Project: Hush! - A scratch-built passive radiator/case

[Monkey Puzzled]

Hello all. I've been a long-term occasional lurker here, but never got round to posting. Could mods please leave this in the water-cooling section as it's more a home-built passive radiator than a case project, and input and suggestions from the watercooling section would be useful.

Okay, so for a bit over a year I've been working on building a scratch-built radiator/case for passive-cooling. The idea is for a low-noise solution to running an overclocked machine, and so it's been optimised for passive cooling - the radiator has 2.5 fins per inch and runs fanless. It has a total cooling surface area of just under 43,000cm^2 of copper. For reference, a TRUE has 7,500cm^2.

The radiator is currently plumbed up and running my i750 at 4.2GHz (1.43vcore) and 4850 GPU with a Heatkiller 3.0 block and an MCW-60, with an 18w DDC pump with XSPC top and 1 metre of ½” tubing. Temps are pretty good – runs around 30C idle on the CPU and gets to around low 50’s C gaming. Not sure of ambient temps; standard indoors of around low 20’s celcius I expect – need to get some temperature probes and do some proper testing.


The posts below are adapted from progress updates along the way over the past year from several other forums, so they may contain info and plans that’s amended along the way. They should be reasonably clear, but as I’ve just pasted the updates and not replies explaining bits about it some bits may be unclear. If that’s the case please ask and I’ll explain.


Here's some teaser pics:


Taster pics:







Hope you enjoy the read. Input and suggestions welcome. 
----------------------------------------

[Monkey Puzzled]

Hello all. I've started making a large passive radiator/case. I've been working on it for a while, but since the design is quite complicated I thought I'd wait until it got to a recognisable, likely to work(!) stage before starting the build log.

I have had an innovatek Konvekt-o-matik before , but you really need a shedload of them to handle high heatloads. Plus, they're expensive (~£80/6tubes), bulky and the design is flawed in my opinion - aluminium, made for 8mm ID tubing, and by their design the more of them you add the larger the pressure loss - 8mm inlet splits to 8mm tubes running in parallel.

So I figured I'd make my own;

copper,
designed for 7/16" or 1/2" tubing,
minimise pressure loss by matching resistance of the tubes to 1/2" tubing,
massive amount of surface area (since it's the equivalent of an Aga it needs to have headway for extra heatload),
wide fin spacing and compact enough to be self-contained within a case.

The case is going to be approximately 45cm wide x 40cm deep x 46cm tall + height of castors. So it's slightly smaller than a mountain mods UFO (45x45x45)

The top, far side of the case (the non-window side panel on a normal case) and the bottom will be made up of a large finned copper radiator, hopefully with enough surface area and passive airflow to run completely passive, with air rising up through the case. I wish I was competent enough with google sketchup to draw a detailed plan, but it just seems a nightmare because of the design, so I'm afraid I've largely stuck to hand-drawn sketches. Here's a very rough idea of what the passive radiator element of the case will look like.



The design has changed since this initial drawing - instead of all the pipes connecting into a pipe at the end, it uses a manifold similar to standard off-the-shelf radiators.



Initially I was going to use 30 metres of standard 15mm outer diameter half-hard pipe used in home plumbing and flatten it as in normal watercooling radiators, but after buying a small sample (a gentle elbow bend) and trying to flatten it I realised it's very tricky to get an even inner channel, and it would take forever to flatten 30m of the half-hard stuff...





So, time for plan B. I managed to find a site selling 10m x 6mm outer diameter, 4.8mm inner diameter soft microbore copper tubing, so I picked up 6 rolls for ~£42 inc p&p





I then unrolled them, measured and cut into 48 lengths - 16 x 120cm, 16 x 125cm and 16 x 130cm - the difference is because the 48 tubes will be arranged as so, so the outer tubes need to be a bit longer:






The microbore copper is so soft it's very easy to bend and straighten, though it's tricky to get 130cm lengths completely straight. It work-hardens and quickly becomes difficult to bend, though it can be annealed again by sticking it over a gas hob.

I managed to source some cheap copper sheet for around a quarter of the going rate from shopping around, and for a small extra charge the seller was even willing to cut it with a metal shear into lots of 395mm x 50mm strips for the heatfins.


Here's a picture of the copper. It had been left lying around in a scrapyard for God knows how many years, and was a bit scratched here and there, but should clean up nicely enough.



In the pic are a 39.5cm x 45.5 cm x0.9mm copper sheet for the side wall the tubes go through, 63 of 39.5cm x 5cm x 0.9mm copper strips to be used as heatfins, and 16 of 7.5cm x 39.5cm strips (additional 39.5cm x 5cm heatfins have since been cut from them). There's also some copper bosses (solid cylinders) that I nmay put to good use.


After cleaning







[/QUOTE]

[Monkey Puzzled]

]





After inserting all the tubes I could then start putting the heatfins on.

I ran into a bit of a problem when putting the fins on - they're a tight fit and the smallest angle off horizontal, or pipe angle off vertical means they can be tricky to get on. It's a bit of a trade-off between tight fitting fins with minimal gaps (less solder and better heat transfer) and ease of putting the thing toether. The fins are pretty tight, and in gently hammering them ona few of the tubes were pushed too far down, which you can se in the pic. Luckily, using the pipe cutter without the cutting blade inserted it's possible to grip the tube and push it back through the heatfins.

PIC OF PIPE CUTTER MINUS BLADE



There's only 11 heatfins on in various temporary positions (finished article will have 74 fins);

PICS WITH HEATFINS













I've just got the solder and flux and need to make a watertight box tonight to deoxidise the copper in a vinegar bath before, and after soldering them with a propane/butane blowtorch, since it'll oxidise all the surrounding copper. Anyone know of a place I can get lots of cheap, relatively strong acid? Not sure where to get it and any regulations - was told by a car bits shop that they can't sell battery acid (~35% sulphuric acid) these days. I could probably make do with distilled vinegar but it'll be expensive for the amount I need and take a long time as it's weak.

[Monkey Puzzled]

I then needed to drill 48 x 6mm holes in each of the 74 copper fins and the copper wall. This took a while.





I initially tried using G-clamps and my bench drill and ran into several problems. As the holes were drilled, they pushed a cusp through, deforming the clamped stack. Whilst the cusp/sleeve from drilling is actually useful for soldering and heat transfer, it introduces inaccuracy in the drilling. So I made a jig for putting the copper strips in for drilling.

JIG FOR DRILLING











Sadly the bench drill I have is only 180W, and so lacks the torque to drill large metal holes, so I switched to using an 810W hand drill in a heavy duty drill stand, which allows accurate vertical drilling.





PICS OF DRILLING

After doing some reading up on natural convection and passive heatsink design, I came across a problem sheet set for engineering students on how to optimise the fin spacing for a passive radiator, from a book by a couple of heat transfer professors, and even better, the software it ran on was freely available on the web. So, using the software, I adjusted the parameters to model my heatsink as best I could.

Passive heatsinks rely on natural convection, and this requires the free movement of air over the fins. The fins are much more effective spaced much further apart than in air-cooled heatsinks (~2mm for a Thermalright Ultra Extreme) or even the most sparsely-finned watercooling radiators (~1fin/3.125mm or 8fpi for an RX XSPC radiator).

The simulator models a given width, height and depth of passive heatsink at a given input heatload, and plots the heat transfer of a given fin, and the total heat transfer of all fins combined, at varying fins spacings.





So, whilst the heat transfer for a single fin increases up to a certain point, increased fin spacing means fewer heatfins overall. There's a balance between the two, giving an optimal spacing for a given heatload:

From the simulator there's also another interesting trend - as the input heatload decreases, the optimal fin spacing (fin-pitch) increases, which means I need to optimise the heatfin spacing for the air-water delta T I want to aim for...

The simulator mathematically models a heatsink made from two copper plates held at a set temperature with heatfins that run perpendicular between them, so it's not exactly what my heatsink design is, and in adapting it to model my design I'm not entirely sure how to adapt mine to it, since my design has 48 6mm outer diameter, 4.8mm inner diameter tubes running through the heatfins. I'm unsure as to whether I should adapt it so I equalise the inner heatpipe surface area (4.8mm)to the end-heatplate- to-heatfin surface area in the model, or the outer heatpipe (6mm) surface area... Hope that makes sense!

I altered the parameters to assume just two end heatplates as in the original design, to give a conservative estimate of the performance, and the heatfin spacing (1 heatfin per 10mm, so about 9.1mm between each heatfin). This gave around a 300w heat transfer for a delta of 10C between the air (20C) and the heatpipe/water temperature (30C). But as I say, this is hopefully the worst-case scenario (though the model uses copper-copper joins rather than soldered joints...). A point to note is that the model only calculates the heat transfer from the heatfins - it excludes the heat transfer from the copper tube surface area (~11,000cm^2) and the copper wall (~3,600cm^2).

PIC WORSTCASE SCENARIO (spacing of 10mm on the x-axis)



Adjusting the model to equalise the end plate-to-heatfin surface area in the model with the tube-to-heatfin surcace area gives silly numbers (~560W heat transfer at a 0.5C Delta T, 400W for a 0.4C Delta T).

The real performance will probably lie somewhere inbetween - whilst the best case scenario is probably largely correct in terms of more accurate surface area for the water to transfer heat to the pipes and fins, the model assumes continuous copper joints, and inaccuracies in hole size and loss from soldered joints (~96% tin/ 3.5%Silver/0.5% copper solder) will no doubt lower performance.

Anyhow, enough hypotheticals, here are some pics of where the project is up to at the moment.








After drilling the heatfins I found the holes were marginally too small, using a digital Vernier, in th order of a few hundredths of a mm.

PIC OF DIGITAL VERNIER



So I decided to erode them down slightly by putting them in a drainpipe full of vinegar and harpic toilet cleaner (since it's hydrochloric acid based). I must say, looking through household detergent ingredients for the strongest acid in the supermarket made me feel like a terrorist!



PIC OF TUBES IN DRAINPIPE


PIC OF TUBES OUT OF DRAINPIPE




I then used a microbore pipe bender to make the bends in the pipe. It's a handy little tool, but unfortunately they put far too much paint on it, meaning the measuring bits and the 6mm tube channel was too small, meaning the tube wouldn't get equal pressure around it when bending and would deform too much for my liking. After a quick bit of paintstripping with Nitromors I bent the tubes for insertion into the copper wall. This was a bit fiddly; in order for the bends to line up exactly with the drilled holes I needed to know amount of length the bend took. After a few annoying mishaps, annealing and restraightening I got the tubes bent accurately.

[Monkey Puzzled]

After inserting all the tubes I could then start putting the heatfins on.

I ran into a bit of a problem when putting the fins on - they're a tight fit and the smallest angle off horizontal, or pipe angle off vertical means they can be tricky to get on. It's a bit of a trade-off between tight fitting fins with minimal gaps (less solder and better heat transfer) and ease of putting the thing toether. The fins are pretty tight, and in gently hammering them ona few of the tubes were pushed too far down, which you can se in the pic. Luckily, using the pipe cutter without the cutting blade inserted it's possible to grip the tube and push it back through the heatfins.

PIC OF PIPE CUTTER MINUS BLADE



There's only 11 heatfins on in various temporary positions (finished article will have 74 fins);

PICS WITH HEATFINS













I've just got the solder and flux and need to make a watertight box tonight to deoxidise the copper in a vinegar bath before, and after soldering them with a propane/butane blowtorch, since it'll oxidise all the surrounding copper. Anyone know of a place I can get lots of cheap, relatively strong acid? Not sure where to get it and any regulations - was told by a car bits shop that they can't sell battery acid (~35% sulphuric acid) these days. I could probably make do with distilled vinegar but it'll be expensive for the amount I need and take a long time as it's weak.

[Monkey Puzzled]

Roughly how the removable motherboard tray will sit (looks a little tighter height-wise than it will be due to the top pipes being a little bent down atm). The PSU will sit behind at the bottom, fan facing down.



A sheet of 2mm styrene that was lying around:



Heatgun:

Bending:


A quick and dirty tray to bath it in. So quick and dirty that it leaked:o, so it has an outer box lined with plastic sheet a mattress came in - now it's in the styrene tray to avoid cutting up the plastic lining:




Deoxidising:





Shiny:


Making solder slinkies.





After cutting to solder rings - not sure how many are in the bag, probably a thousand or so.



Blowtorch - gets up to temperature okay.


Various bits for putting fins on and soldering:


Fin in place with 9mm thick wooden spacers for straightening the fins (they need to be hammered gently into place and deform a little in the process)


Solder rings put around the pipes - it's quicker to put them on like this, with a fin above, near the pipe ends, as it made putting the fins on a lot quicker.


Solder rings after tightening with needl-nosed pliers (surprisingly quick and easy).


A freshly soldered fin on top: The fins are a bit rainbow coloured from scale slowly being removed by the weak acid bath - hopefully it'll all go - the bottom fins are quite pink. It was pretty disconcerting to see the shiny copper scale up and get covered with burnt flux, as it didn't seem to get removed at first. I'm a bit concerned about the acid possibly attacking the solder joints - I may switch to cleaning it only at the end when all in place and using a fine wire brush to clean the pipe before soldering.







Unfortunately I couldn't correct the bend that got pushed too far through (3rd from the left at the bottom). A few of the first fins are a little bent as well, though they'll be mostly out of sight behind the motherboard tray.

[Monkey Puzzled]

I changed the method by which I put the fins on;- before I had deooxidised the copper pipe prior to putting new fins on, now I instead changed to using strips of 120 grit wet'n'dry with sellotape on the back to strengthen, which I used to sand about 4cm of the pipes, flux and slide fins on (I had already filed the holes of):

PIC OF SANDED PIPES





All the heatfins are now on, and the pipes have been trimmed down.

PICS OF MONSTROSITY


















I've decided to make the manifold/plenum from copper - it'll have an inner box made from thin copper (0.152mm thick) joined to the end fin by solder paste, and strengthened with some of the thicker 0.9mm thick copper made from 4 leftover copper strips I had.

My friend Robin has some nice tools and machines, so I asked him to maked the inlet, outlet, fillpoint and drain port for the radiator from the 25mm diameter copper bosses I had.

PIC OF LATHE



After making 11.8mm holes with the lathe the copper bosses were tapped on the lathe with a BSP 1/4 tap, cut with a circular saw mill bit and then fly-cut on the mill to give a beautiful smooth shiny surface, which doesn't really come across in the photos:

MACHINING









PORTS DONE -







I'd bought a roll of thin copper when I started the project:

PIC OF COPPER ROLL



PIC OF ROUGHLY HOW THE BOX WILL GO TOGETHER
The sides look angled and messed up at the moment, since only some of the flaps of copper have been soldered together. I need to be able to open the box at this stage in order to press it against the end-heatfins to make good contact when being soldered to the side of the box with the slits (the pipe ends will protrude through these slits). The rest of the thin-walled box will then be closed up and soldered into place, and reinforcing 0.9mm copper strips soldered to the outside of the box and the inlet, outlet, fillport and drain port soldered on.

[Monkey Puzzled]

Time for a proper picture update - now has end-boxes and leak-tested.

Taken to be bead-blasted yesterday - not perfect, and not shiny due to being blasted with tiny beads, but instead now a matt pink. Still, gives itmore surface area I guess. Still needs some tidying up.

The aluminium plates have been cut and the ends of the copper fins slotted into place. They'll be glued into place with araldite 2011 epoxy. The plates will have steps cut to sit flush with the upright aluminium angle.

Here's lots of pics - they're a bit muddled up atm, soi'll probably clean up this post a bit later....

[Monkey Puzzled]

before bead-blasting; leak-and flow-testing:






Pic showing aluminium plate with slots cut and copper heatfins sat in slots:

[Monkey Puzzled]

Update time!

Lots of countersinking and tapping.



The motherboard tray will sit further back towards the back wall, but the countersunk screws poking through from the back aluminium frame just happen to be the right distance apart to go through the motherboard tray holes....








one of the 19mm anti-vandal switches (photographed badly!). It's anodised-black aluminium with white light. I'm still undecided as to whether to anodise all the aluminium black or have it mirror-polished aluminium..



Needs a bit of filing down to tidy up but getting there.



Not fully screwed together and epoxied into place yet so sitting on a box atm. The aluminium frame sits very snugly in the slits for the copper fins.



The copper back wall got warped from bead-blasting. Should bolt to the aluminium easily enough though.

[Monkey Puzzled]

Sadly my friend cut one of the logos the wrong way up.

[Monkey Puzzled]

[Monkey Puzzled]

[Monkey Puzzled]

Time for an update!


(Photos taken on my n97 so not the best quality I'm afraid.)

First, I decided to attach the GPU block to the GPU. I bled the loop to take the tubes offf the block, and decided to open the CPU block whilst I was at it...

And this is what I found - lots of resin gunk must have come off when I degunked the top fill-port thread (more of that in a minute) and so the previous temperature testing was done with water only flowing into about 1.5 of the 13 impingement slits, cooling only two narrow strips about 5mm wide.





I then added the graphics RAM heatsinks. I was going to cannibalise the single-slot stock cooler on my 4850 as it has nice big sections of widely spaced tall copper pins to cool the VRMs. Except after dremeling about a mm into the stock cooler I realised the stock cooler was just aluminium anodised with copper-coloured dye. Damned charlatans!

So instead I opted to cannibalise an old VRM heatsink from my old x1900xt (the pinky-crimson coloured bits) as they're widely spaced fins an nice and tall, combined with an aluminium heatsink that's supposed to go onto 8800gtx VRMs I think.

[Monkey Puzzled]

I then replumbed the loop, started the machine up and did some more temperature testing....

With a degunked HK3.0 waterblock and with the 4850 GPU also in the loop, with the core i750 at 4.2Ghz using 1.43volts vcore it was maxing out at about 55C running intel burn test with the maximum stress I could (about 3.5GB of RAM). During gaming (Arma II) it maxed out at around 50C.

Previously, with just the CPU in the loop it had been reaching 72 in intel burn test in my toasty room.

It idles slightly higher now after degunking but with the GPU in the loop, at around 30-32C compared to 26c before.



Sadly, a minor disaster struck. When I cast the thick polyester resin over the top and bottom manifolds/plenums, some resin had managed to find its way under the mold top glued to the top fill-port, and a little bit into the screw thread for the fill cap. I'd tried to dissolve it out using acetone on cotton buds, but with no joy. I stupidly decided to try scraping it out, and in doing so the thread got a bit worn and slightly mangled. Then when I tried to screw in the top fill-port plug it was really tricky to get it to seal, and I needed to screw it in really tight. Then I screwed it in too tight, heard creaking, and the copper fill-port (the 1" diameter x ~5mm cooper ring with the thread inside) sheared at the soldered joint attaching it to the 1mm thick copper wall of the plenum, like so:

PIC SHEARED FILLPORT



It will be fixable, and I'll recast resin into the bits where shards have come off...

So then I did a horrible thing to a very pretty waterblock. As I didn't have access to a lathe or a 11mm odd drill bit, I decided to cannibalise an EK x1900xt fullcover copper waterblock to make a replacement fillport...

pic cannibalised waterblock



After a bit of filing and sanding I was left with this; a shiny new fillport.

[Monkey Puzzled]

However, there's now resin covering the plenum, so it can't be soldered on. I didn't want to risk it coming off again so I bought a 2mm countersunk drill bit and some m2 countersunk screws, and set at it with my weedy bench drill.





Here's a comparison of the thread on the old and new fill-port:



With 6 screws attaching the fillport to the 1mm thick copper wall it should be able to resist being screwed in.













So that's roughly where it's up to at the moment. I’m currently in the process of filing and sanding down the aluminium frame to a mirror-finish; I’ll post an update with pics soon.

Any questions welcome.

[jakefalcons]

wow that work is awsome! keep up the good work!

[prava]

There are no words to describe how awesome your word have been. I'm afraid even Xtremesystems is not extreme enough for what you did.

Kudos, muy friend, because what you did was insane

[CrazyNutz]

Nice work Very Inspirational

[nikhsub1]

Totally friggin awesome.

[skinnee]

bravo!

[PatRaceTin]

Great project

How about cooling performance compare to normal WC Rads?

[Aquatuning]

Now thats cool!

[Utnorris]

Very interesting concept. Can't imagine how much it will weigh completely filled. If you wanted, you could add a large fan on top to circulate more air around the entire case, but based on the temps you have shown I am not sure if it would be necessary or beneficial. Nice work and keep the updates coming.

[mlwood37]

erm . Now that's a MONSTA rad pmsl.