Spacing Structural Glass Fins and Glazing Bars

Ten Gaps, Eleven Fins, and the One at the End You May Not Need

With a maximum of 1.5 metres approved and 14.4 metres to cover, the division gives 9.6, which rounds up to ten gaps, and ten gaps need eleven fins. The centres you will actually set out to are then 14.4 divided by ten, or 1.44 metres — not the 1.5 you were given. That distinction is worth writing on the drawing, because the maximum is a limit and the built centre is a dimension, and a site engineer setting out to the limit will run out of wall.

The plus-one assumes a fin standing at each end of the run, and that assumption is wrong often enough to check every time. If the wall dies into masonry returns or steel columns at both ends and those elements restrain the pane edge, the end supports already exist and the order is nine fins, not eleven. One return and one free end gives ten. A run that meets a second glazed run at a corner shares its end fin with that run, and the fin counted twice on two separate schedules is the single most common error on a multi-run job — count each straight run separately, then reconcile the shared fins once before the order goes out.

The other trap is arithmetic rather than judgement. When the spacing divides exactly into the width, the answer is the number of gaps plus one and nothing needs rounding at all: eight metres at a one metre centre is eight gaps and nine fins. A quotient that is already whole must not be rounded up, and the way it gets rounded up anyway is almost never something anyone typed. Two point one metres at three hundred millimetre centres is seven gaps, but in the binary arithmetic a spreadsheet, a script or the field on this page all run on, that division lands a hair above seven — and rounding the hair up hands you a member you do not need and a set-out dimension that no longer meets the return. Exact multiples are the one case to confirm by hand.

Keep the two counts apart on paper as well. Eleven fins bound ten panes, and fins and panes get transposed on schedules with tiresome regularity because both are described as a number of glass pieces from the same fabricator. Label the schedule by member type, not by count.

What Arrives on One Fin, and What Arrives on the Crane

At 1.44 metre centres and 4.2 metres of storey height, each intermediate fin collects a strip of wall 1.44 metres wide over its full height. At a design pressure of 1.5 kilopascals that is a uniform line load of about 2.16 kilonewtons per metre down the fin, roughly 9.1 kilonewtons in total, arriving as something near 4.5 kilonewtons of reaction into the slab and the same again into the soffit above. Those two reactions are the numbers the base shoe, the head restraint and their fixings have to be designed for, and they scale directly with the spacing — which is the clearest reason not to widen the centres after the anchorage has been signed off.

Dead load usually travels a different route, and it is worth being explicit about which route this wall uses before anything is sized. In the common arrangement the panes bear on setting blocks in a base channel and their weight goes straight into the slab, leaving the fin carrying its own self-weight plus wind. In a suspended wall hung from the structure above, the panes hang from the head and the fin carries them in tension with the slab detail there only to restrain, not to support. The two walls look identical in elevation and have opposite load paths.

The pane weight is where the spacing stops being an engineering abstraction and starts being a logistics problem. A pane of 1.42 by 4.2 metres in laminated glass of two 12 millimetre plies comes to roughly 358 kilograms. That is not a manual lift under any threshold anyone publishes; it is a vacuum lifter or a glazing robot, which brings its own floor loading, its own route through the building, and its own minimum door width. A fin of the same height, 300 millimetres deep and made of three 15 millimetre plies, is about 142 kilograms and is handled quite differently again. Weigh both before committing to an access method, because a spacing chosen purely for wind can produce a pane that will not fit through anything on the ground floor.

The Joint Is the Connection

Wind gets from the pane into the fin through one of two things: a structural silicone bond onto the fin face, or bolted patch and countersunk fittings through the glass. Silicone is the more common answer on a fin wall and the more misunderstood, because it looks like the weather seal beside it. It is not a caulk; it is a designed adhesive joint whose width is calculated, whose substrates are qualified, and whose product is specified against ASTM C1184, Standard Specification for Structural Silicone Sealants. The design method sits in ASTM C1401, Standard Guide for Structural Sealant Glazing, and where the pane is an insulating unit the secondary seal has its own requirement in ASTM C1249.

The bite width is sized by the tributary method — half the pane's short span carried by each edge joint, divided by the sealant's long-term allowable stress. On a fin wall the pane's short span is very often the fin spacing itself, so the number this whole page is about walks straight into the glue line. Take a 1.42 metre pane at 1.5 kilopascals against the widely used 138 kilopascal long-term allowable: the required bite comes out near 7.7 millimetres, before the sealant manufacturer's own minimum and their own tested allowable are applied on top. Those two always govern over the arithmetic, and the manufacturer's adhesion and compatibility testing has to be done on the actual glass, the actual interlayer edge and the actual fin.

What silicone will not do is carry the pane. Weight belongs on setting blocks, and a joint asked to hold a pane up in shear for twenty years is a joint that will creep and then let go. Bolted fittings avoid that argument and introduce a different one: every hole is a stress concentration in a material with no ductility, the hole positions are fixed in the factory, and there is no site adjustment left once the glass arrives. Choosing between the two is a design decision, but it is one the spacing constrains, because the wider the bay the larger the reaction each connection has to take.

Lay the Wall Flat and the Load Stops Balancing

A vertical wall is loaded by something that pushes and pulls in roughly equal measure and then goes away. Turn the same assembly horizontal and nothing about the load case survives. Gravity is now permanent and one-directional: the dead weight of every unit, whatever snow arrives, and an access load wherever anyone can reach or fall onto the glass. Wind uplift still acts, but upward, in the opposite direction to everything else. The governing combination changes, and so does the spacing that comes out of it — bars under a glass roof sit closer together than fins in a wall of comparable pane thickness, and that is why.

Snow is the load that catches designs out, because the site figure is not the roof figure. Ground snow load is converted to a flat roof design load through exposure, thermal and importance factors under the snow provisions of ASCE/SEI 7, or through the equivalent shape and exposure coefficients of BS EN 1991-1-3, Eurocode 1: Actions on Structures — Snow Loads. The thermal factor is where a glazed roof over a heated hall genuinely differs from a metal deck over a warehouse, and the value belongs to the code table rather than to judgement. Then there is drift: a rooflight in the reentrant between two blocks, or one running up to a taller parapet, collects a surcharge that the flat-roof figure does not contain at all, and that is exactly the position most glass roofs occupy.

Load duration matters here in a way it does not on a wall. A laminated pane behaves as a composite section only to the degree its interlayer transfers shear, and that shear stiffness falls with rising temperature and with lengthening load duration. A pane checked against a three-second gust and the same pane checked against a month of lying snow are two different sections. BS EN 16613 is the standard for determining those interlayer properties, and the interlayer manufacturer's tabulated data at the relevant temperature and duration is the input the check actually needs.

Finally, two documents can veto the makeup that made a spacing work. IBC Section 2405, Sloped Glazing and Skylights, governs what is permitted over occupied space and where retention is required beneath it. In the United Kingdom the separate question of whether a roof assembly is non-fragile is settled by test, under ACR[M]001, Test for Non-Fragility of Roofing Assemblies, published by the Advisory Committee for Roofsafety, and it is a whole-assembly result rather than a property of the glass alone. Neither document sets a bar spacing. Both can rule out the pane that let you widen one.

Bar Centres, and the Fact That Every Bar Is Also a Gutter

The arithmetic overhead is the same shape and the consequences are not. A rooflight 9.6 metres across, with a maximum bar spacing of 0.9 metres set by the pane's capacity under the combined snow and dead case, divides to 10.67, which rounds up to eleven gaps and twelve bars at 873 millimetre centres. Subtract the cap width and the joint to get the ordered pane width, exactly as on the wall.

What differs is everything the bar has to do besides hold glass up. It is a drained and ventilated channel, so any water that gets past the cap gasket and any condensate that forms on the underside of the pane run down inside the bar to a termination at the eave. It carries a bedding gasket the panes sit on, setting blocks at the low edge to stop them creeping down the slope, and a pressure cap clamping the joint from above. Every extra bar in the count is therefore another two runs of gasket, another length of cap, two more end closures and one more drainage termination to detail and to test. The linear items on a glass roof scale with the bar count, not with the glazed area, which is why a modest tightening of centres costs more than the aluminium suggests.

Deflection is the serviceability limit that decides whether any of that works. A bar that sags takes the pane out of plane with it, opens the cap gasket line at midspan and turns a drained joint into a leaking one, so the limit is set by the system supplier and is usually tighter than a bare strength check would demand. AAMA TIR-A11, Maximum Allowable Deflection of Framing Systems for Building Cladding Components at Design Wind Loads, is where the convention for cladding support members is written down, and the supplier's own tested limit for the bar section governs over any general rule.

The count itself is narrower than it looks. It is the number of parallel bars across one flat width and it is nothing else: hip bars, the ridge, the eave beam, the perimeter members and any transom lines where the panes step up the slope are all separate quantities taken off the layout. The bars also have to land on something — reconcile the bar positions against the rafter or purlin grid underneath before the layout becomes a fabrication drawing, because a bar that falls between two supports is a change to the steelwork, not to the glazing.