Latest Thoughts on Skimmer Design

by Chris Paris for Reef Aquarium Information Depot
Updated November 29, 1998

I am involved in designing a new skimmer to replace my small downdraft skimmer. Right now, this page describes many untested ideas and incomplete and inconclusive tests. After I build a skimmer that I like, I plan to incorporate those ideas herein that still seem sound into a document called Designing a High-efficiency Skimmer. I apologize in advance for the preliminary nature of the ideas on this page. I hope that you find it useful anyway.

What good is a better skimmer?

We already have good skimmers, including the ETS from AETech and the HSA from Marine Technical Concepts. Advanced reefkeepers like Sanjay Joshi are maintaining beautiful reef aquariums using ETS or HSA skimmers, or homemade skimmers based on them. I cannot argue with success like that, but the thought that we have reached the end of the path in skimmer development is not satisfying to a tinkerer like me.

The water on natural coral reefs is extremely low in dissolved nutrients, and relatively rich in particulate food sources. I will not try to prove it here, but I believe that even great reef aquariums have the ratio between dissolved nutrients and particulate food sources backwards. We have water that is dirtier than natural reefs, with far less particulate food available.

People look at Sanjay's tank (at right) and others like it and say, "Hey, we don't need cleaner water." In fact, one of the current fads in the hobby is to run skimmerless tanks, because "Skimmers starve the reef inhabitants." People who remove their skimmers are noticing a new abundance of certain life forms, especially filter feeding types such as feather dusters. I would expect that change if you shut off your skimmer without changing your feeding rate; skimmers certainly remove particles that are potential food, so removing skimming without changing the particulate input will increase the particle count in the water.

If more particulate food for the filter feeders is what you want--and that is my goal too--then why don't you just feed more particulate food? You don't do that because, traditionally, feeding more reduces water quality (increases dissolved nutrient levels). I think that there is a natural chemical or biological force that drives aquariums (and natural waters) from states of high particulates to states of lower particulates and higher dissolved nutrients. Without a matching opposite force, our aquariums will tend towards having eutrophic water without much particulate food in it. We need a force to stand opposite to the natural trend toward eutrophication. That force is a strong skimmer.

Ron Shimek has presented information from a paper in the scientific literature that showed that a particular natural reef receives about 10 ounces of wet food per 100 gallons per day. The reef aquarists who feed the most of all reef aquarists feed about 1/10 of that amount.

I am curious about how our tanks would do with cleaner water than they have now, but my primary motivation is to be able to feed a lot more while maintaining the same water quality that we have today in our best captive reefs. I want to use a strong skimmer as a driving force that enables me to maintain the dissolved nutrient level no higher than it is in our best captive reefs today, while maintaining a more natural particulate food level.

What animals are unable to live in our tanks because of the factor-of-ten food shortage that Ron Shimek suggests that we have? As lush as Sanjay's reef is, would it be even lusher if he could feed ten times as much food while maintaining the same level of dissolved nutrients? Would we be able to keep any interesting carnivorous fish that are currently considered not reef safe because their supply of tasty food is so small that they are forced to pick at the relatively unpalatable (to certain fish) corals? I don't have answers for any of these questions, but these are some of the questions that motivate me to explore more powerful skimmers.

Better skimmers enable us to feed more

Since well before I got involved in the reef aquarium hobby, authors of hobbyist books and internet postings have generally believed that larger aquariums need larger skimmers than smaller aquariums need in order to maintain the same water quality in tanks of both sizes. These authors based their opinions on an assumption or understanding that larger aquariums would have larger "bioloads" (I think that they really meant larger nutrient input rates.). It turns out that this belief is correct according to at least one theoretical discussion of foam fractionation (Chen94a).

However, at about the time that the ETS skimmer came out and became an instant hit, a few clever hobbyists started questioning the common belief. Anthony Tse, an advanced reefkeeper, explained an alternate position in a USENET article back in 1995. The argument for the position is fairly intuitive, and I held the position myself for a while. The argument goes as follows. Each skimmer has associated with it a certain nutrient level at which it effectively shuts down, which is to say that it no longer can remove "gunk" from the water. Any skimmer will be able to remove "gunk" at the rate at which you add it to the aquarium (in the form of food), but a more powerful skimmer will maintain a lower "gunk" level than the weaker skimmer, for the same food input. This particular statement is true. The final point of the alternate position is that all this implies that you should choose a skimmer size independent of your tank size (really, nutrient input rate), and based only on how low a "gunk" level you want to maintain. This last point does not follow from the previous points, and it is false.

One fault in this argument is that skimmers are not binary devices that go from a state of operating well to the state of being shut down. The rate of "gunk" removal of a skimmer varies linearly with the "gunk" level; skimmers gradually slow down as the water gets cleaner. In the equilibrium state, a skimmer removes nutrients from the water at exactly the rate at which they enter the water. When a weak skimmer is in its equilibrium state, a powerful skimmer is still removing "gunk" faster than the aquarist is adding it, so the water is getting cleaner. When a powerful skimmer is in its equilibrium state, a weaker skimmer cannot keep up with the nutrient input rate, so the water gets dirtier, until it is dirty enough that the weaker skimmer exactly keeps up with the input rate. Chen's paper contains an equation that captures the relationship that I have tried to explain in this paragraph. Unless I've made a mistake, you can read these sentences right out of the equation

\begin{displaymath}{{d\,C^d_0}\over{d\,t}} = -1.59(C^d_0 -

where Maybe I will walk through the equation slower in a revision of this web page, but probably no one cares anyway.

Skimmer efficiency

Hobbyists like to use the term "efficiency" when they compare skimmers. We should be careful to define what we mean by this term. I think of skimmer efficiency as being one of two things: We could define other measures of efficiency, but I will limit this discussion to these two measures. Both of these measures of efficiency are useful, and we always have to specify which we mean, unless it is obvious from context. They mean different things. The first measure indicates how well we are using our electricity. The second measure determines whether a given skimmer can keep the water clean enough, given the amount of feeding we want to do and the "gunk" level that we consider acceptable. A good but small airstone skimmer is probably very efficient by the first measure, but not by the second measure. A powerful downdraft skimmer that is tuned poorly may still remove surfactants quickly, so we could consider it efficient by the second measure, but it is inefficient by the first measure.

When we buy or design a skimmer, we make a choice (hopefully) about what levels of efficiency we want. A skimmer's rate of removal per unit of time is limited by the amount of energy we throw at it. A well-designed 50 horsepower industrial skimmer is going to remove stuff faster than the best skimmer for a home aquarium. That's fine; we don't believe that the skimming problem is worth the cost of 50 horsepower, so we accept lower efficiency in this sense of the term. In contrast to that sense of efficiency, we generally want our skimmer to make the best possible use of the energy that we give it, whereas the big industrial skimmer user may not care about the electrical efficiency of a measly 50-horsepower device. The main limit to the efficiency in terms of energy use is how much money and time we are willing to invest in the design and manufacture of the skimmer. We should use our available time and money to design and build the skimmer that performs as best as possible for for the amount of energy we are willing to put into it. Now let us get to making concrete choices about the skimmer that we want.

Different types of skimmers

I have some comments about the potential efficiencies of the different types of commonly-used skimmers: The skimmer that I intend to develop will have the basic form of the ETS and HSA. That is, there will be a relatively shallow sump of water at the base. The water and air mixture will enter the box through one or more tall tubes, with one or more air injectors at the tops of the tubes. I will do a series of experiments with a skimmer that uses a single Beckett injector. I will rely mostly on an air flow meter to feel my way around the design space. Once I have explored the design space with a single Beckett injector, I will substitute my 1" Mazzei venturi injector for the Beckett, and do the same series of experiments, to optimize the use of the venturi injector. If the 1" Mazzei compares favorably, I may contact the Mazzei company and try to talk them into loaning larger injectors, such as their 1.5" and 2" models. I also have a third injector to try: the eductor from Aquatic Ecosystems.

Parameters that determine surfactant removal rate

Referring back to that scary equation, we can identify some parameters that determine the surfactant removal rate.

Four steps in designing a skimmer

I suggest designing a skimmer by following these steps. First, decide how much power you're willing to throw at the skimming problem. Don't skimp! I know reefkeepers who are willing to burn 1200 watts on lighting, but freak out at the thought of using a skimmer pump that uses more than 200 watts. Is lighting six times more important than water quality? Maybe it is. Maybe going from a 200 watt pump to a 300 watt pump is so far out on the curve of diminishing returns that you won't notice a difference. That is probably true if you don't also start feeding more after you install the more powerful skimmer. But if your goal is to feed ten times as much as you do now, don't expect to be able to do it with a pump and skimmer that are not substantially different from what you have now. I'm not saying that you need a 1200 watt skimmer pump. I'm saying that you should consider your needs based on your goals, rather than based on what you already have, or based on your preconceptions about what is an appropriate amount of energy to put toward skimming.

Second, choose a pump. The pump that I'm using to design my new skimmer is a GRI 520, which pumps 1500 gallons per hour at 5 feet of head pressure. This GRI is a commonly-used pump for the medium to large downdraft skimmers. I might find it sufficient, or I might want to try a larger pump in the future, but only if I convince myself that the GRI is limiting me, instead of my skimmer design limiting me. I think that one large pump is better than several smaller pumps, because of the principle that larger machines are generally more efficient than smaller ones.

Third, figure out how to maximize air-water surface area. Don't be too concerned yet with producing good foam or collecting the foam. We can design the air injection system separately from the foam concentration and collection system. We can measure air flow easily, but it's harder to measure air-water surface area. The best technique I can think of is to notice how easily the bubbles form collectible foam. So we do have to think about producing good foam, but don't make the common mistake of getting excited about seeing a small rate of production of beautiful-looking foam. In a later section of this document, I describe my attempts to maximize air flow.

Finally, after you think that you know how to maximize the air-water surface area, figure out how to make good foam and how to collect it fast. I don't believe that building up several feet of dry foam is a good thing to strive for, unless doing that is the quickest way to get the foam out of the skimmer, and I don't think that it is. I have noticed that, when I allow the foam to build up over a wide area, for example by lowering the water level in my skimmer to several inches below the top of the skimmer box, the wet foam becomes dry foam without rising very much. Within the space of only a few vertical inches, the foam transforms from runny water to foam that is still white, but has a fine-pored spongy look that tells me that it would be reasonable to collect that foam. This observation leads me to an idea that I hope will allow a skimmer to export dry foam with less than a foot of height between the wet water and the top of the foam riser tube. I can do this for any air throughput within certain limits, but wide limits. My new battle cry is "get it dry and get it out!"

Foam riser design

My idea is to use a short foam riser tube with a taper in it. That is, the riser is wider at its base than at its top. I have considered two different tapered foam riser designs.

In one design, the riser is in two sections, as is conventional. Only the lower section is tapered. Because I do not know where to buy tapered round acrylic tubing, I would make the tapered riser section from flat acrylic sheet. The riser would be a four-sided pyramid. The base of the riser would be 12x12". The top of the tapered section of the riser would be 3x3". The height of the tapered section would be only 3". On top of the tapered section, I would attach an untapered round acrylic tube of 2.75" inside diameter. The round tube should not be long, although I do not yet know what length it should be. I would begin with it long, and cut it down after I see how much height is required to ensure that the foam coming out the top of it is reasonably dry. I'm hoping that I can have it no taller than 6".

In the second design, the riser is a single part: an inverted large funnel. Suppiers such as US Plastic Corp. and Cole-Parmer list funnels with a maximum diameter as large as 21", as well as funnels with diameters in the 6" and 12" range. The spout of a funnel is very narrow compared to the foam risers that we are used to seeing on large aquarium skimmers. I believe that the narrow top will help eject dry foam. As the diameter of the funnel decreases, the air velocity increases. At the spout of the funnel, the air forms a veritable jet of air. As foam enters the high velocity region, the foam is ejected forcefully, as if by a canon. I have not yet tried this funnel idea on my experimental skimmer, but I am using a smaller funnel as the second stage foam riser on my small downdraft skimmer. That skimmer is now performing better than it ever has before.

The user can adjust a gate valve on the output of the skimmer in order to change the level of water in the skimmer. Because of the tapered riser, changing the water level in the skimmer will change the surface area over which the air-water mixture will initially begin separating into water and dry foam. The area that is best will be different for every pump and airflow combination. The taper allows the user to achieve the ideal area for a wide combination of pumps and air flows. The shallow shape of the tapered riser allows users of smaller pumps and air flows to achieve the ideal area without raising the water level in the skimmer excessively. Raising the water level puts backpressure on the air flow and reduces it. Raising the water level only a few inches will probably not decrease the air flow a lot. Compare this to the practice of raising the water level in conventional downdraft (or HSA) skimmers by many inches, in order to use the skimmers with undersized pumps. I have to do this when my Quiet One operates my downdraft skimmer. My air flow meter shows a dramatic decrease in air flow as I increase the water level in the skimmer.

In the following text, I will describe some of my observations that led me to the short and tapered riser idea. I have been watching bubbles turn into foam for a while now, and I have noticed something interesting. For a given air flow and skimmer, the behavior of the bubbles as they enter the foam riser tube falls into one of three categories:

Collection cup design

I think that some, and perhaps most, skimmers are incapable of operating as I hope, even with the best water flow, air flow, and bubble size. I believe that my small downdraft skimmer's collection cup design is so bad that the skimmer is not capable of performing well. Accounts of other people's downdraft skimmers suggest to me that their skimmers have similar problems, although not to the incapacitating degree that my skimmer has them. When I use my GRI 520 pump and restrict the air input with the valve that is part of my flow meter, I can achieve any reasonable air flow for this skimmer, and beyond. I can admit less air than what the Quiet One produces, which is not enough, or so much air that the skimmer quickly pumps water out through the top of the foam riser column. I don't mean wet foam. I mean water, and that is with the skimmer's 2" gate valve fully open. Most frustrating is that I can find no intermediate rate of air flow that results in a good flow of dry foam. Even when I choose an air flow rate that gets dry foam to the top of the foam riser, the foam has trouble spilling over into the collection cup.

The problem is that my collection cup is too small relative to my foam riser tube. The outside diameter of the foam riser tube is 4.5". The inside diameter of the collection cup is 6". That means that there is only 0.75" clearance between the two tubes. The lid of the collection cup is only about 1" above the top of the foam riser tube. The result is that, for all but wet foam, the foam simply gets blocked by the lid and the walls of the collection cup.

Why did I build such a small collection cup? Because I was cheap! Large diameter acrylic is very expensive, and I found a good deal on 6" inside diameter acrylic. I suspect that the skimmer manufacturers are being cheap with the collection cup, too, although not as cheap as I was.

The collection cup should be of sufficient size so that dry foam can spill over the top of the foam riser tube without touching either the sides or the top of the collection cup. Dry foam holds its shape well, rather like sausage. Therefore, I will suggest the rule of thumb (rule of scum?) that the clearance between the collection cup and the foam riser should be no less than the inside diameter of the foam riser tube. This rule implies that we should be using huge collection cups, compared to what most of us use now. Large diameter acrylic tubing is expensive, but remember that you can make a square collection cup from flat acrylic sheets.

My idea to use an inverted funnel as a foam riser came after I formulated my ideas about collection cup design, and after the initial writeup of this section. Ejecting the foam out of a funnel spout may allow you to use a smaller collection cup, for two reasons. First, a narrow funnel spout allows you to meet my rule of thumb about collection cup size with a much smaller collection cup. Second, the forceful ejection of foam from the funnel spout should push any obstructing foam out of the way, so it seems less important to have a clear path for the foam.

Maximizing air flow

So far I have done a few experiments to try to maximize the air flow that I can get from a pump. I use a Dwyer model RMC-104-SSV variable-area flow meter, with a metering valve, to make my air flow measurements. The meter I have measures 40-400 SCFH. I have another flow meter, which is smaller, that I used to take the measurements that I'm going to report on right now. I mentioned the Dwyer meter because that is the meter that I will use to tune a skimmer to the GRI pump.

The pump was a Quiet One. I performed the tests on my experimental skimmer (not my small downdraft skimmer about which I have a web page). The test skimmer is in some ways a cross between an ETS and a HSA. My skimmer looks like an ETS except for the following: there are no bioballs in my skimmer; the injector tube of my skimmer does not extend lower than the top of the box at the base of the skimmer, whereas in the ETS, the injector tube extends to the bottom of the box, and there are holes in the ETS's injector tube, within the box, to allow the water to escape into the box; I use a Beckett foam head to introduce the air, as on the HSA skimmer; there is no baffle in my skimmer, because I have found that a baffle is probably not necessary, and a tiny bit of bubble loss out to the aquarium sump does not bother me, because this is an experimental skimmer, rather than something that I plan to use for the long term.

My Beckett injector is in a sealed chamber. This is important. When the Beckett is sealed up, rather than simply hung from the top of an ETS-style injector tube, all air that enters the Beckett's air input holes must travel through the entire skimmer. If the Beckett is hanging in open air, it is possible for air that the Beckett sucks through to separate from the water flow and travel up the injection tube, and go through the intake holes in the Beckett, without passing through the the skimmer, and thus without contributing to the foaming process. In the worse case, which does not occur in practice, the Beckett would be sucking air at a high rate, but no air would actually pass through the skimmer. Sealing the Beckett makes a measurable difference in practice. In my Beckett-converted downdraft skimmer, I measured a 50 percent increase in air flow when I added a seal between the bottom of the Beckett and the wall of the injection tube.

One of the design parameters of this type of skimmer is the length of the injection tube. You can mount the Beckett injector directly on top of the skimmer box, which amounts to having a zero-length injection tube, or you can have a length of rigid PVC or acrylic tubing between the Beckett and the skimmer box. I found that the length of the injector tube significantly affects the amount of air that the skimmer draws. To my surprise, increasing the injector tube length increased the air flow.

Before doing the experiment, I expected that increasing the length of the injector tube would decrease the air flow. When the injector tube is longer, the pump has to generate more pressure to maintain the same flow. Centrifugal pumps don't work that way. The Quiet One in particular is known for having poor performance at high head pressures.

I witnessed increased air flow with injector column height increases only when the injector column was filled to the top with water. If you use a large diameter injector column, then the column will be filled mostly with air, and the output of the Beckett injector will simply fall through the air as a waterfall. If you use a narrow injector column, such as the 1" diameter pipe that the HSA uses, then the column will always be filled with water (actually with the uniform air-water mixture that the Beckett produces). The diameters of tubing used for ETS-style injection tubes are in a borderline range, where by slightly varying certain conditions, the injection column may be either filled with air and a waterfall, or filled with water. In my experiment, I found that I could change the injector column between filled with air with a waterfall and filled with a uniform air-water mixture. I accomplished this transition by very slightly changing the water level in the skimmer, using the gate valve on the skimmer's output. Some people describe the filled condition as "backed up." Some authors have written that ETS skimmers perform better when the injector tubes are not backed up. This claim seems suspicious to me now, because my results showed an increase in air flow flow when I transitioned the injection tube from unfilled to filled. The increase was more dramatic with the longer injection columns. Despite the increased in air flow after transitioning to the filled state, the entrance of the air-water mixture into the skimmer box looked more gentle and laminar than when the injector column had a waterfall in it. Judging visually, I guessed that there was less air flow with the injection column filled. Perhaps this illusion confused those authors who reported "better performance" when the ETS injectors are not backed up.

Let us try to make sense of my test results. I said that the air flow increased with increasing injector column heights. I did not measure the water flow. The pump pushes the water up to the top of the injection column, but then the injection column brings the water back down to the base of the skimmer. Gravity creates a suction in the injection tube to pull the water down. For each increase in injection column height, the pump does more work to lift the water to the top of the injection column, but gravity does work in return by pulling the water down the injection column. (I know that this explanation is abusing the proper terms of physics. Someone help me here.) As the injector column increases in height, the only extra effort that the pump has to do is to overcome a little bit more pipe friction. That requires a minor increase in effort. If this explanation is correct, then the water flow rate should not change significantly with increasing injection column heights. But this explanation is an oversimplification. The fluid moving down the injection tube is an air-water mixture. Its density is less than that of the water that the pump is lifting. Therefore I would expect that the suction on the downside does not fully compensate for the height on the upside. But the velocity of the water on the downside is higher than the velocity on the upside. Does that make up for the density difference? Now look at it in terms of mass flow. The mass flow on the downside must be higher than that on the upside. On the upside, we move a certain mass of water per unit time. On the downside, we move exactly that much water per unit time, but we also move the air that the Beckett sucked in.

It's 1:00 am and I'm into the physics way over my head. I need to stop and get some Professional Help. I now understand my results less well than I thought I did when I started writing this section. I'll tell you where I was going with it anyway.

The pressure at the bottom of the injection tube should be just atmospheric pressure, because the injection tube empties into the top of the skimmer, which is where the surface of the water is, and the skimmer is open to the atmosphere through the foam riser tube. The pressure difference between the top and the bottom of a vertical column of fluid increases as the height of the column increases. Because the bottom of the injector tube is at a constant pressure, no matter how high the injector tube is, the pressure at the top of the injection tube (at the output of the Beckett) must be lower than atmospheric pressure. The pressure gets lower as the column height increases. This means that the pressure drop between the Beckett's air inputs and the Beckett's output gets larger with taller injection columns. The air flow rate through the Beckett depends on that pressure difference, so air flow must increase with increasing column height.

Common sense tells me that, for a given pump, there is some height above which increasing the injection column will begin to decrease the air flow rate, because of the decreasing water flow with increasing height. That critical height may be higher than your ceiling. I will perform more testing soon. In the meantime, my gut feeling is that you should maximize your skimmer's performance potential by using an injection tube, or injection tubes, that are as high as your ceiling and your wife will allow.

I would like to make several other observations that follow from the above.

Right now, I have absolutely no information nor guesses about whether there is an advantage to using multiple Beckett injectors rather than just one. If you use multiple injectors, you have to decide whether to give each injector its own injector column. In any case, you have to decide on the widths of the injector columns. Testing on all these details should happen sometime this century.


Did you read through all that, or just skip to the end hoping to find something good? I don't know yet if it's good, but here is a list of my current recommendations about skimmer design. I will try to incorporate all of them in my new skimmer design.


Shulin Chen, Michael B. Timmons, James J. Bisogni Jr, Danel J. Aneshansley. Modeling Surfactant Removal in Foam Fractionation: I -- Theoretical Development. Aquacultural Engineering, 13 (1994) pp. 163-181.

Shulin Chen, Michael B. Timmons, James J. Bisogni Jr, Danel J. Aneshansley. Modeling Surfactant Removal in Foam Fractionation: II -- Experimental Investigations. Aquacultural Engineering, 13 (1994) pp. 183-200.

Uraizee, F., and Narsimhan, G. A Model for Continuous Foam Concentration of Proteins: Effects of Kinetics of Adsorption of Proteins and Coalescence of Foam. Separation Science and Technology. Vol. 30, No. 6, pp. 847-881, 1995.

Add or View Comments

Chris Paris
Last modified: Sun Nov 29 21:18:54 EST