It requires two sounds in sequence - the low-frequency flex of the pane and then the high-frequency fracture - within a short window.

A single loud noise, such as a dropped object or a door slamming, produces one of those and not the other.

The dual-signature requirement is what makes the technology usable in an occupied building.

Some detectors add pattern analysis, comparing the sound against stored profiles for different glass types.

Sustained loud noise - certain alarms, some machinery - can occasionally satisfy both criteria, which is why siting away from such sources matters.

Sensitivity is adjustable on most units, and the correct setting comes from testing with a simulator in the finished room.

Several metres in a bare room and considerably less in a furnished one - soft furnishings absorb the sound the detector depends on.

Published ranges assume a reasonably reflective room with clear line of sight to the glass.

Heavy curtains, blinds, carpet, upholstered furniture and acoustic ceilings all reduce it, sometimes dramatically.

Line of sight is essential. A detector that cannot see the glass, because a partition or a shutter is in the way, will not hear it properly either.

Curtains closed at night are the common trap: the detector is commissioned in an empty room and protects far less once the space is in use.

Test in the room as it will actually be at night, with curtains and blinds in their night-time position.

It depends on the detector and the glass - laminated glass in particular can fail without producing the expected fracture signature.

Plate glass produces the classic two-stage signature and is what most detectors are tuned for.

Toughened glass shatters into fragments with a different acoustic profile, and most modern detectors handle it.

Laminated glass cracks but is held by the interlayer, so the fracture sound is muted and the pane may not break through at all.

Wired and coated glasses each behave differently again.

Check the manufacturer's stated compatibility for the glass actually installed, and where laminated glazing is used, consider a shock sensor on the frame instead.

An acoustic detector covers several windows from one point; a shock sensor detects the attack on one pane or frame, including attempts that do not break it.

Acoustic detection is cheaper per window in a room with several openings, and needs no wiring at the glass.

Shock sensors are mounted on the frame or the pane and respond to impact rather than sound, so soft furnishings do not affect them.

They also detect an attack that has not yet succeeded, which an acoustic detector cannot.

They need mounting on every protected opening, which is more work and more visible.

Many installations use both on high-risk elevations, and insurers sometimes specify which is acceptable.

With the manufacturer's simulator, in the room as it will be at night, at the furthest protected pane.

Simulators reproduce the acoustic signature the detector is looking for and are the only practical way to test repeatedly.

Test at the most distant and most obstructed window rather than the nearest one, since that is where coverage fails.

Set curtains and blinds to their night-time position first. Testing an empty daytime room proves very little.

Record the test and the sensitivity setting, so a later change can be compared against a known baseline.

Repeat after any change to the room's furnishings. A new set of heavy curtains can put a window outside coverage.