...measured key features of the case, modelled the heat sink in Fusion360 and tendered the SAT file to a bunch of CNC outfits in China - most of them have online quotation systems offering pretty quick turn around for quotes...Impressive. The finish and fit
Could you run us through how you got this made?
I'll post the SAT file in the resources section; it's easiest/ cheapest to pick a CNC shop near you or an international one shipping to you directly - I got mine made in China, and shipping plus taxes almost doubled the price (production and finish was only USD 45), so adding another international shipment hop doesn't really compute...Make a few extra heatsinks! I want to purchase one! PayPal READY!
Meltdown Update #2
With heat now being an apparent issue I went back to the drawing board and revised my original plans for a heat sink bezel.
My initial design posted a while back looked like this:
This was a speculative take on how a heat sink may look like (and looking at the above it appears a bit dense for nat.ural convection). Now with efficient cooling being a bit higher on my priority list, I did some research and calculations on what the fin spacing should actually be to really work well - and the results where somewhat different from my initial design:
At 60 mm fin length (that's length in the direction of air flow, in case of natural convection vertical/ up, so height, really), I'd get to a 5.6mm optimum fin spacing. This has to do with the thickness of laminar flow boundary layer (ie the zone directly adjacent to a surface or "fin", where air doesn't really flow but more or less "sticks" to the surface) being a function of the length of the surface in direction of flow, meaning: The longer the fin, the wider the spacing required for air to move heat efficiently and for boundary layers not to collide (that's in the case of natural convection, where air speed is a given).
For those interested, this paper here by Chun Howe Sim and Loh Jit Seng explains the underlying calculations and assumptions; I have derived a simplified formula from it for 25 deg C ambient temperature at sea level (should be good enough for most domestic sff use cases and save all sort of table look ups) below, where s is the resulting spacing in m, L the fin height in m (length in direction of air flow), and t the difference between ambient temp and maximum expected temperature in K (would be 35 at 25 deg C ambient and 60 degree surface temp):
I have then played a bit with the model and in the end slightly increased the fin spacing to 8.5mm (similar to the spacing of the vent slots in the case) to make things appear a little less busy and reduce machining time:
Further I have dropped my initial idea to fix the bezel from the inside. It turned out that a machined thread within the panel would have required pilot holes of minimum 5 mm depth, so the base plate would have been around 7mm thick, making things rather bulky and expensive- I settled for fixing it from the outside with M3 counter sunk Allen bolts utilising the case's original bezel mounting holes.
Finally I have added some 1.5mm deep pads on the inner side of the bezel to fill-in the openings in the case's front face for full and flush contact with the psu's mounted on the inside:
The resulting assembly should look pretty much like this:
To match my case's colour scheme (black frame with silver covers) I have picked a black anodised finish on milled and sandblasted aluminum. The order to the CNC shop is out - and the part should arrive in a few days time...
Yes - Rhino for most of the modelling (it reads SKP); only exception has been the heat sink bezel which I modelled in Fusion360. Fusion comes with a bunch of handy tools such as screw hole libraries and machining preflights allowing to assess parts for potential manufacturing problemsWhat CAD program are you using? How did you import the SKP?
Is it Rhino3D?
Yes - Rhino for most of the modelling (it reads SKP); only exception has been the heat sink bezel which I modelled in Fusion360. Fusion comes with a bunch of handy tools such as screw hole libraries and machining preflights allowing to assess parts for potential manufacturing problems
The Murata may indeed be the solution should I find a way to shave 2cm off - keep me posted on your findings - am most curios to see how it looks like under the hood!wow. just wow. thanks for building this amazing project. it has been a huge inspiration.
murata makes a 12v650w psu that is 54x40x228: https://www.murata.com/en-us/products/info/power/acdc/2016/0219
there's someone selling one on ebay and i've been thinking about buying it to see if it's easily modifiable. my idea is to see if it's all pcb or if the input side can be shortened, design a new housing with way more exhaust, and line the side opposite the pcb with slim 40mm fans. that would definitely make it thicker than what you're looking for, but maybe it can be modded without making it thicker. i'll buy it and crack it open to see.
Rhino keeps meshes and nurbs in a hybrid environment - when I move mesh data (such as SKP imports) into Fusion, I export them as an OBJ and then place them as a mesh in fusion - that usually does the trick. If your SKP doesnt load into rhino, the file might be corrupt... The selmesh command in rhino isolates all mesh geometry which is helpful to see what-is-what. For nurbs from rhino to fusion I use STEP. Hope it works!I use Fusion regularly, and tried to use Rhino to convert some .skp files, but it didn't work.
Amazing build, I really love the effort put into this and the results achieved. A question: if you can shrink an RPS-200-24 down that much, do you think the same would be possible with an UHP-350-12, to power an entire system off of it? Also, the UHP series are specced well out of ATX spec for ripple (200mv p-2 for the 350-12, 240mv p-p for the 200-12, ATX spec for 12V is 120mv), hold-up time (ATX spec is 17ms, UHP series is 10ms), and probably rise time and some other metrics. Does this worry you at all? Lastly, that bezel looks far better than I imagined, the black finish looks excellent, and minimizes visibility of the contrasting curvatures between it and the case. Good job.
That makes sense. Actually, the documentation for the UHP series states that the passive rating applies only when mounted on a 30*30cm aluminium plate for cooling and that it otherwise needs airflow, so with that in mind it sure makes sense for the PCB (or at least the heat-generating components) to be bonded to the frame. Is the frame aluminium, or is it still steel?I still use the HDPLEX plug between the RHP and the mobo so that adds some protection - the UHP feeds into the GTX 1080 and frankly I just tried it out to see whether it works, and so far I had no problems...
Re shrinking an UHP (that's judgning by looking at the UHP-200 and assuming it's built like the 350): This might turn out a bit more challenging as it's built for conduction with the PCB immersed in a heat conducing, glue-like fluid, and more or less baked to the aluminium extrusion. That's probably why it is rated for fan-less operation, but also means that it's hard to disassemble... I wouldn't bet on getting the PCB out of the extrusion without damaging parts...
The voltage is ok, just wary of the 2.5A rating, cuz here it is 13A from wallCheers! The D-Sub is a very interesting choice...
The c6 holds up just fine- that said, I am running it on 240 v so am well within limits.