Excellent How-To with parts list – RoboTech (aka Lee Garbutt)
I like pumps: big pumps (reactor coolant pumps), little pumps (HPLC injection pumps), high-pressure hydraulic pumps, and low-pressure PC watercooling pumps. But to be useful pumps must be controlled. The purpose of this article is to discuss several different PC watercooling pump control strategies, ranging from very simple, low-cost solutions to more complex, failsafe circuits capable of initiating an emergency shutdown of the computer.
The typical pump in a PC watercooling system is driven by a small, fixed speed electric motor, or by “magnetic drive” – a detached rotor driven thru closely coupled permanent magnets. Most pumps are powered directly from the AC mains while a few are powered from the 12 VDC section of the PC’s own power supply. The following picture illustrates some of the AC powered pumps I use in PC watercooling systems and test rigs.
And here are two examples of 12 VDC pumps, which have become available in the past year.
In either case, controlling the pump generally means two things: turn it on and turn it off. The control circuits used to accomplish this may be as simple as an on-off switch (or no switch at all) or they may be more elaborate, incorporating various relays and sensors. For the most part, we will focus on controlling AC powered pumps but many of the same techniques can be applied to controlling 12 VDC powered pumps.
Here is a brief outline of the topics covered in this guide:
- Routing AC power into the case
- No control – the pump runs 24/7
- On-off control using an outlet strip
- Automatic over-temperature shutdown (software)
- Relays: mechanical, solid state, and PCI card
- Smart circuits – introduction
- Smart circuits – initiating an emergency shutdown
- Smart circuits – adding sensors
- Smart circuits – real world applications
- Sources for parts
- Closing Comments
One of the first problems many would-be water-cooled PC enthusiasts face is how to route the pump’s AC power cord thru the wall of the case. There are a number of options available.
1. Drill a hole in a spare PCI slot bracket. This generally requires cutting the cord, passing the wire thru the hole in the bracket and then reattaching the plug. A round rubber grommet is used to trim the hole and protect the wire passing thru it. In some cases slotting the hole and cutting the grommet can make installation even easier.
One good way to splice the wires back together inside the case is to use insulated spade lug connectors. This also permits easily disconnecting and reconnecting the pump cord again in the future.
2. Mount a line cord receptacle on the back panel of the case. This method requires a little more cutting, drilling and filing but can result in a clean, professional looking installation.
Be sure to insulate the 120 VAC terminals on the inside of the receptacle with short sections of shrink tube.
Another option on this theme is to purchase a Swiftech Pump Relay Kit Version 2. In addition to the relay the kit also includes a power cord and panel mount receptacle with hardware. (Swiftech no longer lists these kits so they are becoming harder to find.)
3. Some people like to keep the outside of the case uncluttered and tap into the AC mains inside the computer’s power supply. Only attempt this option if you are qualified to work on power supply circuits – they may contain dangerous voltages, even when turned off. To use this method first discharge the power supply completely (one method is to unplug the power supply and then use a power supply tester to discharge it). Next remove the power supply from the PC and then remove the cover.
Drill a hole in a convenient location in the power supply case to route the pump wire thru using a rubber grommet. Solder the pump wires to the three terminals on the inside of the power inlet receptacle, observing the proper orientation for the hot (black or brown), neutral (white or blue) and ground leads (green or green/yellow). It’s a good idea to install insulated spade lug connectors onto the pump wire so the pump and/or power supply can be removed later on.
4. And finally, a very convenient option is to use a commercial relay card that incorporates a power cord connector built right into the PCI bracket. This solution does not require any cutting or drilling. Here is a picture of the CritiCool PowerPlant PCI relay card installed in a PC. More on these last two items later when we talk about using relays.
This strategy is by far the simplest and actually one of the more popular. Just plug-in the pump and let it run 24/7 regardless of whether or not the computer is turned on. Most AC pond pumps are designed for continuous operation and consume relatively little electric power. This approach also minimizes the added wear-and-tear placed on pump components from frequent startups.
One potential problem is that if the pump gets accidentally unplugged or turned off, the operator may not think to check it before turning on the computer. Wouldn’t it be nice to have a failsafe circuit that would prevent the PC from starting unless the pump was running? More on that subject later… 🙂
This is another simple and popular approach adopted by many PC watercooling enthusiasts. Most people already have a surge suppressor outlet strip with an on-off switch built-in to power their computer and related peripherals.
- Plug the pump, computer, monitor, printer, etc. into the outlet strip.
- When ready to power up the system, turn on the outlet strip first (some people leave all their peripheral devices turned on so they automatically power up everything except the computer when the outlet strip is turned on).
- Turn on the computer. (Using this sequence insures the watercooling pump is turned on before the computer is powered up.)
- To shut down the system, turn off the computer and then turn off the outlet strip.
OK, so now we know how to manually turn the pump on and off – so what happens if the pump fails? First of all, if you have a well-designed watercooling system that incorporates a properly sized, quality pump, it is unlikely you will ever experience a pump failure. But a watercooling system is more complicated than a simple heatsink and fan, so there is more to potentially go wrong. Murphy is watching…
One method of protecting the CPU from overheating and possible self-destruction is to utilize a software utility to monitor CPU temperatures and automatically shutdown the computer when the CPU temperature reaches a user-defined set point. Many new motherboards incorporate CPU temperature monitoring circuits and some have shutdown routines built-in to the BIOS.
If your particular board has CPU temperature monitoring abilities but does not include a shutdown program then you can use the popular MotherBoard Monitor 5 utility along with another little program like ShutdownNow.
The advantage of using a BIOS (or third party) program to shutdown the PC when the CPU overheats is that it is relatively simple and may cost nothing. The disadvantage is that it is software based – if the system becomes unstable due to increasing temperatures (or other problems) it might night be able to initiate a shutdown and protect itself. This is why some people prefer to have added security with some type of a hardware based fail-safe shutdown circuit.
Relays are one of the most popular devices used for controlling the flow of electricity. A relay uses a small signal voltage (i.e. 12 VDC) to “switch” on or off, a typically higher voltage and/or current (i.e. 120 VAC). Relays are available in all sorts of sizes and styles.
Some relays use a coil of wire to create a magnetic field, which closes a set of mechanical contacts when energized. For our purposes, a 12 VDC signal from the computer power supply is used to energize the relay coil. One or more sets of contacts are generally provided with two operational states: Normally Open (NO) and Normally Closed (NC). One set of NO contacts is all that is needed to switch on and off the hot lead going to an AC pump.
A solid-state relay is another popular type of relay to use. As the name implies, a solid-state relay is based on solid state, semiconductor components and has no moving mechanical parts.
In practice, both types of relays are used in the same way. Either 12 VDC or 5 VDC from a spare 4-pin Molex drive connector activates the relay when the computer is turned on. The coil voltage (mechanical) or input range (solid state) should be specified on the relay housing. Since we are working with DC voltages, it is also important to note the proper polarity (+/-) when using solid-state relays. Here is a basic circuit for using a relay to switch on and off an AC watercooling pump automatically when the computer is turned on and off.
Note that 12 VDC is applied to the coil and a set of normally open contacts is used to switch on and off the pump. Only the “hot” wire (black – US, brown – European) in the three-conductor AC pump cord needs to be cut and wired to the relay. The ground wire and neutral wire can be left connected at all times.
Here is a sketch that illustrates the wiring of a typical mechanical relay as shown above:
Next we have a picture of a solid-state relay wired into an AC pump cord. The DC control power is applied to the small terminals on the left (note + and -) with the AC output contacts on the right.
When 12 VDC is applied to either the mechanical or solid-state relay, the normally open contacts close, allowing current to flow to the pump. When the 12v power is removed, the relay contacts revert back to their normally open state and current to the pump stops.
When using a mechanical relay, it’s always a good idea to include a diode across the DC coil input leads. When energized, the coil creates a magnetic field. When the power is turned off, this field collapses and in the process produces a high voltage spike with reverse polarity. Most electronic circuits don’t like these spikes, including the switching mode power supplies used in computers. The diode acts like a short-circuit and safely dissipates the high voltage spikes. The specific diode used isn’t critical (a 1N004 is overkill but very common). It’s like a $0.10 insurance policy.
Relays are readily available at most electronic supply stores (Radio Shack, Jameco, Digikey, All Electronics, etc). For example, here are the parts needed for a basic mechanical relay:
Radio Shack component parts:
- 10-Amp mechanical relay #275-218 $7.99
- Plug-in relay socket #275-220 $2.49
- Rectifier diodes (25 pkg) #276-1653 $2.49
I frequently use a medium size solid-state relay for controlling both the water-coolant pump and Peltier cooler power supplies. When doing so, just make sure the relay contacts are rated to handle the extra load of the TEC (Thermo-Electric Cooler) power supply. Here is a picture of a solid-state relay mounted underneath a dedicated TEC power supply in a homemade 5.25″ drive bay bracket.
The watercooling pump and TEC power supply are wired together in parallel. The power supply’s terminal strip is used to connect the wires. Here is a diagram that shows more clearly how the wires are connected.
If you don’t feel like purchasing parts and building your own relay circuit, there are commercially available relay kits you might want to consider for your pump switching needs.
This second generation kit includes both a mounted relay and power cord receptacle. The relay contacts are rated at 16 Amps, 250 VAC, with a maximum inrush current of 50 Amps. This makes this unit suitable for switching both the watercooling pump and a Peltier cooler PSU.
Another option is the CritiCool PowerPlant card. It installs in a spare PCI slot and incorporates a relay, power cord connector and manual override switch. The switch is a very nice feature as it allows you to run the watercooling system pump independent of the PC (great for setup and testing). When purchasing a CritiCool PCI relay card, be sure to buy the Rev. 2 model, which fixes some grounding and insulation problems found on the original product.
The switched 115 VAC input on the PowerPlant PCI card is limited to 2 Amps due to an onboard fuse (small, reddish brown object in the previous photo). This should provide plenty of current for most any PC watercooling pump but it is not enough to power a pump and a dedicated Peltier cooler power supply. Some people have successfully jumpered across the fuse or replaced the fuse with a higher value, external one to run higher loads. Doing this will obviously void your warranty and you run the risk of damaging the board.
So far, we have covered the basics of switching a PC water-coolant pump on and off. We also discussed software options for a high temperature emergency shutdown. But what if you want more? As I mentioned earlier, some people are not comfortable relying solely on a software safe guard – they want a more robust hardware solution. What if, the…
- Pump fails (loss of flow)
- Leak develops (loss of coolant)
- Reduction in flow (bio buildup, corrosion or collapsed tubing)
- Radiator quits cooling (fan failure or clogged with dust)
These are just a few of the scenarios that could compromise a water-cooled PC and lead to a system failure and potential component damage. Wouldn’t it be nice if the system were able to monitor itself and act accordingly?
To do this requires sensors.
Sensors are available to monitor any system parameter you can think of: temperature, pressure, flow rate (liquid and air), coolant volume, and leaks, even the pH and conductivity of the water. Unfortunately, sensors are typically expensive and frequently require additional signal conditioning circuits to interface with the PC. Bottom line – complexity and cost will eliminate many sensor based failsafe circuits. Most people are not willing to spend hundreds (or thousands) of dollars on a water-cooled PC data acquisition and control system!
Luckily there are a few relatively inexpensive and simple ways to incorporate one or more sensors into a watercooling system that will allow the computer to perform an emergency shutdown should things go wrong.
Have you ever walked into your room and been greeted by a hot burning smell or water dripping out of your water-cooled PC? In this type of situation most people will quickly pull the plug to shutdown the PC as fast as they can in hopes of averting any further damage.
Obviously this is not the most graceful, nor preferred, method to shutdown a computer under normal conditions. But sometimes the slight risk of corrupting data is justified. Before we go on and discuss various means of detecting problems lets look at two methods for initiating an emergency PC shutdown.
1. ATX Power Supply PS-On Signal: Most newer PCs incorporate power supplies that conform to the ATX specification. As many of you know, pressing the computer’s power-on button sends a signal via the motherboard to the power supply requesting it to turn on. This is accomplished by pulling the power supply’s PS-On signal wire low. As long as this line is held low the power supply will remain on. During shutdown, the last thing the motherboard does is release this line and let it float high, which will turn off the power supply – and the computer.
One way to initiate an emergency PC shutdown is to interrupt the PS-On signal going to the power supply. Doing so will have the same affect as pulling the plug! The PS-On line is part of the power supply’s ATX connector wire bundle. It is normally a green wire and terminates at pin #14 on the motherboard ATX connector.
Cutting the green PS-On wire and connecting the two loose ends to a set of relay or sensor contacts will provide a means of initiating an emergency shutdown. Whenever the contacts are closed (green wire connected) the system is enabled to operate normally. If a problem is detected (we’ll talk about various sensor options in the next section) the contacts open and the power supply (and computer) shutdown.
2. Power-On switch pressed > 5 seconds: Most newer motherboards allow you to force the PC to shutdown by pressing and holding the Power-On button for more than 5 seconds. This feature can also be exploited to force an emergency PC shutdown.
Wiring a set of relay or sensor contacts in parallel with the PC’s Power-On switch is another way to initiate an emergency shutdown. The following picture shows a Power-On Y-cable (black and white wires on the right). One connector attaches to the front panel Power-On switch (or the switch cable) and the other to the motherboard header. The coiled gray wire can be run to a set of sensor or relay contacts.
As long as the contacts stay open, the computer will operate normally. But if a problem arises causing the contacts to close for more than 5 seconds, the PC will shutdown.
Of the two methods presented here, breaking the ATX power supply PS-On signal line is the most failsafe way to initiate an emergency PC shutdown because it bypasses the mobo and Power-On switch circuit entirely. Unfortunately, this method requires that the watercooling system be up and running normally before the computer can be started. To get around this requires adding a special time-delay relay into the shutdown circuit (more on this later).
Now that we know how to initiate an emergency shutdown, it’s time to figure out how to monitor the system and decide when such a shutdown is needed. To do that brings us back to sensors.
Sensors are typically electro-mechanical devices (transducers) that convert real world parameters (temperature, pressure, flow rate) into electrical signals. Logic circuits can then make decisions based on the information provided by the sensors.
Sensors generally provide two types of outputs:
- A continuously varying analog signal
- A simple on/off signal
An analog pressure sensor generates a voltage output that is proportional to the pressure being sensed. As the pressure increases so does the voltage output. A pressure switch on the other hand will turn on or off when a predetermined pressure is reached.
It is actually quite easy to fully outfit a PC watercooling system with sensors to monitor all the various system parameters. For example, sensors are available to monitor:
- Flow rate
- Pressure (and differential pressure)
- Liquid level
- Leak Detection
The following picture illustrates some of the components that might be used.
The problem with doing it, however, is cost. The necessary sensors, signal conditioning and data acquisition boards could easily cost many thousands of dollars, far more than the cost of the computer itself. Not to mention the software needed to analyze and view the data and the time and knowledge needed to set it all up and make it work!
Sensors with simple on/off outputs (switches) are generally less expensive and easier to interface to than sensors with analog outputs. For those reasons, we will primarily look at sensors that either open or close a set of contacts based on a certain set point.
For example, it might be interesting to know the actual flow rate of water moving thru the PC’s watercooling system. In reality, what we need to know is whether or not adequate flow is present. Using a flow switch that is preset to close (or open) as long as the flow rate stays above a minimum acceptable value will not tell us the average system flow is 120 GPH, but it will provide a signal (that can be acted upon) if the flow drops below, say, 50 GPH.
To help put this into perspective, a flow switch can be incorporated into a watercooling system for less than $100 US, while a flow meter (one that does not severely limit flow) can cost several hundred to several thousand dollars.
All of the sensors previously mentioned could be incorporated into a PC watercooling system. But each sensor adds complexity and cost. Most users just want some reassurance that there cooling system is operating normally and to have the system capable of protecting itself, should things go really bad. Each user will have to decide for his or herself how much time and money they are willing to invest to enable their PC to protect itself. Let’s take a brief look at what each sensor type listed above has to offer.
The flow rate of water tells us how much water is moving through the system per unit of time. Flow rate is NOT the same thing as water velocity (or the speed water is flowing past a particular point in the system). Flow rate is very important because heat transfer is directly proportional to flow rate. Once a system is setup and operational, the flow rate should remain fairly constant.
Numerous things can occur over time that may decrease the system flow rate and cause the heat transfer to deteriorate, which will ultimately result in the CPU temperature going up. A pump failure or loss of coolant (leak) may cause a complete loss of flow, resulting in a core meltdown (CPU failure). Other factors like the accumulation of biological growths or corrosion products may compromise flow more slowly.
Installing a flow switch into a watercooling system may seem like a good overall solution. The down side is cost and the resulting flow restriction the flow switch itself introduces. Most watercooling enthusiasts go to great lengths to minimize any type of flow restrictions in their system. Here are two flow switches that are suitable for use in a PC watercooling system.
Omega Visual Flow Indicator with Switch FPR121 (0.5-5.0 GPM) $150
Both flow switches are high-quality, industrial grade units with good accuracy and repeatability. The Gems unit comes with a preset flow rate (0.5, 1.0 or 2.0 GPM) while the more expensive Omega unit is adjustable and includes an LED indicator. The Omega unit contains a SPST switch, which gives you one set of NO and one set of NC contacts. The Gems unit has a single pair of NO contacts that close when flow exceeds the preset flow rate. Either set of contacts can be used directly or through an intermediate relay.
As good as these units are, they still introduce a flow restriction into the system. I measured the flow rate thru a typical watercooling system with and without the flow switches installed. The system included an Eheim 1250 pump, Maze3 waterblock, Black Ice Extreme radiator, and 5′ of ½” ID silicone tubing.
The flow rate was determined by measuring how long it took to fill a 5-gallon graduated bottle. Each test was repeated three times. (The E-1250 pump is submerged in the constantly filling reservoir, whose surface level is at the same height as the outlet tube going into the bottle.)
As you can see, both flow switches had a measurable impact on the overall flow rate thru the system, decreasing flow by almost 20% (22~23 GPH). The actual affect on CPU temperature will depend on the watercooling system hardware used and other parameters like CPU, overclock, ambient air temperature, etc.
At first glance, you might think measuring the system pressure with a transducer would be of little value to help in protecting the system. Actually, pressure can be used as a good indicator of overall system operation.
The water pump creates an area of relatively high pressure (2~3 PSI max. for most pond pumps) at the discharge of the pump and an area of relatively low pressure at the pump’s suction. The difference in pressure between these two areas is what causes water to flow thru the system. The waterblock, fittings, hoses and radiator all produce a certain amount of resistance to flow. The water pressure gradually drops as it moves from the area of high pressure to the area of low pressure. There is a measurable change in pressure (differential pressure) as water flows thru each component.
Monitoring the watercooling system pressure with a pressure switch gives an indication that the pump is running and there is water in the system. If the pump should fail or the system develops a significant leak, then the pressure will drop to 0. Monitoring system pressure will NOT tell us that water is actually flowing thru the system however. If a hose becomes kinked or a waterblock clogged, there may still be good pressure upstream of the blockage.
A big advantage of using a pressure switch is that they are small, reliable and inexpensive (less than $20 US). They are also non-invasive so they will not restrict the system flow rate. Here is one example.
These little switches come in five different pressure ranges with adjustable set points. The switch shown above on the left (9011-903) covers the range of 0.4-1.0 PSI. Each pressure switch contains a SPDT micro-switch, which includes both NO and NC contacts.
Here are two examples of methods that can be used to tap into the system for monitoring pressure. On the right is a standard brass hose barb fitting, which has been drilled and tapped to receive a small 10-32 x 1/8″ barb fitting. The pressure tap on the left consists of a short section of ½” copper water pipe with a short brass barb brazed onto the side (braze or solder the barb on, then drill the thru hole). A short piece of tubing is used to connect the pressure tap fitting to the pressure switch.
I generally tap into the system just downstream of the pump discharge (where the pressure will be the highest). In the example above, the pump feeds water directly to the Maze3 waterblock before going on to the radiator and reservoir.
Keeping an eye on the system temperature seems like a logical choice because after all, the whole point of a watercooling system is to cool the CPU. There are four basic types of temperature sensors (thermistor, thermocouple, RTD, and IC) and they are available in many different sizes and shapes. On-chip thermal diodes and on-board thermistors are two of the most common types found in PCs. Thermocouples are an industrial standard and are being used in more and more test rigs and even some RTDs are finding their way into analytical test benches.
Ideally we would like to monitor the CPU core temperature. As we discussed earlier, this is possible on many motherboards. Aside from problems with calibration and accuracy, the main problem with this approach is that if the computer locks-up so does the temperature shutdown circuit. To get around this we need a circuit that is external to the motherboard.
Implementing such a system requires a temperature sensor be mounted in close proximity to the CPU core and a temperature controller that can be programmed to act if a set point is reached. Sounds expensive right? Maybe not…
One simple but elegant approach to solving this problem is to use a DigitalDoc5+ temperature monitor and fan controller. The DigiDoc mounts in a spare 5.25″ drive bay and includes eight small temperature probes. Each temperature probe channel can be programmed to turn on an associated fan output at a certain temperature. For example: temperature sensor #3 can be programmed to turn on fan output #3 when the temperature reaches 50ºC.
With a street price of around $55 US, the DigitalDoc5+ makes for a very affordable temperature monitoring and control solution. Six of the eight temperature probes are thin-film type sensors, which can be carefully installed between the waterblock base and the top of the CPU package. DO NOT place the probe on top of the CPU core! It must be mounted beside the core. Very carefully trim the plastic film packaging around the sensor tip so that it can be located right up along side and touching the CPU core. A tiny dab of thermal compound between the sensor tip and core will help conduct heat.
To implement a CPU over temperature shutdown circuit requires programming the particular DigitalDoc5+ CPU sensor channel to the desired shutdown temperature, say 60ºC. Instead of using the corresponding fan output to turn on a case fan, we will connect it to a small 12 VDC relay. Now if the watercooling system should in some way fail and the CPU temperature rises to 60ºC, the DigitalDoc5+ controller will turn on a relay, which can be wired to shutdown the computer. And it won’t even matter if the PC is frozen or locked up.
This picture shows channel #1 of a DigitalDoc5+ wired up to monitor the CPU temperature with the corresponding fan output connected to a small 12 VDC relay. (Be sure to install a diode across the relay coil.) The relay contacts are used to initiate an emergency shutdown. In the example above, a pair of NO (Normally Open) relay contacts are wired in parallel with the PC’s Power-On switch. When temperature probe #1 reaches the programmed set point, the DigiDoc turns on fan output #1, which activates the relay and closes the contacts wired to the Power-On switch. Approximately five seconds later the PC turns off.
For PC watercooling systems that incorporate a reservoir, adding a liquid level switch can be relatively easy and inexpensive. A liquid level switch can warn the operator (audible alarm) or shutdown the computer if the water level in the reservoir falls below a certain point. Here are two types that can be used.
Omega LV-510 Vertical Float Switch $21
The vertical style level switch can be mounted into the lid of a reservoir while the horizontal style is designed to mount thru the side. Both types use a permanent magnet in the float to trigger a small reed switch inside the plastic body. These switches are SPST and can be configured as either NO or NC contacts depending on how the floats are oriented.
The small vertical style float switch mounts via a 1/8″ NPT fitting. It can be easily extended so the float can reach further down into the reservoir by adding a brass coupling and nipple of the desired length.
Once assembled and installed into a 500 ml custom reservoir, the level switch with the vertical float looks like this.
The horizontal style float switch is another option, which can be used in larger reservoirs. Here is one installed in the side of a 5-gallon reservoir tank. In this application, the float switch doesn’t actually turn off a PC but instead prevents the pump from starting, or turns off the pump, if the water level drops below a safe operating point.
Instead of initiating a full emergency PC shutdown with a reservoir liquid level sensor, it is often more desirable to just turn on an audible alarm – like this:
The float switch is configured to close its contacts when the liquid level drops below a desired height. A small toggle switch is included so the alarm can be silenced once the problem has been acknowledged. Both the audible alarm (Sonalert) and the Silence/Acknowledge switch can be mounted in a spare 5.25″ drive bay cover or on the backside of the case. Note that all of the components are wired in series.
Detecting a water leak inside a water-cooled PC is sometimes of interest to people. There are several commercial water leak detection devices and kits available at many home improvement and electronics/hobby stores that can be adapted for this purpose.
Personally, I don’t worry too much about leaks inside my water-cooled systems. In my experience, if you use quality parts and assemble the system properly, you shouldn’t have a problem. (I know, key word – “shouldn’t” 🙂 Here are two examples.
Aquanot Flood Alert (Home Depot, $14.95)
Leak detectors sense the presence of water by monitoring the resistance between two metal probes. When the probes are dry, the resistance between them is high and the circuit is not activated. If water drips onto the probes, the resistance goes down and biases a transistor to activate the audible alarm.
As an experiment, I modified the Velleman kit slightly by replacing the 470 ohm resistor (R2) with a 680 ohm resistor and powering the circuit from 12 VDC instead of using a 9V battery. I connected the input of a small solid-state relay in parallel with the piezoelectric buzzer so a set of NO contacts will close when the alarm goes off. Placing these contacts in parallel with a PC’s Power-On switch results in the PC turning off 5 seconds after the leak detector alarm sounds.
The actual sensor I used consisted of two bare copper wire grids separated by a double-layer of Bounty paper towels taped onto a thin piece of plastic. A paper towel sensor can be sized to fit most any convenient location you choose inside your PC case. As long as the paper towel stays dry, the resistance between the two wire grids is high. If the paper towel gets wet, then the resistance drops and the leak detection circuit activates.
There are many sensors available for analyzing water coolant chemistry, but these go beyond what even the most dedicated PC enthusiast is probably interested in doing. Measuring the conductivity (electrical conductivity) of water is a good indicator of how pure the water is – or how contaminated it is.
Turbidity measures how cloudy the water is (suspended solids). Normally the water coolant should be clear. If biological organisms (fungus, algae, and bacteria) start growing or corrosion products start building up, the water will become cloudy.
pH measures how acidic or basic the water is. Ideally we want to maintain a pH of around 8~9 to help retard biological growth and minimize corrosion. And that’s all I have to say about water coolant chemistry… for now. (Watch for a more in-depth article on the subject in the near future… 🙂
Sometimes it might be nice if your computer could just let you know there was a potential problem brewing without executing a complete shutdown. Or maybe you want an alarm to sound after the computer shuts down. One of the most popular types of audible alarms for this purpose is a piezoelectric buzzer.
Sonalert 4-28 VDC, 95 dB, Pulsing Tone (All Electronics #SBZ-381 $4.75)
These little alarms will definitely get your attention once activated. They will operate nominally from 12 VDC but can also be powered from the +5VSB lead (purple wire, pin #9 on ATX connector) on ATX power supplies. Since the +5VSB line stays on even when the PC is turned off, the alarm can be configured to continue sounding after the computer has shutdown.
Ideally you want to use a set of NC relay contacts to activate the audible alarm. Under normal operating conditions, the audible alarm relay will be energized, holding the contacts open and keeping the alarm from sounding. If something goes wrong, another relay connected into the actual failure circuit can de-energize the audible alarm relay and cause the alarm to sound. You might also want to include a simple on/off switch in series with the alarm so you can turn it off.
OK, in this section we are going to apply some of the ideas and theories presented earlier to a few real world applications. Specifically, we will look at some ways sensors can be used alone or in combination to help protect your PC from problems that might occur in the watercooling system. There are numerous ways to go about this with no one solution that is right for everyone. By showing a few examples of how things can be done I hope to provide you with enough information that you can construct a working circuit of your own.
As we discussed earlier, there are two relatively easy ways to initiate a PC emergency shutdown (besides pulling the plug): One is to interrupt the power supply’s PS-On signal wire and the other is to close the computer’s Power-On Switch circuit for more than five seconds. The first method requires breaking a connection and the second method requires holding a connection closed to initiate the shutdown.
- PS-ON signal (green wire) – connection must be broken (open)
- Power-On Switch – connection must be made (closed) for >5 seconds
As I mentioned earlier, breaking the ATX power supply PS-On signal line is the most failsafe way to initiate an emergency PC shutdown because it bypasses the mobo and Power-On switch circuit entirely. It is also the more difficult of the two ways to implement because it requires the watercooling system be running normally before the computer can be started.
If a normally open sensor switch or relay contact is used to trigger an emergency shutdown by breaking the PS-ON line, it must be somehow reconnected before the computer can be restarted. To do this typically requires adding a time-delay relay into the circuit. This special relay will maintain the PS-ON line connected during the initial computer startup period and then turn over control to the sensor(s) monitoring the watercooling system. (Keep reading… 🙂
We also must decide whether the sensor contacts will be used directly to initiate a shutdown or whether they will be used indirectly to trigger a relay. Directly using a sensor’s built-in switch contacts is often the simplest approach.
Using an intermediate relay may be desirable when more flexibility is needed. Relays typically have multiple sets of contacts with both Open and Closed configurations. If the switch contacts on the sensor you are using only close (instead of open) when a problem is detected, you won’t be able to initiate an emergency shutdown by opening the PS-ON signal wire. However if the sensor is used to trigger a relay, then a pair of relay contacts that open when the problem is detected can be used.
Relays can be used to initiate more than one action at the same time. For example, two different sets of relay contacts could be used during shutdown: one set to close and turn on an audible alarm (powered from +5VSB) and a second set of contacts to open and shutdown the PC.
As the name implies, a failsafe circuit is one that will leave the equipment (your PC) in a safe condition should it fail. If a sensor or control relay fails then it should ideally turn off the PC and not keep a true problem from being detected.
Implementing failsafe circuits is common practice and required in industry because people’s lives and major capital investments are frequently at stake. Protecting your PC may not be as critical and the cost and/or added complexity of building a failsafe versus a working circuit may become an overriding factor. Still, it’s something to think about and include when possible.
The first circuit uses a flow sensor switch but it could just as easily be designed to use a pressure switch, or liquid level sensor, etc, which has an available set of NC contacts. The purpose of this circuit is to turn off the computer if no flow is detected (or if the flow drops below 1.0 GPM) in the watercooling loop.
The shutdown is accomplished by holding the Power-On Switch lines closed for more than 5 seconds with a pair of NC contacts. (Remember, the NC contacts will open when the set point flow rate is obtained or exceeded and they will return to their normal closed state when flow drops below the set point.
This same strategy can be used for whatever sensor type (flow, pressure, liquid level, temperature, etc.) you elect to use. To work, it requires that the sensor have a set of NC contacts. If not, a relay can be added to supply the necessary contact configuration.
In the next example, we will use a different flow sensor switch and initiate the shutdown by opening the power supply’s PS-ON line. To do that requires using a set of NO contacts instead of the NC contacts used in the last example. This time, when the flow rate exceeds the flow switch’s set point, the NO contacts will close allowing the power supply to operate properly. If the flow drops below the set point, the NO contacts will revert back to their normally open state, which will immediately turn off the power supply and computer.
At first glance you might not see anything wrong with this circuit, but a potential problem exists. As long as the watercooling system is turned on before the computer so that flow is established and the flow switch’s NO contacts are closed before the PC is turned on, it will work as planned. But if your setup uses a relay to automatically turn on the watercooling pump when you startup the PC, it won’t work! The power supply’s PS-ON line has to be closed before the computer can be turned on. One way to accomplish that is to add a time-delay relay into the circuit.
In the last example, we needed a way to hold the PS-ON line closed long enough for the computer to start up and then automatically turn over control to the flow switch. This can be accomplished with a time-delay relay. Just like their mechanical and solid-state cousins, time delay relays come in many different shapes and sizes.
Time-delay relays, called timers for short, are available in four basic modes of operation. Some are multi-function, allowing the user to select which mode they want the timer to operate in.
The interval timer is the one we need to help make our previous circuit work reliably. When energized, an interval timer will wait a user specified amount of time before shifting its output contacts. Connecting a pair of the timer’s NC contacts in parallel with the flow switch’s NO contacts will allow the computer to start up normally when requested. After an adjustable period of time (say 20 seconds), the interval timer’s output contacts will open up. By this time, flow should be developed in the watercooling loop causing the flow switch contacts to be closed.
The down side to incorporating a time-delay relay into the circuit is cost. Three of the timers in the previous picture are industrial control models ranging in cost from $50~$100 US. The Altronix Corp. multi-purpose timer is a more affordable solution for the PC hobbyist. After considerable searching, it was also the only interval timer I could find that offered both 12 VDC operation and change of state (energize relay to shift contacts) at the end of the timing cycle.
The Altronix multi-purpose timer can be set to operate on 12 VDC, has an adjustable range from 1 sec to 60 min, and includes an onboard relay with both NO and NC contacts. (It was designed to be part of a home-security system.)
I modified the Altronix 6062 timer slightly by adding a diode and large capacitor onto the incoming 12 VDC power leads. The capacitor holds enough charge to keep the timer energized for a few seconds, after the emergency shutdown has been initiated and the power supply turns off.
This gives the computer time to completely shutdown and prevents the timer from immediately re-enabling the power supply when the NC timer contacts close. The diode keeps the capacitor’s charge from dissipating back into the main circuits at shutdown.
We can “fix” the previous flow switch circuit by adding the Altronix time-delay relay. The NC timer contacts are wired in parallel with the NO contacts of the flow switch – like this:
The interval timer’s NC contacts will initially hold the power supply’s PS-ON line closed long enough for the computer and watercooling system to start up and create flow. After 10 or 20 seconds, the interval timer’s contacts will open but the power supply will continue to operate normally, because the NO flow switch contacts are now closed. If flow stops, the flow switch contacts will open breaking the PS-ON line and the computer will immediately shutdown.
Again, this same circuit could be used with another type of sensor (pressure switch, etc.) instead of a flow switch, as long as that sensor has an available set of normally open contacts.
Maybe you would like more than one sensor protecting your PC. In the next example, we will combine a pressure switch, reservoir liquid level detector switch and a DigitalDoc temperature controlled relay into one circuit. If any one of these three sensors detects a problem, it will shut down the computer. Combining multiple sensors into a single emergency shutdown circuit is easy – just remember to wire them together in series.
Note that it is quite acceptable to combine sensor switch contacts and relay output contacts in the same series chain. Maybe you can use the NO switch contacts provided in the pressure switch and float switch but need a relay to add temperature monitoring to your circuit.
If either of the sensor switches or the temperature controlled relay contacts in this circuit open up during normal operation the computer will immediately shut down. As before, the interval timer is used to insure the PS-ON line is connected at startup.
I would like to close by saying again that there is no single, right way to control every pump and meet every user’s needs for an emergency PC shutdown. We have covered a lot of information from the very simple to the slightly more complex. I have presented a few examples but there are many more techniques and options that could be applied.
Most of the examples presented here use off the shelf parts. These can be convenient but also can get expensive. It’s not out of the question to roll your own. I’m sure many of you have a better working knowledge of electronics, etc. than I do, so use your creativity and see what you can come up with. Hint: homemade flow switches and time-delay circuits are two potentially good places to start… 🙂
I hope this guide has answered some of your questions and maybe stimulated your interest.
Thanks for reading,