Automatic Refill System for Marine Aquaria

Reliability is something that I do not hear discussed very much by aquarists. Yet at the same time I hear stories of reefs crashing (both biologically and physically) from failures in simple systems. In particular, I have heard accounts of people's automatic replenishment systems failing such that replacement water is added to the point of overflow! Besides being a major mess to clean up, it could turn your salt water aquarium into a fresh water aquarium if left unchecked.

My replenishment system has been entirely manual up until recently. As water evaporated, I replaced it with kalkwasser-laced fresh water that had been prepared before hand. This was a daily chore. My system currently loses about 800 ml per day due to evaporation. This varies as a function of temperature, humidity, and whether I leave the sump and show tank lids on or off (I've found that temperature is moderated somewhat by allowing evaporation to occur. The process of evaporation actually pulls energy out of the system in the form of heat).

So, how should one automate the task of adding fresh replacement water? One way that I have seen is to have a big container of prepared kalkwasser that can drip into the tank on demand. The demand is regulated by a float valve. Simple, but I don't trust it. The float valve is a single point failure waiting to happen. Should some foreign matter stick in the valve, it would drain the entire vat of replacement water into your tank. If kalkwasser is used, calcium deposits will build up on the valve and eventually cause it not to seal correctly unless it is inspected and cleaned regularly (white vinegar works well to dissolve calcium deposits).

What about an electric valve? The solenoid action of such a valve is powerful and positive (more so than a float valve), but it must be actuated by a water level sensor of some sort­ probably a float switch. This still involves single point failure potential, both in the switch which could get "stuck on" or "stuck off" and in the solenoid valve which could get "stuck open" or "stuck closed".

There are many conceivable combinations of gravity-fed replenishment systems using switches and valves. Alternately, switch-operated pumps which actively force water from a reservoir into the aquarium are possible replenishment mechanisms. The failure modes differ. In the first case, the gravity feed could drain the entire reservoir into the tank were the valves to fail "open". If instead, the switches failed "closed", your circulation pumps would eventually run dry. On the other hand, a switch actuating a replenishment pump could fail "on", draining the reservoir into your tank, or it could fail "off" thereby letting your circulation pumps eventually run dry.

The mode of system failure then becomes a function of whether your replacement system fails "on" or "off". If you prefer to loose your coral and fish to oxygen deprivation, or boiling (as your house burns to the ground due to an overheating pump that finally seizes up and bursts into flame from having run dry for several days), then you want a system that fails to actuate. If on the other hand, you prefer to annihilate your coral and fish with a rapid reduction in salinity, then you will want a system that fails to turn off.

One might conclude that you just can't win. For those who ascribe to the philosophy that "if at first you don't succeed, then skydiving is not for you," perhaps neither are marine aquaria. But for the Type-A personalities that cherish the challenge of creating a reliable automatic water replenishment system, there is hope. This hope comes in the form of "quadded logic and probability theory."

Certainly one should choose the most reliable equipment possible (or affordable), but even the best, brand new items suffer from "infant mortality." That is, new stuff can still break. On the other end of the scale, things quit due to wear or "old age." But let us consider the design of a reliable system using the best components that we can afford. My design is shown on the next page.

First, two reservoirs are involved. The upper one is a temporary holding tank for water coming straight from the source. In my case, this is well water from 200 down in granite rock. When this is filled to capacity, it holds 10 gallons. The filling of this tank could be automated, but since it need only be done infrequently, I have chosen to do this task manually.

The water from the upper holding tank is allowed to flow out under force of gravity through quarter inch tubing into a slightly larger reservoir (11 gallons). This assures that the lower reservoir can never overflow.

Before the water can enter the lower reservoir, it must pass through two deionizing (DI) resin filters. The first gets any really nasty things out of the untreated well water, and the second "polishes" the water. Eventually, with only the force of gravity acting upon the water, all of that contained in the upper reservoir drains into the lower reservoir, getting purified by the filters as it goes.

The output of the lower reservoir is blocked by a series of four 120 VAC, 60 Hz normally closed solenoid valves. If the power were to fail, these valves would remain closed and prevent the lower reservoir from draining into the aquarium sump. These solenoid valves are in turn activated by a series of four normally open float switches contained in a wave-shielding baffle that prevents them from activating momentarily with surface disturbances.

A key observation is that fact that the solenoid valves are organized in two parallel paths comprised of two serially-connected valves. The switches are also connected in a similar fashion.

Assuming that all of the valves are of equal reliability, and all of the switches are of equal reliability, this arrangement will result in a system that will continue to operate even in the presence of a failure in a single switch or valve. In addition, that failure can be of either the "stuck on/open" type, or a "stuck off/closed" type. Merely placing multiple switches or valves in parallel would not afford this level of reliability. For example having four valves in parallel is no more reliable (in fact its less reliable) than a single valve when considering the "stuck on/open" case since the failing valve would fill your aquarium with fresh water even if the other parallel valves continued to function correctly.

The architecture of the valves and switches shown in the adjacent figure represents "quadded logic." Quadded logic is a redundant logical structure which is protected against any type of single faults and many multiples ones. It is provided with four versions of each input, and every logical valve state appears in quadruplicate.

In this case, the switch logic may be represented as (A·B) + (C·D) which is read as "A and B, or C and D". What this means is that both switches A and B must be working for solenoid valves A and B to be energized. If either switch A or B fails to close, then water will not be allowed to flow through the A/B valve path. By the same token, if either valve A or B fails to open, no water can flow.

On the other hand, if either switch A or B fails to reopen once closed, then its companion switch will assure that valves A and B do not remain open. And if either valve A or B fails to close once opened, its companion valve will shut off the water flow.

What is true for the A/B water path is also true for the C/D water path. Therefore each path is twice as reliable as a single float switch and valve. Placing both paths in parallel (A/B in parallel with C/D) increases the reliability of the system by a factor of four. Another way to put this is that the probability of a failure in this system is only 25% of that for a single valve and float switch even though there are more parts that could conceivably fail.

Now lets consider the factors involved when physically creating a highly reliable evaporation water replenishment system that uses a number of series-parallel connected switches and valves to form a "quadded logic" network immune to multiple sensor and valve failures. Remember, in spite of the fact that there are more components to fail, the methods described above in networking the components, actually results in a system of higher reliability than a single sensor-valve pair.

To recap the basic system configuration (see Figure 1) described thus far, a reservoir is filled with untreated water (well water in this case) which is allowed to drain through two deionization filters connected in series. The output of the deionization filters in turn drains by gravity into a holding reservoir. The use of series deionization filters allows the filters to be swapped regularly without migration of chemical species from a spent deionization cartridge ever making past a fresher cartridge and into the holding reservoir.

The output of the holding reservoir is channeled through a redundant bank of series-parallel food-grade stainless steel and delrin electric valves where it is finally allowed to enter the aquarium skimmer sump. In practice, the reservoirs are constructed from plastic drinking water containers. Two five-gallon (nominal) containers (totalling just over 10 gallons) are plumbed in parallel to accept the incoming water from the well. These are currently filled manually (requires about 45 seconds) on a 10-day schedule, thereby implying that the losses due to water evaporation system wide, and that due to skimmer expulsion amount to approximately one gallon per day.

Evaporation is minimized in both sumps by surface-air contact reduction. In the Jaubert sump, a glass cover prevents evaporation while in the "open" skimmer sump, the water surface contact area amounts to only a few square centimeters through the use of several thousand floating polyethylene spheres. The spheres allow water samples to be taken, and chemicals added without allowing air-water surface contact. The Show tank also has a set of acrylic lids which can be used to limit evaporation further, however in practice, these are rarely left in place so as not to occlude light penetration by salt spray that inevitably coats the underside of these covers. The oxygen level in this system stays close to saturation due to the down-draft skimmer, the air mixing Bernoulli vents in the show tank, and the approximately ten-foot churning water fall from the show tank to the basement detritus filter and on to the Jaubert sump­ these in addition to the eight square feet of show tank surface area exposed to the open air.

The upper reservoir is located directly above the holding reservoir as shown in Figure 2 below. The holding reservoir is likewise a pair of five-gallon plastic drinking water containers, so there can never be an overflow situation in the lower containers. The two sets of reservoir containers are separated by the deionization filters. Upon filling the upper reservoir, water automatically drains under the force of gravity through the deionization filters into the lower holding reservoirs (which are themselves plumbed in parallel) over a period of hours.

A bank of four micro switches attached to the skimmer sump are connected to a set of four series-parallel electric valves as shown in Figure 1. Four independent floats sense water height. These floats are isolated from wave action in the sump (of which there is little) by placing the switch-float assembly shown in Figure 3, in a small sump compartment on the side of the skimmer sump. This compartment has subsurface flow holes leading directly into the skimmer sump.

As water levels decrease for any reason, the floats begin to descend and one-by-one the sensor switches are activated. Ideally, all floats would follow the water level perfectly and would actuate all four sensor switches simultaneously, but due to friction and differences in switch position calibration, each micro switch will actuate at a slightly different level and time. These differences, though noticeable in terms of which switches are activated at any given time, actually represent only millimeters in water level difference.

Remember, the purpose of the redundant sensor switches and floats is to prevent failures in either from precluding the addition of fresh evaporation-replacement water to the system, or conversely from stopping the flow when a sufficient amount of fresh water has been added. When the right combination of switches is finally activated, the electric valves allow water to enter the system. As the water enters, the floats begin to rise. One-by-one the micro switches are closed as the water raises the floats. When the right combination of sensor switches is activated, the water flow from the reservoir is terminated.

Figure 4 shows a close-up of the micro switches with the floats in various (exaggerated) positions. Note that each micro switch has a paddle that is pushed up by a high molecular density (HMD) plastic knob at the end of a carbon composite rod. The HMD tip is very slick and slides easily without binding on switch paddles. The paddles are a glass-fiber reinforced epoxy resin.

Figure 5 shows a single float mechanism removed from the switch-float assembly. The carbon composite rod is stiff and will not bend or bind in the acrylic channel. A rubber stop is shown at the base of the rod. This stop is adjustable to set the maximum height to which the float can extend upward against the switch paddle. Not shown is a second rubber stop that prevents the float from falling out of the switch-float assembly were the sump to be drained. It too is adjustable. As series of three independent sealed acrylic float chambers gives buoyancy to the float mechanism. Were any one to flood, the other two would allow the float to continue to work. Each of the four float mechanisms is identical.

The valves are housed in an electronics box (Figure 6 above) which has two circuit breakers and a fuse to protect various segments of the circuitry inside. The main circuit breaker energizes the valves and sensor switches. When this is off, no reservoir water can flow. Similarly, in the event of a power outage, the valves will automatically close thereby preventing too much fresh water to enter the aquarium.

A small muffin fan has been included to keep the valves cool since it is conceivable that one or two valves could be energized almost continuously in the event that the sensor switches were misaligned or if one should fail "open". The second circuit breaker controls the fan power.

A series of four square indicator lights shows which valves are activated at any given time. If the two upper lights are both on, reservoir water will flow. Similarly, if the two lower lights are on, reservoir water will flow. Any other combination (other than all four being on) will not result in water flow.

A pair of round red and green indicator lights shows the mode of the valve box. If power is present and the system is energized to sense water level, the green light will be on. A second mode of operation is controlled by a toggle switch above these lights. Under certain circumstances it could desirable to manually force water to flow from the reservoirs regardless of the skimmer sump level. Such an occurrence might be warranted if reservoir water is to be diverted into some other container, for example when new salt water is being prepared prior to a water change. Activating the toggle switch forces all valves to open (independent of sensor switch position). All four of the square valve lights come on and the round green indicator is extinguished. The round red light comes on to indicate the manual override mode. Water will flow from the reservoir for as long as the toggle switch is activated and water remains in the lower holding reservoir.

The reservoirs are housed on metal shelving adjacent to the skimmer sump in the "aquarium command center" located in the basement. To adequately support the weight of these reservoirs when full, metal strapping has been added as a support from the middle of the top shelf up to the floor joists above. A threaded steel rod positively connects the second shelf to the top shelf, thereby transferring its static load to the floor joists as well. The fill hose from the well is large diameter tygon tubing with nylon couplings. This connects the upper reservoirs to the house plumbing via a manual gate valve.

One other float sensor is noteworthy. A polyethylene toilet float has been attached to a delrin shaft that passes through a sleeve to activate a long-paddled micro switch which can apply voltage from a 9 volt radio battery to an oscillating Sonalert as shown in Figure 7. This entire assembly is mounted on a PVC pipe that is in turn attached to an acrylic sheet. As configured, this can be clamped to any container to sense an impending overflow condition. When filling buckets from slow pumps or gravity feed systems, it is easy to become distracted by other matters only to find a growing flood on the floor. Clamping this overfill alert to the container gives adequate warning that the container is full.

When not in use on filling containers, the device is left clamped to the skimmer sump. In the event of a power failure or the tripping of a circuit breaker after a nearby lightning strike during a thunder storm, the water level will begin to rise in the skimmer sump as the water in the lines to the show tank drains back into the sump when the circulation pump quits. Actually this will happen any time the main circulation pump quits, as might happen if the impeller jammed or sheared. The pump could still be running, but no water would be pumped, so the pressure head could not be maintained and the water suspended in the pipes leading upstairs to the show tank would flow backwards into the skimmer sump.

Having the main circulation pump quit for any reason will not result in an overflow because the sumps are all sized to accommodate the volume of water suspended in the pipes by the circulation pump, however it is a good thing to quickly know that there is such a problem. Having the overfill sensor clamped to the skimmer sump results in an audible "beeping" that is loud enough to be heard upstairs whenever circulation stops for more than about 10 seconds. Since the unit is battery operated, warning is given even when power has been interrupted to the aquarium circuit.

Return to Michelson Aquarium.