Cable Net Structures

What is a cable net?

A structural surface made of high-strength cables tensioned in two directions, forming a grid. The infill — glass, ETFE, membrane or metal panels — is carried at the intersections, so the net is not a support for a structure; the net is the structure.

Net geometry: single, cross or diagonal net, with anticlastic double curvature. Grid spacing: 1,000 – 2,000 mm. Cable diameter: 10 – 40 mm.

The distinction from its neighbours in the range:

• 05 Cable structures The cables are the structure, carrying a covering over open space.

• 06 Cable-supported curtain wall Vertical cables replace mullions; glass is point-fixed to them.

• 07 Cable-supported skylight Cables reinforce a steel lattice so its members stay slim.

• 08 Cable net structures Cables in two directions form the surface itself.

Where a vertical glazed net sits. A tall glass wall on a two-way net is legitimately both 06 and 08. Read 06 if the question is envelope performance — air permeability, watertightness, U-value. Read this page if the question is the net: its curvature, its dynamics, and whether the infill can tolerate its movement. The street elevation of the Moghaddas Ardebili Office, a glazed cable-net wall, sits in that overlap.

Why curvature makes a net stiff

Stability comes from opposing curvature. A flat net has no stiffness of its own, and without enough tension it will oscillate.

This is the first thing to understand, because it constrains the architecture before anything else does.

A cable has no bending stiffness. A net made of cables therefore has no stiffness of its own — it gets stiffness only from geometry and pre-tension. The geometric mechanism is anticlastic curvature: the surface curves downward in one direction and upward in the perpendicular one, like a saddle.

That shape creates two families of cables doing opposite jobs:

• The sagging family hangs downward and carries load the way a suspension cable does.
• The hogging family arches upward and is tensioned against the first.

Pre-tension between them keeps both families taut. When load arrives, it tightens one family and slackens the other — but because both started tight, neither goes slack, and the surface resists rather than simply moving until it finds a shape.

• Anticlastic — a saddle One family of cables sags and carries load; the other arches and is tensioned against it. Both stay taut, so the surface resists rather than simply moving.

• Flat No opposing family to pull against. All its stiffness comes from pre-tension, so it needs far more, and still moves much further under the same load (dashed).

• Synclastic — a dome Both directions curve the same way, so neither can be tensioned against the other. It needs internal pressure or ballast to hold its shape.

Curvature exaggerated. Thick accent lines are the sagging family; thin lines the family that arches against it.

Two consequences worth stating plainly:

A flat net has only one source of stiffness. With no opposing family to pull against, all of its resistance comes from how hard it is pulled. It needs far more pre-tension, it moves much further under the same load, and under fluctuating wind a flat net without enough tension oscillates. Flat cable walls are built — that is System 06 — but their deflection limits and their anchorage are the price of that flatness.

A synclastic surface — domed, curving the same way in both directions — does not work either, without internal pressure or ballast to push against. That is why inflated ETFE cushions exist as a separate system. Anticlastic geometry is self-stabilising; domed geometry is not.

So the amount of curvature is a structural quantity, not a styling choice. Flatter surfaces need higher pre-tension, which means bigger cables and a much bigger anchorage. An architect who wants a nearly flat net is asking for the most expensive version of the system.

Geometry, force and anchorage are one calculation

Change the curvature and the forces change; change the pre-tension and the shape changes.

That sentence separates this system from everything else in the range, and it has a specific practical meaning: the three cannot be designed in sequence. They are solved together or not at all.

On a framed façade you can change the glass, then check the mullion, then check the bracket — each step taking the previous as fixed. On a cable net there is no such chain:

• Increase the curvature and the cable forces drop, so the cables get smaller and the anchorage gets lighter — but the surface bulges further from the plane the architecture wanted.
• Flatten it back and the forces rise sharply, so cables, fittings, edge structure and foundations all grow.
• Raise the pre-tension to reduce deflection, and the shape itself shifts, along with every reaction at the boundary.
• Move one anchor point and the whole surface re-forms. There is no local edit.

What this means for the design process:

• The architect's control is over boundaries and curvature, not over the surface between them. Choose where the edge sits and how much saddle the design will accept; the surface follows from form finding.
• Value engineering does not work item by item. Reducing cable diameter without changing geometry is not a simple saving; it is a different structure, and it has to be form-found and checked again from the start.
• The calculation has to exist before the primary structure is fixed, because the boundary reactions are an input to somebody else's design.

Grid spacing, panel size and the loop between them

Grid spacing: 1,000 – 2,000 mm. Maximum panel size: 2,000 × 3,000 mm typical.

These two numbers are linked, and they close a loop that catches people out.

The grid sets the panel. The panel size sets the glass thickness, because a larger pane spanning between four point supports needs more glass. Thicker glass is heavier. That weight is carried by the net, so the cables grow, the pre-tension rises, and the anchorage grows with it.

Run it the other way and it works in your favour: a tighter grid means smaller, thinner, lighter panels, smaller cables and a lighter edge structure — at the cost of more nodes, more fittings and a more visible grid.

• 1,000 mm grid Smaller, thinner, lighter panels, smaller cables and a lighter edge structure — for more nodes, more fittings and a more visible grid.

• 2,000 mm grid Fewer nodes and fittings — for larger panes of thicker, heavier glass, which grow the cables, the pre-tension and the anchorage.

The same 8 × 6 m of net, drawn to one scale. Grid spacing on this system: 1,000 – 2,000 mm.

The optimum is not obvious and it is not a matter of taste. It is found by running the loop, and it is worth running early, because it determines both what the surface looks like and what the building behind it has to resist.

Infill compatibility: where cable nets go wrong

The infill must tolerate the net's movement. Rigid panels on a flexible net need articulated fixings.

Deflection limit: L/50 – L/100 under design wind. On a 20 m net that is up to 400 mm of movement. The infill has to live with that.

Every fixing at every node has to rotate, because the net changes shape while each panel stays flat. A fixing that cannot rotate transfers bending into the panel at the one point it is weakest — its fixing hole. On a net with hundreds of nodes, that is hundreds of articulated connections, and it is why cable nets cost what they cost.

• The net curves Under wind the cables change shape — up to L/50 on this system.
• The panels do not Each panel stays flat, so the angle between neighbours changes at every node.
• The fitting takes the difference A node fitting that cannot rotate puts bending into the panel at its fixing hole — its weakest point.

Deflection exaggerated so it can be seen.

Infill options: laminated glass, insulating units, ETFE, membrane, metal.

The honest summary: membrane and ETFE suit a cable net naturally, because they are tension surfaces themselves. Glass fights it — rigid infill on a flexible structure — and that fight is the engineering. It is entirely doable, and the Moghaddas Ardebili Office is a delivered glazed example, but a glazed cable net is a different order of project from a membrane one and should be budgeted as such.

Dynamic behaviour: flutter and resonance

Dynamic behaviour, not just static load. Flutter and resonance are checked as part of the design.

The design wind load on this system is assessed statically and dynamically, which is unusual enough on a façade system to be worth explaining.

A cable net is light, flexible and pre-stressed. That combination can oscillate under wind speeds well below the static design load:

• Flutter is self-excited. The surface moves, the movement changes how air flows over it, and the changed flow feeds the movement. It can build rather than damp out.
• Resonance happens where the frequency of gusting or vortex shedding approaches a natural frequency of the net. A heavy, stiff structure has natural frequencies far from wind excitation; a light pre-stressed one may not.
• Vortex shedding from edge members and from the building around the net can excite it even where the net itself would be stable in clean flow.

Pre-tension is the primary defence. Raising it raises the natural frequencies and increases the surface's ability to resist without moving. It is also what makes the anchorage larger — so the dynamic requirement and the cost of the anchorage are the same conversation.

Seismic behaviour, by contrast, is generally favourable: the system is very light, so inertial forces are small, and it is flexible enough to accommodate frame movement that would break a rigid façade.

Design is to EN 1993-1-11 as a tension structure and, where the net is a façade, EN 13830 for curtain walling, with wind actions to EN 1991-1-4.

Anchorage: ring beam, edge truss, or slab-edge bracket

The critical element. Cable tension is transferred into the primary structure through connections that govern the whole system's safety.

Critical is the right word. The anchorage is where every cable's tension ends up, and an anchorage that gives way releases every cable that runs to it.

Three types, and the choice is largely geometric:

Ring beam. A closed compression ring that the net's tension pulls against. The horizontal forces balance within the ring and never reach the foundations. Where the geometry allows it, this is by far the cheapest anchorage, and it is a concept-stage decision, not a detailing one.

Edge truss. Where the boundary is not closed, a truss along the edge collects the cable forces and carries them to discrete supports. Heavier than a ring, but it works on an open boundary.

Slab-edge bracket. For nets integrated into a building frame, tension goes directly into the floor structure. Simplest to build, and it puts permanent tension into slab edges that were probably designed for gravity. The structural engineer has to know about this before the frame is priced.

• Ring beam A closed ring in compression. The net’s horizontal pull balances inside it and never reaches the foundations — the cheapest anchorage, where the geometry allows it.

• Edge truss Where the boundary is not closed, a truss along the edge collects the cable forces and carries them to discrete supports. Heavier than a ring.

• Slab-edge bracket Tension goes straight into the floor structure. Simplest to build — and permanent tension in slab edges that were probably designed for gravity.

The edge structure is often larger and more expensive than the net it supports. That is not a design failure; it is the nature of a system that converts everything into tension and then has to stop it somewhere. Budget it as part of the façade rather than discovering it in the frame package.

What drives the cost

In order of leverage:

1. Whether the force loop closes. A ring beam that balances the net's tension against itself keeps its horizontal pull out of the rest of the building. If the forces must go into the frame or the ground, everything that resists them grows — permanently. This is decided by plan geometry at concept stage, usually by the architect, and is the same principle set out on cable structures.
2. How flat the surface is. Flatter means higher pre-tension means bigger everything. Curvature is stiffness you do not have to buy; give it up and you buy the equivalent in steel and concrete.
3. The infill. Membrane and ETFE are light and tolerant. Glass and insulating units are heavy and rigid, need an articulated fitting at every node, and their weight resizes the cables and the anchorage beneath them.
4. Grid spacing. More nodes means more fittings and more labour; fewer nodes means bigger panels, thicker glass and heavier cables. The optimum is found by running the loop, not by preference.
5. How many node fittings are unique. As with panel sets on free-form structures, one fitting type used four hundred times costs a fraction of twenty types used twenty times each.
6. Dynamic analysis and mock-up. Both are real scope on this system rather than overheads, and the mock-up usually sits on the critical path before fabrication is released.

What to decide, and when

1. 01

   Concept

   The boundary — where the edge of the net sits, and whether it can be closed into a ring. How much curvature the architecture will accept. The infill, at least in category, because membrane and glass are different projects. The wind environment, including what the surrounding buildings do to the flow.

2. 02

   Developed design

   Form finding, which produces the geometry and the pre-tension. Grid spacing and panel size, resolved as a loop with the infill weight. Boundary reactions issued to the structural engineer before the primary frame is fixed. Static and dynamic wind assessment.

3. 03

   Technical design

   Node fittings and their articulation, edge structure and anchorage details, the tensioning sequence, perimeter movement details, and the mock-up scope.

4. 04

   Before tender

   A form-found geometry, a defined anchorage strategy and a confirmed infill. A cable net tendered on a drawn shape rather than a found one will be priced with a large risk margin, because the tenderer knows the shape will change.