Underfloor Heating Loop Balancing: Set Every Flow Meter Right

The screed went down in March. The boiler is sized, the pump is running, every actuator clicks open when its thermostat calls — and the client's back bedroom sits at 17 °C while the hallway nobody occupies is uncomfortably warm. The installer turns the bedroom thermostat up. Nothing happens, because nothing about a thermostat can move water that has already chosen a different route.

That job does not have a heat problem. It has a distribution problem, and the fix is a set of numbers that should have existed before the pipe was clipped down: a target flow rate, in litres per minute, for every single loop on the manifold.

Almost every guide to underfloor heating balancing will tell you those numbers usually land somewhere between 1 and 3 litres per minute. That is true, and it is useless, because it is an average of other people's houses. What none of them show you is how to get the number for your room — the arithmetic that turns a heat load in watts into a reading on a flow meter. This article writes that arithmetic out in full, proves it against a manufacturer's own published commissioning table, then runs a complete six-loop manifold through it at two different design conditions.

What Balancing Actually Fixes: One Pump, Six Loops, Unequal Resistance

A manifold is a parallel circuit. One pump pushes water into a flow bar, that water splits between however many loops are connected, and it recombines in the return bar. Nothing in that arrangement forces the split to be fair.

Water divides between parallel paths in inverse proportion to their resistance, and a loop's resistance rises with its length. So a 26 m hallway loop and a 92 m living-room loop hanging off the same manifold do not receive similar flows — the short one takes far more than its share and the long one is starved. The hallway overheats. The living room never reaches setpoint. Both faults have one cause.

This is why turning up a thermostat cannot fix it. A thermostat only decides whether its actuator opens; it has no authority over how much of the pump's output comes through that particular port once it is open. If the loop is hydraulically starved, opening it for longer just starves it for longer.

The two adjusters, and which one you actually use

Almost every modern manifold gives you two devices per loop, and they are not interchangeable:

• The flow meter (topmeter) on the flow bar — a float in a graduated tube that both displays and sets the litres per minute passing through that loop. This is the balancing device. It is the one you turn.
• The return valve on the return bar — the port that carries the thermal actuator and takes the on/off command from the room thermostat. This is a control device, not a balancing device.

Confusing the two is the most common way a manifold ends up "balanced" by throttling return valves, which produces a system that appears to work until the actuators start cycling and the settings quietly stop meaning anything.

Which standard governs which decision

Two parts of the same European standard series cover this work, and mixing them up sends people looking for design rules in the installation document.

One detail from the 2021 revision matters if you are working from an older design method. Compared with the 2009 edition, EN 1264-3:2021 added four new pressure-loss subclauses — 4.1.3.1, 4.2.3.1, 4.3.3.1 and 5.2.1.1 — so hydraulic resistance is treated more explicitly in the current dimensioning basis than it was before. If your reference material predates that revision, the pressure-loss treatment it gives you is the older one. Both parts are published by BSI as British Standards; the catalogue entry for part 3 lists the full change record.

The Equation Behind Every Loop Flow Table

Here is the whole thing. A loop has to deliver a certain number of watts into a room, and it does that by arriving warm and leaving cooler. How much water it needs depends on how much it is allowed to cool down on the way round.

You know Q, because it is the room's heat requirement. You choose ΔT, because it follows from your heat source. So rearrange for the thing you do not know:

The two constants, and why you can stop worrying about them

Both constants vary with temperature, which raises an obvious question: at which temperature do you evaluate them? The honest answer is that for this purpose it barely matters, and it is worth seeing why rather than being told to ignore it.

Across the entire band a floor-heating system will ever run, c_p moves by 0.14 % — it is effectively a constant, and using 4.1789 kJ/kg·K everywhere introduces an error far smaller than your heat-loss calculation already carries. Density falls further, 0.60 % between 20 °C and 40 °C, which is why treating one litre as one kilogram is a shortcut rather than an identity. At commissioning accuracy, on a float meter you are reading by eye against a printed scale, it is a perfectly good shortcut. It is worth knowing it is one.

The calculation in one line

Put the conversion together and you get a formula you can keep:

Proof the Equation Is Right: Rebuilding a Manufacturer's Commissioning Table

A formula in an article is worth very little on its own. So here is a test that anyone can repeat.

Uponor's Underfloor Heating Installation Guide prints a commissioning table on page 51 giving the design flow rate for a loop, indexed by loop length and by the floor's heat output. It is presented as a lookup — no derivation, no explanation, just numbers an installer is expected to trust. Elsewhere in the same document, on page 33, the company states the design basis behind its tables: a ΔT of 7.5 °C, with the 16 mm pipe laid at 200 mm centres.

That is everything needed to check the arithmetic. Pipe at 200 mm centres means one metre of pipe serves 0.2 m² of floor, so a loop's served area is its length × 0.2. Multiply by the output in W/m² to get the loop's watts, then run the equation at ΔT 7.5 K.

Nine values out of nine, every one 3.5 % below the published figure — and note that it is the same 3.5 % every time, not a scatter. A constant offset across nine independent rows is the signature of a deliberate design margin rather than accumulated error: the manufacturer is rounding upward — the safe direction, because a loop given slightly too much flow underperforms invisibly while one given too little leaves a cold room.

The conclusion matters more than the exercise. Those published tables are not proprietary engineering; they are this equation with the rounding already applied and the design assumptions buried. Which means the moment your job differs from the table's assumptions — a different ΔT, different pipe centres, a room whose output is not 50, 70 or 100 W/m² — the table stops applying and the equation still does.

A Worked Six-Loop Manifold, Balanced at Both 5 K and 10 K

Here is a complete job. Six loops on one manifold, real heat loads, real loop lengths. Everything in the two right-hand columns is the equation from earlier — nothing is looked up.

What the two columns are telling you

Same house. Same 7,680 W. The pump has to move exactly twice as much water in the right-hand column as in the left, and this is the single most consequential fact on this page.

Design ΔT follows from the heat source, and so does the pipe you specify to carry it — a point we take further in our comparison of PE-RT and PEX for floor heating. A boiler system is typically designed around 10 K — flow 50 °C, return 40 °C — while a heat pump is typically designed around 5 K, flow 40 °C and return 35 °C, because the narrower drop lets the machine run at a lower flow temperature and a better coefficient of performance. Those are conventional design figures rather than requirements of any standard, and it is worth noting the manufacturer table dissected above sits between them at 7.5 K.

The consequence is not conventional at all, though: swapping a boiler for a heat pump on an existing manifold doubles the flow demand on every loop. A circulator that was comfortable at 0.667 m³/h now needs to deliver 1.333 m³/h against a higher resistance, because pressure drop climbs roughly with the square of flow. Retrofits fail here regularly, and the failure looks exactly like the one in the opening paragraph — some rooms fine, the far ones cold.

The loop nobody's rule of thumb saves

Now look at the bathroom, and divide each loop's flow by its length to see how hard each metre of pipe is working:

Five loops cluster tightly around 75 ml/min per metre. The bathroom is 29 % above them all — it needs the most flow of any room except the kitchen, delivered through one of the shorter loops, because bathrooms carry high heat loss and a higher target floor temperature over a small area.

This is the case the popular advice cannot handle. "Keep your loops within 10–15 % of each other in length" is sound guidance and it is discussed properly in the next section, but it is a rule about length, and the bathroom's problem is watts per square metre. Its loop is already a sensible length. Making it longer to match the others would make things worse, not better, because you would be adding resistance to the loop that already needs the most flow.

What the bathroom actually needs is to be recognised at design time as the hydraulically demanding circuit and treated accordingly: kept short, kept close to the manifold, and given the pump head it requires. If you only discover it at commissioning, your options have narrowed to the pump.

Setting the Flow Meters Without Chasing Your Own Adjustments

With a flow schedule in hand, the physical work is straightforward — provided it is done in the right order. The order is what stops you chasing adjustments around the manifold for an afternoon.

Why the second pass is mandatory

Loops on a shared manifold are hydraulically coupled. Open loop 4 and you change the pressure available to loops 1, 2 and 3 — the flow you carefully set on loop 1 is not the flow loop 1 has once the whole manifold is open. Uponor states this plainly in the same commissioning chapter: repeat the process for each loop, then go back and carry out fine adjustments, because each loop will have a mutual effect on the others.

In practice one full pass plus one fine-tuning pass gets a six-loop manifold close enough. Skipping the second pass is the most common reason a manifold that was "balanced" still has a cold room.

When you run out of adjuster

The flow meter has finite travel. On the Uponor topmeter, three full turns from shut is fully open and it cannot be adjusted further — and every float-type meter has some equivalent limit. This turns an annoyance into a genuinely useful diagnostic.

If a loop is wide open, every other loop is set correctly, and it still will not reach its target, the loop is not the problem. You have run out of hydraulic authority, and the answer is upstream: pump speed, or a design that asks more of that circuit than the pump can deliver. Uponor's guidance says the same — if the topmeter or lockshield is fully open and design flow is not achieved, adjustment of pump speed may be necessary.

There is a related ceiling worth knowing before you design, not after — and it is a ceiling that moves, which is why it catches people out. Float-type flow meters do not share one scale. Uponor's TM topmeter is specified for "setting and visual indication (0–6 l/m)", while aftermarket replacement topmeters are widely sold at 0–5 L/min and 0–4 L/min units are common on budget manifolds. Nothing on the manifold tells you which you have until you read the printed scale.

That turns the worked schedule into two different jobs. At ΔT 5 K the kitchen loop needs 5.06 L/min: readable on a 0–6 topmeter with about 15 % of the scale to spare, and simply not settable on a 0–5 or 0–4 meter. The same schedule is commissionable on one manifold and impossible on another, and the difference is a component spec nobody checks at design stage.

The manufacturer's own tables show the discipline — the 16 mm table rebuilt earlier tops out at 3.0 L/min, and even Uponor's higher 20 mm table peaks at 4.6 L/min for an 80 m loop at 100 W/m², still inside its own meter's range. Check the scale on your specific manifold before the loop schedule is fixed. A loop you cannot read is a loop you cannot balance.

Designing Loop Lengths So the Manifold Can Be Balanced At All

Everything above assumes the loop schedule is balanceable. Plenty are not, and that is decided on paper long before anyone reaches for a screwdriver.

The similar-lengths rule, and what it actually buys you

The trade rule of thumb is to keep loops on one manifold within roughly 10–15 % of each other in length. Where lengths are close, the loops present similar resistances, the flow splits nearly evenly on its own, and the balancing valves end up doing very little work — some designers will leave manual balancing valves effectively wide open when lengths are within about 10 %.

Treat that as design practice rather than a requirement of any standard. The mechanism behind it is real and is what you should actually reason from: the wider the spread of resistances, the harder each valve has to throttle, and the more of the pump's head is being deliberately destroyed to make an unequal system behave. A manifold whose loops run 22 m to 92 m can technically be balanced. It just wastes pump energy permanently to do it, and it leaves you no margin when something changes.

The ceiling on any one loop

Pipe diameter caps how much floor a single port can serve, because a longer loop at a given flow means more pressure drop. Published design tables from named manufacturers put maximum loop lengths in the 80–120 m band depending on pipe size and system output — Uponor's selection tables run 75 / 100 / 120 m rows for 20 × 2 mm PEX and 80 / 100 / 120 m rows for 16 × 2 mm multilayer and 15 × 1.5 mm PEX, and the company's design guidance gives 100 m as the maximum for 16 mm and 80 m for 12 mm, including the tails to and from the manifold.

That last clause catches people out. A room 15 m from the manifold spends 30 m of its loop allowance just getting there and back, before a single metre goes into the floor. We cover the pipe-size side of this decision in more detail in our guide to choosing underfloor heating pipe, and the size and pressure-rating relationships in PEX pipe sizes.

Splitting a room, and what that does to your order

When a room would exceed the ceiling, split it across two ports rather than forcing one long high-resistance circuit. That is why the worked schedule shows the living room as two 46 m loops instead of one 92 m loop — and it is the moment loop design turns into a purchasing decision.

Count ports from the loop schedule, never from the room count. Six rooms in that example produce six loops only because the living room split absorbed one port that the bathroom did not need; a different layout gives a different answer. Manifold sets are commercially available from 2 up to 12 outlets, with modular types built from 2, 3 and 4 port sections threaded together up to a 12-outlet maximum, so the practical constraint is usually that you must decide before you order rather than that the hardware cannot accommodate you. If you are choosing between manifold types and materials, our PEX manifold guide covers that side; the coil and fitting ranges themselves sit in our product catalogue.

Before the screed goes down

A short list, because everything on it is impossible to fix afterwards:

• Loop lengths recorded per port, measured off the coil rather than estimated from floor area.
• Target flow calculated per loop, at the ΔT your actual heat source will run — not at a generic figure.
• Every target inside the flow meter's readable range. If one is not, the design changes, not the commissioning.
• The high-output rooms identified — bathrooms and rooms with large glazing — and kept short and close to the manifold.
• Leak test completed to EN 1264-4 before the screed, since that standard carries the installation and leak-test procedure. Ask your coil supplier for the batch test report covering the reel you are about to bury, and on a project large enough to justify it, order a sample coil ahead of the main delivery so the fittings and manifold adaptors are proven against the actual pipe before site.

The Decision Checklist

Balancing is not a knack. It is a schedule of numbers, produced before installation and confirmed at commissioning, and the arithmetic behind it fits on one line.

The single most useful habit is writing the flow schedule down and leaving it at the manifold. A balanced system with no record is one thermostat change away from being an unbalanced system that nobody can diagnose.

Frequently Asked Questions