Detailed How-To with lots of pics – Tom
There are three things you should note before reading any further:
I am not an expert in this field, I am just a guy who likes to make stuff.
I nor pclincs nor Overclockers.com will be held responsible for any damage or injury caused by the construction and / or operation of a system similar to those described here. It is your responsibility to protect yourself and others in the same way that it is also your responsibility to stay within the law.
I am not an expert in this field, I am just a guy who likes to make stuff.
I nor pclincs nor Overclockers.com will be held responsible for any damage or injury caused by the construction and / or operation of a system similar to those described here. It is your responsibility to protect yourself and others in the same way that it is also your responsibility to stay within the law.
This article describes the construction and operation of a single stage vapor phase refrigeration system for the cooling of micro processors found in modern home computers.
These are the basic components of a direct-die, vapor-phase refrigeration system. Direct-die refers to the fact that the evaporator is mounted directly onto the CPU/GPU as opposed to a water chiller, where the evaporator cools liquid which in turn cools the CPU/GPU.
In the diagram above we see the main components of the system. The purpose of these components is outlined beneath:
To compress the gas state refrigerant returning to the compressor. If you are working with a set refrigerant, the size of the compressor is the main factor to dictate the temperatures you can achieve
Condenser: To remove a sufficient amount of heat from the high pressure gaseous refrigerant that it will condense into its liquid phase. The condenser must be sufficiently sized to give 2-3ºC of sub-cooling to the refrigerant when the system is under full load.
Filter / Drier: No matter how good our brazing or vacuuming of the system may be, trace amounts of moisture and debris may be present which, if allowed to enter the capillary tube, would cause a blockage. The use of a filter drier prevents such blockages from occurring.
Capillary Tube: The Capillary Tube is used as a simple yet highly effective metering system to control the amount of liquid refrigerant entering the evaporator. By varying the internal diameter and length of the capillary tube, a system can be tuned to deliver optimum temperatures at a given heat load. There is an excellent resource which gives capillary tube length and diameter for given refrigerants and heat loads, it can be found
Evaporator: The evaporator in a direct-die system is clamped onto the CPU or GPU. Inside the evaporator, the mainly liquid refrigerant is injected by the capillary tube and as it hits the low pressure begins to violently boil. A great deal of thermal energy is used as the refrigerant changes phase and it is the purpose of the evaporator to ensure that this energy passes with the least resistance from the CPU to the refrigerant.
Flexible Suction Line: As a direct-die system has the evaporator bolted directly onto the processor with the rest of the phase change system situated beneath or to the side of the computer, the suction line needs to have a good degree of flexibility.
Pressure and Temperature: The boiling point of pretty much everything depends on the pressure it is subjected to. For example, water under normal atmospheric pressure (14.7 PSIa) boils at 100ºC, but if we reduce the pressure to 4.5 PSIa, the water will now boil at 70ºC; as we decrease the pressure further, the water will boil at lower and lower temperatures.
Refrigerants are identical, just like water as we decrease the pressure the boiling temperature decreases. The behavior of this is documented on a P/T chart such as this one (source: http://www.airliquide.com) for the very popular DIY phase change refrigerant, propane -r290.
We can see that the boiling point under atmospheric pressure of propane is -42ºC; charts like this are available for all refrigerants, although there exist more than one on-line calculators which offer more versatility and accuracy.
Measuring Pressures: There are many different imperial and metric ways to record pressure – the ones commonly used are:
- InHg – Inches of mercury
- mmHg – millimeters of mercury
- PSIa – Pounds per square inch (absolute)
- PSIg – Pounds per square inch (gauge reading)
- Microns (Hg) – Microns of mercury
One atmosphere ~ = 14.7 PSIa = 0PSIg = 760,000 microns Hg = 760 mmHg = 30 InHg
PSIa and PSIg are the same scale, but PSIg is the gauge reading which is 14.7 PSI lower than PSIa, which is the real absolute pressure. Thankfully it is usually sufficient to only be aware of PSIa, PSIg and InHg.
All pressures are measured on a standard refrigeration gauge manifold such as this:
The gauge to the left is connected to the low side of the system and measures suction pressure. To the right is the high side gauge which is used to measure the discharge pressure.
Superheat: This is probably the most difficult thing to comprehend but put simply it means how much the refrigerant is heated above its evaporating temperature. The amount of superheat present within a system can be used to identify the state of the refrigerant returning to the compressor.
Imagine we have built a system using propane (r290) as a refrigerant. The low side pressure of the system under load is 8.5 PSIg – measuring the temperature of the suction line 6″ from the compressor we get -28.5ºC. Whilst this system appears to be running well, it is not.
Looking on the PT chart or using one of the excellent tools available HERE.
We can see that a low side pressure of 8.5 PSIg gives a saturated suction temperature of -29.3ºC; we subtract our real temperature 6″ from the compressor from this figure to give us our superheat. In this example, the superheat is 0.8ºC, which is far too low. When this system is running under load, the very low superheat indicates that liquid refrigerant is returning to the compressor; if the system drops to idle, the amount of liquid returning would increase even further.
We want to achieve a superheat of 6-12ºC which will ensure:
- Optimum temperatures and capacity
- That liquid refrigerant does not return to the compressor, this will cause the compressor to compress liquid, causing eventual failure of the compressor
- That the gas returning to the compressor is not too hot, if this should be the case the compressor could overheat. Ideally a compressor should have a top case temperature of 50-70C
I consider sub-cooling to be the opposite of superheat. When the refrigerant passes through the condenser, its temperature is dropped until a point at which the refrigerant condenses into liquid. The degree of sub-cooling is the temperature difference between the temperature pressure condensing point and the true temperature at the outlet of the condenser.
This chart is very useful as it features a good amount of refrigerants; credit Russell_hq – xtrememsystemsit for this.
Probably the most important and expensive item used is a vacuum pump. The main purpose of a vacuum pump is to decrease the pressure inside the system to such an extent that water will boil at room temperature. This water vapor can then be flushed out of the system, leaving a totally dry environment for our refrigerant. This is the vacuum pump I use:
It is an Edwards Speedivac designed for low pressure laboratory work – I found it in a skip. It is a peculiar pump and works exceptionally well – however most people use conventional refrigeration vacuum pumps such as this:
These can be bought for ~£150 on eBay. For a cheaper alternative, many people use a home built vacuum pump made from one, or preferably two, small refrigeration compressors. This picture belongs to Russell_hq from Phase-change.com and shows how he made his cheap vacuum pump:
In the above picture the discharge port of the first compressor is brazed to the suction port of the second. A Schrader valve is attached to the suction of the first compressor and this is your vacuum connection.
If you do take this approach, do not use compressors which use Polyol Ester (POE) oil as the oil will turn acidic when exposed to air and very rapidly destroy the compressor. If you see a compressor used in an r134a or r404a system, it will contain POE oil. Compressors designed for r12, r22 and r600 are ideal. Thankfully the majority of older “scrap” compressors are of this second type and subsequently your ghetto “vacuum pump” will last much longer.
Whilst crude, these home made vacuum pumps have been proved to boil water at room temperature. The professional units are of course better, but for a few hours scavenging and a little brazing, you can save yourself £150.
Whilst I am by no means an expert at brazing, I am capable of producing strong and reliable joins which, for the most part, is what you will want to achieve in a phase change system. It is often easy to confuse brazing and soldering but put simply, brazing uses filler rods with a high melting point and high tensile strength whereas (soft) soldering uses filler rods with a low melting point and low tensile strength. As we will be putting up to 250 PSI into our system, we require brazed joints for their higher strength. Note that the process of silver soldering is in fact brazing.
To braze you will need a minimum of three pieces of equipment:
- A good, powerful torch
- Some high temperature flux
- Some brazing rods
For a torch, I use a £17 “Clarke” kit which I bought new from “Machine Mart”; included are 4 nozzles, one hose and one regulator. This is connected to a large 20 Kg propane tank. For the whole set up including a propane tank refill, I paid ~£40, which I feel is a very good price, especially as the tank is about ¾ full after almost a years use.
For flux I use a high temperature white powder variety which I mix into a paste by adding water. It is a standard “silver solder” flux I bought for ~£7 from a local tool supplier in Stoke On Trent.
For brazing rods I recommend Silfos CP1 rods – these contain 15% silver and have been great for me, requiring only a small amount of flux and always flowing smoothly around the join. A 1 Kg pack of these cost me £10 on eBay. Whilst these rods are great for copper to copper and copper to brass joins, for joining to stainless steel you will require 40% or higher silver content. Whilst I have some of these high content rods, they have yet to be used, so if you plan on joining copper to stainless steel you may have to do a little more reading.
The high temperatures involved in brazing will lead the production of an oxide around the heated part – this can be a problem as it will contaminate the finished system. The degree of oxidization you will experience depends on the temperature the piece is heated to and the time it spends at this temperature. To avoid oxidization, many people choose to use a process of purging, which involves surrounding the heated part in a blanket of a gas to expel the oxygen in the atmosphere. I have brazed without purging and completed systems which have worked well, so it is not essential but well advised if you have the equipment to do so.
Common gasses used to purge are nitrogen, carbon dioxide or argon – these are by far the safest and will work very well. There are also many people who choose to purge with flammable gasses such as propane or butane. Any one who chooses to purge with these flammable gasses should exercise extreme caution and be fully aware of the risks, yet despite these risks there may be a chemical advantage to purging with flammable gasses. From my own experience with carbon dioxide, argon and butane, I can say that all have produced good results but I would not like to promote that people purge with a flammable gas due to the dangers involved – it is your choice.
This is how I purge using Argon from a cheap MIG welding bottle. A 1m length of capillary tube is used to transfer the Argon into the heated part, it’s basic but seems to work well.
There are a few tools which are of great use in building a system – these include a basic pipe bender (pictured below) which was bought from Machine Mart for roughly £10; it can bend three sizes of pipe from ¼” to 3/8″. This can create bends of up to 90 degrees.
When we need to cut copper pipe we need to produce clean edges and generate no swarf, or dust, in refrigeration piping – a hacksaw is definitely not suitable. This is a simple pipe cutter which was purchased from Machine Mart for roughly £10 – it can cut all of the different sizes of pipe we will need.
Here is a flaring tool set – it is a standard automotive set which can produce double flares. Whilst we won’t be needing to produce any flares in this build, the flare tool can also be used when we need to slightly increase the diameter of one piece of pipe to fit inside another so we can braze well. Again this was about £10 from Machine Mart.
To build a system we will need copper piping. The accepted pipe sizes are ¼” for discharge and 3/8″ for suction in most of the small systems we work with. The copper piping is purchased from a local refrigeration supplier, the ¼” pipe comes in 15 meter coiled lengths and the 3/8″ in 5 meter straight lengths. I believe for one length of each the price was £20; this is enough to build several systems and this is what you get for your money:
Next we will need our other components – these include Schrader valves on ¼” stems, filter driers with ¼” and capillary ends, 3/8″ male to male flare fittings and ¼” equal tees.
The last major piece of material we will need is our metering device – in a simple system this will be capillary tube. I choose to use 0.026″ ID tubing; if you are planning to build a system for high loads you may find it better to use 0.031″ ID tubing. In the UK we are limited to these two, in the US and other parts of the world 0.028″ ID tubing is available, but sadly in the UK it seems very hard to find 0.028″. Capillary tube comes in 30 meter coils for roughly £15 from a local refrigeration supplier – this amount will be sufficient to build several systems.
For people who wish to source all components themselves, this may be the most difficult part to find. When I got into phase change in summer 2004, I bought a large quantity of Parker refrigeration hose and fittings. Today the majority of people choose to use corrugated stainless steel hose. The benefit of the steel hose is that the joints can be brazed and are therefore less likely to leak around the joint; however with repeated flexing from multiple installations, the steel can sheer and leak.
With the Parker refrigeration hose, repeated flexing will not have any effect on the line, but the hose uses flare fittings and if not done well these can leak; also with the fittings, the line can be a little cumbersome if space is limited, especially around the evaporator. For a single stage cooler, I don’t see any advantage of one hose of the other except that stainless hose is generally cheaper – today I use the Parker hose because I have it.
This is how the Parker hose is bought and beneath is an assembled length with fittings. In the first picture, there is also a braided flexible discharge line above the Parker hose – ignore this.
This shows a line assembly I have made up. The fittings are 3/8″ BSP female flares, although a wide range of fittings are available:
This shows two pictures of ready made, stainless steel, flexible suction lines. This first line includes an outer braid which allows for higher working and burst pressures. The picture of this hose is taken from runmc’s online store at www.under-the-ice.com:
This picture of an assembled unbraided line is taken from Johann’s online store at www.ukphasechange.co.uk:
If you wish, you can make your own stainless assembly by buying the corrugated line from somewhere such as www.bes.ltd.co.uk and using 40% silver rods to braze to the copper line. For those attempting a first time build, a pre made line may save you a lot of hassle.
The evaporator is such a major component of a direct-die system that I felt it warranted its own section in this guide. As mentioned before, the purpose of the evaporator is to allow the transfer of thermal energy from the hot CPU into the cold refrigerant vapor. As I see it, these are the main performance factors which you should hope to achieve in your evaporator:
- Low resistance conduction path(s) from your heat source to the areas in contact with refrigerant vapor. The amount of power which can be transferred per unit temperature difference in a conductor depends on the thermal resistance of the material, the width of the conducting channel and the length of the conducting channel. Whilst there are complex mathematical packages available to simulate this I feel it is best just to think about it and through trial and error evolve your designs to perform more effectively.
- Large surface area which is in contact with the refrigerant vapor. This will allow more power to be transferred for a given temperature difference between the evaporator and the refrigerant.
- Turbulence in your design. Whilst refrigerant will normally boil very violently, I believe that the vapor in the centre of the flow path may remain reasonably unaffected, especially if your flow channel is large. Creating harsh angles or rough surfaces in you design will cause disruption to the normal flow pattern and mix the inner and outer refrigerant allowing more of the inner layers contact to the evaporator surface.
- A small difference between the pressure of the refrigerant entering the evaporator from the metering device (typically capillary tube) and the suction line. Higher pressures in the evaporator with respect to the suction line will lead to higher evaporator temperatures and the possibility of more liquid refrigerant passing to the suction line.
- Partial evaporation before the refrigerant reaches the point where the conduction path and surface area combine to allow maximum energy transfer. Typically this is at a point closest to the heat source. The maximum capacity of the refrigerant is reached when ~30% of the liquid has turned to vapor; whilst this may have been utilized in commercial refrigeration for many years, to the best of my knowledge the first person to apply this to CPU evaporator design was Chilly1 of www.xtremesystems.org.
Nearly every change you will make to an evaporator design will trade off one (or more) of these factors against another; for example, higher surface area will almost always mean a more resistive conduction path and greater turbulence will cause increased pressure difference. You should aim to produce a good balance and target your design to your application.
Occasionally someone will mention evaporator “mass”. In my opinion this is completely irrelevant to performance and merely a by-product of the other factors and usually comes from low resistance conduction path(s).
These are some versions of my own evaporators, they are by no means the best but with the exception of the first they all perform well. Of those of these I have run on completed systems the performance of the “V2″ is the highest.
This shows the latest revision of Chilly1′s spiral evaporators which, from all accounts, performs exceptionally well. Due to the spiral, it can be mounted in many positions without loosing performance.
This shows an evaporator designed by Bowman1964 many years ago, the design and performance was great, however it must have been quite difficult to mount.
This is one of the evaporators made by Baker18, the design of which was later used by Asetek for the Vapochill LS. This was an incredible design.
Evaporators designed by Kayl of www.xtremesystems.org, simple in design but with great performance.
But why not use the tools and materials you have available to design and build your own, its much more fun!
There are two types of compressor commonly used in computer refrigeration – these are rotary compressors and piston type compressors, with the latter the preferred option for small systems where noise is important.
A compressor with an indicated horsepower (IHP) of 1/8 – ½ is ideal for our typical applications – this roughly equates to a displacement of 4 cc to 18 cc with typical refrigerants. Obviously the larger the compressor, the lower temperature you can achieve at a given heat load or the larger your capacity at a given temperature. One point to mention at this point is that below -50ºC, insulating some configurations well can be very difficult, so if you are building a high usage system choose your compressor wisely.
This is a picture of a Samsung Rotary compressor taken from their website at www.samsung.com:
Here is a picture of a selection of Danfoss piston type compressors taken from their website at www.danfoss.com:
Another important factor to consider when choosing your compressor is the refrigerant you wish to use and the refrigerant your compressor is currently using, or is designed to use, should you buy a new compressor. This is very important as it is often the best way of determining what oil your compressor will be currently using.
Different refrigerants can require different oils – in small single stage refrigeration systems two types are predominantly used – Polyol ester which is a synthetic oil and mineral oil.
I’ve made the list beneath which shows oil compatibility for a few common refrigerants – to the best of my knowledge it is correct:
- R134a Polyol Ester
- R413 Mineral oil or Polyol Ester
- R404a Polyol Ester
- R507 Polyol Ester
- R410 Polyol Ester
- R22 Mineral oil or Polyol Ester
- R402 Mineral Oil
- R290 Any Oil
- R600(a) Any Oil
Why does it matter? In order for a compressor to operate reliably, oil will need to pass over the compressor crank case (as it returns via the suction line) to keep the moving compressor parts well lubricated. If the oil is not carried around the system in the refrigerant, the moving compressor parts will not receive lubrication, leading to premature failure of the compressor. A compressor burn-out is probably the most destructive event which could happen to a refrigeration system, as it will contaminate every part of your piping, condenser and evaporator.
The oil compatibility refers to how well the refrigerant will carry oil around the system. As r290 (propane) carries all oils well, it is possible to operate reliably a refrigeration system with the “wrong” oil by adding a little r290 into your systems refrigerant. If you need to change the oil in a compressor to accommodate your new refrigerant, it is not difficult. This is a 1 litre can of standard grade Polyol Ester (POE) oil which cost just over £12 from my local refrigeration supplier.
An important thing to remember is that POE oil is highly hydroscopic – this means that it will actively absorb water from the surrounding air and environment. Because of this we must not leave the oil in contact with the air for long; if we do, we should change the oil in the compressor as when the oil absorbs moisture it begins to turn acidic. The moisture and acidity of the oil may lead to erratic or poor system operation and premature failure of the compressor.
Mineral oil does not suffer this problem anywhere near as severely, but you should always change out the oil if you believe contamination has occurred. This is my procedure for changing old oil to new oil of the same type:
- Whenever I have worked on a system which has caused it to be open to the air, I complete all brazing and seal the unit
- Then by slackening off a suction line flare fitting and tipping the unit upside down, I drain the old oil out via the service port of the compressor
- I then reseal the suction line flare fitting and vacuum the “without oil” unit for about two hours before bringing the unit back to slight positive pressure with a tiny squirt of r134a. I use this because it is cheap and as refrigerant it contains no moisture
- Now I have made a nice dry environment for the new oil; I use a large cattle syringe to inject my fresh POE oil. For my 7cc Danfoss compressors, this is 450ml
- Now the system is ready I begin to triple vacuum my unit, we will talk about this later
Point About r600(a) Compressors: The compressors designed for and running r600(a) are very common in the UK for domestic fridges and freezers. Because of their high volume and low cost production, I believe they are generally not as well made as the more expensive r134a compressors.
Whilst it is almost always safe to use r134a compressors with higher pressure refrigerants such as r404a, I believe that the highest pressure refrigerant which should be used with a r600(a) compressors is r290. This is convenient as r290 will work happily with the existing (probably mineral) oil.
Often the hardest part is finding people who will help and supply your equipment. I would view all of these people and companies as trust worthy.
Specialized computer refrigeration that are (or might as well be) in the UK: UKphasechange.co.uk.
This is Johann’s website and online shop. He lives in London and is very prominent on other forums and produces good work. His shop sells hard-to-get-hold-of items, such as small condensers and ready made suction lines, in addition to virtually everything else you’ll need to build a single stage system. I’ve noticed that he’s recently signed up here although he has yet to post.
This is runmc’s online shop. He is very prominent at his own forums as well as other main forums. Whilst he is not located in the UK, his online shop ships and prices to the UK in the good old Great British pound.
I am going to build a small GPU cooler to attach to my GeForce3.
We are going to use a compressor identical to this one:
It is a Danfoss FR7.5G and is designed for r134a. Because of this we know it use POE oil. Before I started I brazed a Schrader valve onto the service port of the compressor – this is the port on the left hand side of the picture.
The compressor has a displacement of 7cc, and from looking up the datasheet should be charged with 450ml of oil. To gain more performance from our system, we will be using the higher pressure refrigerant r404a. Being a well made Danfoss compressor, it will handle this higher pressure refrigerant without issue. For a condenser we will be using a condenser identical to this – it is made by Lu-ve and came out of a beer cooler:
First we make our evaporator. It was machined from copper bar on a lathe and follows a simple reverse stepper design. These are all of the parts used for the evaporator:
I machined the male-to-male flare fitting to be a good fit on the end of the copper ‘L’ piece. I don’t have any pictures of machining this evaporator but you can imagine how it was made. Thanks to Dave (dualist on www.pclincs.co.uk) for all of his machining advice.
I assemble all of the parts and wrap the capillary tube as close to the evaporator as possible to reduce flashing at the end of the capillary tube. By wrapping our capillary tube around the evaporator exit, we achieve great sub-cooling, which ensures that the refrigerant remains liquid until the very moment it enters the evaporator.
At this point I make a good base on which to build to system. I used a thick piece of MDF and attached some castor wheels off an old computer case, thus the skateboard cooler was born!
Now I mounted the condenser and compressor onto the base using M10 bolts for the compressor and round headed wood screws for the condenser:
Next I used the pipe bender, flaring tool and tube cutter to make the pipe work for the system. Take time when doing pipe work, especially marking it all out – exact fits are hard to achieve if you work quickly.
To use the pipe cutter, place where you wish to cut and slowly rotate around the tubing, slightly tightening after each rotation. After a few turns you will have cut through the pipe:
Here is my suction line coming together – I need to do a little more pipe work and she is ready to braze:
I used the flare tool to slightly expand some ¼” tube to fit snugly inside the larger 3/8″ suction line tubing. I then carefully bend the ¼” tube to for a ‘U’ using a piece of plastic bathroom pipe as a former. Here is the completed fixed section of the suction line:
I repeat the pipe cutting and bending over the next few hours to complete all of my piping. The discharge line has a loop in it to dampen the vibrations coming from the compressor to the condenser:
Now I braze the parts together. Sadly there are no pictures of the brazing as it is tricky to use a dodgy digital camera when you have a flame at over a thousand degrees in one hand and a brazing rod in the other. I use a minimum amount of flux and I have a tendency to over braze. Whilst it often looks a bit ugly, I rarely have leaks:
When you braze a part near to a Schrader valve, remember to remove the valve core, or you will have a mess. With all this coming together well, I need to make a fan shroud to attach my noisy 120 mm fan to the condenser. Again I use MDF and, to create a good seal around the edges, I use some 3mm self adhesive Armaflex tape.
With this done, I attach the suction line and evaporator to the system after I have brazed them also. In this build I used argon as a purge gas as I had been using it in the MIG welder before hand.
In the picture above the last piece to fit is the filter drier into which the capillary tube will be brazed.
After I braze, I remove any excess flux and oxide on the outside surface of the piece. In this case I also polished the base of the evaporator. Whilst shiny outer surfaces look nice, as long as the flux is removed there is no harm in it looking a bit grubby.
The next picture shows the system after this has been done:
Now this is done I am ready to begin the fun part.
First I slacken off the flare nut which joins the fixed and flexible suction lines, then I turn the whole system upside down (it is damn heavy) and drain out the old POE oil through the service port by removing the valve core of the Schrader valve. After the old oil is out, I tighten up the flare fitting and with the valve core still out. I put the unit into vacuum to remove any traces of moisture before I put my new oil in. I followed the steps I have listed before hand.
After roughly two hours of deep vacuum, I bring the unit up to just over atmospheric pressure with a squirt of r134a and at this point I disconnect the hoses and use the syringe to put 450ml of new POE oil in the compressor via the service port.
This shows the system in its initial deep vacuum:
After this is complete, I put the Schrader valve back in the service port line and prepare to triple evacuate the system. Before I went any further I changed my gauge manifold set from the set show in the previous picture to my other set which is calibrated with vapor pressure temperatures for the refrigerant I will be using – r404a.
The reason I used the other CFC calibrated set before is that they have better graduation on the low side manifold gauge which allows me to have a better idea what pressure the system is at when in vacuum, which is important if you need to remove a bit of moisture.
This is an updated section on charging – I’ve added better pictures and more information. These do break from the chronological order of how the unit was built but I hope it will provide more information.
First I connect up the system to the manifold, tank of refrigerant and vacuum pump. This shows the low side of the manifold connected to the service port of the compressor and high side connected to the Schrader valve just before the filter drier.
This is a view of the equipment used and shows how they are connected:
First I put the system into deep vacuum. I always vacuum from the low side; to do this I close the high side valve on the manifold and just open the low side valve. Obviously at this point the refrigerant tank valve is closed. The vacuum pump draws the system into very deep vacuum for about an hour – if the system had not been charged before, I would have extended this to about three hours. The arrows show the direction and location the vacuum pump is pulling from:
After this is complete, I give a tiny (very tiny) shot of refrigerant into the high side. To do this I close the low side valve on the manifold, close the ball valve on the line to my vacuum pump, open the valve on the refrigerant tank a very slight amount for a very small amount of time.
Then by opening the high side valve on the manifold, I let this miniscule amount of refrigerant into the high side.
Now I close the high side valve on the manifold, open the ball valve on the vacuum pump hose and open the low side valve on the manifold.
This will draw the refrigerant from the high side to the low side via the capillary tube and flush out any moisture vapor trapped in the system. If this was the first time the unit had been charged, I would have repeated this process two or three times, but as this is merely a recharge of a perfectly operating system, this is not needed.
I let the vacuum pump pull the system down for about another 30 minutes before I begin to charge the system.
This is the charging setup:
I have mounted a thermocouple on the surface of the evaporator and another at the end of the solid suction line 6″ from the compressor port. The K-Type thermocouple meters are actually food probes I “acquired” from my old place of work. They are rated down to -50ºC with a quoted full range error of +/- 0.5ºC, which is very good.
To charge the system I close off the low side valve on the manifold, close the ball valve to the vacuum pump and turn off the vacuum pump (as I won’t be needing it again). Now I open the valve on the refrigerant tank and gradually let refrigerant into the high side until it reaches about 20 PSIg. Then I gradually let refrigerant into the low side until the high and low pressures are equal. You can charge just through the high side and let the pressure equal through the capillary tube but this takes more time.
With this 20 PSIg static charge I start the systems compressor up. After about 20 minutes the temperature of the evaporator falls to -52ºC.
Whilst this is an impressive figure it is utterly useless to hold a load as would be indicated by the huge superheat at the suction line. For this example I have chosen to charge the system using superheat. As mentioned before a superheat of 6-12ºC will guarantee that no liquid is returning to the compressor and that the system capacity is at a safe maximum.
The worst case for liquid returning to the compressor will be when the system is under no load, so to stay safe I chose to tune the system for 6-12ºC superheat under no load. With the system running and under no load, I gradually open just the low side manifold valve to allow more refrigerant into the system. When I do this I do it incredibly slowly so that liquid does not enter the compressor.
Over the next several minutes I slowly see the temperature of the low side decrease and the superheat become smaller. I constantly look at the vapor pressure temperature on the low side manifold gauge and compare this to the temperature of the suction line. When the difference between the vapor pressure temperature and the suction line temperature is 6 to 12ºC, we are done.
These are my finished no load pressures. The vapor pressure temperature (outer ring on low side) is roughly -54ºC:
This shows the evaporator and suction line temperatures:
If we take the difference between the -47.8ºC and -54ºC, we get a superheat of 6.2ºC which is satisfactory. The no load evaporator temperature is -50.8ºC (I apologize for the quality of the evaporator temperature display, it seemed ok when I took it.)
Next I will see how well the system will handle load. I am using my load simulator to give an exact 50W load – I have chosen 50W as most of my graphics cards are old and have low power dissipation – I like to raise the dead!
In a previous example I had used a 150W load from the simulator, which I later found to be stupid, especially on cards such as GF3′s and 9600se’s which dissipate in the region of 30W. I would say that a 150W load for a CPU unit and a 100W load for a modern high end GPU unit would be realistic.
This is the load simulator:
I have used just one of the 0.5 Ohm resistors through which I pull 10A via a high current 5v DC power supply, giving me the 50W I was after. After one hour of running the 50W load simulator, the temperatures were quite good:
The superheat has increased substantially, but this was to be expected. The evaporator temperature has hardly changed at all, which I am very happy with.
OK – 50W is a very small heat load, but it should be fine for anything up to the X800 series of graphics card.
Now I have finished charging and I must remove the hoses from the system – if not done carefully this can cause the loss of refrigerant which, on a small system such as these, can ruin our hard work in charging.
What I do is turn off the system and wait for the low side pressure to rise to about 10 PSIg. With this slight positive pressure my system will not draw in air when I am removing the hose and only loose a tiny amount of charge due to the small 10 PSI pressure difference.
I quickly remove the low side hose.
Now just the high side is connected; over a few hours this and the low side will equalize at the final static pressure, so I go to do something more useful for a few hours. When I come back the high and low side had equalized at about 45 PSIg – this is as low as it will get, so I quickly remove the hose – and we’re finished!
Here’s a nice picture to end with and gives the reason as to why insulating a phase change cooled graphics card can be difficult:
The aim of this cooler was to gain me the GeForce3 world record in 3DMark2001, which in March 2006 I achieved. This is the GPU cooler in action.
With thanks to my friends at www.pclincs.co.uk for all their encouragement and help. I would also like to thank the many people at www.xtremesystems.org, www.xtremeresources.com and www.phase-change.com for all of their help over the years.
This guide will be updated and hopefully improved in the future as there are many more aspects which should be covered – but I hope this provides a good basis for your projects. Although as I’m very busy at the moment I would prefer if you could voice any questions or comments in this forum thread.
Please respect that I do not build or modify vapor phase change units for other people as I barely have enough time to complete my own projects. If you are in the UK there are numerous people who will do this for you – you can find them on the forums listed previously.
I would be happy to answer any questions or comments – Best regards,
Tom (SoddemFX) – Staffordshire, England