Testing the first ground loop

The initial tests of the first ground loop are very promising! The system has been running for a few days with no major problems or leaks. The 1-wire digital temperature sensors and the iButtonLink LinkUSB reader are performing nicely.

Bear in mind that this is only one well, so the system is running with 1/3 to 1/4 of expected final cooling capacity. It is handling the heat absorbed by the (so far poorly-insulated) system from the 90+ degree air, plus the 45 Watts of heat generated by the water pump and the blower. These heat sources will eventually be spread across all wells in the system, so it is encouraging to see a 5-6 degree F drop in cabinet air temperature below ambient. The heat from the server is not in the system yet, but it certainly seems that as we scale the system to 3 or 4 wells, it might handle that additional heat.

Notice near the peak temperature of the second day in this chart (click for a full-size version), I turned off the water pump for a few hours to demonstrate the cooling effect that the system has on the air temperature:



Other points worth noting: The air temperature (blue dotted line) reaches the temperature of the water passing through the coil (not shown, but roughly equal to the blue dotted line) even with the blower at the lowest setting. This tells me the fan-coil unit is plenty large for what it's cooling right now.

However, the water temperature (not shown) never gets very close to the soil temperature. This tells me the water isn't spending enough time under ground and/or the PEX tubing in the ground loop is too good of a thermal insulator. PEX is about twice as good as soil at conducting heat, so I don't think the problem is so much the choice of material as it is the length of time the warm water is exposed to it.

You can see quite a temperature swing in the soil as we dump heat into it. So heat IS successfully transferring from the cabinet air to the soil. As we add more wells, this temperature swing will be proportionately reduced as the wells share the absorption of heat energy. So with four wells, instead of being raised from 71 to 83 (a 12 degree increase), we can expect the soil to be raised from 71 to 74 (a 3 degree increase).

First ground loop

Since my friend Matt gave me lots of 3/4" PEX for this project, I'm going to see if I can get away with using it for the ground loops. Its walls are pretty thick and PEX is a pretty good insulator, so I'm not sure it's that great of a choice for the ground loop. Copper would be nice, but unless I can get a really good deal on some surplus K-type copper pipe, the cost will be prohibitive.

The first well (borehole) will be 6 feet deep and will contain two PEX loops, for a total of about 24 feet of buried pipe. Geothermal systems don't usually use multiple loops per well, but with such shallow holes and with the small temperature differentials that we'll have (since we're not using a heat pump), we'll need the water to be exposed to the ground longer than in a normal system. Multiple passes through each well will help the water get closer to the soil temperature.

I formed U-bends with PolyAlloy elbows. It's unfortunate that I have to use crimp fittings at the bottom of these loops, but I just can't figure out any better way to form a U-bend that will fit in my 5-inch diameter holes. I tried heating and bending the PEX into a 180-degree U with a heat gun, but I just made a mess.

The PEX has an arc to it because it comes on a roll. I had a tough time getting it to straighten out, so I ended up strapping the pieces to a long piece of wood for support. I realize that this will rot over the years and cause a little sink hole, but I'm not too worried about that.

In this picture you can see the "Tee" fittings that contain the mid-loop temperature sensors, as well as a temperature sensor attached to the bottom to measure soil temperature at 6 feet. There are similar sensors attached at 3 feet and near the surface. Click for a full-size image.



Here is the ground loop ready to bury:



Here is the ground loop inserted into the ground:

Learning to plumb

Thanks to my friend Matt, I have lots of 3/4" PEX on-hand for this project! That will be perfect for all the above-ground plumbing, and I'm going to try it for the first "test" ground-loop. If PEX works out for the ground loop, I'm all set with plenty of materials for the rest of the project.

PEX is Cross-Linked ("X-linked") PolyEthylene, and is used in radiant-floor heating and potable water systems. You may have noticed it in newer houses where the hot and cold water are run through red and blue plastic tubes instead of the old familiar PVC or copper. Those colored plastic tubes are stronger, more flexible PEX.

PEX can't be easily solvent-welded like PVC can, so plumbers use crimp rings to join PEX to fittings. PEX is more flexible than PVC but it's still pretty hard, so hose clamps have a tough time with it, even for temporary installations. I used hose clamps in a few places to do some tests, and they basically exploded when I tried to torque them down enough to stop the leaks.

I started out with an inexpensive PEX crimper from Lowe's Home Improvement designed to be compressed with 10" channel-lock pliers.

There's nothing wrong with this crimper if you only need to do a couple of crimps. But I can tell you, they're stretching it a bit when they say you should use 10" pliers. Yeah, I was able to get good crimps with them, but it often took several tries (a big no-no with crimp rings) and I had to squeeze so hard that my hands were sore the next day. I really should have been using 12" or larger pliers. Also, it's a little awkward gripping the heavy crimper with pliers and making sure they don't slip off the small lip on the crimper.

After doing 7 or 8 crimps with this thing, I RAN back to Lowe's and happily paid the extra money for their professional crimper! I'm going to have way too many crimps in this project to put up with a cheap tool.

I also had to learn how to sweat copper fittings, which is something I've managed to avoid so far in life! ...Adding yet another skill and some new tools for this weekend project...

I must admit, I called on my friend Matt to do the "risky" soldering (on the fan-coil unit and the balancing valve, neither of which I particularly wanted to damage), but I did manage to use my new torch and skills on a couple of fittings!

Fan-coil unit

I think the 9"x9" coil I bought cheap on eBay would have been sufficient for this project, and I liked the nice fat copper tubes in it (which would have provided low resistance to the flow of water). But my friend Matt had a spare in-wall fan-coil heater unit with a nice efficient variable-speed blower conveniently mounted in a complete housing. It was just too nice to pass up! There will be a couple of challenges with this unit, but overall it seems like it will make some things much simpler.

First of all, this is a heater unit, so it's not designed to collect condensation from the coils. I don't expect MUCH condensation in this system since the temperatures I'm dealing with will be much higher than those in an air conditioner. But on humid days, opening the server cabinet will let moisture into the system, and it will be very possible that the coil will be below the dew point. So I needed to modify the fan-coil unit to collect and drain condensation.

This wasn't difficult at all. It looks like this unit might come in a version that handles condensation, because it was really only missing a couple of "gutters" and a drain pipe. I split a piece of 2" PVC pipe on the table saw, capped the ends, used a dremel tool to split off the tops of the end-caps, forming a PVC trough. I drilled a hole in one of the caps to add a 1/2" exit drain (I didn't use a reducer because that would have centered the 1/2" drain pipe in the middle of the 2" pipe, whereas I needed the drain pipe to be aligned with the bottom of the 2" pipe).

Here's a close-up of the back-side of the coil after removing the back of the unit (click for a full-size image):



Here's the same view after adding some gutters and a temperature sensor:



The other potential issue with this fan-coil unit is the small-diameter copper tubing, which adds a lot of resistance to the flow of water. I want to keep energy consumption to a minimum, so I really want to stick with the 20-Watt pond pump I purchased for the project and avoid upgrading to a more powerful circulator pump like the 70-Watt Taco 007 pump (about $60) used in many radiant-floor heating and solar hot water systems. So I'm being very careful not to add any unnecessary resistance elsewhere in the system.

With the efficiency of the blower in this unit, I think I'll be able to run it at the lowest setting, which only consumes about 22 Watts.

Eventually, this fan-coil unit will be ducted to the server cabinet and everything will be thoroughly insulated. But for initial testing, I have ducted its exhaust to its intake with 25 feet of 6" insulated flexible duct.

This gives a closed-loop air system with some extra exposed duct to approximate the heat load of the metal server cabinet.

In this photo, you can see the insulation board front that I made for the unit, the flexible duct, and you might be able to make out the wires leading to the temperature sensors that I added to the water-in and water-out ports.

You can also see the 1/2" PVC drain mentioned earlier for the condensate. Note that I formed a trap at the bottom so water will collect there, blocking the flow of air. We don't want the condensation drain to be an air leak in the system!

1-wire digital temperature sensors

I read about 1-wire sensors from Dallas Semiconductor (now produced by Maxim) back in college in the mid-90's, but I've never had the chance to play with them. The principle is that many sensors can be strung along the same wire like Christmas lights, creating a mini network. A computer or microcontroller can poll the sensors using a single port, eliminating the need for multiple analog-to-digital converters or complex switching circuitry to monitor many sensors.

Each sensor has an electronic serial number, so the computer can tell them apart. The sensors all share the same data wire, and they all "suck" power from that same wire (referred to as "parasitic" power), making it very easy to connect many sensors.

This is exactly what this project needs! With 1-wire sensors, I can log all the system temperatures automatically in real-time. And at less than $5 per sensor, I can put them all over the place and monitor any interesting temperature in the system that I want!

I wanted to keep the temperature sensing cheap, simple, and compatible. And I didn't want to spend an entire weekend getting the sensors to work. So I chose the iButtonLink "LinkUSB" 1-wire reader. This is a USB device, so it works on any PC without adapters, but it emulates a serial (COM) port and is compatible with software designed for the original Maxim 1-wire readers.

Nothing is ever easy, so it actually took me a day or two of e-mails with iButtonLink to get the device working properly. It turns out that there are slight differences between the LinkUSB and the older native Maxim serial reader, and recent updates to the Maxim-provided drivers crippled the LinkUSB. They helped me find the older (version 4.02) drivers that work fine here:

http://files.dalsemi.com/auto_id/licensed/files_1_wire_drivers_v402.zip

...and the Windows command-line utility bm_demo.exe here:

http://www.ibuttonlink.com/ExperimentalSoftware.aspx

I wrote a simple scheduled batch file using my favorite add-on Unix utilities to poll the sensors, append the raw data to a log file, and convert the data to a comma-delimited (CSV) format with headers for Excel to read. The Unix utilities grep, sed, tail, and cut are really helpful tools in Windows batch files! This could be done with native batch commands, but I'm more comfortable with the original Unix versions.

Next, I used Excel's "Get External Data" feature to point a worksheet at the CSV file and auto-update every minute.

Now, the tables, graphs and diagrams in the Excel spreadsheet will be able to auto-update throughout the day showing the current data!

This is quite a step up from my original plan of manually connecting a digital meter to various individual sensors every time I wanted to take a measurement.

Here are some photos of the sensors going together:

Sensors in various stages of progress:




Sensors mounted in PEX plugs, almost ready to add to a "Tee" fitting. I used plumber's epoxy (a 2-part epoxy resin that you knead together and form like putty):



A sensor plug crimped onto a short piece of PEX, ready to crimp onto a Tee:



Sensor visible through open end of Tee:

Backtracking on choice of temperature sensors

I made a bad choice on the temperature sensors. When I received the thermocouples and digital meter from the coffee bean roasting place, I was able to read the detailed specs in the manual. Since the operating temperature range is so wide (several hundred degrees), the resolution is VERY low: +/- 4 degrees F! That's not nearly precise enough for this project.

Also, since the thermocouples generate a voltage in the microvolt range which is read by the very sensitive digital meter, any moisture around the sensor or the fabric-insulated wire or connector can throw the reading off by another 5 to 10 degrees!

And finally, now that I've decided to use extensions on the auger drill to get 6' deep, the 3' leads on these sensors won't be long enough to measure soil temperature at the bottom of the well. Extending the wires shifts the calibration by tens of degrees, so that's not an option.

Mind you, there is nothing wrong with these sensors, and they seem to be working within spec. The sensors aren't designed to be very precise or to work in high-humidity environments (let alone buried in wet soil). I probably should have discovered that before ordering the thermometer, but the detailed specs weren't posted on their website and I ordered them in a hurry.

So these are just not the sensors for this project, and now I'll have to move on to the 1-wire digital sensors I mentioned earlier, which, as you'll soon see, will prove to be a huge benefit to the project!

Copper, HDPE, or Pex?

Most geothermal ground loops use HDPE (Polyethylene) tubing with something like 1.5" or 2" diameter. It is extremely durable, with a 50 to 100-year service life.

Direct Exchange (DX) ground loops, which circulate freon from the heat pump (instead of water-based coolant from a heat exchanger) through the ground loop, use copper tubing. When burying copper in the ground, you have to be careful of acidic or "aggressive" soil which can corrode the copper. A pH below 5.5 is a concern, and would need to be dealt with by using a protective coating, a sacrificial anode, or perhaps treating the soil with lime (though I'm not sure that's an "approved" fix).

I've also seen pre-made DIY geothermal kits that use Pex (cross-linked Polyethylene) tubing with 1/2" or 3/4" diameter.

I haven't decided yet which one I should use. Since my wells will be short, I need to dump the heat out of the tubing into the soil quickly, in just a few feet. So it seems I should maximize thermal conductivity with copper tubing. Copper's thermal conductivity is many times that of Polyethylene, and a smaller diameter tubing will expose more warm coolant to the cold walls of the tubing in the short distance available in our wells (the ratio of tubing wall surface area to liquid volume is higher in a smaller-diameter tubing than it is an a larger-diameter pipe).

I could turn down the flow rate to give the water more time to acclimate to the underground temperature, but I've read that I should keep the flow turbulent to provide better heat transfer, so I can't go too low. Also, the slower the fluid, the longer it's exposed to ambient temperatures on its way back to the server cabinet.

I found a nice calculator to determine whether flow will be turbulent or laminar based on the size of pipe and flow rate here: http://www.gcisolutions.com/flow.html

The soil near our building is not very acidic (6.5 pH), so I don't think corrosion will be a problem. I could bend 1/2" or 5/8" copper tubing into a helical coil and get 20 or 30 feet of tubing per well for maximum heat transfer.

I'm just not sure yet if this is all necessary, or if the rate of heat transfer is limited more by the soil (which typically has half the thermal conductivity of polyethylene anyway). Maybe copper tubing would be a waste of time and money.

Drilling deeper

The two 18” extensions for the auger drill arrived today, so the drilling operation is continuing.

The pin that came with the extensions (actually just a plain bolt and nut! Not hardened steel like the pin that came with the auger) is too big to fit the auger drill, so I had to use a smaller bolt. After about 10 minutes of drilling with the first extension installed, the smaller bolt is showing signs of shearing stress, so I decided to stop drilling until I can get a better pin.

As it is, I was already worried for a while that I wouldn’t be able to get the auger back out of the ground because it was getting stuck on the way up, and the gas-powered auger has no reverse (I don't know if more expensive auger drills do, but this is the $139 version from Harbor Freight; pretty sweet for the price). If I sheer off a pin or break an extension, I’ll be digging a ditch by hand to get my parts back out of the well!

By the time I discovered the damaged bolt I had already bottomed out the drill with one extension, so there was definitely some progress. We expect to reach the deep horizon well by tomorrow… wait, that’s a different drilling project.

Since the extensions don’t have flighting (the blade that augers the soil upwards), the extended bit can’t carry the soil out of the hole. But the shop-vac does a great job sucking dirt and rocks out of the hole between drilling sessions, so I don’t think that will be a problem. I just need to be able to get the auger in and out of the hole without getting stuck. And I might need an extension for the shop-vac.

We are at about 53” now.

First tests

While waiting for the auger extensions to arrive, I drilled a test well to 30" depth, installed some K-type thermocouple probes, and back-filled the hole. I don't have much experience with auger drills, and I almost expect a broken wrist by the end of this project!

I picked up this inexpensive digital temperature gauge and a handful of removable thermocouples at www.sweetmarias.com. Evidently, they use K-type thermocouples to monitor the temperature of coffee beans while they're roasting. K-type thermocouples have a very wide temperature range, so they're handy for that. I'm discovering that I probably shouldn't expose the sensors directly to damp soil because it makes the readings a little flaky. They're only supposed to be used up to 80% relative humidity, so I should plan on sealing them in something waterproof I guess. Also, I'll need to extend the 36" leads now that I've ordered extensions for the auger bit!

The K-type thermocouples are convenient and cheap, but since the temperature range is so wide, the accuracy isn't very high (+/- 4 degrees F). I know I should consider 1-wire temperature sensors and a real-time monitoring/logging system, but that's a project, and I don't want to get bogged down in something that might distract me from the task at hand. If there's time...!

After waiting a day for soil temperatures to equalize, I was pleased to discover that the building provides enough shade throughout the day that the ground temperature at a 2-inch depth is about 10 degrees cooler than the ambient air. At 30", it's only in the mid-to-upper 60's in 87 degree weather.




Here is the cabinet that the server will live in. It will be a tight squeeze because the server is pretty deep (front-to-back). If it's as tight as I expect, I'll probably need to duct the hot air directly out the back wall of the cabinet behind the server and pipe it through the fan-coil unit outside the cabinet. I'm more comfortable with the coil outside anyway so we can keep the server completely isolated from the coolant and condensation.






I considered a radiator from a PC water-cooling system, like the 2-fan or 3-fan units from Black Ice. But they are designed to radiate heat, not cool the air, so they aren't designed to direct condensation away from the coils. Instead, I settled on a 9"x9" surplus fan-coil unit.




The cabinet is only 8 cubic feet, so if I can get a couple hundred CFM from my fans, I'll be turning over all the air and passing it through the coils every few seconds.

Rough estimates

The server has five hard drives and a Core2Duo motherboard. I'll measure the actual current draw later, but for now I'm estimating that it draws less than 150 watts (9 watts per hard drive, plus 80 to 100 watts for the motherboard, CPU, and power supply inefficiency).

The server has capacity for five more hard drives, so we should count on another 50 watts for future growth.

The battery backup for the server should be OK outside the climate-controlled cabinet so there won't be any heat gain from that (I know they really shouldn't be run in a hot environment, but I've seen them do fine in worse conditions, so I'll count it out for now).

There might be a 5-port gigabit unmanaged Ethernet switch that I would want to keep in the cabinet with the server. I'm guessing about 10 or 15 watts for that, but again I'll take actual measurements later.

The circulating pump for the coolant will dump its heat into the coolant, so we should count on another 35 watts from that.

So my initial estimate of the internal heat load is about 250W, or 853 BTU/hr.

There will be heat gain from outside the cabinet during hot weather. Using the heat load calculator at www.cabinetcooler.info (manufacturers of a pretty slick compressed air-powered cabinet cooler!) as a rough estimate, our cabinet (uninsulated) will have a 388 BTU/hr external heat load if we keep the inside at 85F and the outside reaches 125F. I'm planning to insulate and seal the cabinet so it is closed off from the outside air and to minimize this heat load.

So as a first estimate of our cooling requirements, we need about 1250 BTU/hr. Plus any loss from the pipes that run from the cabinet to the ground loop. But I haven't thought through how to calculate that yet, so I'm going to pretend it doesn't exist for now. We'll need to re-visit that later.

From what I've read, a "real" vertical ground loop geothermal system gets about 1 ton (12,000 BTU) of cooling per 200 feet of well. That's 60 BTU per foot. I don't really expect this to scale down directly for a micro-geothermal system, but as long as I space the wells far enough apart, I think it should be pretty close. The first foot or two of each well probably won't have the same 60 BTU per foot capacity, so we'll need an extra well or two to make up for the "end-loss" at the top of the well. But I'm also counting on there being some margin of error built into the "1 ton per 200 feet" guideline.

With a 1250 BTU/hr requirement, we would need 20.8 feet of well. My auger drill has a 32" bit, and I think I can get away with two 18" extensions before the bit torques itself into a pretzel. So I should be able to reach 6 feet with some shop-vac action) which would require 3.5 wells. I'll plan on 4 or 5 for now until I fine-tune some of these calculations and do some tests. With the space available to us, we can fit 5 or 6 wells easily.

So the plan is shaping up: One server with 10 hard drives and an Ethernet switch in an insulated steel cabinet, being cooled by a small fan-coil unit circulating coolant through 4 or 5 wells, 4" in diameter and 6' deep. I think it's do-able!

-Tom Rusnock

The project

Ok, here's the plan: We have a server that mirrors our online data backup server for off-site replication. We don't want to keep it in the same building as the main data backup server in case there's a fire (naturally). So we would like to house it in our second building, a 25'x25' outbuilding on a concrete slab, which is not climate controlled.

Cooling with ambient air will not be sufficient, because it can reach 125 degrees (F) in this building on hot days. I also don't want to draw moist outside air into the server. I would like to keep the server at a maximum of 80 to 85 degrees, with no sudden temperature changes (no, servers don't need to be kept at 65 degrees as some people believe. As long as the temperature is steady, they're perfectly happy at 85 degrees. Swing the temperature 20 or 30 degrees too quickly too often, and you'll see frequent failures.

The server will be in a 24"x24"x24" steel cabinet, and it makes sense to try to climate-control *just* the cabinet to minimize cooling costs. I think it will be difficult to cool a very small space with a standard A/C unit without causing huge changes in temperature as the A/C compressor cycles on and off throughout the day.

I thought about Peltier junctions, but they are horribly inefficient and the energy required would cost more than an air conditioner.

I want a solution that is reliable and inexpensive to run, and it would be nice if it was a fun and unique project too. After a little research, I found several people who had done successful "micro-geothermal" projects for cooling computers or small spaces, and I'm really interested in that idea. Instead of running an A/C unit full-time, or purchasing an expensive outdoor server cabinet climate control system, I will attempt to circulate coolant through a fan-coil unit in the server cabinet, then through a ground loop to take advantage of the relatively constant underground temperatures to dissipate the heat generated by the server and associated electronics.

This project is my attempt at cooling our server with a geothermal ground loop. I refer to it as a "micro" geothermal system because it will not involve a deep, professionally-drilled well, or a large horizontal ground loop covering hundreds of square feet of ground and requiring heavy equipment to excavate. We'll only need about 1500 BTU of cooling, so instead I will attempt to use several 4' to 6' deep wells drilled by hand with an auger drill near the foundation outside the building.

I *could* dig a large trench near the building for a small horizontal ground loop, and indeed that is the backup plan if the small wells don't do the trick. But I really like the analogy that a few 6' deep wells will draw to a "real" vertical geothermal system. This design will also be scalable, in that I could drill additional wells to add cooling capacity without disturbing the running system.

So there it is... one of those projects that costs 10 times more in time and about as much in materials as it would cost to do it the "standard" way with an air conditioner. But I like the uniqueness and "greenness" of the project, and I'm confident that this will work. I like the idea of a project I can learn something with, which is also a "green" solution that will produce some real-world data and establish a usable baseline for others to follow in building practical and economical very small-scale geothermal systems.

-Tom Rusnock