Noise Sensor: How Do You Monitor Noise Levels in an Office?

Two teams in an open-plan office keep complaining about noise, and nobody knows whether the phone area, the printer or the ventilation is to blame. A noise sensor answers exactly that question. It stays mounted in one spot, measures the sound level in decibels around the clock and sends the readings wirelessly at fixed intervals. After a few weeks you can see which hours and which zones are actually loud, instead of relying on impressions and complaints.

A handheld decibel meter shows you a moment; a permanent sensor shows you patterns. Reading them correctly takes a few basics: which frequency weighting the numbers use, how LAeq differs from Lmax, why decibels can’t be averaged like ordinary numbers and where the sensor should hang. The examples use the Milesight WS302, a battery-powered LoRaWAN sound level sensor for indoor spaces.

How does a sound level sensor work?

At its core, every sound level sensor contains a microphone that turns tiny air pressure fluctuations into an electrical signal. The electronics continuously calculate the sound pressure level from that signal, filter it according to the selected frequency weighting and summarize it over time. What leaves the device is not audio but a handful of numbers.

The WS302 reports exactly three values: the current sound pressure level (SPL), the equivalent continuous level Leq and the maximum level Lmax, each with a resolution of 0.1 dB. By default it transmits every ten minutes, and the interval can be set anywhere from 1 to 1,080 minutes. According to Milesight, the measurement range is 30 to 120 dBA or 35 to 120 dBC, the frequency range is 30 Hz to 8 kHz and the accuracy is ±3 dB between 0 and 45 °C (32 to 113 °F).

Take that ±3 dB seriously. A sound level meter used for formal surveys works to much tighter tolerances, and Milesight’s own user guide states that the WS302 is not intended as a reference sensor. Its strength lies elsewhere: it measures the same spot for months without anyone standing next to it. If you want to compare trends, such as before and after installing acoustic panels, Monday versus Friday or mornings versus afternoons, that consistency matters more than the last decimal place. If you have access to a reference instrument, you can also store a fixed calibration offset in the WS302, which it adds to every reading.

The sensor is configured over NFC with the Milesight ToolBox smartphone app, and its data travels through a LoRaWAN gateway to a network server. Two ER14505 lithium cells power it. Milesight quotes a battery life of around five years at a ten-minute interval with fast time weighting disabled (SF7, EU868), and less with a weaker radio link and a higher spreading factor.

How loud is too loud in an office?

Hearing-protection rules only start at around 80 dBA, far above what a normal office produces. The real question is whether people can think, read and take a call without effort. A useful benchmark comes from Germany, where workplace rule ASR A3.7 sets limits based on how much concentration and speech intelligibility a task requires. For highly demanding work, such as creative thinking, precise writing or understanding complex texts, the rating level must not exceed 55 dBA. For general office tasks and work with customer contact, the limit is 70 dBA. A rating level is not simply a sensor reading, though. It is determined with a defined procedure over the duration of a task and includes penalties for tonal content, intelligible speech and impulsive noise, because a conversation you can follow at the next desk is more distracting than steady ventilation hum of the same level. A noise sensor gives you a strong indication of whether an area is near a limit, not a formal compliance measurement.

For orientation, the WS302 user guide lists a quiet office at about 50 dBA, normal conversation at about 60 dBA and busy traffic at about 70 dBA. If a focus area sits around 60 dBA for hours at a time, that’s a clear signal to look closer. ASR A3.7 also gives a reverberation time of about 0.6 seconds as a target for open-plan offices. A noise sensor doesn’t measure reverberation, but it does show whether acoustic treatment actually lowers the level during a normal working day.

dBA vs dBC: which weighting do you need?

Human hearing isn’t equally sensitive at all frequencies. At the same sound pressure, low tones sound much quieter to us than tones in the range of roughly 1 to 4 kHz, where important parts of speech sit. A-weighting mimics that by attenuating low and very high frequencies. C-weighting is nearly flat and lets low frequencies through largely untouched.

For an office, the answer is simple: use dBA. The ASR A3.7 limits are A-weighted, and Milesight recommends A-weighting for normal environments such as offices, hospitals and homes. According to the user guide, C-weighting is meant for very loud environments above 100 dB, such as factories, construction sites and dance halls, where low-frequency content plays a large role.

There is one special case. If a compressor hums in the equipment room next door or an air handler drones, comparing the C-weighted and A-weighted values can reveal a lot of low-frequency sound. The classic mistake is switching the weighting in the middle of a measurement series. With low-frequency noise, the C value is higher than the A value, so the chart appears to jump even though nothing in the room has changed. The WS302 includes the selected weighting in every report. Make sure your decoder reads that field and your dashboard labels the unit correctly.

Besides frequency weighting, the sensor offers fast time weighting with a 125-millisecond time constant. It follows strongly fluctuating sound more closely, but it shortens battery life and is off by default. For office trends, Leq is the more meaningful value anyway.

What does LAeq mean?

LAeq is the A-weighted equivalent continuous sound level. It is the steady level that would contain the same sound energy over a period as the actual, fluctuating sound did. With a ten-minute reporting interval, LAeq tells you how much sound energy arrived in those ten minutes in total, quiet stretches and loud moments combined.

Lmax, by contrast, is the highest level measured in the same period. A slammed door, a burst of laughter or a dropped binder pushes Lmax up but often barely changes the ten-minute LAeq. Conversely, a ventilation system with a steady hiss can have an unremarkable Lmax while raising LAeq all day. The table below shows which question each of the three WS302 values answers.

For assessing an office, LAeq is the most important value, because rating levels such as the one in ASR A3.7 are built on the equivalent continuous level. Lmax complements it. If LAeq is moderate but Lmax regularly sits 20 dB higher, you are dealing with individual disruptive events rather than a constantly loud room. That calls for a different fix, such as soft-close door hardware instead of an acoustic ceiling.

Why can’t you average decibels?

Decibels are a logarithmic unit. Every increase of 10 dB means ten times the sound energy, and every increase of 3 dB roughly doubles it. That’s why two equally loud sources at 50 dB each add up to about 53 dB, not 100 dB.

The same logic applies to averages. Take two periods of equal length, one at 40 dB and one at 60 dB. The arithmetic mean would be 50 dB, but the correct energy average is about 57 dB. The loud period dominates because it carries a hundred times as much sound energy as the quiet one. Simply averaging a day’s Leq values in a spreadsheet systematically underestimates the level, and the more the values vary, the bigger the error.

Here is the correct method. Convert each level to energy (10 to the power of level divided by 10), average those values, plainly for equal intervals or weighted by duration for unequal ones, and convert the result back to decibels with 10 times the base-10 logarithm. With the WS302’s ten-minute values, that gives you an LAeq for an hour, a working day or a specific window such as the lunch break. Don’t fold Lmax values into that calculation. For a longer period, the highest single Lmax is the right summary,.

Where should you mount a noise sensor?

Mounting location affects your readings more than any setting does. Milesight recommends mounting the WS302 at a height of 1.2 to 1.5 meters (about 4 to 5 feet), roughly ear height for people at their desks. Where possible, it should be at least 1 meter from walls and other reflective surfaces and about 1.5 meters from doors and windows. Keep it away from noise sources and keep the microphone opening clear. In small rooms, the manufacturer suggests ceiling mounting.

In practice, the sensor usually ends up on a wall or column, fixed with screws or the included double-sided tape. That’s a compromise: directly on a wall, reflections make it read slightly differently than it would in free space. For trends this is fine as long as the location stays the same. Choose a wall that belongs to the work area you want to assess, though, not the one next to the printer, the coffee machine or an air vent. Otherwise you measure the machine, not the room. According to the user guide, the tape doesn’t hold reliably on wallpaper or rough, crumbling plaster, so use screws there.

One sensor is rarely enough for an open-plan office. A zone layout works better: one sensor in the quiet area, one near the phone or team area, one in the meeting room. That way you can see whether noise travels from one zone into another. The typical mistake in before-and-after comparisons is moving the sensor at the same time as installing an acoustic fix. Then you can’t tell whether the new panels or the new mounting spot lowered the level. Write down position, weighting and reporting interval before you start the baseline measurement, and leave all three unchanged until the comparison is finished.

When should a noise level alert trigger?

An alert is only useful if someone can act on it. Before setting a threshold, decide who receives the notification and what happens next: does a door get closed, does a conversation move to a meeting room, does the seating plan for a zone change? Without that answer, alerts quickly become background noise themselves.

The WS302 has its own threshold function. You set a level, and the device sends a report immediately once the sound pressure level stays above the threshold for more than one minute. It won’t send another alarm until the condition has cleared and been triggered again. Optionally, the LED shows the state right on the device, blinking green below the threshold and red above it. In a meeting room, that visible feedback can do more than an email because it reaches people where the noise is made.

For office monitoring, base alerts on LAeq rather than on short peaks. A single loud laugh shouldn’t trigger anything, but a focus area that sits at 60 dBA for half an hour should. If you process the data in an IoT platform, a rule that fires only after several consecutive Leq values above the limit and clears only below a slightly lower reset level works well.

Also plan for the case where no data arrives at all. A dashboard that keeps showing the last quiet reading after the sensor has gone silent creates a false sense of calm. Always display the time of the last report and mark values as stale when two or three intervals are missing.

Does a noise sensor record conversations?

A noise sensor such as the Milesight WS302 does not record conversations. It processes the microphone signal inside the device and transmits only level values. Milesight states explicitly that its noise monitoring devices only monitor sound levels and never record any sound. You can verify the logic technically, too: the sound portion of a regular WS302 report is exactly seven bytes, containing the weighting mode and three level values. LoRaWAN, with its small data packets and long pauses between transmissions, couldn’t carry an intelligible audio track anyway.

Even so, be open about the installation. A device with microphone slots on the wall raises questions, and a notice or a short mention in a team meeting heads off suspicion: what is measured, how often, who can see the data and what it is used for. If your organization has a works council or employee representatives, or if local privacy rules apply to workplace monitoring, involve the right people early.

Even without audio, a level chart describes a place, not a person. A spike at 2 p.m. tells you it was loud, not who was loud. Use the data to improve rooms, zones and routines, not to police individuals.

Where can you buy a noise sensor?

Start with the reported values. Besides the current level, a good sensor should deliver an equivalent continuous level and a maximum, otherwise you can’t tell sustained noise from isolated peaks. A-weighting is essential for offices, and switchable C-weighting helps with low-frequency hum. Look for a measurement range that reaches down to quiet areas around 30 dBA, and be realistic about accuracy: a sensor rated at ±3 dB is right for trends and comparisons but doesn’t replace a sound level meter for formal surveys. Then come the practical points: an adjustable reporting interval, a threshold function on the device itself, battery life of several years and a mounting method that suits your walls. With LoRaWAN, you also need gateway coverage in the building.

If you’re looking for a suitable noise sensor for your office, you’ll find our models in the noise sensors category.