How Water Meters Work in Buildings

At three in the morning, a shopping mall in Petaling Jaya is empty. No tenants, no cleaners, no toilets flushing. Downstairs in the meter pit, the incoming meter is still creeping forward, and nobody notices, because the only person who ever looks at it is the meter reader who comes once a month and writes down one number. Understanding how water meters work is what turns that creeping dial into a finding: either there is a leak, or something is running that should not be. And in a surprising number of buildings, the meter is not creeping at all when it should be, because the leak is too small for it to feel.

A water meter is one of the few pieces of equipment in a building that quietly decides how much money you hand over every month. It is worth knowing what is inside it.

What is actually inside a mechanical water meter?

Two families cover almost everything you will find in a Malaysian building.

Positive displacement meters work by counting. Water is forced into a chamber of a precisely known volume, a nutating disc or an oscillating piston sweeps that chamber empty, and a magnet couples that motion through the meter wall to the register dial. Every wobble equals a fixed number of millilitres. Because it counts discrete parcels rather than measuring speed, this type stays honest at very small flows, which is exactly why it is the traditional choice for 15mm and 20mm connections. The cost is fragility: grit, scale and pipe debris chew up the moving surfaces, and a partly blocked strainer will stall it.

Velocity meters, usually multi-jet, work by timing. Water enters a chamber through a ring of ports, the jets spin a small impeller, and the register counts revolutions. Turns per litre is a calibration, not a physical constant, so the reading depends on the flow being fast enough and well behaved enough to spin the wheel properly. Multi-jet meters are cheap, tolerant of dirty water and dominate the 15mm to 50mm range in commercial buildings here.

Above roughly 50mm you meet the Woltmann or helix meter, a propeller sitting in the flow, used on incoming mains and large risers. It handles enormous flow rates without much pressure loss and is thoroughly bad at small ones.

Every one of these has moving parts sitting in tropical, chlorinated, sometimes sandy water for a decade. They wear. Worn meters almost always slow down, which means they under-read, which means the error drifts in the utility's disfavour on the incoming main and in your favour on a tenant sub-meter.

How ultrasonic meters measure water with no moving parts

An ultrasonic meter puts two transducers in the pipe wall, angled at each other, and fires a sound pulse in each direction. The pulse travelling downstream is carried along by the water and arrives fractionally early. The pulse travelling upstream fights the current and arrives fractionally late. That difference in transit time, measured in nanoseconds, is proportional to the average velocity of the water. Multiply by the known bore of the meter and you have flow rate; integrate over time and you have volume.

Nothing rotates. Nothing wears. The practical consequences matter more than the physics:

- It stays accurate at flows far below anything a mechanical meter will register, which is the whole ball game for leak detection.
- It measures in both directions, so backflow shows up as backflow instead of quietly cancelling out or being ignored.
- It usually arrives with a digital output already built in, rather than needing one bolted on.
- Battery-powered units are typically rated for well over a decade and are commonly sealed to IP68 so they can live in a flooded pit, which describes most Malaysian meter pits after a heavy afternoon.

The trade-off is purchase price. On a tenant sub-meter serving a small unit, the arithmetic may not favour it. On an incoming main, a chiller plant make-up line, or any point where you intend to hunt for leaks, it usually does.

Why does a water meter under-read at low flow?

Every meter is specified over a flow range, not at a single point. Under the international standard the manufacturer declares four figures: a permanent flow rate the meter is designed to run at all day, an overload rate it can survive briefly, a transitional rate, and a minimum rate below which accuracy is simply not guaranteed. In the normal zone a meter is typically held to around plus or minus 2 percent. Between the minimum and the transitional rate the allowance widens, commonly to about plus or minus 5 percent. Below the minimum, there is no specification at all.

"No specification" is doing a lot of work in that sentence. Below its starting flow, a mechanical meter does not read badly. It frequently does not read anything. The impeller sits still, or the disc creeps so slowly that friction holds it, and water passes through the body completely unrecorded.

Now think about what a leak looks like. A weeping flush valve, a pinhole in a buried service line, a gate valve passing on a disused riser: these are not dramatic events. They are small, constant trickles. A leak of half a litre a minute sounds trivial and adds up to about 720 litres a day, or roughly 21 cubic metres a month, running twenty four hours a day, seven days a week, forever. On a large multi-jet or a Woltmann meter sized for peak building demand, that flow can sit permanently below the starting threshold. The leak is invisible on the meter that is supposed to catch it, and it is invisible on the bill, because a bill is one number a month with nothing to compare it to.

This is the single most useful reason to care about meter type and meter sizing. The measurement gap is precisely leak-shaped.

Pulse outputs, and what automatic reading actually gives you

The cheapest way to get a mechanical meter talking is a pulse output. A magnet on the register passes a reed switch and closes a contact once per fixed volume: one pulse per litre, per 10 litres, per 100 litres, depending on the model. A logger or controller counts the closures.

Two things go wrong with this in practice. The first is pulse weight. Pick 100 litres per pulse on a small tenant meter and your overnight data is a flat line with an occasional step, which tells you almost nothing about the shape of consumption. Pick one litre per pulse on a big main and you generate pulses faster than a slow input can reliably count. The pulse weight has to suit the flow you actually expect, not the pipe size. The second is that a reed switch is a mechanical contact and behaves like one: it bounces, it can be triggered from outside with a magnet, and it eventually fails. Solid-state Hall or inductive pickups avoid most of that, and ultrasonic meters usually offer a proper digital register over Modbus or M-Bus, where you read the totaliser and often the instantaneous flow rate directly instead of inferring it.

The real prize is not automatic billing. It is interval data. Once you have a reading every fifteen minutes instead of every thirty days, the night-flow profile appears, and the night-flow profile is the leak report. A building with genuinely zero occupants at 3am should trend towards zero flow. If the floor of the overnight curve is 300 litres per hour and stays there for weeks, that is not usage. This is the pattern CobiNeural watches on the water side: continuous metering, a learned baseline for each meter, and a WhatsApp or email alert when the overnight floor lifts and stays lifted.

Sizing the meter for the flow, not for the pipe

The most common sizing mistake in commercial buildings is choosing the meter to match the pipe. A 100mm main gets a 100mm meter, the installation is tidy, and every low flow event for the next fifteen years disappears.

Size instead against the demand profile. That profile is shaped by how the building is plumbed, and in Malaysia most buildings pump up to roof tanks and gravity feed down, so the incoming flow is not smooth at all. It is a series of on and off surges as float valves open, with genuine dead periods in between. If you have not read how water gets to your tap in a building, that storage and pumping arrangement is where the pattern comes from.

Two practical rules follow. Undersizing costs pressure and shortens meter life, so do not simply pick the smallest body available. And where one meter cannot honestly span both a fire-season peak and a 2am trickle, either use a meter with a genuinely wide range, which is where ultrasonic earns its price, or use a compound arrangement that switches a small meter into circuit at low flow.

Why the bulk meter and your sub-meters never agree

They never do, and expecting them to will waste your afternoon. The gap has several sources stacking on top of each other.

Real losses come first: anything leaking between the utility's meter and your sub-meters is water you were billed for and never measured downstream. Tank overflows belong in this bucket, and so does every unmetered tap, cooling tower make-up line, landscape point and washdown hose that nobody sub-metered.

Then measurement error, pulling in both directions at once. The incoming bulk meter is large and blind to small flows, so it under-records the quiet hours. The sub-meters are small, more sensitive, and each carries its own tolerance. Add twenty sub-meters each within a few percent and the errors do not cancel neatly, they accumulate into a real number.

Finally, timing. If the bulk meter is read by the operator on the 8th and your sub-meters are logged at midnight on the 1st, you are subtracting two different weeks from each other. On a fast-moving building that alone can produce several percent of apparent loss.

Anyone who has reconciled electricity sub-meters will find this all painfully familiar, and for the same reasons: the argument in why your sub-meter and TNB meter never match is the same argument in a different fluid. The useful goal is not a perfect balance. It is a stable, known gap. Track the difference between the bulk meter and the sum of sub-meters every month, accept whatever percentage it settles at as your normal, and treat a change in that number as the alarm. A balance that has quietly drifted from 6 percent to 14 percent is telling you about a new leak long before anyone finds a wet ceiling.

One last piece of context on the money. In Selangor and KL, Air Selangor bills water in tiered blocks, with separate schedules for domestic and commercial accounts, and sewerage is charged separately by IWK. Because the rate per cubic metre steps up as you consume more, a leak in a large building is billed at the top block, not the average one. Check the current structure and bands directly at airselangor.com rather than relying on last year's figures.

If you want to see what your building's water actually does between midnight and 5am, talk to us about water monitoring.