Another planned passive radiator

[Weldingheart]

Hi, first post here, though I’ve been following SPCR pilgrimages for few years

I got some extruded aluminum heatsinks and planned to build a fanless radiator.
don’t have any experience in water cooling and got very few basic on thermal engineering &
none on fluid mechanics too, so this would be an experimental project and for the sake of some fun
(silence is fun,definetely).
Please give me some input to optimize this

It will be tasked to counter OCed Venice 3000+ in this case
(not something that passive Ninja can not perform I suppose),
I also plan to cool ‘medium’ power GPU such as 6600GT if possible

System:
Abit AN8 Ultra
Athlon 64 3000+(Venice)-plan to OC
Galaxy GF6600GT
1GB Gskill PC4000
Samsung SP 120GB
Antec Neo Power 480W
Compucase(HEC)6A19 Black (0.8mm SECC…)
1 Aerocool Turbine fan as exhaust (the only quiet 120mm fan available here…
the others being Thermaltook,Vantec Stealth,Delta)

Construction:
Basically it’s in the same vein as Repack-cooling Silentio (review here)
and will be mounted on the side panel ala Zalman TNN (I feel like thermaltake at this point )

The radiator is built from nylon plate+copper heatspreader+heatsink,
water channels would be built by milling groove in nylon plate,
there will be another groove around the channels for rubber seal,
Copper plate will be attached to top the water channel, and aluminium heatsinks on top of it,
these three parts will be coupled by several bolts securing the water container
also providing clamping force for copper-aluminium joint.
Inlet will be located on top&outlet on bottom,
quick connect fittings will be installed on inlet&outlet 90 degree from radiator’s plane.
Radiator 1 and radiator 2 will be connected by 3/8” PU tubing.
CPU WB>rad1>rad2>(reservoir)>pump
Planning to include GPU WB, but doubt this setup will handle it.

Parts
The radiator will be consisted of 4 aluminium heatsinks,
copper heatspreader plates, and nylon water channel boxes.
Here are the parts:

-Extruded aluminium(might be 6063) heatsink measure:
140 x 210mm, 20fins (50x210x1.5mm each) with 5 mm spacing between,
it seems adequate to work by natural convection in this apllication.
The heatsinks are refurbished ones; got drilled mounting holes through the base
and milled base profile(stepped)for previous mounting purpose.
I will mill and grind the base so that its surface will become flat,
allowing more contact with copper plate.

Each fin will possess about 21000 sq-mm(32.55 sq-inch),
roughly each heatsink would result 420000 sq-mm (651 sq-inch)
so that total area of dissipation of this passive rad is around 1680000sq-mm(2604sq-inch)
(Alphacool CORA 462 passive radiator measure 2376 sq-inch in Madshrimp’s review ),
well there’re lots more than just surface area for a radiator.

Hole locations may prevents the heatsinks to be coupled directly with
water channels (at least causing limited design on water channels),
thus I would coupled heatsinks and water box with copper heatspreader plate between it
to allow more flexible water channels design,
also to prevent galvanic effect as the heatsinks are not anodized.
I guess actually the copper plate would only hinder heat transfer
by adding thermal resistance/junction, instead of coupling without it.

-1mm thick copper plate(supposed to be C1100), ,
bridging the 'water chamber/channel and aluminium heatsink,
measure 360x220mm:

Here lies the barricade of heat transfer:
1.it’s hard to create fine surface to both copper plate and aluminium heatsinks,
resulted airgaps between two substrate
(just wonder how much tube of AS5 to cover the entire contact )
2.it’s hard to provide even contact over all contact surface of both,
unless there must be so many bolt to provide enough clamping force.

-nylon ‘water box’,(I planned to go aluminium, but it will be very costly,
beside it would even add some more for the anodizing)
Measure 360x220x16mm, there’re two of them, each will accommodate 2 heatsinks.
It will contain the water channel(or maybe a pond) machined in it&
water transfers heat to the copper plate.

-quick connect fitting(brass)with 3/8 barbs for inlet&outlet

Design
I’m not so sure about the design, some spontaneous design comes to mind:
A : rough copytake of Silentio

B : presumably give lower flow resistance? and should allow more contact

C : much more water-copper encounter and less resistance,
but absence of fastening bolts in the ‘reservoir’ area might result poor copper-heatsink contact?

(sorry for crappy pics, they were made in MS Word)

legend:
Big dots on diagonal corners are inlet and outlet
Small dots are bolts
Red line is rubber seal
Dashed line is for heatsink location

I also thought that cross section profile of water channel , is the volume of water matter in this case?
Since only the ‘in contact with copper’ side is utilized for heat transfer.


Pics

Copper plate on background, extruded heatsinks,& the white thingie below the heatsinks
is nylon plate for water box, intel northwood cooler for size comparison


heatsinks-copper plate-nylon plate,
3/8” quick connect fitting will be located opposed diagonally on top and bottom

http://img206.imageshack.us/img206/3125/2rows5fp.jpg
Comparison to mATX mobo,the two rads will be connected via tubing,
total measurement would be 460x360x71mm,
its height is slightly taller than some mid tower

http://img228.imageshack.us/img228/3248 ... lic3ws.jpg
Another alternative;as I luckily found a scrap of 20mm acrylic
(it’s thicker than that nylon plate on the right)
I think this method offer no benefit since it make a thicker rad
with less height (I’ll go for this if I can find another 4 heatsinks and scrap thick acrylic for another cheap)

Another components will be(the only items available in my country):
-CPU WB: Asetek Antartica/Universal/DD Maze 4/Koolance/local made waterblocks like this, and this one
-Asetek GPU WB (?)
-pump:Eheim 1048 or Hydor L20
-10mmOD-8mmID(3/8”?) PU tubing,
-3/8” quick connects fittings
-1/2” to 3/8“converters

Summarizing my enigma:
1.which channel will provide optimal performance allowing for copper-aluminium contact,more water to copper contact, and low flow resistance?
2.suggested bolt to bolt distance to provide even clamping for copper-aluminium?
3.which channel profile should be applied?
4.which WB should I get from above choice?
5.any comment and suggestions to optimize my planned WC system will be appreciated

Thanks in Advance

Regards
-Weldingheart-

NB:sorry if there's some strange words&sentences,I don't speak english very well

[haelduksf]

I'm not sure that I can answer *all* your questions, but I'll give it a shot.

1. I think that your idea B would be best. I'm far from sure of this, but it seems like it will distribute the heat better than A, and C seems to me like it might create pockets of warm water that would be difficult to circulate with such a weak pump. I think, however, that you might want to make your vertical sections much thinner than your horizontal, so that you have some decent water pressure in each vertical channel.

2. I'm not sure of this. I would assume that you wouldn't need a huge number of them; just enough to make sure you have even, consistant pressure. You might also want to use some sort of thermal interface compound between your copper playe and your aluminum heatsinks- I don't think that just bolting them together will give you very good heat conductivity.

3. The 1st, IMHO. more water contact per volume.

4. Tricky. I would personally go for the Asetek, just because the Koolance blocks (I think) mostly have aluminum in them, and if your radiator is going to be all-copper on the inside, your block should be too, to prevent corrosion. The Maze4 is not really worth considering, unless you want to add a lot more GPU/Northbridge/memory/whatever waterblocks in the future. That second custom block looks quite nice too, but more restrictive than the Antarctica.

The asetek GPU block is OK, but not the greatest. What video card are you cooling?

5. This looks like a really interesting project- good luck!

[Weldingheart]

Hi,Thanx for the responsehaelduksf

I'm not sure that I can answer *all* your questions, but I'll give it a shot.

1. I think that your idea B would be best. I'm far from sure of this, but it seems like it will distribute the heat better than A, and C seems to me like it might create pockets of warm water that would be difficult to circulate with such a weak pump. I think, however, that you might want to make your vertical sections much thinner than your horizontal, so that you have some decent water pressure in each vertical channel.

is half wide of the horizontal channel will be sufficient?so there will be more vertical channel created then?
Is Eheim 1048 enough to handle flow restriction on this setup?

2. I'm not sure of this. I would assume that you wouldn't need a huge number of them; just enough to make sure you have even, consistant pressure. You might also want to use some sort of thermal interface compound between your copper playe and your aluminum heatsinks- I don't think that just bolting them together will give you very good heat conductivity.

Well I just found Shin Etsu 150gr tube(could use the remaining for toothpaste ) for $1.5, hopefully thin layer of it may help, I also plan to fine grind the aluminium to about 600grit sandpaper

3. The 1st, IMHO. more water contact per volume.

4. Tricky. I would personally go for the Asetek, just because the Koolance blocks (I think) mostly have aluminum in them, and if your radiator is going to be all-copper on the inside, your block should be too, to prevent corrosion. The Maze4 is not really worth considering, unless you want to add a lot more GPU/Northbridge/memory/whatever waterblocks in the future. That second custom block looks quite nice too, but more restrictive than the Antarctica.

The asetek GPU block is OK, but not the greatest. What video card are you cooling?

5. This looks like a really interesting project- good luck!

The VGA is Leadtek 6600GT,instead of watercool it I think I may just putting Zalman 80HD/Aerocool VM-101 on. not sure about add GPU to the loop

Think to go for Asetek Universal since it has only 2 fittings.

Thanx,

Regards
-Weldingheart-

[mattthemuppet]

with A you'd have one side (inlet) of the rad with a large temp difference (and therefore heat transfer) and the outlet side with a small difference, so you wouldn't neccessarily be making the most of the radiators. B looks better, though I wouldn't have the wide vertical channels at either side, rather two wide horizontal channels top and bottom, connected by a large no. of thinner vertical channels. That should (in my head at least) result in there being no short cuts through the radiator so the heat gets evenly spread out.

With channels design I think there's a trade off between surface area:volume and dead flow. From what my brother (aerodynamical engineer) has been going on about for years it seems there's a layer of water at the sides of everychannel that doesn't move - the highest water velocity is in the centre and the lowest at the sides/ edges. If you have a very high surface area in the channel and a small cross section, you may have alot of dead flow so that the water circulating isn't actually in contact with the sides of the channel. That's a bit woolly and the limits of my knowledge, but something to look into at least.

not sure what else to offer, but it'll sure be an impressive project - one that I've been thinking about for awhile as a mental exercise. Good luck!

[Marci]

1: (C). (A) would restrict flow due to bends. (B) would lose turbulence and flow. Method C but not so much an open cavity as a pin array (similar to apogee's baseplate, or MCW6000...) so greater surface area of copper to water contact. Pins would also create turbulence, lowering thermal resistance. You can add in formers to redistribute turbulence and flow around the cavity...

2: Considered soldering? Cover the entire mating surface of the copper with solder. Sit sinks in position on top with weight on top of sinks. Heat copper up evenly til solder melts...

3: Channel profile... personally, the first... if going with (C) as above. Keep channel depth into nylon thin.

4: Defo the Antartica out of that list, partnered with 1048 pump.

[Weldingheart]

Thanks for the inputs guys!

with A you'd have one side (inlet) of the rad with a large temp difference (and therefore heat transfer) and the outlet side with a small difference, so you wouldn't neccessarily be making the most of the radiators. B looks better, though I wouldn't have the wide vertical channels at either side, rather two wide horizontal channels top and bottom, connected by a large no. of thinner vertical channels. That should (in my head at least) result in there being no short cuts through the radiator so the heat gets evenly spread out.

With channels design I think there's a trade off between surface area:volume and dead flow. From what my brother (aerodynamical engineer) has been going on about for years it seems there's a layer of water at the sides of everychannel that doesn't move - the highest water velocity is in the centre and the lowest at the sides/ edges. If you have a very high surface area in the channel and a small cross section, you may have alot of dead flow so that the water circulating isn't actually in contact with the sides of the channel. That's a bit woolly and the limits of my knowledge, but something to look into at least.

not sure what else to offer, but it'll sure be an impressive project - one that I've been thinking about for awhile as a mental exercise. Good luck!

Interesting, so that profile 2 may be the least efficient?
I'll consider the "modified B" channel design.

1: (C). (A) would restrict flow due to bends. (B) would lose turbulence and flow. Method C but not so much an open cavity as a pin array (similar to apogee's baseplate, or MCW6000...) so greater surface area of copper to water contact. Pins would also create turbulence, lowering thermal resistance. You can add in formers to redistribute turbulence and flow around the cavity...

2: Considered soldering? Cover the entire mating surface of the copper with solder. Sit sinks in position on top with weight on top of sinks. Heat copper up evenly til solder melts...

3: Channel profile... personally, the first... if going with (C) as above. Keep channel depth into nylon thin.

4: Defo the Antartica out of that list, partnered with 1048 pump.

@Marci:
So that it better to create C for the channel/chamber with pins on copper plate do you mean?
I think it could be done by soldering some 10mm length-2mm diameter copper wire/lead on the copper plate to create pins array,but I 'm little bit afraid the pins may be broken and go with the flow in case the solder joint is not sufficient.
Also for soldering the aluminium heatsinks to copper plate,I don't get such a large furnace to handle that, using torch may warp/distort its flatness instead of it's hard to heat a heatsink that size to permit soldering-thought that the aluminium base must be plated by copper/nickel to allow soldering?

Another question: is that any advantage to install inlet&oultlet 45deg (like old MCW5000) instead of 90deg for the rad,?it supposed to lower turbulence at the points below in/out I think?

Well it's time to search for the rest of the WC parts before making the rad.

Regards
-Weldingheart-

[mattthemuppet]

I couldn't really say which would be better with any certainty, but it would definately be something worth looking into. I'll be ringing my brother at the weekend, so I'll ask him what he thinks - he's considerably more qualified than me to give an opinion!