How Science Works | Acoustics — Sound, Vibration, Resonance, Spectra and the Evidence We Hear

Reading route: Source and wave → Pressure, intensity and decibels → Resonance and rooms → Spectra and measurement → Learning route and checkpoints.

1. The scientific job is to explain a source–path–receiver chain

A useful acoustic explanation begins with a vibrating source, identifies the medium and path, describes boundaries or losses, and ends with a specified receiver. A loudspeaker cone, a vibrating string and a column of air generate sound differently. Yet each can produce a disturbance that travels through a medium and reaches a microphone. The receiver’s location is part of the explanation, not an afterthought.

That chain helps separate questions which otherwise become entangled. Why did the source produce this spectrum? Why did the room emphasise these frequencies? Why did the microphone report this level? Why did one listener describe the result differently? The basic wave mechanism is introduced in OpenStax: Sound Waves; the distinctions between sound, instruments and signals are developed by UNSW’s Music Acoustics group.

2. A sound wave transports disturbance without carrying all the air with it

In a simple sound wave in air, pressure and density oscillate around their background values. Air parcels move back and forth over small distances while the disturbance propagates over a much larger distance. Sound reaching the far side of a room does not require the same parcel of air to travel from the speaker to that listener.

The restoring response of compression and the inertia of the medium together support propagation. In ordinary fluids, these pressure waves are longitudinal: parcel motion is along the direction of propagation. Solids can support additional kinds of mechanical wave because they resist shear. The particular medium therefore changes which motions are possible. See the mechanical-wave description.

For a first explanation, distinguish three velocities in words: the source’s movement, the small oscillatory motion of the medium, and the propagation speed of the disturbance. Assigning all three the same value is a warning that the mechanism has been compressed too far.

3. Frequency, period, wavelength and amplitude do separate jobs

Frequency counts cycles per second; period is the time per cycle. Wavelength measures the spatial separation of matching phases of a travelling wave. For a simple travelling periodic wave, c = fλ. Amplitude describes the size of the oscillation, but the quantity oscillating must be named: pressure amplitude and displacement amplitude are not identical measurements.

A frequency change can alter perceived pitch, but a sound may contain many frequencies rather than one. An amplitude change can alter loudness, but frequency content and the listener also matter. UNSW’s sound and signal guide and discussion of loudness and spectra keep these physical and perceptual descriptions distinct.

4. Worked example: translate between time and space

Original hypothetical example. Assume the relevant sound speed is 343 m/s and a source produces a 500 Hz sinusoidal signal. The period is 1/500 = 0.0020 s, or 2.0 ms. The wavelength is 343/500 = 0.686 m. These are different views of one repeating disturbance: time between matching phases at one location, and distance between matching phases at one instant.

Now double frequency while retaining the assumed medium and sound speed. The period and wavelength both halve. Merely increasing the amplitude would not, in this linear model, halve either quantity. The example gives a diagnostic test: a learner who says “louder sound has a shorter wavelength” has mixed amplitude with frequency.

The value 343 m/s is an explicit assumption for this exercise, not a universal speed for sound in all air, water or solids. A real measurement requires the relevant material conditions. This numerical example is calculated rather than measured; no acoustic experiment by eduKate is being reported.

5. Pressure is a local oscillation; intensity is energy flow

A microphone commonly responds to pressure variation at its location. Acoustic intensity describes energy transferred per unit area per unit time. For a progressive plane wave in a simple lossless medium, the mean intensity is related to squared root-mean-square pressure by I = p_rms²/(ρc). That relationship carries assumptions about the wave field; it should not be applied blindly to every standing or near field.

The squared dependence means that doubling pressure amplitude produces four times the travelling-wave intensity under otherwise unchanged conditions. The relation between pressure, intensity and acoustic impedance is explained in OpenStax: Sound Intensity. Before converting a pressure measurement into power, ask what area and propagation geometry the conversion assumes.

6. Decibels describe a ratio, so the reference must be named

A decibel is logarithmic. Sound-pressure level in air is conventionally expressed as L_p = 20 log₁₀(p_rms/p_ref), using a reference pressure of 20 μPa. A sound-pressure level of zero decibels therefore means pressure equal to the reference, not absence of sound. A negative value means below that reference.

Power or intensity ratios use a factor of 10 rather than 20. The two forms are consistent because intensity is proportional to pressure squared in the appropriate travelling-wave situation. The reference and the type of level are indispensable. A bare number followed by “dB” may refer to an acoustic measurement, an electrical ratio or a digital recording scale. UNSW’s acoustics FAQ explains why the label alone is insufficient.

7. Worked example: one pressure reading and two combined levels

Original hypothetical example. Let p_rms = 0.020 Pa with a reference of 0.000020 Pa. Their ratio is 1,000. The level is 20 log₁₀(1,000) = 60 dB SPL. Doubling pressure to 0.040 Pa raises the level by approximately 6.02 dB, not by a factor of two on the displayed decibel number.

Next suppose two uncorrelated sources each produce 60 dB SPL at a specified receiver under the same measurement convention. Their mean-square pressure contributions add. Converting out of logarithms, adding, and converting back gives approximately 63.01 dB SPL. Simply writing 60 + 60 = 120 dB would be an enormous error.

For unequal uncorrelated levels of 82 and 78 dB at that receiver, the calculation is 10 log₁₀(10^8.2 + 10^7.8) ≈ 83.46 dB. These invented cases illustrate logarithmic addition, not exposure guidance. The physical distinction between pressure addition and average intensity addition is discussed in UNSW’s explanation of combining sounds.

8. Coherence changes what addition means

If two equal coherent sinusoidal pressure signals arrive in phase at one point, their pressures add directly. Their combined amplitude doubles there, producing a roughly 6 dB increase in pressure level. If they arrive out of phase by half a cycle, they can cancel ideally at that point. Other positions can have different phase differences and therefore different results.

Neither case contradicts the approximately 3 dB increase for equal uncorrelated contributions. The assumed correlation is different. A statement about “two identical speakers” is incomplete until it identifies their signals, phases, separation and receiver location. The intensity framework supplies the squared-amplitude connection; the examples here show why a physical situation cannot be replaced by a memorised decibel rule.

9. Distance weakens a simple spreading field, but rooms change the rule

For an ideal point source radiating a fixed acoustic power uniformly into free space, that power crosses spheres of increasing area. Since sphere area grows as distance squared, intensity falls as inverse distance squared. Doubling distance gives one-quarter of the intensity and a level decrease of approximately 6 dB under these assumptions.

A real source can be directional, nearby fields can be complicated, and reflected sound can contribute strongly indoors. A rule derived for spherical spreading should not be imposed on a room measurement without checking those conditions. OpenStax explains the geometrical spreading relation. The relevant diagnostic question is whether the receiver is mainly sampling direct sound or a combination of direct and reflected contributions.

10. Resonance is selective response, not energy appearing for free

A resonant system responds strongly over particular frequency ranges because its inertia, restoring effects and losses favour those motions. Sustained response requires energy supplied by a source. Damping limits how large the response becomes and how long it persists after excitation stops. An ideal undamped calculation is a limiting model, not a prediction that real instruments grow louder without bound.

Resonance becomes especially instructive when the source and resonator are separated conceptually. An excitation can contain many frequencies while the receiving structure emphasises only some. Conversely, changing the way the structure is driven can change which available modes are strongly excited. Standing-wave modes provide a concrete example of boundaries selecting the possible pattern.

11. Standing waves put pressure and motion in different places

Oppositely travelling waves can combine into a standing pattern. Some positions have large pressure oscillations and others have small ones. For the familiar ideal standing sound wave, pressure nodes correspond to displacement antinodes, and vice versa. “Node” is therefore incomplete unless the quantity is named.

At an ideal rigid closed end, normal air motion is restricted, producing a displacement node and pressure antinode. At an ideal open end, the pressure variation is approximately constrained by the surrounding air, though real openings require end corrections. The distinction is set out in OpenStax’s normal-mode treatment. A microphone and a particle-motion sensor need not report matching spatial patterns.

12. Worked example: the same tube length can support different fundamentals

Original hypothetical example. Use an ideal tube of effective length 0.250 m with assumed sound speed 343 m/s. If one end is closed and the other open, the lowest ideal pressure pattern occupies one-quarter wavelength along the tube. Its fundamental frequency is c/(4L) = 343 Hz.

If both ends are open in the ideal model, the lowest mode instead corresponds to half a wavelength along the tube. The fundamental is c/(2L) = 686 Hz. The same length has not produced contradictory answers: the boundary conditions changed. In a real instrument, the effective acoustic length need not equal the simple physical ruler length.

To diagnose understanding, ask the learner to sketch pressure variation and displacement separately before using either formula. Then ask what would happen if the assumed sound speed changed while length remained fixed. A proportional frequency change follows. The numerical exercise applies the standing-wave model rather than reporting a measured musical instrument.

13. A bottle-like resonator exposes inertia and compressibility

A Helmholtz resonator can be understood as an oscillating plug of air in a neck coupled to compressible air in a cavity. The moving plug supplies inertia; compression supplies a restoring effect. Neck geometry, effective neck length and cavity volume determine the approximate resonance. This model differs from treating a long air column as a simple quarter-wave pipe.

UNSW’s Helmholtz resonance guide derives the relationship and discusses familiar acoustic examples. As a reasoning exercise, hold the neck dimensions fixed and reduce cavity volume. The air spring becomes effectively stiffer and the model predicts a higher resonance frequency. This is a prediction to compare with suitable data, not permission to infer every property of a real container from one note.

14. A room is part of the acoustic system

The receiver indoors hears a superposition of direct arrival and sound that has encountered surfaces. Reflection changes direction, absorption removes energy from the airborne sound field, and transmission carries energy into neighbouring regions or structures. Geometry determines delays and possible standing-wave patterns. Surface materials contribute frequency-dependent behaviour.

A room can therefore make the same loudspeaker recording sound different at different seats. Calling the source “bad” from one listening position can mistake a path effect for a source defect. UNSW’s discussion of enclosed spaces and the acoustic-impedance framework provide useful starting points. For investigation, keep source settings fixed while changing receiver position, then ask which features follow the room rather than the source.

15. Worked example: interpreting a decay record without inventing certainty

Original hypothetical dataset. Suppose a processed sound-level decay in one frequency band falls approximately 20 dB over 0.40 s after the source stops. If that section represents a single exponential energy decay, its slope is −50 dB per second. Extrapolating that slope to a 60 dB decrease gives a decay-time estimate of 1.2 s.

The calculation does not mean the instrument actually observed 60 dB of clean decay. The usable record may be much shorter because it meets the background noise floor. Nor does one band establish a single universal reverberation time for every frequency and seat. The estimate inherits the selected interval and exponential-decay assumption.

A useful test is to change the fitting interval. If the estimate changes greatly, the record may contain several decay behaviours, insufficient signal range or a contamination problem. State what was observed and what was extrapolated separately. The aim is not to ban a compact decay summary, but to prevent a neat number from hiding the conditions that made it possible.

16. Acoustic impedance governs how pressure and flow meet a boundary

Acoustic impedance relates an oscillating pressure to an associated motion or volume flow, depending on the definition being used. It can be frequency-dependent and include a phase relationship. An abrupt impedance mismatch can reflect part of an incoming wave. Efficient transfer is therefore not determined by pressure amplitude alone.

Definitions must remain explicit. Specific acoustic impedance uses pressure divided by particle velocity, whereas an input impedance for an instrument may use pressure divided by volume flow. Their units differ. UNSW’s What Is Acoustic Impedance? explains why pressure and flow can be out of phase. A number copied without its definition can make two correct measurements appear to disagree.

17. Doppler shift depends on relative motion and the wave medium

A moving source changes the spacing of wavefronts ahead of and behind it. A moving receiver changes the rate at which it encounters wavefronts. In ordinary sound problems, the medium supplies a reference for the propagation speed, so source and receiver motion enter the classical Doppler calculation in different ways.

Do not infer that the source’s vibration frequency necessarily changed because the receiver hears a changing frequency. The propagation and reception geometry may explain it. OpenStax’s Doppler treatment describes the distinction. The corresponding shift for light belongs to relativity and is not obtained simply by replacing the sound speed while ignoring the different physical framework.

18. Worked example: an echo measures a path, not just an object

Original hypothetical example. A short sound pulse returns after 0.120 s. Assume propagation speed 343 m/s, a stationary reflecting surface, and a direct outward-and-return path. The total travelled distance is 343 × 0.120 = 41.16 m. The one-way distance is half that, approximately 20.6 m.

Forgetting the return journey doubles the inferred distance. But dividing by two is not sufficient if the recorded feature is actually a second reflection or a longer indirect path. A timing calculation requires a path model. The source time, detector delay and propagation speed must also be known well enough for the required precision.

As a sensitivity exercise, a 1 ms timing uncertainty corresponds to approximately 0.172 m uncertainty in this one-way distance when sound speed is treated as exact. Additional uncertainty in sound speed would contribute separately. This is a calculation about the stated model, not a claim about the accuracy of a particular rangefinder.

19. A spectrum changes the representation of a sound

A waveform shows variation with time. A spectrum describes contributions across frequency. A complex sound can contain a fundamental, harmonics, inharmonic components, noise and transients. The frequency representation is useful because different source and path mechanisms can leave different spectral signatures.

A spectrum still needs a definition of its vertical quantity, units, frequency resolution and time interval. Pressure amplitude, power and power spectral density are not the same plot. UNSW’s What Is a Sound Spectrum? explains the representation. Asking “a peak of what, measured over which interval?” is often the most valuable first question a student can ask of an acoustic graph.

20. Timbre depends on more than the strongest frequency

Two sounds can share a nominal fundamental yet differ in their harmonic strengths, attack, decay and time-varying structure. Those differences contribute to timbre. A single labelled pitch therefore does not contain enough information to reconstruct the sound of an instrument or voice.

UNSW’s loudness and spectra discussion shows why changes in higher harmonics can alter both perceived character and loudness. As a classroom analysis, compare two provided recordings with the same nominal note. Describe the time envelope separately from the spectrum. This prevents “different instrument” from becoming a substitute for explaining which measurable features differ.

21. The source–filter model can be tested, not merely imagined

A useful acoustic model separates an excitation from the frequency-dependent response of a resonating path. The resulting signal depends on both. In voice acoustics, this distinction can connect a source-like excitation with the filtering action of a vocal-tract geometry, without claiming that every detail of living speech is captured by one simple model.

Wolfe, Chu, Chen and Smith report an experimentally measured source–filter model using physical models and measured responses. The educational value is the separation of what was supplied, what the model transformed, and what was recorded at the output. A model earns credibility by predicting those connections, not by merely drawing a box labelled “filter”.

22. Loudness belongs to perception; a microphone measures a physical response

Perceived loudness depends on frequency content and human hearing, not only on one unweighted pressure number. A change of a certain number of decibels should therefore not automatically be translated into the same subjective change for every signal and listener. A physical measurement and a perception report can both be valid while addressing different quantities.

Frequency weighting can be useful when a specified measurement aims to approximate aspects of hearing sensitivity, but the weighting must be stated. It does not turn a sound-level meter into a universal listener. The distinctions are explained in UNSW’s account of intensity, pressure and loudness. This article does not supply exposure limits or diagnose hearing from a recording or device reading.

23. A recording is shaped by the measurement system

To interpret a recording quantitatively, ask what the microphone responds to, where it was placed, how its gain was set, which frequencies were retained, and whether processing changed the signal. Automatic gain adjustment can make a changing sound appear more constant. Clipping can alter a waveform and create additional spectral components. A displayed digital level is not automatically a calibrated sound-pressure level.

UNSW explicitly notes the dependence of computer-based acoustic tests on the sound card, headphones and calibration in its measurement cautions. For a science lesson, use that as a reason to distinguish relative comparisons from absolute claims. A supplied recording can support a frequency-pattern exercise without being suitable evidence for the sound pressure present in the original room.

24. The time window changes what a frequency analysis can resolve

A spectrum made from a short section of a changing sound answers a different question from a spectrum of the whole recording. A longer interval can separate close stationary frequencies more clearly, while averaging over rapid changes. A shorter interval can locate change in time while making precise frequency separation harder. A spectrogram repeats such local analyses across time.

UNSW’s spectrum guide introduces frequency representation, and its spectral examples show time-dependent structure. The practical instruction is to write the analysis interval beside the plot. A claim that a tone “disappeared” is weak if the chosen representation simply lost the resolution needed to separate it.

25. A useful investigation separates source changes from path changes

Consider an original diagnostic scenario: one note in a supplied loudspeaker recording appears much weaker at a particular seat. Possible explanations include the source spectrum, a room cancellation, microphone directionality or processing. Repeating the same recording at the same seat cannot distinguish them. The repetition may establish consistency while leaving the cause unresolved.

A stronger investigation changes one relevant feature. Move the receiver while holding the source and processing fixed. Compare a second source position. Inspect the unprocessed waveform and the spectrum. Ask in advance which candidate explanation predicts which change. The objective is not to collect the largest number of recordings; it is to obtain evidence that separates alternatives.

Keep all listening and demonstrations at comfortable, low levels. Deliberately uncomfortable sounds are unnecessary for these reasoning tasks. Supplied data, animations and ordinary quiet observations can teach the same distinctions without treating loudness as an experimental challenge.

26. Repair an acoustic explanation at the earliest weak link

Suppose an echo distance is wrong by almost a factor of two. Check the path account before inventing an unusual material effect. Suppose two combined levels disagree with an expected 3 dB increase. Check correlation and receiver position before discarding logarithms. Suppose a tube resonance misses a calculation modestly. Check effective length and boundary assumptions before concluding the entire wave model has failed.

These examples suggest a disciplined repair order: define the quantity, verify units and reference, inspect the source–path–receiver model, check the instrument, then decide whether richer physics is needed. A more complicated explanation is useful only when it fixes a specific insufficiency. Preserve the previous prediction so that the improvement remains inspectable rather than becoming an invisible rewrite of the original claim.

27. A learning route from vibration to independent evidence

Primary entry: identify something vibrating, something carrying the disturbance and something receiving it. Distinguish a high or low pitch from a loud or soft sound using words and simple representations. Ask whether changing the receiver’s position changes what is heard. The initial goal is a causal chain, not a numerical scale.

Secondary development: connect frequency, period, wavelength and speed. Introduce reflection, interference and standing waves through clearly defined diagrams. Require the learner to name whether a node concerns pressure or displacement. Use logarithmic examples to expose why decibels cannot be added as ordinary linear quantities.

JC and beyond: connect wave equations, impedance, phase, Fourier descriptions, calibration and uncertainty. Compare a predicted spectrum with an observed one, then design a change that separates a source effect from a room or instrument effect. This is a proposed teaching progression, not an assertion about the contents or rules of any particular current examination.

28. Checkpoints with worked reasoning

Does 0 dB SPL mean there is no pressure variation? No. It means the measured root-mean-square pressure equals the stated reference. The logarithm describes a ratio; the reference must travel with the number.

Do two equal sound sources always produce a 3 dB increase? No. That result applies to equal uncorrelated contributions under the specified averaging and measurement conditions. Coherent signals can reinforce or cancel differently at different positions.

Why divide an echo travel distance by two? In the simple direct-return model, the measured time includes the outward and return journeys. More complicated paths require their own geometry rather than automatic use of the same division.

A room recording has a strong spectral peak. Does that identify the source’s strongest emitted frequency? Not necessarily. The room and receiver can emphasise a frequency. The conclusion needs a source–path–receiver analysis.

Can two sounds with the same fundamental sound different? Yes. Their other frequency components, envelopes and time-varying behaviour can differ. One number does not reconstruct a complete sound. These answers apply the wave, level and spectrum models explained above.

29. The final test: explain what the recording can and cannot establish

A complete acoustic explanation should identify the source, medium, path, boundary behaviour, measured quantity and receiver response. It should distinguish a calculated prediction from a measurement, a relative signal comparison from a calibrated level, and a physical quantity from a perception report. That makes the claim reproducible in meaning even before another person repeats the experiment.

For independent practice, take one provided waveform, one spectrum and a short description of the recording setup. Write three claims the record supports and three it does not. Then identify the smallest extra measurement needed to settle one of the unsupported claims. That exercise teaches the core scientific habit of acoustics: hear the phenomenon, but make the evidence chain explicit.

Sources and connected routes

The scientific foundation comes from OpenStax University Physics sections on sound waves, intensity, standing-wave modes and the Doppler effect, and the authored resources of UNSW Music Acoustics on spectra, impedance, resonance, loudness and experimentally tested source–filter models. The calculated examples and hypothetical diagnostic cases are original teaching constructions, not eduKate laboratory results. Follow the links in the relevant sections for derivations and experimental context.

Neuroscience develops the biological processing of signals; Materials Science develops structure and material response; and Optics provides a useful comparison with another wave-based measurement discipline. Analogy helps organise questions, but sound and light do not become the same physical medium.

Return to How Science Works or the How X Works Hub.