6 Millimetres of Deflection Corrected by a 22mm Plywood Overlay Under Ceramic Tile

A dial gauge set at midspan on a 3-metre joist bay recorded 6mm of downward movement under a 90kg point load. The Tile Council of North America gives ceramic tile a deflection limit of L/360, so a 3000mm span allows 8.3mm. Natural stone is held to L/720, which leaves only 4.16mm on the same span.

On paper, that 6mm figure passes for ceramic and fails for stone. It also leaves very little spare margin for a large format tile bonded rigidly across two joists, where remaining live-load movement can appear later as stress at the grout joints.

The chosen correction was a single layer of 22mm structural plywood, WBP grade to BS EN 636. It was laid perpendicular to the existing 18mm boards and fixed at 150mm centres along the joist lines with 50mm serrated screws. After the overlay was installed, the same load test brought the measured deflection below 3mm.

Fasteners had to reach the joists. Screws ending in the old boarding would have added a heavier floating layer with little gain in stiffness. The 50mm length was enough to pass through the overlay and the boards, then bite at least 20mm into the joist timber, which is where load transfer occurs.

Why stone was removed from the finish list

Deflection comes from span and load, while surface flatness is a separate matter. The L/360 and L/720 ratios assume a continuous rigid finish bonded to a substrate that bends under live load.

Ceramic and porcelain tolerate the wider ratio because the fired body can accept a limited amount of substrate movement before the bond line shears. Dimension stone, especially calcite-based marbles, fractures at lower strain, which is why the stricter L/720 figure exists.

At 6mm, a stone finish was off the table unless the joists were sistered or a structural beam was added. Either route meant opening the ceiling below. The 22mm overlay cost less, took less time, and kept porcelain available.

The overlay did not shorten the joist span. It increased the section modulus of the deck carried by that span, and the tile bonds to that deck. That distinction controlled the whole sequence: stiffen the deck first, then make choices about membranes, adhesive beds, and tile format.

Ditra after the plywood, never before it

Schluter Ditra is a 3mm polyethylene membrane with a dovetailed stud profile. It decouples the tile layer from the substrate, so in-plane movement in the deck does not pass shear straight into the tile bond.

On this floor, the membrane was bonded to the plywood with unmodified thinset troweled using a 3mm notch. The tile was then set into a second bed of thinset keyed into the studs. While the membrane handles horizontal movement, it gives no added resistance to vertical deflection.

Grout cracks after a Ditra installation are usually deflection-related rather than evidence that the membrane failed. A floor flexing 6mm vertically will still move that amount with Ditra above it. The membrane accommodates thermal and moisture movement in a screed, along with seasonal swelling in a timber deck, yet the plywood overlay is the part that changed the load test result. That is why it came first.

The membrane earns more of its cost over heated screed. Ditra-Heat carries the loose cable of an electric underfloor heating system inside the stud matrix, fixing cable spacing without clips and placing the heating element directly below the tile while keeping the uncoupling layer intact.

A cementitious screed expands and contracts with each heating cycle. In that assembly, decoupling does real work at the bond line. On a timber deck already corrected to below 3mm of movement, the same membrane functions mainly as insurance.

Flatness still needed a separate control

While the plywood overlay corrected deflection, lippage remained a separate risk. Adjacent 18mm sheets can differ by a millimetre at a joint, and a large format porcelain tile spanning that joint can show the step along one edge.

Tile levelling clip systems handle that surface-plane problem during the cure. Raimondi and Rubi systems use the same basic method: a base plate slides under two adjacent tiles, then a wedge or threaded cap is driven down against the tile faces so the two surfaces are pulled into a common plane as the adhesive sets.

On a 600mm by 1200mm porcelain tile, clips are placed at roughly 300mm intervals along each edge. That works out to eight to ten clips per tile. Once the thinset reaches initial set, usually the following day, the caps are removed with a lateral kick along the grout line.

Those clips reduce lippage during curing, although full adhesive contact still has to be built into the bed. Large format tile needs support under the whole body, which is why back-buttering matters. The installer burns thinset onto the back of the tile with the flat side of the trowel, places it into a combed bed, then slides it 10mm perpendicular to the comb marks to collapse the ridges before setting the clips. A void left under a large tile can become the fracture point under a chair leg years later.

The parquet adhesive that stayed hypothetical

Herringbone parquet is set into a moisture-cured polyurethane adhesive that skins in roughly 40 minutes at 20 degrees Celsius. That working time controls how large a block a single fitter can lay before the bed becomes unworkable.

This floor received porcelain, so the parquet adhesive question remained hypothetical. It matters only because the same joist bay could have carried engineered oak, with subfloor preparation moving onto a completely different path.

Engineered oak would have changed the priorities

Engineered oak does not require the L/360 rigidity that ceramic tile demands. A floating engineered board tolerates much more deck movement because the boards mechanically lock to their neighbours and move as a raft over acoustic underlay, with no bond to the substrate. The 6mm deflection that ruled out stone and raised concern for porcelain would have been irrelevant under a floating oak floor.

Oak asks first for flatness and moisture control. A common tolerance for a floating floor is 3mm under a 2-metre straightedge. Manufacturers of many 14mm to 20mm engineered boards, including Kahrs and Boen, void warranties above a defined substrate moisture content.

Over a timber deck, that points to a vapour-control underlay. Over a screed, it points to a relative humidity reading below 75 percent taken with an insulated hygrometer probe, rather than a surface meter.

Under floating oak, the 22mm plywood overlay required for this tile floor would have been wasted money. Deflection at the joist matters less for that finish than the final 3mm of surface variation and the amount of water still leaving a screed. The same joist bay could therefore demand strengthening for porcelain and moisture control for engineered timber.

What the gauge fixed and what it left open

The original 6mm reading was static, taken with a known load at midspan on a dry afternoon. It described the deck in that condition. It did not describe how the joists would behave after two winters of heating dried the timber, or how firmly the 150mm screw pattern would hold after the plywood reached equilibrium with the room.

After the repeat test showed less than 3mm of movement, the porcelain installation could proceed in the intended order: plywood, membrane, adhesive, tile, grout. The finished surface hid the structural change completely, leaving the unresolved tension below the grout line where a rigid finish crosses living timber.