HOW-TO BY ODD ONE – Building a cooling system controller

WARNING!

This project involves direct connection to AC mains wiring. This is bad enough in the U.S. where house mains wiring carries 120 volts. Elsewhere in the world the wiring carries 220 or 240 volts AC. Home AC wiring carries enough voltage to kill no matter where you are.

If you do not have a decent familiarity with AC electronics, the electrical code of your area, and mains wiring safety procedures, DO NOT BUILD THIS PROJECT. You could very easily cause damage to equipment and property as well as injury or even death if you try this project and do something bad. Due to the safety concerns, this article was deliberately written to be a bit too complex to follow for hobbyists that lack the required background knowledge.

This article is written with U.S. AC mains (120VAC) and electrical code requirements in mind. The project must be adapted to suit non-U.S. AC characteristics and connectors – instructions on how to do this for anywhere outside the U.S. will NOT be provided.

By attempting to build this project, you assume all risks and liabilities. The author of this article and any website that posts it cannot be held responsible for your actions! Proceed at your own risk!

Introduction

Overclocking requires technology and technique, an arcane art-technology fusion that involves creative solutions to the technical problems of pushing computing beyond its limits. Sometimes this manifests itself in the creation of a novel solution for a simple and common problem – dealing with the excess heat an overclocked processor generates.

Air-cooling is the most common means of pulling excess heat off a processor, but cooling with air has limits. When those limits are exceeded, it’s time to deal with more aggressive cooling avenues, such as thermoelectric or Peltier cooling modules and recirculating-water cooling systems. These systems are often fairly complex and have the need for their own power. This is where the cooling system controller comes in.

The cooling system controller is a simple device that can manually (or automatically, with some additional circuitry) switch power to the components of a cooling system. Water pumps, high-airflow AC-powered cooling fans, power supplies for thermoelectric cooling (or TEC) modules, the computer itself, practically anything that needs AC can be managed through the cooling system controller. And the controller can provide power to these items from a separate supply line, without the need to tap power from the computer’s power supply. This project even includes its own 5VDC/12VDC switching supply to run 12-volt DC case fans with no extra drain on the computer’s power supply.

The controller may be built with as few or as many features as your specific setup requires, simply by adding additional components. Best of all, the controller can be built to mount entirely within the case, leaving nothing outside to show its presence aside from a few switches mounted in a drive-bay cover plate to turn the cooling system components on and off, and an extra power cord out the back.

The Design In Theory

The cooling system controller is little more than a collection of solid-state relays to switch mains AC, conventional DC relays to switch 12-volt power to DC components such as fans, and an on-board switching power supply to feed DC devices directly. Thanks to the design, any circuit that can switch 5 volts at 100 milliamps or less (usually much less) could control AC loads in the 25-ampere range.

By using parts rated for 240VAC, the controller can be built to suit most any power source worldwide, requiring only a change in power supply cable. (Of course, the items you plug into the controller will be another matter.)

The controller consists of three sections – the source section that holds the RFI/EMI filter, master power switch, and protective fuse and fuseholder, a small switching power supply for fan power and to supply the turn-on voltage for the solid-state relays, and the relays themselves active to switch AC current.

The source section as mentioned is a smaller box connected to the controller by a length of three-conductor 14-gauge AC-rated power cable. This box is mounted to the rear of the case. It holds the RFI/EMI filter that protects the controller and everything plugged into it from any AC line noise as well as preventing line noise backing out onto the AC mains.

A master power switch permits one-switch disconnect of the controller and all items plugged into it. Finally, a fuse provides short-circuit protection in case of a serious problem.

The controller itself houses the other two sections. The power supply used in this case was a 17-watt switching power supply manufactured by Research Ltd. and sold by All Electronics as their part number PS-9505. This unit is small and provides 12VDC @ 1A and 5VDC @ 1A. The 5VDC will be used for controlling the states of the solid-state relays, and the 12VDC will be used to power DC fans.

The solid-state relays were selected to handle the expected loads they would face. The pump and AC fan relays are each 5A 240VAC models made by C.P. Claire, and the Peltier supply relay is a beefier 25A 240VAC heavy-duty model made by Opto.

All of these are TTL compatible, which means that they will switch on with 5VDC, which is perfect for this application. All of these relays were also purchased from All Electronics as their item numbers SSRLY-2405 and SSRLY-25, respectively.

For safety, the outputs of two 5A relays are fed through 2A 240VAC fuse-style circuit breakers to provide a safety margin of over 200%. This reduces the capacity of the pump and AC fan outlets to about 480 watts each. Even this is overkill given that most pumps used for water-cooling consume less than 50 watts, and AC fans small enough to fit in a computer case only pull about 20 watts or so each.

The way all this is wired is quite simple. All outlets’ neutral lines are tied to the neutral AC supply line, and all outlet grounds are tied to the AC supply line’s ground lead as well as to the enclosure itself. The AC hot line is then connected to each solid-state relay’s AC input or line terminal. The relay’s output or load terminal is then connected to the circuit breaker and then to the hot terminal on the appropriate AC outlet.

All AC connections are point-to-point, routed to prevent chafing if needed, using 14 gauge stranded UL/CSA-rated hookup wire for safety. The negative terminals of all relays’ control connections are connected to the ground from the switching supply, and each relay’s positive terminal is connected to a specific lead on the control cable.

Another control cable lead connects to the +5VDC output from the switching supply – switching this +5VDC onto any of the leads wired to one of the relays will turn that relay on. The relays need a tiny amount of current, often less than 10 milliamps, to switch on, so most any means to switch 5VDC can be used to switch several hundred watts of AC power.

To provide both AC and DC fan power, a small traditional mechanical 5VDC SPST relay is used. Its coil connects to the positive control terminal on the solid-state relay used for AC fans, and to the switching supply’s ground. We do this so that when the fan relay receives its 5VDC to switch on electronically, the traditional relay does the same mechanically. The normally-open connections switch the +12VDC lead from the switching supply to a terminal on the two-position feed-through barrier strip, whose other terminal connects to the switching supply’s ground.

Next, the Prototype

Email Odd One


The Prototype

The original prototype is shown in figure 1.

Figure 1
Figure 1 – The first controller prototype. The white connector is the control connector

This prototype was housed in a basic enclosure purchased from Radio Shack. The controller prototype worked, quite well in fact, but was hardly the right size and shape to fit within a typical case.

So a second device was designed specifically to fit inside the case, which in this case is an Inwin Q600. This particular case features a large, slide-out motherboard tray that has two inches of clearance between itself and the 3.5” drive bays’ cage, which is more than enough space to mount the 1 ¼” tall controller.

Figure 2 shows a top-down view of the wiring inside the controller prototype. The entire controller is wired point-to-point. Point-to-point wiring is hardly pretty, but it’s effective – however, the rebuilt controller’s wiring will be much more neatly laid so as to minimize magnetic field generation.

Figure 2
Figure 2 – The controller’s internal wiring

The Controller’s Enclosure

As noted before, this was not an enclosure that would fit inside a case very well at all. So an enclosure was designed that would be large enough to allow room for the controller’s internal components and room for AC plugs to clear, but small enough to fit inside a case.

The dimensions that worked out best are 8.5” wide x 1.25” tall x 5” deep. Borrowing from basic chassis designs, a template was drawn out based on those target dimensions. The end result was what is shown in figure 3.

Figure 3
Figure 3 – The controller’s enclosure dimensions

The material for the enclosure is .026” thick aluminum sheet, which is easy to come by in the U.S. The front and back are bent to form a ‘U’, and the ½” tabs are all folded 90 degrees to form reinforcing.

The tabs along the long edges of the front and back are each drilled to permit mounting screws that will tighten the lid into place.

Figure 4
Figure 4 – The aluminum sheet marked for cutting

Figure 5
Figure 5 – The aluminum sheet, cut

The next stage in the project is to mark and cut holes for all AC outlets, circuit breakers, and the two cables that will act as the AC supply and control lines.

Given how many items are being fit into how small a space (two grounded 120VAC North-American home outlets, one non-grounded outlet, two circuit breakers, a two-position feed-through barrier strip, and two rather thick cables in an 8 ½” wide by 1 ¼” tall space) layout is crucial.

So the layout was sketched out and finalized on paper and the measurements transferred to the cutout. Then the marked cutouts are pilot-holed and cut to the correct size.

(Author’s note: A sheet metal hand-nibbler tool is the best tool for this purpose, as it can cut square holes with ease. This particular tool is shown in figure 6. It is made by GC Electronics and can be purchased from major electronics retailers for about $20.) This is shown in figures 6 through 8.

Figure 6

Figure 6 – The front-panel hole measurements transferred to the enclosure cutout. The tool shown is a hand nibbler, which is extremely useful for the rectangular holes required for the AC outlets

Figure 7
Figure 7 – The difficult-to-cut holes with pilot holes drilled to permit the nibbler to start on the material

Figure 8

Figure 8 – Holes, cut and tested to ensure a tight but usable fit. The undrilled area will later be filled with a heavy-duty two-position terminal strip for switched 12VDC for DC fans

Next comes the hard part – forming the cut metal into the bottom of the enclosure. Given the difficulty of hand-forming this type of enclosure, it would be advisable to seek a pre-fabricated all-metal enclosure as compared to fabricating one yourself.

This particular design, however, does not exist in a pre-made form and had to be hand formed. The end result of the forming efforts is shown in figure 9.

Figure 1
Figure 9 – The formed and drilled enclosure half, ready for the next phase of the project

Next: Controller Circuitry

Email Odd One


The Controller’s Circuitry

Now that the enclosure half is formed, parts can be placed and their mounting holes drilled as necessary. The best way to do this is to test-fit all components and play with their positions and layout. Once the best layout is found, the mounting holes can be marked, the parts removed, and the holes drilled.

For this controller, due to the point-to-point wiring to be used, two 5-position terminal strips were mounted on opposite ends – one for the AC connections and the other for the DC connections. Figure 10 shows the enclosure with all holes drilled and the two strips installed.

Figure 10
Figure 10 – The enclosure with holes drilled. The two white posts will hold the switching supply board, and the objects on each side are the terminal strips

With the holes drilled, the components can be mounted. Since AC outlets often come with terminals and not leads, it is better to solder leads onto these before installing the outlets into the front of the enclosure. Figure 11 shows the controller’s components mounted, but no wiring run.

Figure 11
Figure 11 – The relays and switching supply installed. The switching supply is mounted on nylon standoffs and wrapped with fiberboard to prevent shorting the back of the board on the enclosure

With the components mounted, install and wire the front-panel components, starting with all ground AC lines, then all neutral AC lines, then all AC hot lines, and finally the DC connections to provide control voltage. The AC and DC wiring is run as far away from each other as practical to prevent noise introduction into the DC control lines.

The end result is shown in figure 12, undergoing testing. The orange cable is the AC supply line that will connect to the small filter/switch box to be built next, and the purple cable is the control cable, a segment of category 5 UTP data cable.

In the control cable, two pairs of wires will act as control leads. These are soldered to the positive control terminals on each relay. (Since this version has three relays, one wire will be left unused and was folded out of the way.)

One of the two remaining pairs is connected to the +5VDC and ground lines on the switching supply, and the last pair was connected to the +12VDC and ground lines on the switching supply. This permits supplying +5VDC/+12VDC to any control circuits for LEDs, fanbus/baybus connections, etc.

However, any items drawing off the +12VDC line will reduce the amount of current available at the 12VDC fan terminals on the controller, so it’s best to use that switched +12VDC from there or tap off the computer’s power supply for additional fans.

Once all wiring is run, it’s time to triple-check the work and test the controller. This will involve connecting the AC supply cable to AC mains and testing for proper switching.

WARNING – This means that AC mains current will be present in exposed wiring – test with extreme caution. It is important to note that some solid-state relays (especially larger ones like the 25A unit in this project) will not switch unless they have a load to switch.

So plug something into each outlet and connect the +5VDC lead on the control cable to the turn-on lead for the relay connected to that outlet and check to ensure the device plugged into that outlet is powered when the relay’s turn-on lead is tied to +5VDC and loses power when that connection is broken.

Figure 12
Figure 12 – The wired controller being tested. Note the meter showing 120VAC. USE CAUTION while testing the controller, as AC mains current will be exposed!

At this point, the lid is cut out and folded to fit over the enclosure. A layer of poster board is also trimmed out to cover the wiring and prevent short-circuits against the lid. The lid is a simple ‘U’ of metal cut to 5” wide x 11 1/16” long, with both ends folded at 1 ¼” inches in from each short end. The ‘U’ thus slips over the enclosure perfectly and is secured with eight self-drilling screws.

Last, the Finale

Email Odd One


The AC Source Enclosure

The controller is now largely complete, but the umbilical cable that is to feed AC into it needs to go somewhere.

So another enclosure is fabricated to hold the RFI/EMI filter with built-in EIA AC power socket, fuse holder, and master power switch.

This enclosure is designed the same way as the enclosure for the controller, only smaller at 3 ¼” x 3 ¼” x 2 ½”. This size was selected to fit case cutouts and mounting locations for 80mm case fans, so that an unused case fan vent on the case’s back panel could be cut out and serve as the plug-in point for the controller.

The cutout, drilling, and fabrication process is the same as before, and is shown in figures 13-17.

Figure 13
Figure 13 – The controller lid and AC source enclosure base marked on the remainder of the aluminum sheet

Figure 14
Figure 14 – The AC source enclosure, trimmed. Note that the design is basically the same as the controller’s enclosure in figure 5

Figure 15
Figure 15 – The mounting holes for the fuseholder, switch, and filter

Figure 16
Figure 16 – The AC source enclosure drilled and folded

Figure 17
Figure 17 – The components test-fit into the AC source enclosure. Note the four holes, which align to mounting holes in a case’s stock 80mm fan mounting locations

Once the AC source enclosure is fabricated, the power cable from the controller is measured and cut to the required length for the case in question, and fed into the AC source enclosure through a strain-relief bushing.

The neutral line from the filter is connected to the neutral line on the power cable, and both grounds are tied to each other and to the enclosure by means of a machine screw (not shown.)

The filter’s hot lead is connected to one terminal of the switch, and the other switch terminal connects to one terminal on the fuseholder. The other fuseholder terminal connects to the hot lead on the power cord.

The lid for the AC source enclosure is fabricated the same way the lid for the controller was formed (again, except for measurements, which in this case is 3 ¼” wide x 5 5/16” long, folded at 2 ½” in from each short end to form a ‘U’) and attached with self-drilling screws.

An additional 3-prong (grounded) AC receptacle is mounted into the lid, so that the controller can be plugged into the AC source enclosure inside the case.

Since some cases employ a slide-out or tilt-out motherboard tray, this design permits unplugging the controller from the AC source without having to detach either if the controller is mounted on the motherboard tray.

Another test is warranted with a traditional EIA power cord to power the controller.

With everything successfully tested and buttoned up, the controller project is completed. The end result is shown in figure 18. The controller undergoing its final pre-installation test, powering the components of a real cooling system from outside the case, is shown in figure 19.

Figure 18
Figure 18 – The completed controller (lower right) and AC source (upper left). The orange power cord from the controller, shown here with an AC plug for testing, will be plugged into the AC source box, which will in turn be mounted on the rear of the case. The purple cable is the control cable

Once the controller’s enclosure is finished, the top panel can be drilled for mounting standoffs or a mounting spacer plate.

Figure 19
Figure 19 – The controller undergoing real-word testing with a live cooling system’s components

It’s a mess, but a working one – the computer is a Pentium 3 700E running at 1035 MHz. The controller (center) sits atop a 230-watt computer power supply used to run the Peltier module. (It is the leftmost plug on the controller.) The center plug is the water pump’s power cord, and the white plug is powering two 120mm 120VAC case fans. At upper left is the AC source box.

(Author’s Note: The black bucket with the white lid is in fact a portable 130W RMS speaker system. The author is locally famous for his “speaker in a bucket” designs, which perform comparably in both sound quality and power capacity to speakers costing several hundred dollars each. And they just look funky. [grin])

Controlling The Controller

With the controller installed inside the case and the AC source box mounted to a cut-out back-panel case fan opening, the only thing left to do is fabricate the switch panel to run everything.

Since the control cable carries power as well as control lines, a set of SPST on/off (NOT momentary!) switches would be perfect. This cable can be trimmed to the desired length once a suitable location for the cooling system control cluster is found.

The wiring for each switch is simple – one terminal to the +5VDC lead on the control cable, the other to the control lead for the relay to be switched. Since the relays are sensitive, any switch that can switch 100 mA at 5VDC will work fine. (Add 20 mA to that if using a LED to indicate that switch’s position.)

For more eye-pleasing appearance, panel-mount LEDs can be employed, with each LED’s positive terminal wired to each relay’s control lead, and negatives tied to the ground lead on the control cable.

(NOTE: Most LEDs will require a dropping resistor be connected between the LED’s anode (longer) lead and the positive voltage source to reduce the voltage across the LED. Check the LED’s specifications and visit this webpage for information on calculating the resistor value you’ll need.) Please link the phrase “visit this site” to http://www.btinternet.com/~g4wif/q_tech8.htm

Probably the most common place for such a switch cluster is a blank drive bay faceplate, although some have installed similar switch clusters on the front case panel itself.

Conclusion

The cooling system controller is a simple, flexible, expandable design that works well.

More importantly, it makes previously difficult-to-impossible cooling designs, such as totally within-the-case water/Peltier cooling systems with NO outside lines or wiring aside from a single power cable.

It can also be employed to tame some of the cable mess common to aggressively cooled computers.

It’s a simple solution to a logistically difficult problem – powering and managing a complex active cooling system for a high-performance computer. And, it was a fun and challenging project to design and implement.

However, it’s not a project for the unwary. If you have the expertise with AC electronics and need to simplify your cooling system, this project is for you. If not, it isn’t.

Email Odd One


Be the first to comment

Leave a Reply