Screeding Over Underfloor Heating Pipes

Sixty-Five Millimetres From What?

The tubing crew has gone. What they left is a mat of 16 mm pipe clipped down at 150 centres, insulation boards that rock slightly underfoot in the corner where the boarder ran out of full sheets, a manifold holding pressure on a gauge nobody has photographed, and a drawing with two words on it: 65 mm screed. That figure is the first argument of the day, because sixty-five millimetres is a dimension with no datum attached. Measured up from the insulation it is one order. Measured up from the crown of the pipe it is another, and across a modest ground floor the gap between those two readings is most of a mixer load.

The dimension the heating system actually controls is the cover over the crown, and it is controlled for two reasons at once: it is the depth that has to carry load in bending over a void of pipe, and it is the depth heat has to spread sideways through before it reaches the surface as something other than stripes. EN 1264, the water-based surface embedded heating and cooling standard, is where that cover requirement lives, alongside whatever the screed system supplier prints for the specific product. The total depth you actually place is that cover plus the outside diameter of the tube, because the tube is sitting underneath it — and where the pipe is carried on a raised rail or a castellated panel rather than clipped flat to the insulation, the rail height goes on top of that again.

None of which helps until you know where the base is. A power-floated slab, a beam-and-block deck with a levelling sand blinding, and a room where the insulation has been laid over a slab that fell 8 mm across its length are three different quantities for the same drawing. Set a laser to finished floor level, take the base down at a grid of points, and write the worst of them down. That reading is not pedantry — it is the single biggest term in the waste allowance, and it is the one that decides whether the last bay gets screeded on the day or on a Saturday.

The Class Is a Letter and Two Numbers

Where the screed comes from a manufacturer rather than out of a gauge box, stop specifying a ratio and start specifying a class. EN 13813 designates screed materials as a binder code followed by strength classes — CT-C25-F4 and its relatives — and that designation, verified by the EN 13892 test methods, is a far more useful thing to hold a supplier to than a proportion nobody will witness being batched.

Over a heated floor the flexural figure earns its place ahead of the compressive one. A floating screed sits on compressible insulation, spans between clip rails, and has a lattice of round voids running through its lower half at regular centres. That is a thin plate in bending with a deliberate line of weakness under every pipe run, which is precisely the loading a flexural class describes and a compressive cube does not. It is also why cover over the crown is a structural number and not only a thermal one.

Bays, Edges, and the Places It Will Break Anyway

A heated screed is a floor that is deliberately made to change size. Every bay layout decision follows from that. The screed system supplier sets the maximum bay area and the aspect ratio it will tolerate; heated floors are commonly bayed more tightly than unheated ones for the obvious reason. Where a bay runs long and thin it wants breaking regardless of area, and every doorway gets a joint whether the geometry demands one or not, because a narrow throat between two rooms is where a floor cracks if you do not pre-empt it.

Perimeter edge strip is not trim. It is what allows the whole bay to grow when the floor first runs at design temperature, and it has to be continuous, full depth, and carried up past finished floor level to be cut back after the covering goes down. A strip that has been trimmed flush with the screed before tiling, or stopped short around a kitchen island plinth, hands the floor a hard point to push against. Where a flowing screed is being used, the strip and the membrane also have to be taped into a genuine tray, because a self-levelling material will find every gap under a door frame and every service penetration and go looking for the rest of the house.

Pipe crossing a joint is the detail that gets skipped. Sleeve it, cross it square, and keep the crossing as short as the geometry allows. The manifold fan is the same problem at higher stakes: a dozen flow and return tails leaving one cabinet at close centres is the hottest square metre in the building, and left unsleeved it will run visibly warmer than everything the design promised and crack the screed in front of the cupboard.

Surface regularity is the last thing to agree before anyone starts, because it decides how the screed is finished and by whom. BS 8204-1 sets the SR classes measured under a two-metre straightedge — SR1 at 3 mm, SR2 at 5 mm, SR3 at 10 mm — and the flooring specification, not the screeder's pride, is what selects one. Elsewhere the same argument is had in F-numbers to ASTM E1155. Either way, agree the class in writing, because an SR1 floor and an SR3 floor are different labour on the same day's materials.

1. Mark the bay lines on the insulation before any screed arrives, with the doorway breaks set out against the door schedule rather than the room names.
2. Run the edge strip continuous round the perimeter and round every column, upstand and plinth, full depth and standing proud of finished floor level.
3. Tape the membrane laps and the strip junction into a closed tray if the material is flowing, so nothing can drain out under a frame.
4. Sleeve every pipe crossing a bay line, and sleeve or insulate the manifold tails for their first stretch out of the cabinet.
5. Photograph the pressure gauge, with the day's date visible, before the first barrow lands and again when the last bay is closed.

Compaction Is the Job

A semi-dry sand-cement screed is not poured, it is placed and beaten. The consistency is the old one: a handful squeezed in a gloved fist holds together, keeps its shape, and gives up no water. Anything wetter is easier to spread, weaker when it hardens, and shrinks more on the way — which on a heated floor means curled bay edges and cracks over the pipe lines within a season. The temptation to add water arrives at about two in the afternoon on a hot day, and it should be answered with a covered barrow and a shorter run, not a hose.

The pipe zone is where compaction actually fails. Material bridging across the crown of a tube leaves a void under it that nobody sees, because the surface above closes up perfectly. It becomes a hollow patch under a tile, a debonded area on an unbonded screed, and eventually a crack running dead straight down the floor at exactly the tube centres. Work the material in around and under the pipe first, in one operation, before bringing the whole bay up to the datum strips and compacting to level.

Flowing calcium sulfate screeds trade this problem for a different one. They compact themselves and encapsulate the tube better than any amount of beating, which is part of why they can be laid thinner; in exchange, the tube has to be genuinely restrained against flotation, the tray has to be watertight, and the surface laitance has to be dealt with by the programme rather than discovered by the flooring contractor. Neither material is the easy one. They are difficult in different places.

Three Clocks, and Nobody Wants to Wait for Any of Them

Curing, drying and commissioning are three separate processes with three separate purposes, and conflating any two of them cracks floors. The first is hydration: the screed needs its mix water to react, not to evaporate, so it gets covered and kept damp for the period the material's own literature states. A screed that felt dry on the third day was not cured quickly, it was cured badly, and its strength class is now a claim rather than a property.

The second clock starts when the first one finishes and belongs to whoever is laying the floor covering. Cementitious screed dries slowly and from one face, and the rule of thumb the UK flooring codes are usually quoted for — roughly a day per millimetre for the first fifty millimetres in good conditions, considerably slower beyond that — is a planning figure and nothing else. The only acceptable evidence is a measurement: in-situ relative humidity probes to ASTM F2170, the calcium chloride method to ASTM F1869, or a surface hygrometer box as described for the UK flooring codes of practice such as BS 8203. ACI 302.2R exists precisely because slabs receiving moisture-sensitive flooring generate more claims than any other detail in a building.

The third clock is commissioning, and EN 1264-4 is where its shape comes from: the floor is not heated until the screed has reached the age the standard and the manufacturer specify, longer for cementitious material than for calcium sulfate. Then it runs at a low flow temperature for a stated period, steps up to the maximum design flow temperature for a further stated period, and every one of those dates and temperatures is written down and signed. Take the durations from the standard and from the screed manufacturer, not from a forum and not from the programme.

That cycle proves the system and drives out residual moisture. It is not the drying test, it does not replace the drying test, and it is emphatically not a service to the plasterers, who will ask. Running an untested, uncommissioned heated floor to dry a building out is how a screed gets taken from ambient to design temperature in one afternoon, and how a floor that would have lasted forty years acquires its first crack before anyone has moved in.

Write the three dates into the programme as three separate line items with three different owners. Where they appear as a single bar labelled screed, somebody will price the bar and somebody else will compress it.

What the Cracks Are Telling You in February

Straight cracks running the length of the floor at regular spacing, and the spacing matches the tube centres: that is cover. Either there was never enough screed over the crown, or the material bridged during placing and left a void that the first heating cycle turned into a hinge. It is diagnosable from a tape measure alone, and it is the failure this entire page is arranged to prevent.

Lifted, hollow-sounding bay edges and corners are drying shrinkage acting on the top face — too wet a mix, too fast a dry, no curing, or all three in sequence. Fine map crazing across the surface is a different animal: a laitance-rich skin brought up by over-working the surface. It is cosmetic on its own and fatal to an adhesive bond, which makes it the flooring contractor's problem and therefore, in practice, an argument.

Then there is the crack that appears the first genuinely cold week of the year on a floor that passed everything. Look for a joint that was drawn and not cut, a doorway with no break in it, an edge strip trimmed off flush before the tiler arrived, or an unsleeved manifold fan. A heated screed that has been ordered to the right depth and gauged to a real ratio still fails at whichever restraint nobody thought about, and every one of those four is decided before the mixer starts, not after.