How Frequency Spectrum Analysis Improves Noise Control
One decibel reading tells you nothing about the fix. Here is how to read a spectrum, what each shape means, and how it changes the treatment.
Frequency spectrum analysis breaks a noise into its component frequencies so you can see where its energy actually sits, instead of guessing. That single piece of information decides everything else, because the treatment for a 60 Hz problem and a 3 kHz problem have nothing in common. Measuring first is what separates noise control that works from noise control that was expensive.
A sound level meter gives you one number. That number tells you a room is at 58 dB and nothing about why, or what to do. A spectrum tells you the 58 dB is mostly a tone at 120 Hz coming from a transformer, which is a completely different afternoon.
What the analysis produces
A Fast Fourier Transform converts a recorded waveform into a plot of level against frequency. Every complex sound is a sum of simpler tones, and this is the operation that separates them out.
You will usually see the result in one of three forms. Narrowband FFT gives fine resolution and is what you want for identifying a specific tone. Octave or third octave bands group the spectrum into standard ranges, which smooths out detail and is what most acoustic standards are written around. A spectrogram plots frequency against time with level as color, which is the right view for noise that comes and goes.
Weighting matters too. A weighting discounts low frequencies to approximate how the ear responds at moderate levels, and it is the default in most regulations. For diagnosis, use Z weighting or unweighted data, because A weighting will hide exactly the low frequency energy you are trying to find.
What the shape of the spectrum tells you
Different sources leave recognizable signatures, and reading them is most of the skill.
A sharp peak at one frequency with harmonics above it means a rotating or reciprocating machine. Fans produce a blade pass frequency equal to rotational speed multiplied by blade count, so a 1,200 rpm fan with 8 blades gives a tone at 160 Hz. Finding that peak identifies the culprit without opening anything.
Broadband energy rising toward the low end is traffic or general building rumble.
A broad hump between 100 and 300 Hz in a small room, present only at certain listening positions, is a room mode rather than a source. Treating it means bass absorption or moving the speakers, not blocking anything.
Energy concentrated between 1 and 4 kHz is speech, and it is the range that matters for privacy because that is where intelligibility lives.
A 50 or 60 Hz tone with harmonics is electrical, either a transformer or a ground loop, depending on whether you can hear it with the equipment switched off.
How it changes the treatment
| Where the energy sits | What works | What does not |
|---|---|---|
| Below 100 Hz | Mass, decoupling, isolation at the source | Panels, curtains, foam of any thickness |
| 100 to 500 Hz | 4 inch absorbers, bass traps, added mass | Thin panels, soft furnishings |
| 500 Hz to 2 kHz | 2 inch absorbers, sealing gaps, barriers | Very little fails here |
| Above 2 kHz | Almost any soft material, a closed door | Rarely the actual problem |
| A single narrow tone | Fix the source, or a tuned absorber | Broadband treatment, which wastes most of its effect |
The most common expensive mistake in home acoustics follows directly from skipping this step: buying 2 inch panels for a problem that lives at 80 Hz. A porous absorber only works well down to about a quarter of the wavelength, so a 2 inch panel is effective above roughly 500 Hz and does essentially nothing an octave lower. Our explainer on whether soundproofing absorbs or blocks noise covers why thickness and mass are not interchangeable.
Doing it yourself
A phone app with a real time analyzer will get you most of the way. The microphone in a modern phone is reasonably flat from about 100 Hz to 8 kHz, which covers the majority of domestic problems.
Be aware of its limits. Below 100 Hz phone microphones roll off and understate the level, sometimes badly, which is unfortunate because low frequency is where the difficult problems live. The absolute numbers will also be several decibels out without calibration. Relative comparisons between positions and between before and after are reliable, and that is usually what you need.
A calibrated USB measurement microphone at USD 80 to 150 with free measurement software removes both limitations and is worth it if you are treating a room properly.
Method matters more than equipment. Measure with the source running and again with it off, so you have a baseline. Take readings at several positions, because low frequency levels vary enormously across a small room. Measure for at least 30 seconds on anything variable, and use a spectrogram rather than an average for intermittent noise.
Where spectrum analysis feeds automatic systems
Adaptive noise control systems run this continuously. An active cancellation system analyzes the incoming spectrum many times a second and adjusts its filter to target whichever frequencies currently carry the most energy, which is why modern headphones handle a change from aircraft cabin to street noise without intervention.
The same analysis drives room correction in AV receivers, which measure the response at the listening position and apply the inverse. It works well for the peaks caused by room modes and poorly for the nulls, since filling a cancellation with more energy at the same frequency simply feeds the cancellation.
Frequently asked questions
Why is frequency analysis important for noise reduction?
Because the correct treatment depends entirely on where the energy sits, and a single decibel reading tells you nothing about that. Low frequency problems need mass, decoupling or a fix at the source, while mid and high frequency problems respond to absorption and sealing. Measuring the spectrum first is what stops people buying thin acoustic panels for a bass problem, which is the most common wasted purchase in the field.
Can I do spectrum analysis with a phone?
Yes, well enough for most domestic diagnosis. A real time analyzer app on a modern phone is reasonably accurate between about 100 Hz and 8 kHz. Below 100 Hz the microphone rolls off and understates the level, which is a real limitation since that is where the hardest problems are. Treat the absolute numbers as approximate and rely on comparisons between positions or before and after.
What is the difference between octave band and narrowband analysis?
Octave and third octave analysis groups energy into standard frequency ranges, which smooths the data and matches how acoustic standards and regulations are written. Narrowband FFT gives much finer resolution and is what you need to identify a specific tone, such as a fan blade pass frequency or an electrical hum. Use bands for compliance and narrowband for diagnosis.
Should I use A weighting when analyzing noise?
Not for diagnosis. A weighting deliberately discounts low frequencies to approximate the ear’s response at moderate levels, so it hides the low frequency energy that causes most difficult noise problems. Use Z weighting or unweighted data when you are trying to find the source. A weighted figures are appropriate when comparing against a regulatory limit, because that is how the limits are defined.