Building a Curved Drywall Wall

Stud Centres Are a Surface-Quality Decision, Not a Strength One

Nonstructural steel framing is specified by ASTM C645 Standard Specification for Nonstructural Steel Framing Members and installed to ASTM C754 Standard Specification for Installation of Steel Framing Members to Receive Screw-Attached Gypsum Panel Products; in the British and European system the equivalent component standard is BS EN 14195 Metal Framing Components for Gypsum Board Systems. None of them will hand you a stud spacing for a radius, and that is not an omission. On a straight wall the spacing is governed by the height, the board thickness and the limiting deflection. On a curve it is governed by how flat the board is allowed to go between bearing points, which is a finish question, and it will almost always give you a tighter answer than the structural one.

Run the offset. A sheet bridging 400 mm centres on a 1.5 m radius stands a shade over 13 mm inside the true curve at midspan — and because the sheet is screwed flat to each stud, that 13 mm is not spread as a gentle sag but concentrated as a straight facet with a crease at every stud. Halve the spacing to 200 mm and the offset drops to about 3.3 mm; at 150 mm it is under 2 mm; at 100 mm it is under a millimetre and the board is doing something the finisher can actually bury. That is the whole reason curved partitions get framed at what looks like an absurd stud count, and it is worth being able to show the number to whoever is questioning the take-off.

The stud's own flange is a flat too, but it is a small one — a 38 mm flange on a 600 mm radius contributes about three-tenths of a millimetre, which is noise next to the span. What does matter is the track. Two routes are in general use: proprietary flexible or pre-slotted radius track, which is the right answer when the radius is tight or the programme is short, and standard track with one flange relieved with snips at intervals so it can be sprung to the line. If you are cutting your own, put the snips in the flange and leave the web alone — the web is the only thing holding the track's own curve fair, and a track cut through the web goes round in a series of short straights that every stud then inherits.

Set out along the arc, not the chord. A layout marked with a tape pulled straight across the opening bunches the studs at the ends and opens them in the middle, and nobody re-measures framing that looks evenly spaced from the doorway. Mark the divisions on the track after it is fixed to the struck line, walking the tape round the curve.

1. Strike the arc on the slab with a trammel if the centre point is reachable, or from a chord and measured offsets if it is not.
2. Prove the struck line at three or four stations with a plywood pattern cut to the finished radius before anything is fixed to it.
3. Transfer the line to the soffit with a plumb laser at close stations, and check the head arc against its own pattern.
4. Fix track to both lines, relieving one flange with snips if you are not using purpose-made radius track.
5. Walk the tape around the curve to mark stud divisions on the track, never across the chord.
6. Stand the studs, then sight along the completed frame from one end before a single sheet is offered up.

Kerfing: Taking the Stiffness Out of the Back

A kerfed sheet has had a series of parallel cuts run across the back, through the back paper and most of the core, stopping just short of the face paper: what is left is a row of gypsum blocks hinged on a continuous strip of paper. The cuts run parallel to the axis of the curve — vertical, on a wall curved in plan and sheeted horizontally — and they have to be dead parallel to each other, because a kerf set at an angle twists the sheet as it closes and puts a wind in the wall that no compound will take out.

Two separate limits decide the spacing, and it is worth working both because they answer different questions. The first is whether the kerfs can close far enough to turn the corner at all. A kerf of width w cut to a depth of the board thickness t rotates the two blocks either side of it by about w divided by t radians when it closes tight, so the spacing that just turns a radius R with every kerf shut is R times w divided by t. A 3 mm saw kerf in 12.7 mm board on a 1.5 m radius gives about 354 mm. That is the theoretical maximum, and on its own it is a useless number.

It is useless because of the second limit, which is the one that actually governs at any wall-scale radius: the block of board between two kerfs stays flat. At 354 mm spacing on that 1.5 m radius, the sheet stands about 10 mm inside the true curve between every pair of cuts — a facet you could find with your hand in the dark. Work it the other way instead. If d is the deviation your finish can absorb, the spacing is the square root of eight times the radius times d. On a 1.5 m radius that is about 110 mm for a millimetre, about 77 mm for half a millimetre, and about 55 mm if you want a quarter. On a 3 m radius, a millimetre buys you 155 mm — which is why the spacings that get used in practice on gentle sweeps sit in the region of 150 mm and the ones on tight returns sit at a third of that.

Set spacing that way and the kerfs will not close fully, which is the correct and desirable outcome. A sheet whose every kerf has shut hard has run out of adjustment: it is at its limit, it cannot be persuaded another degree, and any discrepancy between the sheet and the frame has to come out of the frame. The closure arithmetic earns its keep at the other end of the scale — column wraps, tight bulkhead returns, and anywhere a narrow blade kerf simply has not got enough width in it to give up the angle required.

Cutting is bench work, not wall work. Set blade depth on an offcut of the same board and prove it by flexing that offcut: a blade grazing the face paper turns the kerf into the crack the wall eventually fails at, and a blade set shy leaves too much core and the sheet fights you. Cut face down on a flat surface, working to marks stepped off one end — a router with a fence, a circular saw with a straightedge, or a purpose-built jig, in ascending order of how many sheets you have. Carry the sheet on edge with the kerfs inboard, then fix it progressively from one end or from the centre out, so it is fed onto the curve rather than sprung onto it.

1. Set the blade depth on an offcut and prove it by flexing the offcut, not by looking at the cut.
2. Mark the spacing along both long edges of the sheet from the same end, so the cuts stay square across it.
3. Cut every kerf with the sheet face down on a flat bench, working to the marks rather than to a gauge.
4. Carry and stack the cut sheet on edge with the kerfs inboard.
5. Screw from one end, or from the middle out, feeding the sheet onto the frame a stud at a time.
6. Fill the kerfs as the board manufacturer directs before the face is closed up.

The Route That Needs No Saw

The alternative is to bend two thin sheets instead of relieving one thick one. The first layer goes on the framing and is screwed; the second is bonded to it with laminating adhesive and held with temporary fasteners until it grabs, joints staggered so none runs through both layers. The result is a continuous core with no hinge line under the finish, and on a wall that will take trolley knocks it is the better wall for that reason alone. It is also the honest answer for the concave face, where kerfs would open rather than close. What it costs is handling, programme, and an argument about thickness: the build-up has to land on the same plane as the flat wall it meets, or the bullnose at the junction sits on a step.

Adhesive is an area item, spread over the whole face rather than dabbed like a direct bond to masonry, and the coverage figure belongs to the product: take it from the technical data sheet for the application method you are actually using, because a notched trowel and a bead pattern do not consume the same quantity. The temporary fasteners holding the second layer while it cures also leave holes on the most visible surface in the room, and whether they come out or stay in is the adhesive manufacturer's call, not the fixer's.

Taking the Bead Round Without Scalloping It

Wherever the arris itself is curved — the head of an arched opening, the top edge of a partial-height radius wall, the line where a curved wall meets a curved bulkhead — the bead follows the arc, and that is where a curved wall gets found out. ASTM C1047 Standard Specification for Accessories for Gypsum Wallboard and Gypsum Veneer Base specifies the accessory and says nothing about how it goes round a radius. Two families do the work: flexible archway bead, vinyl or paper-faced, manufactured to bend; and rigid bullnose stock, which is not, and so has to be cut into short pieces that each chord the curve.

The arithmetic is the same one from the framing and the kerfs, applied to the bead's own length. Each rigid segment is a chord, so the middle of it sits inside the true curve by the segment length squared over eight times the radius, and that shortfall shows up either as a flat in the middle of the piece or as a gap at the joints, depending on how you set it. A 600 mm piece on a 3 m radius is about 15 mm out — a scallop you could read from the far side of the room. Holding half a millimetre on that same 3 m radius means segments of about 110 mm, which is forty-odd pieces in a ninety-degree sweep; even a 10 m radius wants 200 mm pieces for the same tolerance. Those numbers are the reason flexible archway bead exists, and they are worth running before anyone commits to segmenting rigid stock on a tight curve.

The count has a second cost that the deviation does not capture. Every joint between two segments is a place where two nose profiles meet at an angle, on the most exposed line in the room, and every one of them is a joint the finisher has to bury in compound. Ten of those is a detail; forty is a different job with a different price, and it is worth putting in the quotation as a line rather than absorbing it. That is also the practical brake on driving the segment length down indefinitely to chase a smaller offset.

Set the bead to a line, not by eye. There is nothing to sight along on a curve, so string a line or run a plywood pattern and bed each piece to it. Paper-faced bead is set in compound, which suits a curve because the compound takes up the small variations; crimping a metal bead round an arc distorts the nose at every fixing, permanently. At the tangent point carry the bead past the transition rather than stopping short of it, and take curved and straight bead from the same product line — two nose radii differing by a couple of millimetres meet in a visible step exactly where the eye is already looking.

1. Decide flexible against segmented from the offset the radius produces, before the bead is ordered.
2. If segmenting, take the shorter of the manufacturer's rated chord and the length your finish tolerance allows.
3. String a line or cut a pattern to the finished radius and bed every piece to it.
4. Run the bead past the tangent point into the straight, using the same product for both.
5. Count the joints as a finishing item in their own right, not as part of the bead run.