A 950 mm Oak Top That Shrinks Itself Apart in a Heated Restaurant

Step 1 - equilibrium moisture content. Use an EMC table rather than intuition. At 31 C / 45% RH the equilibrium is roughly 7.7-8.0%. At 22 C / 30% RH it is roughly 6.0-6.3%. So the timber was glued at about 8% and will dry in service by roughly 1.5-2.0 percentage points.

Step 2 - movement across the width. Oak moves about 0.27-0.30% of its dimension per 1% change in moisture content tangentially. Taking 0.28%:

950 mm x 0.0028 x 1.8 = about 4.8 mm, so plan for 5-6 mm of shrinkage across the width (the range matters because the restaurant will also see hot, humid summer air, where the same top wants to swell back and beyond its glued size).

Step 3 - why it failed. A 5 mm contraction in a 950 mm width is a strain of about 0.5%. The aprons run across the grain of the top, so they do not shrink lengthwise at all; with PU adhesive on all four sides and screws in round holes, the top is fully restrained at every edge. The restrained shrinkage cannot be relieved, so it goes one of two ways: the top cups because the drier face and the restrained lower face behave differently, or the stress concentrates in the weakest glue line and opens it - which is the 0.6 mm gap observed. The screw cracks at the ends are the same stress at a stress raiser: a 4.5 mm screw in a 4.5 mm round hole leaves no clearance at all, so the screw acts as a clamp while the wood tries to move.

The design did not have a moisture problem; it had a restraint problem. Even correctly dried timber would have failed in this detail, only later.

The principle is: fix the top firmly in one place, let it move everywhere else, and never glue it along its length to a cross-grain member.

Attachment detail.

• Fix along one centre line only, using screws in round holes, so the top cannot shift bodily.
• Every other fixing must allow movement: slotted holes or figure-eight fasteners, or hardwood buttons in a groove in the apron. The slot must be long enough for the full 6 mm plus a margin - a slot at least 10 mm long for a 5 mm screw, running across the grain of the top.
• Screw holes in the aprons should be 8-9 mm for a 4.5 mm screw, with a washer, and the screw should be a part-thread screw so the thread bites into the top and the plain shank slides in the apron.
• Remove all adhesive between top and aprons; a small dab at the centre fixing point only is acceptable.
• End aprons and any breadboard ends must be attached with an elongated mortise-and-tenon, pinned only through the centre, or with slotted fixings - never glued and screwed solid.

Glue-up routine. Bring the timber to within about 1 percentage point of the service EMC by stickering it in a warm, dry conditioning area with air movement for one to two weeks, and re-measure before machining, not after. Machine and glue in one sequence so the joint faces are freshly exposed. Use a D3 or D4 adhesive with adequate open and clamp time at 31 C - at that temperature PVA gels quickly, so work in smaller glue-ups.

Finishing. Apply the same number of coats to the underside as to the face. An unbalanced film lets the top face dry faster than the underside in winter, and gives you the cupping even when the fixings are right.

Specification. Do not specify 'dry'. Specify a target with a tolerance tied to the service environment: 8% +/- 1.5%, measured at the core, because the restaurant's annual EMC cycle runs from roughly 6% in winter heating to about 11% in a humid Cyprus summer. A single figure inside that band, near the middle of the range the top will actually see, minimises the total movement in either direction. Quoting 9% 'because the meter said 9%' on a face fresh off the planer is how the fault started.

Measurement that means something.

• A pin meter reads the surface layer, and a freshly planed face is the driest part of the board. Drive insulated hammer-probe pins to about one third of the thickness, or cut a section and measure the exposed core, on at least six boards from the stack.
• Apply the meter's species correction for oak and note the timber temperature; most meters are calibrated near 20 C and read high on cold timber.
• Record meter type, correction setting and date. A hand-held resistive meter carries roughly +/-1-2% uncertainty, which is exactly why the tolerance has to be a band and not a single number.
• Verify the meter periodically against an oven-dry determination to constant mass, and against a known reference block.
• Re-measure after conditioning, not before. The reading that matters is the one taken the day of glue-up.

Consequence for the client. Because no single MC can be right all year, the specification only works together with the movement-tolerant fixing detail. Dry timber and a rigid fixing still fails; the two answers belong together.