System design · Guide

The noise measurement that shrinks a PAVA quotation

BS 5839-8:2023 hangs every loudspeaker tap, amplifier and alarm battery load on one number: the background noise. Measure it instead of assuming it, and 6 dB less noise means a quarter of the amplifier power.

By Nigel Williams · Published 28 October 2026

A tailor who guesses your size rounds up. A jacket that is too big still goes on. One that is too small does not. Nobody is blamed for the spare inch, so the spare inch is what you pay for.

Background noise works the same way in a voice alarm design. Round it up by a few decibels and nothing fails on the day. The quotation just gets bigger.

One number holds up the whole design

A Public Address Voice Alarm (PAVA) system used for fire evacuation is designed to the code of practice BS 5839-8:2023. It recommends that speech is broadcast above the background noise expected under emergency conditions, and notes that a signal-to-noise ratio of at least 10 dB is usually needed (Clause 20.1.1, Note 1).

Everything else follows from that one figure:

• the background noise sets the speech level
• the speech level sets the power tap on every loudspeaker
• the tap total sets the amplifier, which the code recommends is rated at least 20 % above the design load (Clause 15.3)
• the amplifier load sets the alarm share of the standby battery, sized for 30 minutes of alarm after the standby period (Clause 25.3.5)

The code knows how much rides on it. Its informative Annex A points out that 3 dB more design level doubles the amplifier power, and asks for the most realistic ambient noise level to be taken as the design datum (Annex A).

Where the assumed figure comes from

What the code does not do is tell you how long to measure. It asks for the noise as an LAeq,T, the average level over a period, with the period "appropriate" to the premises and an acoustician involved if in doubt (Clause 20.1.2).

So the figure is often not measured at all. It comes from a typical-range table, a previous job, or a cautious estimate. Each of those rounds up, for the tailor's reason.

There is a second cost. Where ambient noise is, or is likely to be, greater than 75 dBA, the code recommends involving a competent electro-acoustic designer (Clause 20.1.4). An assumed figure that lands above that line can bring a design fee with it.

Worked example: one mall zone, assumed against measured

The figures in this example are illustrative, chosen to show the arithmetic. They are not from a particular survey.

A shopping-centre mall zone has 40 ceiling loudspeakers on 100 V line, with taps at 6 W, 3 W and 1.5 W. The design takes speech at 10 dB above ambient.

Six decibels is two halvings of power: 6 W to 3 W, then 3 W to 1.5 W. The amplifier rating falls from 288 W to 72 W, three quarters less, and the alarm-load share of the battery falls with it.

The number of loudspeakers need not change. Their spacing is set by coverage, not by noise. What changes is how hard each one is driven, and so what has to be bought, racked, cooled and kept charged to drive it.

How to measure so the figure stands up

A lower figure is only worth having if it survives a challenge. These are the habits that make it survive:

• The right meter. A sound level meter that logs LAeq over a set time, of the kind the code names for its own sound level checks (Clause 19.2, Note 3), with a calibration date you can quote.
• The right moment. Measure during normal operation at its busiest, not on a quiet weekend. Then add the noise that only arrives in an emergency: smoke extract, stair pressurisation fans (Clause 20.2.2 g)).
• The right height. Ear height, about 1 m to 1.2 m where people sit and 1.5 m to 1.7 m where they stand, the heights the code uses for intelligibility checks (Clause 20.2.2 f)). Stand at least 1 m back from the meter.
• Enough of it. At least 30 s per reading and three readings per position. If they differ by more than 3 dB, find out why before you use any of them.
• A quick check of the noise's character. Read A-weighted, then C-weighted. A C figure well above the A figure means the noise is mostly low-frequency hum. Record it; take advice before trading any decibels on it.
• The record. Position with a photo, date and time, duration, what was happening, the meter's make and calibration date, and anything unusual, such as a door that banged twice.

"Won't a lower figure leave us exposed?"

Only if the measurement leaves out noise that will be there in an emergency. That is why the emergency plant goes into the survey, not after it.

The code's own route is the realistic figure, agreed before design (Annex A). And the building does not stay as measured: the six-monthly service looks for a change of use that raises ambient noise (Clause 40.1 b)). A louder building is caught at the next visit, not at the next fire.

A measured figure with its record attached is also easier to defend than an assumed one with nothing behind it.

Two examples

The plant room that was switched off. A survey is taken on a mild day with the chillers idle. The figure is low and honest, and wrong for summer. The record notes the plant was off, so the gap is visible on paper and the survey is repeated with it running.

The figure from the last job. A design reuses 75 dBA from a similar mall elsewhere. Nobody measured this one. A 30-minute survey at the busiest hour reads 69 dBA, and the amplifier schedule in the worked example above follows from it. (Illustrative, like the worked example.)

In short

• One background noise figure sets the speech level, the loudspeaker taps, the amplifier and the alarm share of the battery.
• Every 3 dB of design level doubles the amplifier power (Annex A); 6 dB is a factor of four.
• BS 5839-8:2023 asks for the most realistic figure but leaves the measuring method to you (Clause 20.1.2).
• A measured figure with its record attached survives a challenge, and the six-monthly service catches a building that gets louder (Clause 40.1 b)).

The tailor measures before cutting the cloth. The noise survey is the same job.

Next step. Run your zone through the ZOE zone output estimator twice: once with the assumed noise figure, once with a measured one. The difference between the two answers is what the measurement is worth. If you would like the measurement taken for you, contactus@proaudium.com or 01273 034630.