Moiré Effect: The Defect Two Good Specifications Create

Two panels arrive on the same site. One has a 20 mm aperture, the other 21 mm. Both are the same grade, square to the frame, inside tolerance, and no inspector will fault either. Stack them 150 mm apart and the elevation develops a band of light and shade some 400 mm wide that slides across the building as you walk past. Nothing in either specification is wrong. The defect lives in the difference between them, and no line on either document describes it.

Executive Summary

A woven panel is a periodic structure, and two periodic structures seen one behind the other produce a third pattern that neither contains. This is the moiré effect, and in architecture it is unusual among failure modes because it is created by two decisions that are each individually correct. The band is not there because a supplier cut a corner, but because the outer and inner apertures were chosen a few percent apart, and a few percent is exactly the wrong relationship.

The geometry inverts the instinct. The closer the two apertures sit to each other, the larger the band they generate, so the pair that looks like a safe match produces the most visible defect. Two layers of mesh at 20 mm and 21 mm throw a band twenty times the size of the opening; the same 20 mm layer against 40 mm throws a band one and a half times the average pitch, which barely registers as a pattern at all.

The reference specification is the woven stainless mesh range product page.

Send both apertures and we will check the pair before the mock-up

Quick Answer:

What causes the moire effect on a mesh facade is a mismatch between two periodic patterns, not a defect in either one. Where a woven panel sits in front of a second panel, a window grid or a perforated screen, the two repeats beat against each other and produce an interference band whose width is the average pitch squared, divided by the difference between the pitches. The band is widest when the two apertures are nearly equal, which is why 20 mm against 21 mm is a worse specification than 20 mm against 40 mm. English technical writing usually drops the accent and spells it moire effect; the geometry is the same either way.

Key Takeaways

• The band belongs to the pair, not to either panel, so neither datasheet can warn you about it.
• Near-equal apertures make the largest bands: a 1 mm difference generates a band 20 times the hole size.
• The pattern outlives the mesh. A 20 mm panel stops reading as mesh at 69 m; its band stays visible past a kilometer.
• Open area does not predict the band. Two panels at almost the same figure can throw bands twenty times apart.
• A 1:2 ratio between the layers, or 20 mm of difference, keeps the band small enough to ignore.

A Third Pattern That Belongs to Neither Layer

The effect takes its name from a French word for a rippled silk, because the same shimmer appears when one fabric is laid over another and pressed. What the eye receives is not either surface but the envelope of both: where the opaque lines coincide, light passes; where they interleave and cover each other, it does not. Because the misalignment accumulates slowly across the elevation, the result is a broad band of alternating tone rather than a fine texture — and on a curved soffit, such as the vault of a covered walkway or tunnel, the same band bends with the geometry instead of running straight.

Read the panel as a lattice rather than a surface and the mechanism becomes plain. Each layer is a grid of openings at a fixed pitch, and the two grids are almost but not quite in step. Walk along the facade and the relationship drifts, closing to a perfect overlay at one point and opening to maximum offset half a band later. The moire pattern is the map of that drift, at a scale far coarser than either panel’s own texture. A single panel shows nothing, because it has no second period to beat against, and the same holds for a render, which models the panel as a texture and has no idea what stands behind it. A mock-up has to reproduce the whole stack at the real spacing, or it will pass a design that fails in the street.

The Worse the Match, the Bigger the Band

Specifiers assume that two similar panels will sit quietly together and that mismatched ones will fight. The opposite holds. Write the average pitch as p and the difference as δp, and the band width comes out as p squared divided by δp. As the difference shrinks, the band grows without limit, because the grids take longer and longer to drift through one cycle.

That is not a rule invented for this article. It is standard geometric optics, and the technical literature states it plainly: Geometrical Effects, the fourteenth chapter of Berthier and Schöllhorn’s Light-Arthropod Interactions (Springer, 2026), derives the interference of two superposed grids and draws the practical conclusion. “Note (1) that the bigger the period p, the bigger that of the Moiré pattern Δ, and (2) the greater the discrepancy δp between the periods of the two grids, the smaller this periodicity.” The same chapter gives the condition for no pattern at all: “If the surface is parallel to the grid, the two periods are equal, and there is no formation of Moiré.”

Run the aperture ladder we weave through it and the numbers are uncomfortable. A 20 mm layer against 21 mm gives a band of 420 mm, which on a four meter elevation is a stripe visible from across a road. Against 22 mm it gives 220 mm; against 25 mm, 101 mm; against 30 mm, 62 mm; against 40 mm, just 45 mm. Every step away from a near match shrinks the pattern faster than intuition suggests, until it stops reading as one and becomes a slight unevenness in tone.

Where the Band Survives

The second surprise is how far away the effect still works. Acuity at the 20/20 reference is one minute of arc, which is the angle a 20 mm opening subtends at about 69 m. Past that, a 20 mm panel stops resolving as individual wires and the mesh closes up into a flat gray field. A 102 mm panel holds its texture to roughly 351 m, five times further.

An interference band follows the same rule at its own scale, which is much larger. The 420 mm band from the 20 mm and 21 mm pair does not fall below one minute of arc until 1,445 m; the 220 mm band survives to about 758 m; even the modest 45 mm band reads out to 155 m, most of the way down a street. At what distance the facade reads as a solid is therefore a question with two answers: the mesh disappears at 69 m, and the band it throws is still moving at a kilometer.

This is what makes the defect expensive. Being coarse, it survives the loss of detail that hides nearly every other finishing fault: a scuff, a weld mark, a slack bay all vanish at 50 m, while a moiré band grows more pronounced, because the eye is no longer distracted by the wire and sees only the slow shift in tone. On a tall facade cladding mesh system, or a sunshade mesh layer standing off a glazed wall, that is the one thing a passer-by will notice, and the one thing that cannot be fixed without taking the elevation down. Facades built around planted structure meet the same constraint from the inside, where the depth between mesh and foliage sets both the band and the planting.

The Pairs That Avoid the Band

The cheapest fix costs nothing: stop specifying two apertures that sit close together. A ratio of 1:2 is the cleanest relationship available — 20 mm against 40 mm, 25 against 50, 30 against 60, 50 against 100. Whole-number ratios keep the two grids in step at every second opening, collapsing the beat to the scale of the mesh.

Where a double layer mesh is required for a reason other than appearance, physical spacing is the next lever. Band width does not change with distance, but its visibility does, because separating the layers softens the transition between light and dark. The same 101 mm band that is obvious at 60 mm of separation is much less so at 200 mm, and a facade that gives up 200 mm of depth at the mullion often solves what no panel change can.

The third lever is rotation, and it gets forgotten. Turn one layer a few degrees and the beat rotates with it, pushing the band into a diagonal that reads as intentional or scattering it into a fine moiré the eye ignores. It cannot be tuned on a drawing; it has to be seen on a sample at the real spacing, from the distance the public will use. Anyone who has worked through the structural logic of a mesh railing will recognize the principle: a panel is never judged alone, only against what sits behind and beneath it.

Two Specifications for One Mesh

The deeper problem is that architecture and manufacturing describe the same panel with different numbers, and the two do not convert cleanly. A facade schedule is written in open area, because that single figure drives daylight, solar gain, privacy and wind load. A weaving works is described by aperture and wire diameter, because those are what a loom is set to. Both are legitimate, and neither can be recovered from the other.

Take a 50 mm panel woven from 1.2 mm wire. It is 95.3 percent open. Ask a 20 mm panel to reach the same figure and the wire would have to fall to 0.48 mm, thinner than any strand that holds tension across a bay. The two cannot be matched on open area at all, and a schedule that says 95 percent open without naming the aperture has quietly specified a panel that can only be built one way. A mesh has to be specified by aperture and confirmed by open area, in that order.

The consequence for the band is direct. Two panels can share an open area only if the wire scales with the aperture, and when it does the pitch differs by the same factor, so the pair behaves perfectly. Two panels within a millimeter of each other on pitch, and therefore almost identical on open area, produce the worst band available. Open area tells you how a facade will perform and says nothing about how it will look. Two specifications for one mesh is the minimum needed to describe the object, and the selection criteria that matter are never a single number.

What a Woven Panel Adds to the Calculation

Handweaving changes the reliability of all of this, though not the geometry. Each intersection in a handwoven stainless mesh is a node that cannot slip, so the pitch you specify is the pitch that arrives: tension applied at one opening stays at that opening and does not migrate along the weave. That matters here more than it does on a fence, because the calculation is only as good as the pitch it is given. A welded or expanded product that stretches unevenly has a varying pitch, and a varying pitch produces a band that wanders instead of one that marches.

Grade is a separate decision. 304 stainless covers most inland elevations and most interior work. 316 stainless, or the inox grades continental suppliers quote, is bought where chlorides are present: coastal sites, pool halls, anywhere de-icing salt lands on the panel. The difference is how long the surface stays passive, and it has no bearing on the interference pattern, which is why the two questions belong in that order. It is also why a shop drawing that says SS and nothing beside it is fine for a fence and inadequate for an elevation, where the netting trade’s habit of quoting grade, pitch and wire separately is the only version that works.

We weave panels to 9 m by 18 m, 162 square meters, large enough for one panel to span a story without a horizontal joint interrupting the weave. On a woven metal facade that matters, because a joint is a phase break: the pattern on one side starts from a different point in the cycle and the band appears to jump. Where the panel count cannot be reduced, panels are woven to the opening rather than cut from a roll, so the pitch stays continuous. A metal fabric facade built that way has a band you can predict from the drawing; one assembled from cut lengths has a band you meet on site. Both routes are covered by the custom panel work we do for irregular openings.

Send the two layer pitches and we will run the pair before you build

FAQ