature.


As for the rest of the build, I set the 8700K system up with an old aircooler I had laying around to play with the memory OC:


5GHz only takes 1.25v
biggrin.gif


Now I'm just waiting on my STH10 case, and the TEC waterblocks.
 
#18 ·
ok so the 8 pin pci-e connector from the power supply is 3 12vollt power and 5 grounds. are you using all 3 of the 12volt power wires in your controller to reduce stress on the wires?

the reason I ask this is because by looking at the circuit board it looks like the wires are split 4 and 4 if this is correct you have a ground wire in the power strip on the circuit board.

what is the amp rating of the controller circuit you have made for each TEC?

by the way they really do look nice.
 
#19 ·
Quote:
Originally Posted by toolmaker03 View Post

ok so the 8 pin pci-e connector from the power supply is 3 12vollt power and 5 grounds. are you using all 3 of the 12volt power wires in your controller to reduce stress on the wires?

the reason I ask this is because by looking at the circuit board it looks like the wires are split 4 and 4 if this is correct you have a ground wire in the power strip on the circuit board.

what is the amp rating of the controller circuit you have made for each TEC?

by the way they really do look nice.
Edit: Oops, I see what you mean now, good catch! I was thinking the screw terminals were for a bench PSU and the 8-pins were for the TEC, but I think I had it backwards.

You are correct, but technically if he is using an ATX PSU he has a sense wire merged on the 12v side since it is 2x sense wires, 3x 12v, 3x GND on a PCIE plug (two of the "GND" wires are supposed to be sense wires but many manufacturers just GND them with a short jumper wire).
 
#20 ·
I use 8 pin EPS cables for this reason because it is 4 gnd and 4 +12v, but my PSU only has 2 of them and I need 1 for CPU, luckily I only have 1 TEC.

If he is using PCIe he might need to just use 6 pin cables. My 8 pin EPS cables get slightly warm to the touch with up to 18amps, so 4.5a/wire, I think they are 16 or 18awg. If he is using multiple PSU's and using 8 pin EPS cables he should be fine.
 
#21 ·
Quote:
Originally Posted by LiamG6 View Post

If he is using PCIe he might need to just use 6 pin cables.
correct.

I was just wanting to make sure that he was aware of this possible issue, and that he should only be connecting the 6-pin part of the 8-pin pci-e connector to the controllers he has created.
thumb.gif


man my communication skills really do suck.
redface.gif
 
#22 ·
we understood exactly what you said, don't worry about that tools, and it is a fair thing to point out because logic says the move from 6 pin to 8 pin pci-e would add one gnd and one + but it's not the case.

Now there is a caveat to my post about my PSU only having 2 EPS connectors too because if I make custom cables I can actually make up to 6 EPS cables because it uses the same modular connection on the PSU as the PCI-e cables but they just have different pin out, PCI-e splits gnd wires and only uses 3 of the + wires leaving the 4th + wire vacant, so you can modify your PCI-e cables to be EPS cables if your PSU has same sort of modular setup as mine.

https://content.hwigroup.net/images/products_xl/203083/2/seasonic-platinum-series-860w-v2.jpg
 
#24 ·
Quote:
Originally Posted by LiamG6 View Post

I use 8 pin EPS cables for this reason because it is 4 gnd and 4 +12v, but my PSU only has 2 of them and I need 1 for CPU, luckily I only have 1 TEC.

If he is using PCIe he might need to just use 6 pin cables. My 8 pin EPS cables get slightly warm to the touch with up to 18amps, so 4.5a/wire, I think they are 16 or 18awg. If he is using multiple PSU's and using 8 pin EPS cables he should be fine.
Yep, I noticed this after I had already placed the order. It works fine with a 6-pin. I fixed it in the design files so any future PCB orders won't have this issue. I'm also working on a version with SMD components and the ATMEGA32U4 from the Teensy 2.0 integrated directly.

I'm going to be using a single Corsair AX1500i for this build, so it should be plenty of power. The TECs should use 12.5A each, so that's just over 4A per wire with 6-pin PCI-E.
Quote:
Originally Posted by whk0 View Post

Should consider add MOSFET drivers, those MOSFET have very high input capacitance (and 4 in parallel!).
I am using a MOSFET driver in the PCB I ordered. (TI UC3705T) This diagram is what I got built, minus the fixed PCI-E pins:
 
#25 ·
howdy all & jedi96
smile.gif


Firstly, nothing but praise & non-critical & critical suggestions here;
thumb.gif


jedi95 - truyly awesome work on the controller PCB, love it !!! but for your next version, please please check out the minimum components needed to make the Arduino function or you will have major problems, might even kill the microcontroller ATMEGA328 chip, this pic below shows minimal components, including 5v power supply but as your using a computer PSU - You can omit the 5v voltage regulator circuitry components;

FVTVSOTHA4MBIDF.png


That schematic gives the ATMEGA328/P stable operation and;

> 16Mhz timing crystal (for stable operation - critical)
>> Programming Pins (so you can change Arduino Code without taking the Chip off the board)
>>> A reset button if the Arduino hangs/crashes (unlikely but always good to have this safety net)
>>>> 0.1uF Capacitor on the RESET line (critical for programming the code)
>>>>> Power ON & Pin-13 LED's, Green & Blue (optional suggestions/configurable to your taste)
smile.gif


As opposed to your one; (LOVE that youve used CADSOFT EAGLE !!!!!!!!!!!)
biggrin.gif




Just those little additions/corrections will get you a very stable Arduino Based microcontroller, using the SMD Version of ATMEGA328P-AU will also greatly reduce PCB space but make sure not to have high voltage anywhere near the arduino low voltage traces, try separating them with an Air-Gap groove or another method for separating HV from LV circuitry, if you haven't already, not really medical grade life-saving apparatus here, but its good practise when making your own PCB's
smile.gif


i'd be more than happy to help with any PCB designing and can source very very awesome LOW prices small-runs PCB's, professionally made of course. i make home-brew PCB's when i want to test proof of concept, but when im happy with the results, i usually get them printed off in batches of five x 50x50mm or panelise the 50x50mm onto a 100x100mm square, yielding 4 PCB's per time,, x 10 pcs = 40 pcs total, all V-Grooved so you can easily break them apart with no tools needed, PM me for a quote for future PCB revisions, guarantee youll shlap yourself twice for the price realisation lol
biggrin.gif


Obviously larger the PCB - Higher the cost, so by using SMD Components, which i only use these days, you can make really intricately small/tiny PCB's, panelise them onto a larger PCB then order in small batches of those lol But for 1-off prototypes, minimum order is Five of 50mmx50mm, thickness adjustable and can use upto 2 oz copper thickness too but thats a bit prices, colour of PCB,, free/included lol

Also - im guessing the temp//humidity sensor is for measuring your ambient temp/humidity ? or is it for measuring the temp/humidity for the CPU where the cold block resides, if it is the later then moving it off the PCB via wires and sticking it to the side of the waterblock/motherboard then you'd get a more accurate reading for calculating dew point as thats the place thats most prominent for dew to start and condensation ruining your system

Lastly, where did you get that software, id really be grateful if i could try out the same, plus Arduino code if thats okay with you,, of course you can license it as non-commercial & share alike with attribution,, im on the same road but right at the start of my journey as ive just come out of a long hibernation away from all tech lol

Example of a few of my PCB's and how small one can get a simple board, this is just a silent Low-Battery-Indicator; smallest PCB ive made so far; (plus using 1206 SMD components, the photo of the penny showing a couple 1206 sized resistors, a 0603 resistor and a 0603 sized capacitor, 0.1uF)



A Nokia 5100 LCB PCB;



& my own version of an Arduino-Uno, breadboard thin-style,, with under-IC lighting (Pimp my Arduino, lolz)



i wrote an in-depth Article for how to breadboard up Arduino circuitry, Right Here smile.gif worth a read at least once before you make any more PCB's, adding a pre-made teensy/Arduino-Mini/pro etc is easy as you got all this circuitry already on-board, plus you can now get those type boards dirty cheap from places like THIS smile.gif

almost 20,000 of those sold in that ebay listing and ive personally ordered 2 dozen of them too, not one has failed since March 2016 lol

So, including an SMD ATMEGA328P-AU would save some space, but i'd say not a great deal compared to how small the pre-made one's footprint is,, but it would also mean sourcing all the minimal-operation components too and soldering them yourself, which i prefer though, standard soldering iron, standard tip here, nothing fancy, a no-name brand adjustable soldering iron too lol

Dont have to go insane on SMD Tools, just have a decent magnifier if going smaller than 0603 SMD Parts, 0402 or even 0201 lol 1206 ive found to be real easy, without the need for a magnifier, from the pic above that was a x40 magnifier,, plus a decent set of precision-tweezers with very fine points (careful, very sharp points)

Overall - this is truly beautiful, well done, hats off to you, totally awesome job (this is the point you take a bow, lol) & please please Lemmi know where i can get my hands on the software lol thanks loads in advance
smile.gif
thumb.gif
smile.gif
 
#26 ·
Quote:
Originally Posted by bobcat1969 View Post

Firstly, nothing but praise & non-critical & critical suggestions here;
thumb.gif


jedi95 - truyly awesome work on the controller PCB, love it !!! but for your next version, please please check out the minimum components needed to make the Arduino function or you will have major problems, might even kill the microcontroller ATMEGA328 chip, this pic below shows minimal components, including 5v power supply but as your using a computer PSU - You can omit the 5v voltage regulator circuitry components;

FVTVSOTHA4MBIDF.png


That schematic gives the ATMEGA328/P stable operation and;

> 16Mhz timing crystal (for stable operation - critical)
>> Programming Pins (so you can change Arduino Code without taking the Chip off the board)
>>> A reset button if the Arduino hangs/crashes (unlikely but always good to have this safety net)
>>>> 0.1uF Capacitor on the RESET line (critical for programming the code)
>>>>> Power ON & Pin-13 LED's, Green & Blue (optional suggestions/configurable to your taste)
smile.gif


As opposed to your one; (LOVE that youve used CADSOFT EAGLE !!!!!!!!!!!)
biggrin.gif




Just those little additions/corrections will get you a very stable Arduino Based microcontroller, using the SMD Version of ATMEGA328P-AU will also greatly reduce PCB space but make sure not to have high voltage anywhere near the arduino low voltage traces, try separating them with an Air-Gap groove or another method for separating HV from LV circuitry, if you haven't already, not really medical grade life-saving apparatus here, but its good practise when making your own PCB's
smile.gif


i'd be more than happy to help with any PCB designing and can source very very awesome LOW prices small-runs PCB's, professionally made of course. i make home-brew PCB's when i want to test proof of concept, but when im happy with the results, i usually get them printed off in batches of five x 50x50mm or panelise the 50x50mm onto a 100x100mm square, yielding 4 PCB's per time,, x 10 pcs = 40 pcs total, all V-Grooved so you can easily break them apart with no tools needed, PM me for a quote for future PCB revisions, guarantee youll shlap yourself twice for the price realisation lol
biggrin.gif


Obviously larger the PCB - Higher the cost, so by using SMD Components, which i only use these days, you can make really intricately small/tiny PCB's, panelise them onto a larger PCB then order in small batches of those lol But for 1-off prototypes, minimum order is Five of 50mmx50mm, thickness adjustable and can use upto 2 oz copper thickness too but thats a bit prices, colour of PCB,, free/included lol

Also - im guessing the temp//humidity sensor is for measuring your ambient temp/humidity ? or is it for measuring the temp/humidity for the CPU where the cold block resides, if it is the later then moving it off the PCB via wires and sticking it to the side of the waterblock/motherboard then you'd get a more accurate reading for calculating dew point as thats the place thats most prominent for dew to start and condensation ruining your system

Lastly, where did you get that software, id really be grateful if i could try out the same, plus Arduino code if thats okay with you,, of course you can license it as non-commercial & share alike with attribution,, im on the same road but right at the start of my journey as ive just come out of a long hibernation away from all tech lol

Example of a few of my PCB's and how small one can get a simple board, this is just a silent Low-Battery-Indicator; smallest PCB ive made so far; (plus using 1206 SMD components, the photo of the penny showing a couple 1206 sized resistors, a 0603 resistor and a 0603 sized capacitor, 0.1uF)



A Nokia 5100 LCB PCB;



& my own version of an Arduino-Uno, breadboard thin-style,, with under-IC lighting (Pimp my Arduino, lolz)



i wrote an in-depth Article for how to breadboard up Arduino circuitry, Right Here smile.gif worth a read at least once before you make any more PCB's, adding a pre-made teensy/Arduino-Mini/pro etc is easy as you got all this circuitry already on-board, plus you can now get those type boards dirty cheap from places like THIS smile.gif

almost 20,000 of those sold in that ebay listing and ive personally ordered 2 dozen of them too, not one has failed since March 2016 lol

So, including an SMD ATMEGA328P-AU would save some space, but i'd say not a great deal compared to how small the pre-made one's footprint is,, but it would also mean sourcing all the minimal-operation components too and soldering them yourself, which i prefer though, standard soldering iron, standard tip here, nothing fancy, a no-name brand adjustable soldering iron too lol

Dont have to go insane on SMD Tools, just have a decent magnifier if going smaller than 0603 SMD Parts, 0402 or even 0201 lol 1206 ive found to be real easy, without the need for a magnifier, from the pic above that was a x40 magnifier,, plus a decent set of precision-tweezers with very fine points (careful, very sharp points)

Overall - this is truly beautiful, well done, hats off to you, totally awesome job (this is the point you take a bow, lol) & please please Lemmi know where i can get my hands on the software lol thanks loads in advance
smile.gif
thumb.gif
smile.gif
For the fully integrated PCB I'm actually using the ATEMGA32U4 rather than the ATMEGA328P. This avoids the need for the USB -> serial chip. Additionally, it prevents the controller from being reset when using the serial console. As for the schematic, I simply re-implemented the design used by the Teensy 2.0 device:
https://www.pjrc.com/teensy/schematic.html

This is what I have at the moment:


The only major change to the board outside of the controller is switching to 12V gate drive for the MOSFETs. The driver IC I'm using supports a separate gate drive voltage from the PWM input. The current PCB experiences a bit of noise on the analog temperature sensor when the PWM duty cycle is not 0% or 100%. My best guess is that the 5V being supplied over USB is dropping and spiking a bit when the MOSFETs are being switched on/off, but I don't have an oscilloscope to confirm this. Using the 12V from PCI-E should isolate this noise from the analog sensor input, in addition to getting lower MOSFET on resistance. I'm going to assemble another one of my existing PCBs sometime this weekend with the gate drive voltage change to see if this works as expected. (will need to run a short wire)

I'm not too concerned with being able to assemble the SMD stuff myself, since the intent behind the SMD design is to make it easier to assemble in volume. I don't have the time to manually assemble 100s of these myself. Assuming my build goes well and works as expected, I have some interesting plans for these PCBs in the near future.
wink.gif


The temp/humidity sensor is for determining the dew point of the ambient air. This is to ensure that the chilled water temperature stays above the dew point and doesn't cause condensation inside the PC.

The control software is something I made in C#. It's fairly simple:
https://github.com/jedi95/TECControl

This will only work with the firmware I have flashed to the board. Below is the Arduino code that handles serial communication. I use a constant "magic value" at the start of each command to ensure that the right values are being read into the right variables. This code is part of the loop() function. There is an ugly signed/unsigned conversion required because the serial library in C#.NET only supports unsigned bytes.

Code:

Code:
int cmdLength = Serial.available();
  if (cmdLength >= COMMAND_SIZE)
  {
    //Read from serial
    char data[COMMAND_SIZE];
    Serial.readBytes(data, COMMAND_SIZE);
    unsigned char magic = (unsigned char)data[0];
    signed char newTarget = (signed char)(data[1] - 128);
    signed char newOffset = (signed char)(data[2] - 128);
    bool newSafety = (unsigned char)data[3] != 0;

    //Make sure we got valid data
    if (magic == COMMAND_MAGIC)
    {
      targetTemp = (float) newTarget;
      dewPointOffset = (float) newOffset;
      enableSafety = newSafety;
    }
    else
    {
      //Invalid data, empty the buffer
      while (Serial.available())
      {
        Serial.read();
      }
    }
  }

  Serial.print(REVISION);
  Serial.print(",");
  Serial.print(PWM, 0);
  Serial.print(",");
  Serial.print(Tcold, 1);
  Serial.print(",");
  Serial.print(Tcase, 0);
  Serial.print(",");
  Serial.print(Tdp, 1);
  Serial.print(",");
  Serial.print(h, 0);
  Serial.print(",");
  Serial.print(targetTemp, 0);
  Serial.print(",");
  Serial.print(dewPointOffset, 0);
  if (enableSafety)
  {
    Serial.print(",true,");
  }
  else
  {
    Serial.print(",false,");
  }
  Serial.println(actualTarget, 1);
 
This is an older thread, you may not receive a response, and could be reviving an old thread. Please consider creating a new thread.
Top