Eng-Tips is the largest engineering community on the Internet

Intelligent Work Forums for Engineering Professionals

Register Log in
  • Congratulations waross on being selected by the Tek-Tips community for having the most helpful posts in the forums last week. Way to Go!

Convective heat transfer from heatsink calculations? 1

Status
Not open for further replies.
ParthZala

ParthZala

Mechanical
Dec 12, 2017
14
0
0
CA
Good day guys,
I am designing a cooling system for computer with large number of small processor. Simply, There will be 24 boards working together.
Each board have 60 heat sinks and generates 1600 W heat. So, we have 26.67 W heat, under each heat sink. This will be a liquid immersed cooling system. I have selected fluid. I calculated required speed of fluid to take away this heat for single heat sink, but numbers coming up here are impractical. I feel, am using wrong method.

I have these properties of fluid, density, specific heat, viscosity at 40 celsius, thermal expansion coefficient.
Heat sink is made up of aluminium and heat generated beneath the heat sink.
I need to find fluid velocity over heat sink.

Can anyone help on calculating this problem? You can also help me to setup the problem in ansys, if it can solve the problem with given information.
​
 
Replies continue below

Recommended for you

Sort by date Sort by votes
24 x 27W = 650W plus whatever your system chips reject.
Where is your 1600W from?
IF your total heat is 1600W you will need at least 2400W of cooling.
You should just need to circulate the oil, but your fluid will need to be cooler than your aim temp.

I built a single board/single CPU machine that was in an oil tank.
That was about 300W from the CPU and another 100W from power regulators and other chips.
I used a very thin synthetic oil, very thin. I had about 900W of cooling and was aiming for 10C.

= = = = = = = = = = = = = = = = = = = =
P.E. Metallurgy, Plymouth Tube
 
Upvote 0 Downvote
24 boards with 60 heat sources each.

Each source is 27W, so 60*27W=1620W per board.

You're probably limited to about 150C junction temperature, and maybe have 125C case/sink temperature. The fluid needs to be flowing vigorously, so probably not cooler than about 25C, resulting in about 100C delta temperature. Assuming a sink surface area of 2 square inches, you need a fluid convection coefficient of about 250 W/m^2-K. You may find that water is probably the only fluid that comes close.

TTFN (ta ta for now)
I can do absolutely anything. I'm an expert! faq731-376 forum1529 Entire Forum list
 
Upvote 0 Downvote
Hi guys,

Each board is generating 1600 W and each board has 60 processor sets. Each set has individual heat sink. Means each heatsink get 26.67 W heat. and total number of boards are 24. So total heat is 38.xx kW. Each heatsink has surface area of 6345 mm^2 area. This is insane, but is manufactured item. Now I am using mineral oil as water will fry that board. Theoraticaly I will need to calculate heat transfer co-efficient (h) for this condition and then velocity of fluid. Calculating h is the problem. I am confused what method equations I should use. I tried to calculate with this method

It's from Fundamental of Heat and Mass Transfer 6h Edition, By Incropera, chapter 7, problem 7.30.
I used values for Mineral oil.
 
Upvote 0 Downvote
So, problem is we got many unitless numbers in CFD. Like Nu, Pr, Pe, Re etc. We need to use different equations based on the conditions to get those numbers. All numbers are not required to solve the problem, so it is also important to used correct number and relations to solve it.
Now, in given example, I follow the method, but answer, I am getting is, 1305.309 m/s. It;s m/s. [ponder] more than mach 3.
It must be wrong. I need help to solve this. I tried to do ansys analysis, but I am not that good in ansys.
 
Upvote 0 Downvote
Have you looked at this from and energy density point of view? How large is your box? and you have 38kW or rejected heat, is this practical in the package that you are looking at?

On problem that you have is that the heat sinks were designed for air cooling, and you want to use a fluid, so the dimensions of the heat sinks are wrong. Have you looked at using very cold air to cool these? 40kW of cooling is the same either cooling air or a liquid. So supply 12T of refrigerated air at some suitably low temperature (chillers get less efficient if you go too cold, so there is a practical limit) and let the heat sinks do what they were designed to do.

You could look at high purity water.
X-ray tubes are water cooled, and they have 50kV across them.
It takes effort, the system must be sealed, the water to start with must have very high resistivity (1Mohm-cm or 0.1 microS/cm) and it gets recirculated through ion exchange resin to keep it low conductivity. This takes great care, but it works very well, and if you keep the water cold (5C) you will get great cooling.

In these multi processor system there are two approaches. One is to use Cu cored boards and mount them into cooled racks, obviously this is highly specialized. The more practical method is to cool plates that are them mounted directly to the processors (just good heat transfer compound, no heat sinks used). A large Cu plates with passages for the coolant (actually a serpentine of Cu tube soldered to a Cu plate) is used to cool the processors. These systems are usually also sealed and dehumidified so that they can run colder coolant without the risk of condensation and the board are conformal coated. This lets you run high flows through each cooling loop, and each plate is only handling 1.6kW in heat load. You control the flow in each plate based on the coolant outlet temp

= = = = = = = = = = = = = = = = = = = =
P.E. Metallurgy, Plymouth Tube
 
Upvote 0 Downvote
not sure how you are calculating the velocity, but using Ed's approach, 38.88kW with a 5C net rise only requires 78.6 gpm. Assuming a 1 ft x 1 in cross section results in only 640 mm/s, and if you allow a 25% capture area, you're still only at about 2.5 m/s

TTFN (ta ta for now)
I can do absolutely anything. I'm an expert! faq731-376 forum1529 Entire Forum list
 
Upvote 0 Downvote
Thank you guys, for your effort, but I am still confused. @EdStainless, isn't it that liquid will most probably take away more heat compared to air.

@IRstuff I calculated that for mineral oil and with temperature difference of 15 degree celsius. Can you please explain how you calculated?
 
Upvote 0 Downvote
But if the heatsinks were optimized for air then they will not be very efficient with a liquid.
And most liquids (besides water) actually have fairly low heat capacity and thermal conductivity (of course better than a gas, but harder to move fast).

= = = = = = = = = = = = = = = = = = = =
P.E. Metallurgy, Plymouth Tube
 
Upvote 0 Downvote
seems all pretty consistent to me, assuming the oil flows through the heatsink reasonably well.

oil_cooling_casqjl.gif


TTFN (ta ta for now)
I can do absolutely anything. I'm an expert! faq731-376 forum1529 Entire Forum list
 
Upvote 0 Downvote
Status
Not open for further replies.

Similar threads

C_V_C_V
  • Locked
  • Question
Controlled Constant Heat Transfer.
Replies
8
Views
22
LittleInch
LittleInch
VermontMaker
  • Locked
  • Question
Cast aluminum plate heat transfer 1
Replies
16
Views
58
MintJulep
M
FEA_User
  • Question
Abaqus heat transfert
Replies
0
Views
38
FEA_User
FEA_User
Ohuti
  • Locked
  • Question
Thermal oil calculations
Replies
4
Views
28
LittleInch
LittleInch
YungPlantEng
  • Locked
  • Question
Calculating characteristic length for heat transfer along nozzle injector
Replies
0
Views
12
YungPlantEng
YungPlantEng
Back
Top