How Does Reverse Osmosis Work? Stages, Pressure and the Waste Ratio

If you are choosing between a $300 under-sink unit and a $2,000 whole-house build, how reverse osmosis works stops being trivia. The physics settles three things you live with for years: whether the system hits its rated output in your house, how much water runs to the drain for every gallon you drink, and what it will not remove no matter what the box promises. I am Chriss R., I write this site. Feed pressure catches out more RO buyers than anything else, and it is the one number almost nobody checks before ordering.

How does reverse osmosis work in practice

Put salty water on one side of a thin film that passes water molecules and blocks dissolved salts, and fresh water on the other. Water migrates toward the salty side on its own. It keeps going until the extra column of water pushes back hard enough to stop it. That push-back is osmotic pressure, and in drinking water it is small: roughly 1 psi for every 100 ppm of dissolved solids. City water at 400 ppm resists you with about 4 psi. That is the whole opponent.

Reverse osmosis runs that backwards. Your house pressure, typically 40 to 80 psi, shoves water at the membrane from the concentrated side. Subtract the osmotic pressure and whatever back-pressure the storage tank is applying, and what is left is net driving pressure. That surplus is the only thing making water cross the film. Everything else in the cabinet is support staff.

One detail confuses almost everyone. The membrane is not a dead-end filter. Water flows across its face, not just into it. A fraction crosses to become permeate, the water you drink, and the rest sweeps the surface and leaves through the drain line with the rejected salts in it. Stop that cross-flow and the membrane fouls solid within days. The drain water is not a design failure. It is the cleaning mechanism.

What each stage in the stack does

A typical under-sink system is four or five stages, and exactly one of them is the membrane.

• Sediment prefilter, usually 5 micron, so grit does not shred the carbon block or plug the membrane face.
• Carbon prefilter. This one is not optional on city water. Thin-film composite membranes are attacked by free chlorine, and continuous exposure ruins one in months rather than years. Chloramine strips off more slowly and wants catalytic carbon with real contact time.
• The membrane itself, rated in gallons per day at fixed test conditions. Residential elements are usually 50, 75 or 100 gpd.
• A flow restrictor on the drain line, sized to the membrane. It is a small, cheap part that sets the recovery rate, and the wrong one ruins either your ratio or your membrane.
• Storage tank, on most systems, because a 75 gpd membrane makes under 7 ounces a minute at its rating and closer to half that in a real kitchen. Nobody stands at the faucet for that.
• Post-carbon polish, which cleans up taste and catches the small uncharged molecules the membrane let slide.

Stage count is marketing. A seven-stage system is often a five-stage system with two cheap inline cartridges bolted on the end. Three things matter instead: the membrane rating, the certified claims on the data sheet, and whether the cartridges are a standard size you can still buy from more than one supplier in five years. Proprietary quick-change cartridges are cheap to install and expensive to own, which is where the real ownership cost hides. I put those numbers side by side in the filter replacement schedule and yearly cost.

What the membrane rejects and what slips through

The membrane sorts by size and by electrical charge. Charged ions get rejected hard. Small, uncharged, dissolved molecules are a different story, and this is where a lot of published advice goes wrong.

Two disagreements are worth naming. First, you will read that RO removes VOCs and also that it does not. Both statements describe real test data. A membrane alone gives mediocre VOC rejection because those molecules are small and carry no charge, while a complete system with good carbon in front and behind performs well. So a VOC claim is usually a carbon claim wearing an RO badge. Check which standard the certification names, because NSF/ANSI 58 (reverse osmosis), 53 (health effects), 42 (aesthetic) and 401 (emerging contaminants) are not interchangeable. A unit certified only to 42 has been tested for taste.

Second, bacteria. The pore size argument is correct and still not the whole picture. A pinhole, a tired o-ring, or growth in the storage tank downstream all bypass the physics, and none of it shows up in a taste test. If your well has ever failed a total coliform test under the EPA rule, treat that as a separate problem with a separate fix, and start with your county health department or a state-certified lab. The trade-offs between shock chlorination and a UV system are covered separately, and UV certified to NSF/ANSI 55 is the usual answer, not RO.

Why feed pressure decides almost everything else

Membrane ratings are laboratory numbers. The common residential test condition is around 60 psi of feed at 77 degrees F with a defined test water. Your kitchen is none of that. Drop the feed to 40 psi and a 75 gpd membrane will often deliver closer to half its rating. Drop it further and rejection sags too, because the salts keep diffusing across at their own pace while less water is being pushed through to dilute them.

Then there is the tank fighting back. A standard tank is precharged around 7 psi empty, and the air bladder pressure climbs as it fills. The automatic shutoff valve closes when tank pressure reaches roughly two thirds of feed pressure. At 60 psi feed that is about 40 psi of back-pressure at the end of the fill, so the last third of the tank fills slowly and with worse rejection than the first third. At 45 psi feed the system spends most of its life near stall.

Well owners get hit hardest. A pump on a 30/50 switch spends part of every cycle at 30 psi, which is below where most membranes do useful work. A booster pump runs $80 to $250 for the hardware, more if an electrician has to put an outlet under the sink, and it is the best dollar-per-gallon upgrade available to a low pressure house.

Cold water costs you as well. Water gets more viscous as it cools, so a winter feed at 50 degrees F can cut production by something like a third against the 77 degree test condition. Nothing is broken. The tank simply refills slower in January, and every year people replace a perfectly good membrane over it.

The drain ratio, told honestly

The number quoted as the waste ratio is drain volume to permeate volume. A conventional tank system with a matched flow restrictor runs somewhere between 3:1 and 5:1 in a real house. A 75 gpd membrane is normally paired with an 800 mL/min restrictor, which is about 300 gallons a day to the drain against 75 produced, and that is where the 4:1 headline comes from. Add a permeate pump, which borrows the drain flow’s own energy to fight the tank back-pressure, and 1.5:1 to 2:1 is realistic. Tankless direct-flow units often advertise 1:1 or better.

Watch the vocabulary too. Some manufacturers quote recovery, the share of the feed water that ends up as permeate, instead of a drain ratio. Twenty percent recovery and 4:1 describe the same machine, and 50 percent recovery is 1:1. Neither figure means much without the feed pressure attached to it.

Here is where published numbers diverge, and the mechanism is boring rather than sinister. Manufacturers measure at ideal conditions: full test pressure, warm water, an empty tank, low feed TDS. Your worst ratio happens exactly when the tank is nearly full and the driving pressure has collapsed, which is the state the lab test skips.

drinking and cooking = 3 gal/day
drain at 4:1         = 12 gal/day
per month            = 360 gal
at $6 / 1,000 gal    = about $2.16
at $12 / 1,000 gal   = about $4.32

Municipal water plus sewer runs roughly $4 to $14 per 1,000 gallons depending on the utility, the sewer surcharge and which tier of the rate block you land in, so a 4:1 ratio on a drinking-water system costs a few dollars a month. Not nothing, but far smaller than most people fear when they hear “wastes four gallons.” Two situations change that math. A drought-restricted district, where the volume itself is the problem whatever it costs. And a septic system with a marginal drainfield, where 360 extra gallons a month lands in a tank that is already borderline.

Where an RO system earns its money

Most households do not need reverse osmosis. It is a narrow tool that does one job extremely well, and nearly all the disappointment I read about comes from people buying it for the wrong problem.

If your actual complaint is chlorine taste, scale on the shower glass, or sediment after a main break, RO is a slow and expensive way to answer it. An under-sink carbon block at $40 to $90, priced by housing size and cartridge quality, handles the taste. Scale needs a softener. The comparison of filters, softeners and RO sorts that out by symptom before you spend anything.

Verify your own rejection with a TDS meter

A handheld TDS meter costs $20 to $30, the spread being mostly temperature compensation and calibration quality, and it turns a marketing claim into a measurement. Fill a glass from the cold tap upstream of the system. That is your feed. Then run the RO faucet for 30 seconds before you sample, because water that has sat in the tank overnight reads high. Salts keep diffusing across the membrane while the system is idle, an effect usually called TDS creep, and sampling the first ounce out of the faucet will make a healthy system look sick.

feed TDS      = 420 ppm
permeate TDS  = 18 ppm
rejection = (420 - 18) / 420
          = 402 / 420 = 95.7%
healthy range: 95% or better

Take the reading when the system is new and write it on the cabinet wall in marker. That baseline is what makes later readings mean anything. A slow decline over two or three years is a membrane reaching the end of its life. A sudden drop to 60 or 70 percent is usually a bypass: a failed o-ring, or an automatic shutoff valve leaking feed water into the permeate line. Check pressure before you buy a membrane, because low pressure lowers rejection by itself and a new element will read just as badly.

Know the limit of the tool. A TDS meter reads electrical conductivity and converts it. It is blind to most organics, blind to dissolved gases, and blind to PFAS at parts-per-trillion levels. It tells you the membrane is alive. It does not tell you the water is safe, and no honest person should sell it to you as a health test.

What to check before you spend the money

Buy a hose bib gauge first. They run $10 to $20, the difference being whether the gauge carries a lazy hand that records the overnight peak, and reading your static pressure takes two minutes at the laundry spigot. Under 40 psi, budget for a booster pump in the same purchase rather than discovering the problem after installation. Measure feed TDS the same afternoon. Above roughly 1,000 ppm the osmotic pressure you are fighting stops being trivial, membrane life shortens and the drain ratio gets worse.

Then get under the sink with a tape measure. A standard 3.2 gallon tank is roughly 11 inches across and 15 inches tall, and it holds only about 2 to 2.4 gallons of usable water at household pressure, not the number on the label. It also has to stand upright somewhere, which a tankless unit does not. I am not a licensed plumber, and drain saddles, air gaps and permitted connections are precisely where local code varies, so that part is worth a professional’s twenty minutes. The physical fit and the real installed numbers are in the under-sink RO cost and sizing guide.

Last, decide in advance how you feel about flat-tasting water. RO strips calcium and magnesium along with everything else, and whether that matters for health is a genuine disagreement: drinking water contributes a real but modest share of most people’s mineral intake, and food supplies the rest. If you have a medical reason to care, ask your doctor rather than a filter salesman. Taste is less debatable, and most people notice it. A remineralizing cartridge runs $30 to $80 depending on the media and the housing, and I go through what remineralization actually changes rather than what the packaging promises.