Jump to content

Quantum Systems - Project Alpha^-1


Quantum Systems
 Share

Recommended Posts

image.png.56b312b536e1c463c2d91e5bf7db6271.png

Design process.

It's been some days since my last update, but had to wait for some parts to proceed to the next steps. 
I currently got the cables and started on the tubes.

My friend tried to solder all the copper with some professional help, but it would not bother unfortunately (see my previous update). Currently the copper cool block will probably be installed near the end of the build since I need some more time to get in proper shape.

BUT problems aside: I promised some of you that I would explain how this thing works. Since we won’t use fans but use a natural convection-way of cooling trough the copper.

To get all your variables right, you need to do some calculations. It might get a bit technical here and there but we are all modders here, so shouldn’t be a big deal right?

I’m not going to clarify all my calculation steps and tire everyone with loads of numbers, but I have made an Excel document which allows me to play with some variables, like sizes and thickness of the heat fins. Note that all my units are based on the metric system, since I know a lot of you live on the other continent and use Imperial. Yes, TLDR, it’s for the enthusiast.


Heat Transfer Equation:
image.png.43af8217d790e82bcfbaac124db1851e.png
The base formula I used to calculate my cool-block:

Q            = Heat transfer (W or J/s)
rth           = Heat Transfer coefficient (W/(m
²K)) (or thermal resistance)
A             = Heat transfer area (m²)
dT           = Temperature, difference between ambient temp and liquid temp (K)
t              = time in seconds (s)


What we really would like to know is; how much Area we must overcome to have the cool-block loose the amount of heat (Q) which is generated by the hardware. So rewrite the formula which will clarify the surface area:

image.png.e7b8852af4a08b20ba547cb58767371e.png

Q
Looking at the TDP of the hardware gives you an indication of the heat (Q) that will be available inside the loop, which will be roughly 350W. CPU at 100% is approx. 120 Watts, GPU at 100% is approx. 230 Watts


rth
You must be aware that every radiator actually uses 2 physical ways to transfer the heat from the liquid to the ambient air in this situation:

- Forced convection; the ‘moving’ water touches the copper heat fins, in which the movement is caused by a pump. (Active cooling) Also caused by fans on radiators, which we don’t have in this situation.

- Natural convection; the ambient air cools down the copper heat fins on the outside. (Passive cooling).

My way of calculating this is nearly not as accurate as I should be doing, but also to be honest, to fully calculate a heat-fin construction I had to use differential equations. This is knowledge that I never fully understood. So I used an easier model to catch a grip. I will explain later why.
 

To calculate the thermal resistance of the cool-block, we need to know what elements the heat will go through to get to the other side and eventually being cooled by ambient air. From warm water, into the copper and on to air… So ‘forced convection’ on the liquid side due to flow from the pump
Will give a r=1/
α. The copper is a static or passive material which gives a r(th)=d/λ. Than back to air which will be natural convection, but still caused by a minor flow (warm air going up, cool air going down). This will also give a r=1/α. These values λ and α I grabbed from the internet, there are loads of tables. I used 370 W/(mK) for λ(copper). For α(‘laminar flow’-water) I used 500 W/(m^2.K) and for natural convective air I used 15 W/(m^2.K).


T
The delta-T refers to a temperature difference between the liquid inside the loop and the ambient temperature on the outside. To stay safe, I calculated a max liquid temperature of 40 °C and an ambient air temperature of 20 °C, which is a pretty common temperature in my house.

t

I neglected the time factor, by looking at only 1 second. Everything x1 remains the same.

A
Than we have one factor left: the heat transfer area (A) in square meters. This is all we want to know: how much area of copper cool fins do I need to make/attach, to generate enough surface area so we can cool it down. My calculation pointed out I needed roughly 1.16 m², that doesn’t seem so bad now does it? Well 1.16 square meters in a 24L case was going to be interesting. I needed an area where I could fit all these fins, on which I chose to place them on the backside of the case (back panel modification).
So now we have a heat transfer area, but we need to know how thick/wide/long the heat fins must be. The only 1 thing I want to control is the thickness. The thickness is really relevant regarding your budget. (I think we all know what coppers costs these days.)
The length of a fin is pretty much limited, which cannot be longer than the size of the copper base or the size of the back plate of the Masterbox Q300P. We also don’t want them to ‘stick out to much’ as we still care about the looks and I want to balance that part.
So the only 2 variables we have left is the width of a fin and the quantity of fins. According to my calculations I needed around 14 fins on the liquid side (I chose 16 to be safe and to fill up the empty gaps). On the passive air side I need around 83 fins. So by playing with the width it would automatically calculate how much fins I needed.
Fun fact; cool block plate thickness doesn’t really make much difference in the required transfer area, which I did not expect on forehand. This is actually good for me, because we now can use thin sheet copper: 0.3 mm! This was a lot cheaper than using let’s say 1mm sheet thickness.


Why copper and not aluminum?
I had the option to choose between copper and aluminum, since both materials have a low heat-resistance factor. Aluminum is cheaper but harder to process and a nearly 2x higher resistance to heat compared to copper. Copper is more expensive but easier to process and has a lower heat resistance factor compared to Aluminum. So I decided to go with copper. The counter part is made from Acrylic as shown in my first build update.

This is the 2D base design I used to fit/measure/calculate. Front and side view:
 

image.png.930dd2cfd8c210427bcaf3eb6f76b468.png



I first made a design (without any limits or boundaries) to find out how I would route the channels for the water/coolant. Then I entered the boundaries of the case and back plate to fit the whole radiator in the case.
I repeated this process for over 5 times, since the space is very limited and I had to think about all the manufacturing steps to see if my model was compatible with a CNC machine.

This was the 3D model I used on the first CNC attempt:

image.png.f204578b43f5a0e7eb5bf24b0caf1c9c.png

Obviously I did not think about all the manufacturing steps which resulted in ruining my first acrylic block on the first touch of the mill in the CNC. Turned out the CNC program wasn’t able to read the 3D model properly to generate a proper path, which wasn’t visible in the 3D model M-code feedback. I had to split up all the steps per depth (-Z direction) and specify the boundaries and path for the mill to run. Also bought a new mill, specialized for acrylic and came with good documentation to calculate all the process parameters. And that actually did the trick, as smooth as butter the mill started following the pre-programmed path.

Theoretical results:

To be entirely sure that we had the right assumptions made during our calculation steps earlier in this post, I now need some sort of confirmation. For this I used SolidWorks Simulate, which is an awesome program that allows me to see if the heat can be dispersed by my designed copper cool-block + heat fins. As I said earlier; I used an easier method to calculate my heat transfer area. So I had no actual confirmation if this would work, theoretically speaking.
 

image.png.3ff09540a87e9c1fb8426eaac93a2180.png

This was the outcome and I am very happy with the results.
I set my ambient temperature to 20
°C or more specifically: 293 K. Where the convectional factor is 15 W/(m^2.K) (the ambient air that will use natural convection to cool the warm object)
On the inside of the board there will be 350 W Heat Power, not specifically coolant, but more 350 W of heat.
Now what does the picture tells us? That the max. ‘outside’ temperature on the cool-block will be 32.5
°C and the lowest will be 28.3 °C.
What I am trying to say is: You don’t want the coolant to heat up to 60
°C where the copper will reach about the same temperature and you burn your hand if you touch it.

 

Conclusion:
The calculation method seems to be a bit overkill and we can probably do it with less copper, but the results are nice and since the whole designing is done. I rather stick to the initial design, without having to change it all in the last month.

If you read this all the way to the end: I hope you enjoyed it and I hope this made any sense to you, because it’s not easy to translate English with scientific terms. But if you do have questions, don’t hesitate to ask! 😄
In the upcoming weeks I will be working on the cool-block and share some more pictures.

20201201_180307.thumb.jpg.48d8d44ce4e0764cc2be102406acdaae.jpg

Edited by Quantum Systems
Added variable 't' (time)
  • Like 1
Link to comment
Share on other sites

  • 3 weeks later...

379719446_bannerworklogtext7.thumb.jpg.a3dba5b203ddb28e7664198983774066.jpg

As the title suggests, I am fully working on the radiator. The water cooling loop is pretty much done, but I keep having some trouble with the distro block. 
I was going to use a PTFE gasket, which is very resistant to high temperatures and (liquid) chemicals, but it was to stiff. I need something more soft to

seal of the entire plate. So I went back to a plate gasket from silicone. I've cut the gasket to the right size and doing another leak test; still leaking air.
But! It improved a lot, so I just have to optimize and figure out some stuff on where it escapes.
I also wanted to try out something different than the traditional rubber cords, because full-face gaskets has its applications within the industry.
The down-side from what I learned, is that a full-face gasket needs more grip on the bolts. That could be tricky and ruin the thread inside the distro|
(metal vs acrylic). 

Upcomming week I will be disassembling the build and make the distro leak tight, its haunting me now 🤣.
Than everything back in the case and organize the cables and than focus on the (outside) body work to finalize the build! I will share some updates on the distro next week as part
of this build-update.

20201210_213915.thumb.jpg.6388b755e7836fd2137805a967d816cf.jpg

80 bolt holes..
20201214_223640.thumb.jpg.b66d57c8253e6300973e5701406bfebb.jpg

 

Silicone gasket, cut out of a 2mm sheet. The sticker was used as a mal. 
20201222_141208.thumb.jpg.15d5caeb15ee48ebdee5f30390223af3.jpg
20201222_142641.thumb.jpg.8d93896e05a4dba68a24470f5df6aa47.jpg
20201222_145657.thumb.jpg.45064f1e00bafe680c7ca9d6eeb924e4.jpg
20201222_164128.thumb.jpg.5f2bce1307790e3bcfc7224a833438ea.jpg

Little sneak peek on the tubes. The tape is there to protect the tubes.
I have to get them out again.

20201219_153631.jpg

Edited by Quantum Systems
  • Like 1
  • Upvote 1
Link to comment
Share on other sites

  • 2 weeks later...

Small update on previous post:
Finally leak tight! 

Currently building everything back in, getting the cables sorted (as I wasn't satisfied with the first outcome). Piece by piece finishing the inside.

Up next, the final touches on the outside as I will be modifying the panels for better air contact. I will do the remaining heat-fins after.

IMG_5135.thumb.JPG.ce59c59267a79d852ea97ddebc4319da.JPG

 

IMG_5137.thumb.JPG.0c60b2e17f9c83018bb09a621bcb5ef9.JPG 


This is the Hope controller, made by Nerd Does (UK). He made a aRGB controller with 8 channels which can have over 1024 leds per channel. I wont be using all the channels
but it comes with software to arrange the lighting per led through a pretty simple UI called Hi-Jinx. To connect this controller I had to get al the connectors off the cables

to get the SM-connectors on. Must say, way better connection than the default ones being used within the hardware business. It will pretty much be hidden inside the case.
 

IMG_5136.thumb.JPG.038bdaffdbba9fcb6ee2a9cab079964b.JPG

 

IMG_5142.thumb.JPG.4e720c3f25f5c7cab4e9916759e66eda.JPG The Hope aRGB-controller inside the case

 

IMG_5145.thumb.JPG.90d967b104eb3c737c0881b324d1e7c3.JPG "Underneath the case"

 

IMG_5144.JPG

Edited by Quantum Systems
typo
  • Like 1
Link to comment
Share on other sites

  • 2 weeks later...
  • 2 weeks later...

962376276_bannerworklogtext8.thumb.jpg.e818c13c5eff3377cb7a6db7a2ddca5a.jpg
Last few weeks I've been busy rebuilding the complete build and placing the cables properly. The loop has been flushed, coolant dyed to purple and filled the loop. No leaks!
So now last weekend it was time to try it's first boot. Stupidly enough I placed the RAM back in the wrong slots and the i5 doesn't have enough memory channels to open all 4 of them. But after fixing that, it booted.
The temperature of the coolant was jumping up by a 15 degrees C (20 amb. and 35 inside the loop) with having the PC at idle. That is to be expected since the fins need to warm up to activate the ambient air flow. 
But still 35 degrees on the coolant is pretty nice. The fun part was that the DDC pump, was very loud in the beginning. After adjusting the RPM's in the BIOS, it's super quiet. No sound at all beside some air bubbles moving around. SWEET!

I had some problems with the HOPE controller, but after some trouble shooting with the seller I found I had the power supply connected on the wrong
molex connection (12V output instead of the 5V)😅 oops... So, yeah, two dead LED strips and getting new ones so close to the deadline was not very pleasant.


IMG_5148.thumb.JPG.2122d33189a6d21548f97c0ca448e9fa.JPG


The Heat Fins
The Fins, well that's a whole different story...
Since the soldering was no option on that big copper base, we reached out to 2-component glue and heatsink plaster. We needed around 80 fins to reach the amount of area to dissipate the heat. But how to get the fins on the baseplate without any machine assistance. (And here's where I went wrong in my thought process.) In the ideal situation you get a copper block with a Thickness of 25mm and use a CNC to carve out all the fins. But there's none available so I needed to improvise here. You want the copper to have as much surface contact with the baseplate as possible. So how are we attaching every single fin on a plate?
It took us nearly 2 days, from morning to night time, to get the method right. Here's how we did it (yes I called in some help to assist on this job).

First creating a bigger surface to work with:


20210124_151010.thumb.jpg.5f40e51f6495119dba1a99cdb2f5d8f4.jpg20210124_151032.thumb.jpg.9aa9ae622653c073e079ef90c7449611.jpg20210124_151047.thumb.jpg.c94764fcee59ec65179470343d97950c.jpg20210124_151109.thumb.jpg.aaacaf52d11ecad830d5971ffdaccafe.jpg20210124_151139.thumb.jpg.bd9ff16230dac02c01da0651c08328cc.jpg20210124_151020.thumb.jpg.12f981ff170290c1d682703ade910e85.jpg

Than, made the bundles per area on the copper base plate:

20210130_211607.thumb.jpg.f0edfe9aaef82f56eaa6d5c723a271cc.jpg20210130_211615.thumb.jpg.f21223adde9b1b7aa3c538f409fd4847.jpg

 

Than attaching the bundles on the base plate with heatsink plaster to fill up the gaps and 2-component glue for strength.
20210130_234926.thumb.jpg.8bda6142d0f3a0191eb7f63b584395b9.jpg

End result:

1U7A2015.thumb.JPG.cfbbc5ff0a2da1a88dc389f424fbe14d.JPG

You might wonder now, why is it sealed of? Well, think about a normal radiator in your home, those also have fins which are also enclosed. Closing of the fins causes a vortex to stimulate the natural convection in between the fins.
It's been quite the ride this project, and it gave me plenty of headaches but also valuable learning lessons. I'm pretty sure I NEEEEEED a CNC now 😆 the machines make the curves and the cuts, so much better than doing things by hand. 


Test results
There's good news and there's bad news. The good news is; it does work. The bad news; it doesn't work as efficient as I hoped.
To clarify, there's a tempsensor inside the loop which allows me to monitor the loop temperature closely, it also has a shutdown option when the coolant gets to hot (max 55 deg.C).
During a burn test I see the coolant temps going up to around 47 degrees. Which made the CPU and GPU get somewhere between 70-80 degrees C. 
the i5 7640X was running at 5.0 GHz, although that does not matter. It only heats up the coolant faster than having it at stock speeds. The block reaches a certain point where the temperature difference is enough to stabilize/balance the heat being transferred to the ambient air.
So my priority was finding this point where it stops heating up the coolant and actually keep it steady at a certain temperature.

 

Primary test results: 
- With CPU+GPU on idle the coolant reaches around 35 degrees C
- When burning CPU and GPU the coolant ramps up pretty quick to just below 50 degrees C.
- Still testing different games, as not much was installed yet on the drive, but for now the coolant goes around 43~44 degrees C. The CPU around 55 degrees C, same goes for the GPU.

 

So these temperatures gave me a bit of a scare, but to be completely honest: I was kinda expecting this to happen. When attaching heat fins to a copper plate the heat transfer surface needs to be at an absolute maximum to ensure the fins warm up as much as the copper plate does 
inside the distroplate. So ofcourse the copper plate warms up, but the fins don't warm up as equally as I hoped. A big upgrade here, would be a heat sink that is made out of one piece. Than it would work ALOT better, and my Thermal Simulation would be more accurate.
Downside is the price tag on that $$$ :P 

03-FEB-'21
I'm not planning on leaving out the dirty details, but I know now that the Deadman switch works. After a 4 Hour session of gaming, the coolant went all the way up to 55 deg.C and the temp sensor auto shut-downed the PC. I really think this Deadman switch was a good idea to install, as I wasn't sure how the radiator would perform. So its really depending on the natural airflow before it starts cooling. I did some tests with different fans as well to see how the cooling responds to that.
A room ventilator 1 meter away, made the coolant temp stable at around 34 deg.C and a 200mm case fan attached to the fins stabilized the coolant temperature at 29 deg.C. Since I don't want any fans involved, it is obvious that it needs a bit of airflow to perform better. Maybe not putting the PC next to the room heater would make a difference as well? 🤣
But it's actually fun to find out how this system performs in real life, instead of a theoretical model. I'm not displeased with the results, but I will need to investigate further.

(I will keep updating this when I have done more games/tests.)

Edited by Quantum Systems
Primary test results on cooling
Link to comment
Share on other sites

Create an account or sign in to comment

You need to be a member in order to leave a comment

Create an account

Sign up for a new account in our community. It's easy!

Register a new account

Sign in

Already have an account? Sign in here.

Sign In Now
 Share

×
×
  • Create New...

Important Information

By using this site, you agree to our We have placed cookies on your device to help make this website better. You can adjust your cookie settings, otherwise we'll assume you're okay to continue..