phase change cooling The long road to designing and building custom phase change evaporators

[Lord Nicoll]

[preface; in this post I talk about the long road of idea's and attempts at making custom evaporators for phase change coolers, talking about my rationale and why I did things the way I did, if you think you'd enjoy it why not give it a read, but at this point with all said and done the project was worth it I think. Maybe it'll inspire you or give you a good idea or two about your own projects]

So for the last few months I've been working on new evaporator designs, for use mainly in lower budget single stage units (usually called SS units but for obvious reasons I tend not to call them that, at least not around people who don't know what I mean ahahahha). By lower budget what I really mean is under €150 for everything including 3D printed mounting hardware, holds up well with SS temps, rods and nuts, springs, washers,  evapator block itself. These current prototype evaporators are all 50.8mm, 2 inch round copper, C101 alloy. They use a dual flow path that is forced to crash into itself to cause more turbulance in the flow and help the pressure spike then fall again in the constrictions to ensure refrigerant is evaporated and able to absorb as much heat from the block as possible. As shown here in the first prototype to be machined (not actually the first head tested or even finished, it was left until later as the taller evap heights are harder to make) The design uses 3 concentric channels with a central chamber, On some models the refrigerant is introduced in the centre chamber and sucked out though the outer ring, these are GPU models, and on the CPU blocks the refrigerant is introduced on the outside but either can be used on either, it's mostly the suction port location that makes the centre injection design better for verticle set ups. I rate my evaporators within each diametre grouping, 50mm, 20mm-32mm, 65mm, 30mm-45mm etc. The taller the internal channels the more room there is for higher flow rates of refrigerant to expand, more surface area to absorb heat energy and the the harder and more epensive they are to make. While the surface area is most important on the bottom, the walls do help bring up heat energy to be absorbed. The shortest ones are around 20mm in height, with base thicknesses of 2.5mm to 3mm. The thinner the base the better for evaporatation however too thin and it can deform from the pressure. 3mm is the absolute thicknest it should be, with somewhere around 2.5mm being optimal while still being able to resist the pressure. The tallest evaporator in my 50.8mm family will almost certainly be 32mm tall, as at this height 3mm endmills aren't too happy about cutting that far into a pocket, to get to the desired depth the endmill will be 29mm into the part, which is almost 1:10 diametre to depth cut. At €22 per carbide 3mm end mill it's not exactly easy. The second and third evaporators I made used what I am retroactively refering to as the Gen 1 design, this design just uses straight walled channels. The Gen 2 design uses scalloped cuts in the walls of the channel for extra surface area and further disruptions to the flow paths. The first evaporator can have those scalloped cuts added to it now since It was not completed and is therefore still open. Other than plunging the endmill in partially to the wall, the process is the exact same. Most of my evaporators are TIG welded at the top, unlike most evaporators you buy that are brazed by the end user. TIG welding produces a very strong and durable weld if done correctly, which is not so easy on copper. Welding copper in general is unusualy however I learned to TIG weld on copper so I'm not that bad at it. when I start making more of these I'll likely switch to welding them in a purge chamber for as close to a flawless fusion weld as possible and ensure the copper remains impurity free. 

The tools I use to make these evaporators is almost entirely my mini lathe. I use a toolpost mounted spindle with rotating endmills to cut the channels. Currently I have a really crappy "500" watt spindle with an ER11 collet. The shaft that mounts the collet to the motor is only 5mm and so has a crazy amount of deflection. This is basically unacceptable really, it's why I can only make evaps about 18mm deep currently reliable. I plan to upgrade to an ER16 spindle soon, these spindles are typically claimed to be 1500 watts and I'm included to beleive them, unlike the "500" watt one I have that I suspect is around 350 watts. The ER16 ones are inverter driven and come in both air and watercooled variets so I do suspect they're at least more of the 1500 watts they claim. The main benefit of the larger ER16 spindles is the thicker shaft really, at 5mm it can barely hold up to real cutting loads like you'd see in the channel cutting operation. My minilathe is fairly large as far as minilathes go, a 1100 watt brushed motor, 220mm x 750mm, very long for a minilathe, and about a 100mm x 100mm cross section on the bed, which is better than most of the minilathes, it works out as a 9 inch minilathe more or less, 8.7inches. 

 

Evaporator Prototype A000, model M01(P)

 

Spoiler

This was the first evaporator attempted. It actually still remains unfinished as I decided to convert it from a gen 1 to gen 2 evaporator since I haven't finshed it. I haven't finished it really because I already broke 2 HSS 3mm endmill and 1 carbide endmill getting it to this point and breaking another would make my wallet cry more. Cutting channels this deep in a metal like copper can be pretty dangerous, especially on light duty machines like hobbyist lathes. If the tool crabs into the copper and stalls one of the operations it can severely damage the rest of the equipment or the work peice. Extreme caution should be used if you want to follow the plans here and don't have proper tools. As seen in the photograph the first prototype came out very nice, a little bit of chatter but not an issue for this use, until you remember this isn't anywhere near as deep as it needs to be, and it already has that much chatter, going much deeper with the current set up is not really an option. I did not cut the slots between the channels yet but you get the idea of what these look like, and the fact I hit the dimensions pretty good on the first one gave me slight false confidence in how easy it would be in the next two as you'll see how they both had their own issues. 

 

 

Evaporator Test type A001, Model M01(T)

 

Spoiler

The first evaporator to be fully welded with the injection port and suction line port, this type laid the ground work for the next types to be based off of. The two stalk ports where welded on the back to hold them in place and creat a strong joint, then brazed from the top side to ensure if there was gaps in the welding, due to the thin wall of the copper tubing, it would not leak. 

On this example I messed up the channel spacing (my lathe does not have a DRO so I guess I messed up on the hand dials or calculator) on the channel walls and it shows as inconsistant spacing in the cuts. The channels between the ring channels however was very nicely possitioned on centre line and all where almost perfectly in line. This example will be used in a very small test unit targeting around -25°C at about 120 watts.

 

Evaporator Test type A002, model M01(T)

 

Spoiler

On this example I got the concentric ring spacing correct but somehow managed to mount the toolpost spindle off centre line and not realise it, so while the spacing between rings is much better, the cuts are not on centre line. That is unfortunate as it will likely affect performance of the evaporator.. This evaporator is slated for use in a much more capable test unit testing the performance of the evaporators at near athmoshpheric pressures (around 1 barA), targeting around -48°C and around 200 watts of cooling power. Actual performance depends on the compressor's ability  to keep up, and the evaporators ability to absorb and disipate latent heat for the phase change. Both Evaporator A001 and A002 are the same models and follow the same design, with minor changes in how they where made and the obvious screw ups I made. 

 

Evaporator Prototype A003, model M02(PX)

 

Spoiler

I don't have a picture of this evaporator as it's no where near finished, and it's very very special, probably a one off the likes of which won't be seen again. It's a dual sided evaporator for use with dual PCB nVidia GPUs, the GeForce 9800 GX2 and the GeForce GTX 295, having a dual contact face on both sides and the tubes coming out the top, it is still in the design phase and the evaporator has been foxed to the correct side but has not had any channels machined in yet

 

Evaporator Prototype A004, Model M03(P)

 

Spoiler

A special request from a someone in the European Phase Change building community, they where looking for an evaporator for slot CPUs and older GPUs, like the ATi Radeon 9800 XT, it was made to be a centre injection model with the suction line port on the side of the evaporator. This Evaporator was made to have wider channels tom allow more room for the refrigerant to expand and to ultimately make it cheaper, the load this will see was quite small. A 4mm end mill was tested on this to see how larger endmills perform and the results were mixed, the actual end mill was also a cheap carbide one from aliexpress and no where near as sharp as the expensive 3mm carbide but also much thicker, being 4mm. This meant I could only get 2 channels into the design so to compensate I came up with the scalloped cuts you see, which marked the end of the gen 1 evaporators and signaled the beginning of the gen 2 before the gen 1 even got properly tested. The chuck was held by hand as the tool was plunged in to make the cuts and was indexed entirely by hand. I have since added a proper spindle lock and will be devising a way to index the chuck reliablly. 

 

 

A couple of moments was pretty sketch making this and led to me making the modifactions like a spindle lock, however I wanted to get this done fast as it had a dead line and for the price I sold it at (€50 + €15 shipping) the customer was very happy, as they said, the internal cuts didn't need to be perfect, they wouldn't be seen and it added more surface area so didn't matter, however I am a perfectionist and soon will be attempt another model like this, probably based on the M01 type, for CPUs, around 25mm tall, mostly on principle that I'm not this bad. The cheaper endmill's performance also is a bit more obvious as it didn't cut as nice however all things considered, for 1/10th the price it did a nice job. 

 

Evaporator Brazed onto a lineset
 

Spoiler

This is what the evaporator, suction line and capillary tube look like when all brazed up and assembled together. There are a number of ways to do this and this is the method I prefer but no reason you can't do it another, use other metering devices etc.

And there is it, the extent of my 6 month long expedition into designing, researching, and actually building phase change evaporators. Out of a large hunk of copper I believe was 200mm longs and cost about €100, I managed to get the raw material for 6 evaporators, and a lot of copper chips and swarf. Hopefully when I try to machine the A000 model and finish it I can get a really nice example of what these evaporators can look like when everything goes right and they come out nice. In the future I hope to electroplate them with nickel too, or as I like to call it Nicoll Plating. 

 

[Edit] ; Wow this text editing is awful and annoying the way it has two scrolling windows inside one another, wish there was a full screen version you can enable. Sorry it's taken a like 10 minutes and 10 edits to make this post readable lol

[Somerandomtechyboi]

I might consider phase change in the future for f ing around with xoc but how do you even build one of these things?

 

For all i know the easiest way to get subzero that can actually cool things is just use a feaking ac unit and have the evaporator be submerged and used as a waterloop like what linus did when chilling threadripper

 

Theres also tecs but those are useless outside of nb and ram cooling

[CommanderAlex]

2 minutes ago, Somerandomtechyboi said:

I might consider phase change in the future for f ing around with xoc but how do you even build one of these things?

 

For all i know the easiest way to get subzero that can actually cool things is just use a feaking ac unit and have the evaporator be submerged and used as a waterloop like what linus did when chilling threadripper

 

Theres also tecs but those are useless outside of nb and ram cooling

If I had expenditure to screw around with, I'd buy one of these cases to do some extreme overclocking with, just like the one in the LinusTechTips video where they used a similar case.

 

 https://ldcooling.com/shop/ld-pc-v10-reverse/94-ld-pc-v10-phase-change.html

 

OP: That's really awesome!!! I'd love to see how well this turns out as you progress further and see some results. 

[Somerandomtechyboi]

7 minutes ago, CommanderAlex said:

If I had expenditure to screw around with, I'd buy one of these cases to do some extreme overclocking with, just like the one in the LinusTechTips video where they used a similar case.

 

 https://ldcooling.com/shop/ld-pc-v10-reverse/94-ld-pc-v10-phase-change.html

Yea those are expensive asf, i think ill just stick to tec ram and nb cooling for now and wait for a good deal on a cheap used ac unit or something

 

Priority number 1 is ram cause if i wanna go 700fsb+ on my ud3p once i fex the damn thing i need good rams, my extreem dark top out at 1250mhz 6-6-6-18 so only 625fsb, ~1150 max stable so only 575fsb stable which sucks, then once ram is solved i gotta subzero the nb aswell cause to even get near or over 700fsb most boards will require subzero, some lucky boards do ~700fsb on air but thats just ludicrous, though maybe if the nb from my dud p5q pro turns out to be trash guess ill just chop my p5q and yoink its nb, asus boards are trash so only 550fsb max for the p5q with modded bios to p5q pro turbo, barely beating my g31m s2c at 542fsb so i have no issue chopping one up to put on a superior board

[BuckGup]

Have you tested the performance of it in a system yet?

[Lord Nicoll]

Just now, BuckGup said:

Have you tested the performance of it in a system yet?

I'm waiting on a new vacuum pump to arrived, shipping from china was never fast, shipping from china with COVID is about as fast as a glacier, so as of right now sadly only A004 was used in a system, where with R507C it held -48°C on a 120 watt load, with a 10.8cm³ rotary compressor.

[BuckGup]

Just now, Lord Nicoll said:

I'm waiting on a new vacuum pump to arrived, shipping from china was never fast, shipping from china with COVID is about as fast as a glacier, so as of right now sadly only A004 was used in a system, where with R507C it held -48°C on a 120 watt load, with a 10.8cm³ rotary compressor.

Did you have any condensation issues?

[Lord Nicoll]

17 minutes ago, Somerandomtechyboi said:

I might consider phase change in the future for f ing around with xoc but how do you even build one of these things?

 

For all i know the easiest way to get subzero that can actually cool things is just use a feaking ac unit and have the evaporator be submerged and used as a waterloop like what linus did when chilling threadripper

 

Theres also tecs but those are useless outside of nb and ram cooling

Well, you'd typically begin by making something to hold or house everything. I weld fully enclosed cases but some people just bolt it all to a board of MDF. At that point you start picking components, like compressor, condenser, fans, copper tubing, filter drier, service ports, flexible suction line, metering device (I typically use capillary tubes but EEVs and TXVs are also options) and the evaporator. You'd use CuP 94 to braze all the copper tubes together after cutting and bending them into shape, add the compressor and filter drier, vacuum the system to about 500 microns, do a nitrogen sweep to help remove moisture, vacuum again to about 100 microns vacuum and charge your selected refrigerant, then fine tune the charge in the unit under load. Selecting the temperature and load the unit can take primarily depends upon the compressor, evaporator and metering device, plus the size of the condenser. 

[Lord Nicoll]

4 minutes ago, BuckGup said:

Did you have any condensation issues?

I wasn't the one who built it into a unit so I can't say, however the person who did is a pretty well known HWBot XOCer in the extreme league, so he'd know what he's doing. Obviously our units that are built by botique builders perform much better than the ones you buy, because we build with better evaporators, larger condensers and bigger compressors and while following EPA guides we don't exactly follow the other guidelines, so units tend to use more power and maybe wouldn't be fully certified like consumer ones, however as I said, ours cool higher capacities with lower temps. One of the people on the HWBot discord, Natchfalke, built a 500 watt electronically controlled -105° 2 stage cascade cooler, now that is a mighty beast of a unit. Weights like 86KG, uses like only 1600 watts to cool 500 watts, pretty impressive. 

[Somerandomtechyboi]

1 hour ago, Lord Nicoll said:

Well, you'd typically begin by making something to hold or house everything. I weld fully enclosed cases but some people just bolt it all to a board of MDF. At that point you start picking components, like compressor, condenser, fans, copper tubing, filter drier, service ports, flexible suction line, metering device (I typically use capillary tubes but EEVs and TXVs are also options) and the evaporator. You'd use CuP 94 to braze all the copper tubes together after cutting and bending them into shape, add the compressor and filter drier, vacuum the system to about 500 microns, do a nitrogen sweep to help remove moisture, vacuum again to about 100 microns vacuum and charge your selected refrigerant, then fine tune the charge in the unit under load. Selecting the temperature and load the unit can take primarily depends upon the compressor, evaporator and metering device, plus the size of the condenser. 

Obv i do not understand any of that and im pretty sure getting the equipment alone is expensive asf, and i do not have any experience in air conditioning

 

I think ill just use an ac unit instead and use that as a waterchiller, and an added benifit of being able to cool more than the cpu like cooling the gpu aswell, though i doubt itll stay below 0 throughout the entire loop so prob just keep nb and ram cooling a separate system with tecs, or ill just have nb at the start of the loop and rams as tec cooled just to make sure they stay subzero

 

I do already have a tec but ill need some metal, i can use my scrap psus for metal but defo not ideal, though thinking about it i can actually get away with using scrap psu metal, maybe once my end of semester exams pass ill try fabbing a completely bullsht tec cooler out of scrap metal, and obv my tool of choice for cutting metal is just some random asf scissors cause apparently this scrap metal can be cut with just scissors, though i will buy some basic asf cutting tools once i buy the stuff i need to fix my ud3p, and for welding the metal together ill just use some f ing solder cause im too cheap xD

 

A 14 year old with limited budget gotta improvise so ill just use some existing tools and scrap metal, guess the limited budget seems to have made me like diying sht to the point of being a fire hazard or a cutting hazard just to save some money xD  well atleast i wont end up wasting money on garbage in the future (looking at you overpriced rgb crap) so guess im lucky to have that mindset

 

Ud3p and p5q pro seem to have a very similar layout but ud3p has a slightly lower ram slot position, but prob nothing to worry about cause im only making the ram cooler portion and not the nb cooler just yet because no point trying to go near 700fsb when the ram just tops out

[Lord Nicoll]

1 hour ago, Somerandomtechyboi said:

Obv i do not understand any of that and im pretty sure getting the equipment alone is expensive asf, and i do not have any experience in air conditioning

 

I think ill just use an ac unit instead and use that as a waterchiller, and an added benifit of being able to cool more than the cpu like cooling the gpu aswell, though i doubt itll stay below 0 throughout the entire loop so prob just keep nb and ram cooling a separate system with tecs, or ill just have nb at the start of the loop and rams as tec cooled just to make sure they stay subzero

 

I do already have a tec but ill need some metal, i can use my scrap psus for metal but defo not ideal, though thinking about it i can actually get away with using scrap psu metal, maybe once my end of semester exams pass ill try fabbing a completely bullsht tec cooler out of scrap metal, and obv my tool of choice for cutting metal is just some random asf scissors cause apparently this scrap metal can be cut with just scissors, though i will buy some basic asf cutting tools once i buy the stuff i need to fix my ud3p, and for welding the metal together ill just use some f ing solder cause im too cheap xD

 

A 14 year old with limited budget gotta improvise so ill just use some existing tools and scrap metal, guess the limited budget seems to have made me like diying sht to the point of being a fire hazard or a cutting hazard just to save some money xD  well atleast i wont end up wasting money on garbage in the future (looking at you overpriced rgb crap) so guess im lucky to have that mindset

 

Ud3p and p5q pro seem to have a very similar layout but ud3p has a slightly lower ram slot position, but prob nothing to worry about cause im only making the ram cooler portion and not the nb cooler just yet because no point trying to go near 700fsb when the ram just tops out

Oh you can use a heat exchanger instead of an SS evap head like shown there, the chillers the builds make also outperform a lot of the AC ones and aquarium chillers but they typically can't handle the same load so you gotta space benchmarks out more. You'd really wanna water a TEC, but when wateredcooled you do get some pretty low temps on it. You shouldn't need to weld or solder anything (other than maybe the TEC leads), typically we use bolts for these things

 

[Somerandomtechyboi]

5 hours ago, Lord Nicoll said:

Oh you can use a heat exchanger instead of an SS evap head like shown there, the chillers the builds make also outperform a lot of the AC ones and aquarium chillers but they typically can't handle the same load so you gotta space benchmarks out more. You'd really wanna water a TEC, but when wateredcooled you do get some pretty low temps on it. You shouldn't need to weld or solder anything (other than maybe the TEC leads), typically we use bolts for these things

 

Wiring sht up wise i just jam wires into sata/molex connectors and call it a day, though i may put some tape on

 

Ill be building a testbench later on so yea ill prob have alot of rad cooling incl for stuff like tecs

[Blue4130]

3 hours ago, Somerandomtechyboi said:

Wiring sht up wise i just jam wires into sata/molex connectors and call it a day, though i may put some tape on

 

Ill be building a testbench later on so yea ill prob have alot of rad cooling incl for stuff like tecs

For the love of god, take pride in your work and do it properly.

[Somerandomtechyboi]

56 minutes ago, Blue4130 said:

For the love of god, take pride in your work and do it properly.

l a z y n e s s

 

Also i am pretty much fine with just jamming wires instead of soldering cause i can remove em easily and if i shove em hard enough then theyll be rather secure anyways, though a screw terminal or something like that would def have more peace of mind, but thats for more permanent solutions

 

I dont give a crap about asthetics so i dont mind the color coded cables, scrap metal which hasnt been polished, etc. Though i def hate it if the damn computer is not tidy cause fk tangled cables