Solidworks Case Design & Thermal Analysis

[Crispy002]

As an engineering student with access to Solidworks and its associated add-ons, I've tasked myself with designing a case to the best of my ability with all that I've learned in school (to clarify, I'm still in college). Additionally, I will be performing static and thermal analyses to evaluate the success of my designs. The level of detail I'm aiming for in this project will take some time to accomplish, but I will update as I make progress.

Let me know what you think

Introduction
Heatsink Revision 2

 

[Crispy002]

Introduction

Initial restraints & design goals:

• Must have an aluminum frame
• Must be no larger than 23x23x8"
• Must fit a full sized ATX motherboard and components internally
• Must fit watercooling components

Non-essential goals:

• Passively cooled system (all components)
• Silent operation

Yet to be decided:

• Target aesthetics and/or theme
• Use of existing parts

With these ideas in mind, I spent a fair amount of time sketching my ideas on good ol' fashioned paper. In these sketches, I found that fitting a radiator inside the case substantially increases the minimum size dimensions. However, mounting traditional radiators externally can be potentially dangerous (risk of punctures) and, in my opinion, is ugly.

With that thought in mind, I began with a large waterblock.
Note: All images will display this block as copper due to it being the ideal metal in this application.

Simple enough. Let's slap an acrylic sheet on it though to seal everything up. This WILL require an O-ring to make a full seal, but the necessity has been temporarily ignored. For any doubters, it's possible to make one in this shape.

Those with a keen eye may have noticed the channels are different between the two renders above. Additionally, the observant reader would recognize the screw hole pattern in the acrylic strongly resembles the pattern for an inverted ATX motherboard.

This takes care of the motherboard mounting, but if this large block is to serve as a radiator, it's going to need some fins.

Those who lack formal training in Solidworks might consider the above "good enough" to continue designing the rest of the case. However, failure to analyze the thermal characteristics of the plate could prove disastrous down the road. It'd really suck if it couldn't dissipate enough heat right? Right. So let's figure that out.

I've tested the plate using Solidworks Flow Simulation. There are other fluid modeling programs available that are superior to this one, but the differences are negligible in this application (to my knowledge). Below are three views of the plate assembly during calculation of the steady-state properties of the plate. I've assumed an incoming water temperature of 85F (30C) in ambient air temperature of 68F. The water pressure is set according to my current pump's data sheet.

One of the many results of the thermal analysis. The exit temperature of the fluid was calculated to be 76F. Not bad, but I think some rearrangement of the finned heatsink could improve things a decent bit. That's for another time.

Then came the question of how to make the finned heatsink fit with the case wall. That solution actually wasn't too hard.

On that note, I decided to do an unevaluated model design of the general case components and layout. This is subject to extreme change at this point since nothing has been calculated for stability, strength, etc. All parts shown in the render were designed by me with the exception of the power supply and hard drives.

What remains to be done:

• Complete case conceptual design
• Test alternative forms of copper block
• Static analysis of hard drive rack
• Static analysis of entire case
• Thermal analysis of entire case

Let me know what you think

 

[EdSurly]

Looking forward to reading more.

 

[Wylee]

Really cool. I'm thinking about doing a project like this right now. I'm psyched to see what you do with this.

 

[jrocks84]

I was thinking of doing something similar recently, but decided to stop since I have zero experience with Solidworks along with no time to learn it. I hope this works out for you, and look forward to updates.

 

[AgentQ]

Some observations:

I'm not a big fan of the heatsink protruding from the side of the case, or the mounting of the motherboard directly to your radiator piece.

I'd ditch the separate heatsink, and instead mill some channels directly into the back of your copper radiator piece. Make this piece the top of the case, so heat can rise unimpeded from these channels and out the top. This will also allow you to have heatsink fins that cover the total heat dissipating area. If the piece follows the outline of the top of the case, you should have a considerable amount of surface area.

In your picture, the hottest part (in red, near the inlet) doesn't actually have any heatsink fins. This means the heat absorbed there is being dumped back into your case. Not a good idea. At the very minimum, move the heatsink toward the inlet and make your channels wind through the area under the heatsink.

If you do want to stick with an attached heatsink, you'll probably want one with wider fins. Narrow fins (like you have pictured) are generally better for forced-air cooling. Passive cooling does better with slightly wider fins. There are plenty of closed-form equations to calculate a good fin width/spacing starting point.

Also, your thermal numbers seem a bit optimistic at first glance. How are you calculating your inlet water temperature? Here's what I'd do:

Take the flowrate you're using for your calculations and determine what volume of water will pass through your waterblocks in one unit of time. Next, take the amount of heat you're going to dissipate from the CPU, GPU, chipsets, etc (I'd use 80% of the power you're drawing from the wall, just to be safe. The rest goes through the PSU, smaller components, etc.). Determine how the temperature rise that would occur when the volume of water you calculated above was heated with the power your components are dissipating. The time units will drop out, leaving you with a temperature increase for your cooling loop. Add this temperature increase to your target radiator outlet temperature. This becomes your radiator inlet temperature.

Your radiator must be efficient enough to cancel this thermal rise. If it doesn't do this, the radiator outlet temperature will be higher than your target. This means you have to re-run your calculations with the higher outlet temperature, and re-run your simulations. Repeat until you can take the radiator outlet temperature, add all of your heat, dissipate the heat in the radiator, and come back to the same target temperature you started with.

You probably won't have to do this many times before you get a feel for how efficient your radiator needs to be.

Alternatively, SolidWorks fluid simulation may allow you to add a heat source to the loop somewhere like the higher-end packages. I've never used the SolidWorks thermal/fluid software.

 

[Crispy002]

Thanks for the input AgentQ, you've made several valuable points. I'll update you on my current changes as best as I can describe without pictures (I'm not home currently).

As a result of looking at my initial results, I came to many of the same conclusions you provided. The finned heatsink was poorly located, and the initial fluid temperature seemed too low. As a result of these conclusions, I have since moved the heatsink more towards the outlet and enlarged it slightly. Also, machining from one solid block would be ideal in terms of performance (no resistance at the boundary) but high in cost. To facilitate modeling, I will keep them as separate parts. However, I have instructed Solidworks to treat the two parts as one continuous piece during calculations.

To gain a better understanding of its performance under higher temperatures, I performed an additional thermal analysis with an assumed intake temperature of 100F. Solidworks Flow Simulator DOES offer the capability to model heat generation from computer chips, but my model has not progressed enough to use this feature. Rest assured, I full intend to dial in the analysis accuracy as the model gets completed.

Good note on the heatsink fin width--you caught me on that one. Stay tuned for more updates and analysis to support the above.

 

[Crispy002]

Double post, please delete.

 

[ckryan]

This is pretty fresh.

Keep it coming.

 

[AgentQ]

Thanks for the input AgentQ, you've made several valuable points. I'll update you on my current changes as best as I can describe without pictures (I'm not home currently).

As a result of looking at my initial results, I came to many of the same conclusions you provided. The finned heatsink was poorly located, and the initial fluid temperature seemed too low. As a result of these conclusions, I have since moved the heatsink more towards the outlet and enlarged it slightly. Also, machining from one solid block would be ideal in terms of performance (no resistance at the boundary) but high in cost. To facilitate modeling, I will keep them as separate parts. However, I have instructed Solidworks to treat the two parts as one continuous piece during calculations.

To gain a better understanding of its performance under higher temperatures, I performed an additional thermal analysis with an assumed intake temperature of 100F. Solidworks Flow Simulator DOES offer the capability to model heat generation from computer chips, but my model has not progressed enough to use this feature. Rest assured, I full intend to dial in the analysis accuracy as the model gets completed.

Good note on the heatsink fin width--you caught me on that one. Stay tuned for more updates and analysis to support the above.

Thanks for the updates. Keep us posted as you make more progress.

Sounds like you're on the right track. I would definitely separate the heatsink and your copper piece, though. You should be able to assign a thermal resistance to the interface where those two parts meet. Nailing down the number to use here will be tricky, but luckily the SolidWorks help has a good starting point:

http://help.solidworks.com/2011/eng...tance.htm?id=8ef23a02a46348f1909e9ac36a99f892

You'll definitely want to use some sort of TIM there. Arctic Silver or something similar. This will decrease your thermal resistance at the junction significantly.

I'd e-mail the ICDiamond crew to see if they're willing to share some numbers with you for your project. The engineers there seem to do some great thermal interface material testing (despite the semi-spammy marketing practices of their vendors).

 

[Uncle Humjaba]

Sounds like a project I recently did for my heat transfer class. Did all of the calculations by hand though, no solidworks simulation to help us Keep in mind copper is heavy and expensive.
Behold! My behemoth laptop cooler (channels in top contained paraffin wax, to absorb excessive heat [i.e. from gaming, benchmarking, charging, etc.]). I made it through with an A - if you have any questions let me know. Won't be able to answer specific equation questions (left my book at school) but can point you in the right direction.

Also, if I might make some suggestions - on that big copper slab with the channel cut in it, there is a very poor surface area to material ratio... I'm not sure how far along you are in your studies, but a major component of heat sink design is surface area. With a better design, you could transfer the same amount of heat using less material (and therefore lower cost, lower weight, etc. etc.). If your simulation was set up correctly, it appears as if the majority of your heat was dissipated into the copper before the channels even made it to the fins.

 

[Crispy002]

Thanks for those model pictures and the tips Uncle Humjaba. That must've been quite the project. I'm nearing completion of my second year as a Mechanical Engineer, so my skill-set and design intuition are still developing. I will be taking Thermo and Fluids (naturally, the courses most relevant to this project) in the fall.

Project Update:
I've recently discovered an error in my use of Flow Simulation. The suspiciously good temperatures from the images posted above are indeed too good to be true. I incorrectly configured the simulation to have the entire assembly submerged in 3 cubic feet of room temperature water instead of air*. Re-running the simulation with the appropriate corrections made resulted in complete failure of the design.

In a purely natural convection setup, the fin-width corrected design was yielding a maximum of about 30w dissipation with 100F intake temp. Increasing the size of the heatsink to slightly exceed the size of the plate increased this to about 54w dissipation with 100F intake temp. Needless to say, this won't do.

In light of this new information, I have decided to scrap the idea of a passive external heatsink. My current intention is to test a low CFM design I sketched out today which (oddly enough) is somewhat similar to your design Uncle Humjaba. I expect to have the modeling and analysis completed on the 12th.

Thermal analysis of the enlarged plate with the corrected simulation for those interested. Temperature is in Kelvins.

*Note the time-stamp of 5:37am from the system clock in one of the earlier thermal pictures.

 

[Uncle Humjaba]

Eh, you won't learn much with regards to this in thermodynamics - I just finished that class and I didn't like it. It's got more to do with the transfer of abstract quantities like enthalpy and entropy. Heat transfer is really where it's at, though at my university it has fluids as a pre-requisite, so that may not be until spring for you.

Like I said before, it's all about the surface area. The reason you see coils like that in air conditioners, etc. is because there is fluid moving across them (into the plane of your copper chunk). Google some pictures of water blocks for inspiration - it's not quite what you'd expect. Feel free to PM me if you run into any issues, I enjoy this crap (and this is why I'm going into engineering...)

 

[Crispy002]

Heatsink Revision 2

Had some spare time tonight so I went forward with my second heatsink design. The design is very similar to a traditional radiator, featuring channels with fluid flowing through them. These channels are then actively cooled by forced convection from a low powered fan.

A plate attaches to the top of this assembly, sealing each parallel channel.

The thermal characteristics of this heatsink are considerably better than my previous design, but there's still room for improvement. With incoming water at 100F, the radiator dissipates approximately 220w with a single ducted 120mm fan rated at 20 cubic feet per minute.

Below is an image showing some of the particle trajectories of the air blown by the fan.

The intake and exhaust ports are located underneath the plate. Each pools into a mini-reservoir which feeds the narrow channels. The current geometry provides a fairly even flow rate across each individual channel, but there is room for improvement.

For the first rendition of this design, I'm pleased with the results. There is significantly less material than the previous design, it's easier to machine, and it should be near silent. I'll likely tweak this design some more over the next few days and then either try another concept or continue the actual case design over the weekend.

Let me know what you think

 

[HAF72]

Wow, looks very interesting. Will be keeping up with this thread!

 

[Uncle Humjaba]

Just out of curiosity, are you planning on actually making this one day? Or is it just an experiment?

 

[Crispy002]

Just out of curiosity, are you planning on actually making this one day? Or is it just an experiment?

If the analysis shows the design is functional and the project remains economically feasible, I have every intention of seeing this through to a full-scale build. My campus has all of the machinery necessary to make this a reality and I have the knowledge to operate them.

 

[jrocks84]

Is the new version of the heatsink going to be placed on the top of the case? If it is still behind the motherboard, I would be worried about heat transferring from the heatsink to the motherboard if it wasn't thermally insulated.

 

[Crispy002]

Is the new version of the heatsink going to be placed on the top of the case? If it is still behind the motherboard, I would be worried about heat transferring from the heatsink to the motherboard if it wasn't thermally insulated.

At this time I'm more concerned with achieving the thermal goals than determining how it will fit in the case. However, my current intent is to mount the rev. 2 heatsink internally to the top or front panels. I might even mount two!

 

[yottabit]

I think it's going to be hard for you to beat a traditional radiator in design. You say the design is very similar to a traditional radiator but it doesn't look like any radiator I've ever seen. All radiators (and heater cores) I've seen consist of a tube (usually copper) that is bent back and forth either once or multiple times (like a strain gauge) to form some symmetric pattern. Then there's hundreds or thousands of fins made of either aluminum or copper bonded to the tube.

The trouble with your design is that lots of the copper you are using serves little purpose. You only really need the copper close to the interface with the water. Then you want to reject the heat to the air immediately in as little distance as possible. The designs for your radiator look more like a waterblock to me, and if you do want to go forward with that design you'll want to check the head required to pump water through there at your desired flow rate. It may be quite high.

Also, if you go for a design like that I think you'll have issues sealing it over it's entirety. Machining is not an exact process and when you get into parts with large dimensons (particularly long, thin parts like yours) it's easy for the parts to warp and distort. Even with O-Rings I think it would be difficult to ensure a good seal.

That all being said, I really like your project I'd like to see you investigate using a traditional radiator and maybe see how they compare in SW... I think you might be pleasantly surprised

 

[Crispy002]

I think it's going to be hard for you to beat a traditional radiator in design. You say the design is very similar to a traditional radiator but it doesn't look like any radiator I've ever seen. All radiators (and heater cores) I've seen consist of a tube (usually copper) that is bent back and forth either once or multiple times (like a strain gauge) to form some symmetric pattern. Then there's hundreds or thousands of fins made of either aluminum or copper bonded to the tube.

The trouble with your design is that lots of the copper you are using serves little purpose. You only really need the copper close to the interface with the water. Then you want to reject the heat to the air immediately in as little distance as possible. The designs for your radiator look more like a waterblock to me, and if you do want to go forward with that design you'll want to check the head required to pump water through there at your desired flow rate. It may be quite high.

Also, if you go for a design like that I think you'll have issues sealing it over it's entirety. Machining is not an exact process and when you get into parts with large dimensons (particularly long, thin parts like yours) it's easy for the parts to warp and distort. Even with O-Rings I think it would be difficult to ensure a good seal.

That all being said, I really like your project I'd like to see you investigate using a traditional radiator and maybe see how they compare in SW... I think you might be pleasantly surprised

No heatsink I design will outmatch currently available radiators in price/performance. "Traditional" radiators are the pretty much the best at what they do. Their wide use helps prove this. I'm hesitant to humble myself and put a regular rad in the case since I would probably end up designing a do-it-yourself TJ07.