Reverse osmosis unit

Presses water through a semi-permeable membrane and leaves nearly everything dissolved behind, producing a near-blank water that must be built back up before it reaches a boiler or a cup.

Osmosis moves water across a membrane toward the more concentrated side. Reverse osmosis overrides that with pressure: the feed is pressed against the membrane, water molecules pass through, dissolved ions largely do not, and everything rejected leaves continuously as a concentrate stream to drain. Three consequences define how the device has to be used, and skipping any one of them is the classic failure in this category. FIRST, it is indiscriminate. Hardness, alkalinity, chloride and everything else fall together; you cannot ask a membrane to remove only the part you dislike, so RO does not so much solve a water problem as replace the supply entirely. SECOND, what comes through lands far below the target TDS band, and water with almost nothing dissolved in it extracts badly — there is little for flavour compounds to bind to and carry, so the cup reads thin and hollow regardless of grind, dose or ratio. THIRD, and the one that costs real money, the same water is chemically aggressive: with no buffering it readily takes up carbon dioxide, its pH falls, and it will attack copper, brass and boiler metal from the inside. So RO is never a finished product. It is a raw material that remineralisation or blending completes, and the two decisions are made together or the installation is not finished. That is precisely why the instinct to just fit RO is wrong: it converts a scale problem into a taste problem and a corrosion problem, and only the first of those announces itself. A membrane also needs chlorine removed ahead of it, so RO always sits behind carbon, and it sends part of every litre it treats to drain — which in a hot region is a design consideration and not an afterthought.

  • RO earns its place when the supply is wrong in more than one way at once — high TDS AND high hardness, or an ion you cannot selectively remove. A supply that is merely hard does not need it, and reaching for it there is over-engineering with a drain attached. What the supply test said is the only thing that answers this.
  • NEVER specify RO without deciding remineralisation or blending at the same time, in the same conversation, in the same order. Water with almost nothing dissolved in it extracts badly AND is aggressive to boiler metal — the strip is half a specification and the rebuild is the other half. Installing the first and leaving the second for later is the single most damaging sequence in this category, because the taste fault is noticed within days and the corrosion is noticed within years.
  • Carbon goes in front of the membrane, always and without exception. Chlorine attacks membrane material, and a membrane fed a chlorinated supply fails early and expensively for a reason that a cheap cartridge in front of it would have prevented entirely.
  • Size on production RATE against peak demand, then decide whether you are buying a tank. A membrane produces at its own pace and a bar draws in bursts; if the rate does not cover the peak, a storage tank is not an accessory, it is part of the specification, and it needs a position and a cleaning routine.
  • Budget for the concentrate stream before you commit. RO sends part of every litre it treats to drain, continuously, whenever it is producing. A drain has to exist and the connection is a licensed plumber's design governed by the local authority. In a region where water cost and supply are genuine constraints, this is a live operating consideration and not a footnote.
  • Check what happens to the ICE and to the cold tap. Ice made from unremineralised water dilutes an iced drink with nothing in it, and iced coffee is where dilution does the most damage to a drink's body. Decide the treatment line by line rather than plumbing everything to whichever outlet was nearest.
  • Set the blend or the dose against a measurement at the tap, and put re-checking it in the maintenance schedule with an owner. An RO chain has two things that drift — the membrane as it ages and the rebuild as it depletes — and both are invisible in the glass until the cup changes.
  • Watch the membrane's own end of life, which arrives as a rising TDS reading rather than as a failure. It does not stop; it passes more than it used to, and a chain calibrated around a new membrane slowly stops delivering what it was set to deliver.
  • Confirm feed pressure, the drain, backflow protection and any booster pump with a licensed plumber and, where a pump adds electrical load, a licensed electrician. Feed pressure is a property of the building and it is what a membrane actually runs on; the local authority governs the rest.

Behind carbon and in front of the rebuild, which means it is never the last item in the chain — if it is, the chain is unfinished. Physically it is a back-of-house installation with a permanent drain beside it, and the branch that feeds it is drawn at design stage because deciding afterwards which outlets get treated water usually means opening the joinery again. Everything downstream of the membrane, including the pipework itself, is exposed to aggressive water until the remineralisation or blending stage brings it back, which is an argument for keeping that stage as close behind the membrane as the layout allows.

  • Fitting RO and calling water treatment done. Stripping the water is half the specification; putting a controlled amount back is the other half, and the second half is what protects the boiler as well as the cup.
  • Putting a membrane on a chlorinated supply with no carbon in front of it, and replacing it early for a reason a cheap cartridge would have prevented.
  • Reaching for RO to solve hardness alone, where softening would have done it without a continuous drain, a waste stream and a corrosion risk.
  • Sizing on daily total and discovering the production rate at the first morning peak, when a tank was the thing that should have been specified.
  • Remineralising the espresso line and leaving the ice machine and the cold tap on untreated membrane output.
  • Setting the blend once at commissioning and never verifying it at the tap, on a chain that drifts from both ends.
  • A membrane running on mains pressure alone adds no load, but where the site's pressure is insufficient a booster pump does — and that pump may want its own supply. What the building can deliver, and whether a dedicated circuit is required, is for a licensed electrician to confirm and the local authority to govern. This site states no amperages or phase.
  • A carbon-filtered feed at adequate pressure, with an isolating valve reachable without dismantling anything. Feed pressure and feed temperature both change what a membrane produces, and both are properties of the building rather than of the unit. A licensed plumber confirms what the site can deliver; the local authority governs.
  • A permanent drain connection for the concentrate stream, which discharges continuously whenever the unit produces — this is not an occasional discharge like a drip tray and it cannot be trayed and emptied. The connection, any required air gap and the discharge arrangement are a licensed plumber's design and are governed locally. This site states no pipe sizes and no discharge requirements.
  • The membrane housing and its cartridge stack, plus any storage tank and any booster pump — and the removal arc for every cartridge in the chain. This is the largest treatment installation in the category and the one most often measured after the cabinetry was designed.
  • Access to every cartridge, the blend valve and the tank lid without moving the unit or emptying a cupboard. A treatment chain that has to be dismantled to be adjusted is a chain that is set once and never verified again.
  • Feed water temperature affects membrane output, and in a hot climate a supply run through a warm building performs differently from the same supply in winter. A stored tank in a warm cupboard is also standing water and belongs in the cleaning schedule.
  • RO with a blending valve
  • A set proportion of the carbon-filtered feed is routed around the membrane and recombined with what passed through it.
  • The simplest and cheapest way to land back inside the target band, using water you already have and adding no consumable. What it also does is bring back a proportion of everything in that feed — hardness, alkalinity, chloride — so a supply with a problem you specifically wanted gone will have a smaller version of that problem in the blend. It must be set against a reading and re-checked, because the feed moves and the blend is a proportion of it.
  • RO with a remineralisation cartridge
  • What passes the membrane runs through a bed of mineral that dissolves back into it on the way out.
  • Control over WHAT goes back in rather than merely how much, which is the whole reason to strip the water in the first place — a supply with a chloride problem gains nothing from a blend that reintroduces chloride. The cost is a consumable that depletes, and depletes silently: the water looks and flows identically as the dose fades.
  • Direct-flow point of use vs tank-fed
  • A membrane feeding one machine as it draws, against a membrane filling a storage tank the bar then draws from.
  • This is the recovery argument the boiler entry makes, in a different device: a membrane produces slowly and steadily, and a bar draws in bursts. Direct flow is simpler and caps you at the production rate. A tank covers the peak and is standing treated water, which brings its own management — it is a vessel that has to be kept clean and turned over, not a cupboard.
  • No RO — carbon plus softening instead
  • The alternative chain, chosen deliberately: filter the taste, remove the hardness, and change nothing else.
  • Frequently the correct answer, and rarely the exciting one. Where the supply is only hard and the TDS already sits in the band, RO adds a continuous drain, a waste stream, a remineralisation consumable and a corrosion risk in order to solve a problem softening had already solved. Listing this as a variant is deliberate: the strongest RO decision most operators make is not to fit one.
  • Espresso reads thin and hollow across every coffee on the list, and grinding finer makes it bitter without making it fuller.
  • Membrane output reaching the machine with little or nothing added back — the rebuild stage was never fitted, was set too low, or has depleted.
  • Drink quality across the whole menu at once, and weeks of chasing it in the grinder and the beans because water is the last variable anyone suspects after it has just been improved.
  • Green or blue-green staining at fittings, or a faintly metallic edge on long drinks.
  • Unbuffered water attacking copper and brass downstream of the membrane, because remineralisation was treated as a flavour upgrade rather than as corrosion protection.
  • Money, and the slow kind: fittings first, then valves, then the boiler. It is the most expensive possible way to discover that the rebuild stage was not optional.
  • The tank does not refill overnight, or the machine waits on water during the morning peak.
  • Production rate below peak demand, a fouled membrane, or feed pressure that was adequate at commissioning and is not now.
  • Throughput, and it appears at the peak rather than gradually — the treatment chain is fine all day and fails for the ninety minutes that matter.
  • Logged TDS at the tap climbs steadily over months with no adjustment made to anything.
  • The membrane approaching end of life and passing more than it used to, or a blend valve that has drifted open.
  • Quality first and then scale, in that order — and this failure is only visible at all to a bar that logs a reading, which is the argument for the log.

Related tools

Materials

Reference

Related articles

Equipment, workflow & commissioning

العربية