How to Choose Between Softening, Filtering and Treating

Why this matters

Most water equipment gets specified from a symptom and a catalogue. The customer says "hard water," somebody sells a softener, and the actual failure was silt cutting a seat, or oxygen coming in through an air vent, or a tower running at cycles its chemistry cannot carry. The three families do genuinely different work: filtration removes what is undissolved, softening and its cousins remove or exchange what is dissolved, and treatment changes the chemistry or puts a protective film on the metal without removing anything. Picking the wrong family does not half-work. It fully fails, on a delay long enough that nobody connects it to the purchase.

The steps below are ordered by what skipping each one costs you, worst first. Step 1 sends you to the wrong technology; step 7 only means you find out late.

Step 1: Test the water you actually use, at the point you actually use it

Skip this and every later step is a guess, which is how a building ends up with a softener for a corrosion problem.

The sample that matters comes from the connection that feeds the system, in the season that matters. A utility's annual report is an average across a distribution system and a year; a well moves with the water table; a building's own supply picks up whatever the building's pipe gives it. Where a system is fed from more than one source, or switches sources seasonally, you need both reports, because the switch is often the event that starts the failure.

Ask the laboratory for the parameters your decision rests on rather than a package: hardness split into calcium and magnesium, total alkalinity, pH, chloride, sulfate, silica, iron and manganese, conductivity or total dissolved solids, and turbidity or suspended solids. Add a microbiological count where the system is open, warm and recirculating. Note the units the lab used, since hardness is reported both in grains per gallon and in mg/L as calcium carbonate, and the two differ by a factor of 17.1.

Step 2: Name the failure you are preventing and the surface it happens on

Skip this and you buy equipment that changes a number nobody was suffering from.

"Hard water" is not a failure. Scale on a heat transfer surface is a failure. A blocked spray orifice is a failure. A pinhole at a weld is a failure. Spotting on a finish is a failure. Each of those points at a different surface, a different mechanism, and often a different family of answer, and two of them can be present in the same building at once.

Write the failure down as a sentence with a surface in it. If you cannot, you have a water report and no problem, which is a legitimate finding and a good conversation to have before anyone quotes.

Step 3: Classify each contaminant by its form, and the system by what it does to water

Skip this and you fit the right technology in the wrong place, which is the most common expensive mistake in the list.

Form decides the family:

System type decides where the load lands, and this is the half that gets skipped:

• A closed loop gets its contaminant load once at fill, plus whatever makeup brings in. The dissolved load does not concentrate; it deposits once and is done.
• An open recirculating system, a cooling tower above all, evaporates pure water and leaves everything else behind, so every dissolved species concentrates until blowdown removes it.
• A once-through system, domestic water for example, sees fresh contaminant continuously at the source concentration and never concentrates it.
• A steam system concentrates in the boiler and separates volatiles into the steam, so a species can be a boiler problem, a condensate problem, or both.

Step 4: Decide between removing, converting and inhibiting

Skip this and you over-buy on removal where chemistry would have held, or lean on chemistry where the load overwhelms it.

Remove when the contaminant does harm at any concentration you can reach, or when it will concentrate past any chemistry's ability to hold it. Suspended solids, and hardness in an evaporative system, sit here.

Convert or hold in solution when the contaminant is tolerable if it stays dissolved. Scale inhibitors and dispersants do this; they do not remove hardness, they keep it from depositing, and they have a concentration ceiling above which they stop working.

Inhibit the surface when the water is going to stay as it is and the metal needs defending. Closed loop corrosion inhibitors sit here; they work by filming or by passivating the metal, which is why they require a clean surface to reach.

The families combine, and the order they run in matters more than which ones you picked. Filtration goes ahead of ion exchange or a membrane, because solids foul both. Dechlorination goes ahead of a polyamide membrane and after any point where the chlorine was doing microbiological work.

Step 5: Check what your choice creates downstream

Skip this and you solve one problem and post another one forward, usually to a different trade.

Softening adds sodium in proportion to the hardness removed, sends a chloride-bearing regenerant to drain, and leaves the treated side with no calcium available to form a protective carbonate film, which makes softened hot water measurably more aggressive to copper and galvanized than the hard water it replaced, so the hot-side piping material is part of the softening decision. A membrane sends concentrate to drain at a volume set by its recovery, and delivers aggressive low-conductivity water that attacks copper and galvanized. A biocide programme creates a discharge the drain may or may not be permitted for. Filtration creates a waste stream and a maintenance task on a differential-pressure trigger nobody may be watching.

None of these are reasons not to proceed. They are line items in the specification, and the time to find them is before the equipment lands.

Step 6: Size on load, not on flow rate

Skip this and the equipment passes its startup test and breaks through in service, which reads as a defect and is a specification error.

Flow rate sizes the pressure drop. Load sizes the consumable. A softener is sized on grains between regenerations, which is hardness times volume, not on the peak gallons per minute. A filter is sized on dirt holding capacity at an acceptable differential, not only on the flow it can pass clean. A chemical feed is sized on the makeup volume that dilutes it. Every one of those is a volume, and the volume is the number nobody meters.

Step 7: Set the monitoring point and the number that says it is working

Skip this and the system still works, you just find out about the failure from the customer.

Name one measurement per installed device, the value that means healthy, and who reads it. Treated hardness at a tap for a softener. Differential pressure across a filter housing. Permeate and feed conductivity for a membrane. Inhibitor residual and makeup volume for a treated loop. Cycles of concentration and biocide residual for a tower. Put them on one sheet with the date, because the trend is the diagnostic and a single reading almost never is.

Worked example: one report, one building, three answers

A small commercial building has a 900 gallon closed chilled water loop, a cooling tower serving the chiller, and domestic hot water, all fed from the same municipal supply. The report reads 18 grains per gallon hardness, which is 308 mg/L as calcium carbonate; alkalinity 220 mg/L as calcium carbonate; chloride 40 mg/L; silica 22 mg/L; conductivity about 700 microsiemens per centimetre; iron a trace; chlorinated.

The closed loop. Work out the actual hardness input. Nine hundred gallons is about 3,406 litres, and at 0.308 grams per litre that is roughly 1,050 grams, a little over 2 pounds of hardness expressed as calcium carbonate, deposited once across the entire wetted surface of a building loop. That is not a scaling problem and softening the fill would be equipment bought to solve nothing. The real risks in a closed loop are oxygen ingress, microbes and suspended solids, so the answer is clean, fill, inhibit, fit side-stream filtration, and meter the makeup.

Then check the makeup, because it is what changes the answer. If this loop is passing 46 gallons a week through an automatic fill, it takes about 2,392 gallons a year, which is 2.7 loop volumes, and the annual hardness input becomes roughly 6 pounds and rising rather than a one-time 2 pounds. At that point the loop is behaving like a once-through system, the inhibitor is being diluted as fast as it is fed, and the correct action is to find the leak, not to soften the fill.

The tower. Evaporation leaves everything behind, so the tower runs at cycles of concentration: dissolved solids in the tower over dissolved solids in the makeup. At 3 cycles the tower sits near 2,100 microsiemens per centimetre; at 6 cycles, near 4,200. Unsoftened, that means hardness near 1,848 mg/L as calcium carbonate against 1,320 of alkalinity, water no dispersant holds, so the number is the reason 6 cycles is unreachable on this makeup rather than a description of a tower running there. Soften the makeup and the calcium wall comes down while conductivity barely moves, because sodium replaces calcium equivalent for equivalent, so the same 4,200 setpoint now means a chemistry that can run. What is left is silica near 132 mg/L and chloride near 240, the walls softening does not touch. Total liquid loss is evaporation divided by (cycles minus one); the blowdown valve carries that figure minus drift and any other liquid loss, which is negligible at low cycles and material above about ten, so 3 cycles to 6 cuts blowdown from half the evaporation rate to a fifth, a 60 percent cut in blowdown and about 20 percent in total makeup.

That is the case for softening the tower makeup: it removes the calcium wall so the cycles can rise. It does not remove the other walls. At 6 cycles the limiting species becomes silica or chloride against the tower's metallurgy, and where the tower or its heat exchanger uses stainless, the chloride number is the one to take to the equipment maker. Raise cycles and you save water and concentrate every other species; lower them and you spend water and dilute your own inhibitor faster than you can feed it.

Treat the tower as a public health item as well as a chemistry one. It aerosolises its own water, and a water management programme following ASHRAE Standard 188 is the framework that covers it. Do not clean fill or basin deposits with a high-pressure spray while the fan is running or without respiratory protection selected under a programme meeting 29 CFR 1910.134, because the hazard route there is inhalation of aerosolised biofilm and no glove or face shield addresses it.

The domestic hot water. Once-through, so nothing concentrates, and 308 mg/L of hardness deposits on the heater and the fixtures continuously at source concentration. This is the load where softening pays. It also adds roughly 142 mg/L of sodium, which is 308 mg/L as calcium carbonate times the ratio of sodium's equivalent weight to calcium carbonate's, about 0.46, or near enough 8 mg/L of sodium per grain per gallon removed, which is a conversation to have before installing rather than after, and a common resolution is to soften the hot supply and leave a hard cold drinking tap.

One report. Three systems. Three different answers, and the thing that separated them was not the water at all, it was what each system does to the water.

References

• ASHRAE Standard 188, Legionellosis: Risk Management for Building Water Systems, for the water management programme that covers an open recirculating cooling system
• OSHA 29 CFR 1910.134, respiratory protection, for the programme requirement where respiratory protection is used during cleaning of biologically active water systems
• Independent laboratory analysis or the local utility water quality report for the parameters listed in step 1
• Equipment and treatment supplier data for capacity, limits and target residuals
• See related: What a Softener Removes and What It Leaves Behind; What Suspended Solids Do Once They Are Inside a System