Amongst the lore of our family is a story, one of several, of my Uncle Smiley, who stated: “When we were living in Florida, we bought 100 feet of black garden hose, and spread it across the roof of the house. We never spent another penny heating water.”

When I got a look at our electric bill for August – $193 – I realized that there was probably good reason to seek out a new way to make hot water. So I followed the vision of the abundant garden hose.

How might one connect The Hose to one’s tank heater? I wondered. This inquiry led me to the Heliatos bottom feed connector (PDF), which led to an appreciation of the EZ system.

For about $1100, including the tools and accoutrements to be PEX-capable, I  got it up there on the roof. The roof offers southwest and southeast faces, I chose southwest, because our hot water needs tend to occur in the evening, and because it puts the panels over the garage, where a leak would be considerably less worrisome.

I actually received the 4 panel EZ system in September, but then I had two out-of-town trips coming, so doing new plumbing seemed unwise. I hate getting a call about something gone awry at home when I’m away, and when I do, it’s better when I am not the direct cause. An additional advantage to waiting was that my wife was leaving town for two weeks, November/December, and the prospect of having a no-hot-water interval at home during the install was easier to bear with her somewhere else.

I got the bottom feed connector installed the evening of Thanksgiving Day (after enjoying a lovely buffet at the Hale Koa). I had sort-of-dreaded this step. Existing plumbing doesn’t always follow your plans, especially in the humid salt air of Hawaii. Luckily, this water heater is only about 6 years old, and everything’s still pretty shiny. I had let the tank cool for a couple of days, which gets showers down to about skin temperature, to avoid flushing too much money down the drain, before draining the tank. In retrospect, I should have planned to do a proper clean and flush, by removing the bottom heating element and spraying the mud in the bottom of the bottom of the tank, but there’s no reason I can’t still do that.


The installation of the BFC was, in fact a little dicey due to the fact that the only part of the internal nipple assembly visible outside the insulation was the threads on the union, so I had to devise a way of turning the thing to acceptable tightness without hurting the threads. I put several wraps of teflon tape on the boiler end and turned it in, first with a piece of leather between the pipe wrench and the threads, and later just turning gingerly with the wrench teeth on the threads themselves. Luckily the boiler threads were pretty clean and easy to turn into, and also the type of union used in the BFC has a rubber gasket that acts as the water barrier, so nicking the threads on the union isn’t quite the problem it would be on regular tapered pipe threads. Somehow, the BFC fits inside the water heater closet door, without modifications, something I spent time assessing and worrying about.

The interconnection of panels, per instructions, took probably a couple of hours, and I did the better part of half of the work downstairs, by finishing the left side of each connection, leaving only the right side to complete on the roof.  At one point, my skim of the instructions left me with the impression that the connectors on the last end of the last panel should be connected together, but it turns out that they need to be capped. If I had made that mistake, the panels would have been curiously ineffective, and it probably would have been frustrating to diagnose and fix. The design of the panels relies on the water flowing bottom-to-top through the panels, which is a slightly more resistant flow path than a pipe, so adding a pipe on the end would effectively bypass the panels. I think that ideally things should flow bottom-to-top, since the heat-diffusion-based flow of the water in the panel would assist, and I think that if you reversed the pipes, the detrimental effects might not be noticeable.

I thought and read a lot about panel mounting, and I finally ended up just squirting a pillow of roofing sealant into each mounting bracket and then deck-screwing through the sealant, shingle, and roof boards. I was blessed with two days of hard rain immediately following the install, which produced no drips.

One flurry of squirreliness occurred around the installation of the PV panel. They give you the panel, some wire nuts for the pump motor end, and 25 feet of wire. The connection for the panel end, however is neither well thought out, nor easy. There are two solder points to connect the wire, and two screws holding the little metal tabs to the panel body. The connection point is covered by a little plastic slide cover, which allows enough room for, well — nothing. One could try to simply use the screws with crimp-lugs, but the screws are inclined to thread in once, and any removal and re-install makes a not-very-connected connection against the metal of the tabs. I ended up soldering the connectors, and then smooshing a ball of Coax-Seal into the connection box, to exclude water, which seems to have worked. I also ran a bead of roofing sealant along the top of the panel to direct water around the panel when it rains.

Plumbing the panels with PEX turned out to be relatively uneventful. I decided to route both pipes up the same side of the water heater, because that side of the closet is less likely to get stuff stored in it. Having a brass elbow on either side made a convenient isolated temperature sensor point (more below).  I slid the foam insulation onto the pipe end-wise, rather than opening the perforation on the side, in order to keep it from opening on the turns. IMG_20141129_173632_955






There is an Intermatic mechanical timer which turns on the electric to the water heater from 5 PM to 7 PM, the idea being that if the tank is below about 105 degrees F at 5 PM, the electric will boost the temp for evening showers. If the tank temp is above 105, the built-in thermostat will prevent the electric heaters from turning on.

Thermal Sensors

I had already done various things with DS18B20‘s and Raspberry Pi, it’s an easy, reliable way to instrument something for temperature. For the water heater app, I put sensors on the water lines to and from the panels, and on the wall of the tank, which can be accessed through the heating element access panels on the water heater. YOU SHOULD NOT WORK INSIDE THOSE PANELS BECAUSE YOU WILL LIKELY GET YOURSELF KILLED. Take that advice seriously. You cannot work inside the access panels on the water heater without contacting the electrical connections. Unless you understand completely what I am saying, and how to remedy the situation, don’t screw around inside your water heater. If you’ve got a gas heater, you are on your own.

IMG_20141202_205025_950I used more Coax-Seal to stick the sensors to the tank, and electrical tape to hold them on the first brass elbow away from the bottom feed connector. I originally put two sensors on the outer extents of the hot and cold sides of the BFC, but there was insufficient thermal isolation, and considerable thermal inertia in that big piece of metal. Sensors on the BFC gave the impression that the cold side was warmer than the hot side, whereas the first elbow on either side, set apart from the BFC by about 6 inches of PEX, shows that the hot return from the panels is always warmer than the cold to the panels. This is a great source of comfort. Another probable advantage of measuring at the elbow is that it may benefit from some turbulence on the turn, which breaks up laminar flow and gets a better thermal coupling to the temperature of the water in the pipe.

After doing a project with DS18B20’s and CAT5 networking cable, I shifted to 18 gauge thermostat cable, which I got at Lowe’s in a 500 foot roll for about $80. I placed each sensor on the uncut cable by middle-stripping it as shown in the pic, and then hooked the ends of the sensor’s 3 legs to wrap and solder onto the cable. I insulated the result by threading electrical tape between the conductors, but I think that when I plan ahead more, I will use epoxy on each sensor. I tend to work things so that I can put the flat side of the TO-92 case on the thing to be measured, although any side is probably as good.


A sample graph above — you can see that the lower tank, somewhat counter-intuitively, stays warmer than the upper tank, after the day’s solar heating, and the first shower (~21:00) causes a drop in the lower tank, because that’s where the cold water supply enters the tank. The apparent temperature inversion could be in some part due to laminar flow of buoyant warm water inside the tank, Of course, the hot output is drawn off the top of the tank. The second shower, at about 01:30, is considerably more voluminous (not to mention names), and causes a precipitous drop in lower tank sensor temperature of 18 degrees F,  while the top sensor only declines about a degree or two.  The above shows two partly-sunny days in a row, without any really aggressive solar heating.

The Raspberry Pi reads the sensors once per minute, during nominal daylight hours, and once per 5 minutes when it’s dark. The graph shows some interesting changes, probably due to thermal diffusion, which leads to siphoning, etc. Heat loss from the tank seems to be in the range of 2 to 3 degrees F over 6 hours, which of course varies proportionally to the differential between inside and outside.

The solar loop began operation in early December, which is of course, in the northern hemisphere, probably the least effective time in terms of sun angle on the panels, and also among the least sunny times of year in Hawaii. The sunniest day so far brought the upper tank to about 115 deg. F. It’s likely that by March the yield will be significantly higher. A contigency plan is to add a thermostatic mixer valve on the water heater output, to regulate the scalding potential on the output. Solar water heating in Hawaii can sometimes be too effective…



Comments are closed.