By C.F. “Chubb” Michaud CWS-VI
One thing I have come to appreciate during my three plus decades as a water treatment professional is just how accomplished our industry really is. No matter what it is that is wrong with water—no matter how disgusting or toxic that contaminant may be—no matter how large or small the need, we can fix that water. A sewage plant operating in Orange County, CA converts their wastewater back to potable water standards that exceed the original water quality sent to the homes and businesses in the area. I drank some of this water and I am still here to write about it. Being a resident of Orange County, I will, no doubt, drink it again some day! There are countless island nations that routinely transform seawater into potable water. Unthinkable water from nuclear reactors is purified to stringent discharge standards every day. We have the knowledge. Does that mean we are 100- percent correct 100 percent of the time? Unfortunately, that answer is no.
Our industry professionals have the knowledge and skills to solve any water or waste- water problem. Yet we have failures because the necessary level of knowledge is not shared by all but is assumed by many. Also, the necessary level of professionalism is not practiced by all. A mistake made out of ignorance is still a mistake. All filtration technologies have limitations. Failures occur when those limitations are exceeded—usually through a lack of understanding.
This article will examine a variety of technologies and explain how they work. This will help the reader visualize what can go wrong and why things sometimes fail if some other set of properties come into play. It is not the author’s intent to produce an epic reference design manual but rather to reveal hidden technologies that make different filters do different things. The focus will be on media filters.
Classifications of media filters
Media filters can be classified as either non-reactive or reactive. Non-reactive filters are those that work as barriers for particulate removal. They are non-chemical in nature and are more sensitive to hydraulics (surface flow) than they are to time.
Reactive filters are those that work by altering the chemistry of the water. These include oxidation filters, neutralization media filters and phosphate (sacrificial media), adsorption filters and ion exchangers. These filters are sensitive to retention time. Keep in mind that reactive media are ALSO barrier filters and that they can release particulate if they are subjected to variable flow. The following table lists some types of media filters along with suggest flow rates. It is not an all-inclusive list and some liberties can be taken with special applications.
Table 1. Types of media filters
Non-reactive (barrier) filters
The simplest types of filters to build and understand are non-reactive media (particulate) filters. These may be all of one media or multiple layers with different densities that classify and settle to the same position after backwashing. These are barrier filters and are used primarily to remove suspended dirt and turbidity. The packed media presents a series of channels through which water passes. If suspended particles are larger than the space between the media granules, it is retained in the filter. There are, however, two filter mechanisms at play in such a filter: 1) the barrier (size of the space between particles of the media) and 2) the torturous path, which is the quiet space between granules where smaller suspended matter can settle.
The open space between a uniform stack of filter media granules (picture a stack of canon balls) is approximately six percent of the diameter of the media itself (see Figure 1). This means that when using a media measuring 0.5 mm in diameter (500 micron) such as typical filter sand, the free space between the grains is approximately 30 microns (µ). We have a 30-µ barrier filter. It matters little how fast water is run through it. As this filter begins to load up, the spaces between grains get smaller and it actually becomes a more effective filter. All you have done, however, is to load the surface of the media one or two inches (25.4 or 50.8 mm) down into the bed and created a serious pressure drop which now requires a backwash to straighten things out.1
Slow that filter down to 4-6 gpm/sq. ft. (15.14-22.71 L/m) (or less) and it creates eddies between the grains that contain dead spots—areas of almost stagnant flow that can trap much smaller particles. By slowing down, particulates in the range of 15-20µ can effectively be trapped with this same filter. If multi-media or a layered media bed containing anthracite and garnet (see figure 2) is used, the anthracite layer will capture larger particles and the garnet will stop intermediate particles. Typically the top (anthracite) layer is 0.8-0.9 mm and a somewhat flattened platelet that creates a zig-zaggy path that can trap 25-30µ solids. The garnet is a 0.4-0.6mm granule that is irregular in shape so space between grains is very tight. This layer can trap solids in the 10-15µ range. Using 0.6-0.8 mm anthracite and a #60 garnet (0.25 mm) will create a 5-7µ filter that can remove 90 percent of particles in this range. If this filter is upstream of a reverse osmosis 5µ pre-filter, the filters will last three to four times longer!
How can such a purely mechanical filter fail? Let’s say you have a tri-plex set up that you have carefully designed to run at six gpm/sq. ft. It is working great (with all three filters on line). One filter is taken out of service to backwash it. Flow through the two on line units suddenly jumps to nine-gpm/sq. ft. (34.06 L/m). Worse yet, it is using clean product water from the outlet of the two in service to backwash the third unit (at 15-18 gpm/sq. ft. [56.78-68.13 L/m]) and flow per square foot goes through the roof. Fine suspended particles trapped at the lower flow rate head down stream. This not only overwhelms R/O pre-filters but sends a stream of fine mud up into the third filter.
If the application is for a critical performance media filter, it is much better to design the system with a bank of six or more parallel filters so that when one is pulled off line, total flow increase through on line units is minimized. Also, consider using a separate source of clean water for backwashing. The expense for the storage tank and the extra pump will be worth it.
Reactive media filters—oxidative
Reactive media filters such as oxidation filters often have upstream oxidants added to precipitate iron and manganese so filters act as particulate polishers. These precipitates are very fine and these filters must be run with low velocities as indicated in Table 1. The same rules apply here. If run too fast or superficial flow is suddenly increased by taking one of the filters out of service, trapped iron will blow off and enter the down-stream flow. Take the same design precautions by using multiple smaller units in parallel, rather than only two or three.
What else can go wrong? I recently had a call from a dealer that had installed a well-designed iron oxidation filter in a home; it was working just fine. A water analysis on the effluent revealed a high arsenic level. The influent was below the MCL at only six ppb but the effluent was over the 10 ppb limit. The home had the usual variable flow rate and when the flow would peak, it was blowing precipitated iron through the filter, but not above the 0.3-ppm limit level. The iron, however, was binding to the arsenic and when iron increased during higher flows, it was carrying the arsenic with it. A downstream particulate filter was suggested to capture the errant iron along with the arsenic.
Reactive media filters—adsorptive
Adsorptive media filters are definitely dependent on empty bed contact time (EBCT). If a GAC filter is designed with a short retention time, it can still be a very successful dechlorination system. Design a GAC filter with a short retention time for the removal of trihalo methane (THM) and you fail. The same is true for activated alumina used for fluoride reduction and a host of iron-based media used for arsenic removal. An EBCT of 3-5 minutes is a good design. It still works at higher flows but the reaction time comes so close to the actual retention time achieved that working capacity is seriously compromised. Increasing flow through a bed of alumina from 1.5 gpm/cu. ft. (5.67 L/m) up to only 3.0 gpm/cu. ft. (11.35 L/m) reduces the capacity by 40 percent.2 Most arsenic adsorbers will lose 50-60 percent of their capacity when flow doubles from 2.5 gpm/cu. ft. to 5.0 gpm/cu. ft. (9.46 to 18.92 L/m).3 One has to ask the media supplier what flow rate they used in doing the projections for capacity. With adsorption media, it is always advised to make the filters big enough. It is also good advice to use redundant systems—a worker and a polisher—just in case.
What else can go wrong? The mere fact that a particular media will remove a particular contaminant does not mean that it will work for all applications with equal aplomb. Adsorption of THM by GAC is far slower than the catalytic reduction of chlorine and, even though the same media and hardware are used, their success dictates two very different flow rates. In addition, dechlorination filters are considered fine if the remove 75 percent of the chlorine.4 Toxic organics, however, may necessitate 95 percent or more reduction. Chlorine can be removed with an EBCT of one minute. Chloramine takes about three minutes and THM may require EBCT of 10 minutes. Please keep all of this in mind when you put a 10-inch (25.4-cm) cartridge on line. They work great when polishing R/O effluent at 125 cc/min but may not work at all at one gpm (3.78 L/m) (the equivalent of 40 gpm/cu. ft. [151.41 L/m] through a tank system).
There is a concept called half-length, which holds true for almost all adsorption media. It tells us is that if we can remove 50 percent of a contaminant with x seconds of EBCT, it will take an equal amount of time (another x seconds) to reduce the remaining contaminant by an additional 50 percent. So—two half-lengths make a whole—right? Not quite. You see, the second half-length only removes 50 percent of the remaining 50 percent for a total of 75 percent. The next half-length takes 50 percent of the remaining 25 percent for a total of 87.5 percent. It takes five half-lengths to get to 97 percent reduction and seven to get to 99 percent. As an adsorber removes contaminant from the fluid stream, the driving force that pushes the contaminant onto the media decreases. As the driving force decreases, the rate of removal decreases as well.
If a water supply has a low (acidic) pH, it is common to use a sacrificial neutralization media to save plumbing and fixtures. The most commonly used media are calcium carbonate (calcite) and magnesium oxide (Corosex®). Both of these media are solid forms of a high pH base. Calcite (CaCO3) is converted to the bicarbonate (Ca(HCO3)2) by low pH water and is effectively neutralized. Magnesium oxide, a much stronger base, simply acts as a hydroxide to neutralize the acid. The biggest bang for the buck is in treating very low pH water. As water approaches neutral (pH>6.5) however, the calcite slows down quite a bit and it may take a mix with magnesium oxide to do the trick. I suggest 2-3 parts calcite to one part magnesium oxide. I also suggest an upflow bed to prevent the bed from fusing together or caking. Water remaining in the filter tank at night will jump in pH and may over-correct the situation (resulting in a pH >8.5). Although not harmful, it is outside the US EPA range. For this reason, I favor calcite, which is less prone to over correction.
I did some lab studies a few years back to determine the relative properties of these two media. If you run pH 2.7 water through calcite at a flow of 3.0 gpm/cu. ft., you can neutralize to about pH=6.5. A 50/50 mix of calcite and magnesium oxide results in a pH of 6.7. Slow that down to one gpm/cu. ft. and you get pH=6.8 and 8.6 respectively (over correction). If you feed pH 4.6 (this represents a possible R/O water where the low pH is from dissolved CO2), the response is pH 6.4 through calcite and 6.9 through the blend (the media is slow to respond to weak acid such as carbonic acid). At a feed of pH 7.6, the effluent doesn’t change through the calcite and comes out at pH 8.5 through the blend. After standing for 24 hours, the first water out of the calcite is pH 7.6. The first water out of the blend is 9.3 and straight magnesium oxide is 9.9.5 I suggest a 3:1 blend of calcite and magnesium oxide and a flow of 2.5-gpm/cu. ft. The guesstimated pH of the effluent will be just under 7.0.
Phosphate is a good chelating ion for hardness. Sodium tri-polyphosphate (TSPP) was used for years as a bath additive (for lasting suds) and a laundry additive (to tie up hardness and boost pH for better cleaning). In large doses, hardness will totally precipitate with the phosphate and although the water may be cloudy, it has been chemically softened. At low levels of addition, phosphates do not soften the water but attach to the hardness in such a manner as to interfere with the formation of hardness scale. This is known as ’threshold chelation’ and there is research on it going back to 1945. The scale is softer and tends to not stick to pipes. Low-dose phosphates reduce iron deposits as well. In addition, phosphates will react with copper to form a glassy layer on the copper surface (inside of pipe). This provides not only scale control but corrosion control as well. Development of glassy phosphates (typically hexameta phosphate) provides a convenient method of introducing the low level dose to feed water. Glassy phosphate in the form of large granules or beads is very slow to dissolve in water. Passing water slowly through a bed of these beads will add only 1-2 ppm of phosphate to the water. This is the perfect dose for the aforementioned applications.
If you place a weighed quantity of glassy phosphate in a jar and fill it with water, then weigh the amount of phosphate remaining a month later, you will note that the weight loss is about five percent; a month later you have another five percent. To determine how much crystal or bead would be needed to treat a given stream of water, first calculate the amount of water to be used in a month at a given sustained flow rate. So, one gpm = 1,440 gpd (5450.99 L/d) = 43,200 gallons per month (163529.78 liters per month) and the weight of that water is 360,000 pounds (163,293.25 kilos). To add two ppm of phosphate, divide that weight by 1,000,000 and then multiply that number x 2, which determines the addition of 0.72 lbs. (0.32 kilos) of phosphate. Since glassy phosphate dissolves at the rate of five percent per month, the necessary supply would be 0.72/0.05 = 14.4 lbs. (6.53 kilos) of product to treat a continuous stream of one gpm. How does this translate to a residential need that has intermittent flow and only uses about 300 gpd (1,135.62 L/d)?
Glassy phosphate does not completely dissolve in water because it rapidly reaches a saturation level and the rate of solubility drops dramatically. Beads can sit in water for several years. When glassy phosphate is used for intermittent flow, the total amount dissolved in any given day is pretty much the same as that with a continuous flow of water; the total amount of treatment is similar. It is simply not distributed evenly. Nonetheless, this still works well in preventing corrosion and reducing the hardness of hard scale. Base filter size on the total water demand per month and add about 25 to 30 percent more crystals to the filter. If a household uses 10,000 gpm (37854.11 L/m), which is approximately one quarter of the calculation above, use a cartridge with about five lbs. (2.26 kilos) of bead and top it off every three or four months.
As far as reactive media go, ion exchange is very rapid. It only takes a few seconds to soften water. The typical household system may only be one cubic foot and run as fast as 7 to 10 gpm (26.49 to 37.85 L/m). The typical family only uses about 300 gallons of water per day; a single cubic foot of resin rated at 24,000 grains can treat 1,200 gallons (4542.49 liters) of water with 20 grains per gallon hardness. It may regenerate once every three days. If you look at a commercial application for a five-gpm softener, however, you have to build it larger because five gpm is 300 gph (1135.62 L/h). There is only a four-hour capacity with a one cubic foot system. If this is a 16-hour/ day operation, to make it last the entire day; use four cu.ft. If hardness is super critical, make this system a twin alternating two cu. ft. per side and bump the salt level to 10 lbs. or more to guarantee low leakage.
Not all ion exchange resin systems will be this forgiving. If resin is used for a critical application, such as arsenic, nitrate, uranium, barium, radium or any other health-related need, it must be put together with a generous engineering factor. The engineering factor is an artificial downgrade of system capacity in order to build in a safety factor. Once the cycle is determined, downgrade it by 25 percent to allow for error. Flow rates should not exceed three-gpm/cu. ft. to actually achieve rated book capacities. If levels of contamination are high enough to cause an immediate problem to anyone drinking the water, install a polisher (a duplicate system run in series with the primary).
The water treatment industry has the knowledge to fix any water problem. Oft times the treatment is extensive and expensive. If called upon to treat a critical need water problem, make sure you have your ducks in a. Make sure the technology is proven and the design is adequate to provide the level of safety the buyer thinks he is paying for. Don’t expect a cartridge to solve a whole house problem. Always start with a good water analysis. Most media will do more than one job but they may not do them equally well. Ion exchangers can pick up many ions but they will release them down the road according to selectivity. The same is true for adsorbant media. Contaminants may release if the beds are over run. You have to design your systems around the weakest link.
- Michaud, C.F., Designing for Success, WC&P, April 1992
- Michaud, C.F., Fluoridation: The Good, the Bad and the Ugly—Parts I & II, WC&P, Feb & March 2010
- Michaud, C F., Factors Affecting the Capacity of Arsenic Removal Filters, WC&P, March 2008
- Michaud, C.F.,GAC—To Become the Workhorse of Water Purification—Part III, WC&P, August 1988
- Systematix Co, Buena Park, CA, 90620, Neutralization Cartridges—product literature
About the author
C.F. ‘Chubb’ Michaud is the CEO and Technical Director of Systematix Company, Buena Park, CA, which he founded in 1982. An active member of the Water Quality Association, Michaud has been a member of its Board and of the Board of Governors and past Chair of the Commercial/Industrial Section. He is a Certified Water Specialist Level VI. He serves on the Board of Directors of the Pacific WQA (since 2001) and Chairs its Technical Committee. A founding member of WC&P’s Technical Review Committee, Michaud has authored aor presented over 100 technical publications and papers. He can be reached at Systematix, Inc., 6902 Aragon Circle, Buena Park CA 90620; telephone (714) 522-5453 or via email at cmichaud@systematixUSA.com.