A generic commercial parametric array measured 118 dB of 40 kHz ultrasound at 3.5 meters, at least 17 dB above the general-public guideline it was checked against (Dolder et al., Proceedings of ICA 2019). The same class of device can squeeze audible sound into a beam only 3 degrees wide (Pompei, JASA 1999), but it turns generally less than 0.15% of its ultrasonic power into audible sound (Li et al., JASA 2023), and its output drops by about 12 dB every time the audio frequency halves (Zhong, Kirby and Qiu, JASA 2020). That trade-off, a tight beam bought with a lot of inaudible energy, defines the technology. We aggregated data from the Journal of the Acoustical Society of America, the APSIPA Transactions on Signal and Information Processing, Health Canada, ICNIRP, MDPI Applied Sciences and Sensors, Nature Communications, and manufacturer documentation, which we label as vendor claims. For the factory side of ultrasound exposure, see our industrial ultrasonic noise exposure statistics.
TL;DR
- Westervelt’s 1963 “Parametric Acoustic Array” paper has 812 citations in Crossref as of October 2026 (Crossref metadata for JASA vol. 35, 1963).
- Carrier frequencies in the devices reviewed here span 27 kHz to 90 kHz (Aoki, Kamakura and Kumamoto 1991; Ahn et al., Sensors 2019).
- Pompei cut distortion from up to 50% THD to 5% while keeping a 3 degree beam (JASA 1999).
- Beamwidth can double in a hot, humid climate compared with a cold, dry one (Shi, Kajikawa and Gan, APSIPA 2014).
- Audible output falls about 12 dB each time the audio frequency halves (Zhong, Kirby and Qiu, JASA 2020).
- Conversion efficiency from ultrasound to audio is generally below 0.15% (Li et al., JASA 2023).
- Single-sideband modulation kept distortion below the 5% (-26 dB) threshold over the working range, while plain double-sideband stayed above -20 dB (San Martin et al., Applied Sciences 2020).
- Health Canada allows 110 dB in the 25 to 50 kHz third-octave bands (Safety Code 24, 1991).
- A commercial unit measured 118 dB in the 40 kHz band at 3.5 m, exceeding the ICNIRP public guideline by at least 17 dB (Dolder et al., ICA 2019).
- Nine listeners exposed to 40 kHz ultrasound at up to 120 dB showed no significant hearing changes versus controls (Carcagno et al., Acta Acustica 2019).
1. From a 1963 Theory to the Audio Spotlight: How the Parametric Array Grew
A parametric array is a way of making sound with sound. Instead of moving air directly at audio frequencies, the loudspeaker emits an inaudible ultrasonic beam (the “carrier”) whose strength is shaped by the audio signal. Air does not respond perfectly in proportion at such high pressures (it is “nonlinear”), so as the beam travels, a small part of the ultrasound is converted back into audible sound, a process called self-demodulation. The column of air in front of the device behaves like a long, invisible loudspeaker array, which is what keeps the audible sound narrow.
The idea took 20 years to move from theory (1963) to a loudspeaker (1983), and another 17 years to reach a commercial product (2000, according to the vendor). Westervelt’s 1963 paper is still the field’s anchor, with 812 citations in Crossref, more than double the 287 citations of the 1983 “audio spotlight” paper that first proposed the loudspeaker (Crossref metadata, read October 2026). Holosonics, which sells the Audio Spotlight brand, says that Japanese electronics companies tried and nearly abandoned commercial versions by the end of the 1980s; that is the company’s own history, not an independent record.
| Metric | Value | Source |
|---|---|---|
| Original parametric acoustic array paper | 1963, JASA vol. 35, pp. 535 to 537; 812 Crossref citations | Westervelt, JASA 1963 |
| Theory of self-demodulation of a modulated ultrasonic beam | 1965 | Berktay 1965, as listed in Shi, Kajikawa and Gan, APSIPA 2014 |
| First published parametric array in air | 1975, JASA vol. 57, pp. 562 to 568; 112 Crossref citations | Bennett and Blackstock, JASA 1975 |
| First parametric loudspeaker (“the audio spotlight”) | 1983, JASA vol. 73, pp. 1532 to 1536; 287 Crossref citations | Yoneyama et al., JASA 1983 |
| Early research prototype: projector radius and carrier | 21 cm radius, 27 kHz carrier | Aoki, Kamakura and Kumamoto, Electronics and Communications in Japan 1991 |
| Commercial Audio Spotlight systems in use since | 2000 | Holosonics company history (vendor claim) |
| Newest peer-reviewed design in this roundup | Published May 22, 2026, 60 kHz carrier, 100 mm plate | Kim, Oh, Moon and Rho, Nature Communications 2026 |
Most recent available data for the historical rows is necessarily old: these are the founding papers. Citation counts are a snapshot of Crossref on October 27, 2026 and grow every month.
2. Carrier Frequencies: From 27 kHz Prototypes to 90 kHz MEMS Arrays
The carrier frequency is a compromise. It has to sit well above the upper limit of human hearing: the median upper hearing limit is 17.9 kHz for people aged 20 to 29 and falls to 12.8 kHz for people aged 70 to 79 (Lawton 2001, as tabulated in AGNIR, Health Effects of Exposure to Ultrasound and Infrasound, 2010). But air attenuates sound roughly with the square of frequency (Skinner, Groves and Hinders, Applied Acoustics 2019), and the AGNIR report says ultrasound above about 300 kHz essentially does not propagate through air. The usable window is therefore narrow. 40 kHz is the most common research carrier, and the 2026 Nature Communications paper names the Murata MA40S4S, an off-the-shelf element fixed at 40 kHz, as its example of the emitters used in array loudspeakers.
Moving the carrier upward shrinks the hardware. At 100 kHz the wavelength is only 3.43 mm, which is what let a Korean team fit 672 micro-machined transducers into a 42 mm by 52 mm radiating area for a phone-sized prototype (Ahn et al., Sensors 2019).
| Metric | Value | Source |
|---|---|---|
| Carrier of an early research parametric loudspeaker | 27 kHz | Aoki, Kamakura and Kumamoto 1991 |
| Carrier used in the directivity-control examples, with lower and upper sidebands (the extra frequencies that carry the audio) | 40 kHz; lower sideband 39 to 28 kHz, upper sideband 41 to 52 kHz | Shi, Kajikawa and Gan, APSIPA 2014 |
| Two primary tones in a length-limited array simulation (0.3 m aperture radius) | 47.5 kHz and 52.5 kHz | Skinner, Groves and Hinders, Applied Acoustics 2019 |
| Design carrier of a metamaterial parametric speaker | 60 kHz | Kim et al., Nature Communications 2026 |
| Carrier measured on a 0.6 m by 0.6 m parametric panel | 64 kHz | Zhong, Kirby and Qiu, JASA 2020 |
| Carrier stated for the Holosonics AS-24iX film transducer | About 60 kHz (vendor claim) | Holosonics AS-24iX product page 2026 |
| Measured resonances of a MEMS array built for mobile devices | 83 kHz and 90 kHz | Ahn et al., Sensors 2019 |
| Frequency above which ultrasound effectively stops propagating in air | About 300 kHz | AGNIR RCE-14, 2010 |
The 64 kHz independent measurement and the roughly 60 kHz figure in the vendor documentation are consistent, which suggests that larger film-transducer panels have moved above the classic 40 kHz point. The same 40 kHz arrays are also the workhorse of acoustic levitation, covered in our acoustic levitation statistics.
3. Beam Width and Directivity: How Narrow Is the Spotlight?
Directivity is the reason anyone accepts the drawbacks. Because the audible sound is generated along the ultrasonic beam, its width is set mainly by the ultrasonic aperture and wavelength rather than by the much longer audio wavelength. Pompei’s 2002 MIT thesis describes the parametric array as giving much higher directivity than traditional loudspeakers of similar size, and the APSIPA overview adds that it can transmit an equally narrow beam with a smaller driving unit. Pompei reported a 3 degree beamwidth in 1999, and even a palm-sized MEMS array kept its half-power beamwidth (the angle within which the level stays within 3 dB of the on-axis value) near 7 degrees from 5 to 10 kHz.
Two caveats matter for real rooms. Beamwidth depends on weather: the APSIPA authors found it can double in a hot, humid climate compared with a cold, dry one, while staying almost constant for relative humidity between 40% and 80%. And sound leaks backward: in the 0.6 m panel measured by Zhong and colleagues, the region behind the device where the 315 Hz level exceeded 35 dB was roughly circular with a radius of about 1.3 m.
| Metric | Value | Source |
|---|---|---|
| Beamwidth of a low-distortion airborne parametric array | 3 degrees | Pompei, JASA 1999 |
| Half-power beamwidth of a MEMS parametric array at its 83 and 90 kHz carriers | About 6 degrees | Ahn et al., Sensors 2019 |
| Half-power beamwidth of the same array for audible sound at 5 kHz and at 10 kHz | 7.3 degrees and 7.1 degrees | Ahn et al., Sensors 2019 |
| Null-to-null beamwidth (angle between the first silent directions) of an 8-channel, 40 kHz parametric array | 40 degrees | Shi, Kajikawa and Gan, APSIPA 2014 |
| Change in beamwidth, hot and humid climate vs cold and dry climate | Can double | Shi, Kajikawa and Gan, APSIPA 2014 |
| Audible level at 315 Hz, 0.1 m behind the panel vs 1.0 m in front | 45 dB vs about 61.4 dB | Zhong, Kirby and Qiu, JASA 2020 |
| Beam stated for the Holosonics AS-24iX | 5 degrees; 1.5 m by 1.5 m at a typical listening distance of 2 m (vendor claim) | Holosonics AS-24iX product page 2026 |
The front-to-back gap in the Zhong row works out to about 16 dB (61.4 minus 45), a figure the paper does not state itself. Treat the vendor beam row with care: a full 5 degree cone at 2 m would be only about 0.17 m wide (2 x 2 m x tan 2.5 degrees), so the 5 degree and 1.5 m figures on the same page cannot describe the same edge of the beam (see the Data watch note).
4. Audible Output, Low-Frequency Roll-Off, and Conversion Efficiency
The physics that makes the beam narrow also starves the bass. Theory and measurements agree on a slope of about 12 dB per octave: halve the audio frequency and you lose roughly 12 dB of output, which is why parametric speakers sound thin below a few hundred hertz. The efficiency picture is even starker. Li and colleagues calculated that audio sound power rises with the square of the audio frequency and with aperture size, falls as the ultrasonic frequency goes up, and that the ultrasound-to-audio conversion efficiency is generally below 0.15% for typical configurations (Li, Zhong, Jing and Lu, JASA 2023).
That is why power draw, not loudness, dominates engineering papers. The Korean MEMS team cut its system’s consumption from 6.65 W to 3.25 W with a matching circuit, against the 65 to 100 W nominal it cites for conventional parametric loudspeakers, yet it still produced only about 62 dB of audible sound at 0.45 m.
| Metric | Value | Source |
|---|---|---|
| Measured rise in audible output between 300 and 1500 Hz (40 kHz array) | About 12 dB per octave | San Martin, Tello, Valencia and Marzo, Applied Sciences 2020 |
| Drop in audible output when the audio frequency is halved | 12 dB | Zhong, Kirby and Qiu, JASA 2020 |
| Ultrasound-to-audio conversion efficiency, typical configurations | Generally below 0.15% | Li et al., JASA 2023 |
| Audible (difference-frequency) level from a 125 dB ultrasonic primary wave, MEMS array, 0.45 m | About 62 dB | Ahn et al., Sensors 2019 |
| Audible level at the ear of an artificial head, parametric speaker vs conventional speaker, set for a listening test | 72.3 dBA vs 72.4 dBA | Lee, Shimomura and Katsuura, Journal of Physiological Anthropology 2012 |
| Audible band of the 2026 metamaterial design, aimed at a listener 0.5 m away | 500 Hz to 10 kHz (four octaves) | Kim et al., Nature Communications 2026 |
| Maximum level stated for the Holosonics AS-24iX | 95 dB SPL at 1 kHz; 90 W maximum power draw (vendor claim) | Holosonics AS-24iX product page 2026 |
The Nature Communications authors are candid that their prototype’s level is “relatively limited” compared with conventional loudspeakers and blame the “inherently low nonlinear conversion efficiency”. Better transducers have not changed the underlying physics.
5. Harmonic Distortion and Preprocessing: From 50% THD to Single-Digit Figures
Self-demodulation in air is itself a nonlinear process, so feeding a parametric array plain amplitude modulation gives the loudest output but also a strong second harmonic. Engineers fight this by preprocessing the audio before modulation: taking its square root (square-root AM, SRAM), sending only one sideband (single-sideband AM, SSB), or using modified schemes (MAM). The modulation index, written m, is how deeply the audio varies the carrier; deeper modulation is louder but more distorted. Pompei showed in 1999 that pure AM with narrow-band transducers produced up to 50% total harmonic distortion (THD), and that preprocessing plus wideband transducers brought it down to 5%.
Hardware still limits the math. The 49-transducer, 16 mm array tested in 2020 peaked at 40.9 kHz with a -3 dB bandwidth of only 1.5 kHz, a narrow window for the extra frequencies that preprocessing schemes add to the carrier. The same team noted that carrier levels above 120 dB are common even 2 m from such arrays, enough to overload ordinary measurement microphones; for comparison with conventional small drivers, see our smartphone speaker distortion statistics.
| Metric | Value | Source |
|---|---|---|
| THD with pure AM and narrow-band transducers vs after audio preprocessing and wideband transducers | Up to 50% vs 5% | Pompei, JASA 1999 |
| Distortion level treated as generally acceptable for domestic equipment | Below -26 dB (5%) | San Martin et al., Applied Sciences 2020 |
| Distortion with plain double-sideband AM | Above -20 dB across virtually the whole range | San Martin et al., Applied Sciences 2020 |
| Single-sideband AM (lower or upper sideband) | Below the -26 dB threshold across the working range | San Martin et al., Applied Sciences 2020 |
| Audible level given up by the recommended scheme (modified AM 1, m = 0.6) | 2 dB | San Martin et al., Applied Sciences 2020 |
| Residual in-band distortion after recursive equalization, in simulation | -70 dB or lower | Lee and Gan, IEEE Transactions on Audio, Speech and Language Processing 2006 |
There is no single winner. In the 2020 study, single-sideband schemes had the lowest harmonic distortion but the worst intermodulation distortion (distortion created when two tones mix), while square-root and modified AM did best on intermodulation. Harmonics above the second were negligible in every scheme. Most recent available data for the 50% and 5% figures: Pompei, JASA 1999.
6. Airborne Ultrasound Exposure Limits: What the Guidelines Allow
A parametric speaker works only if its ultrasound is intense, so its exposure rules matter more than for any ordinary loudspeaker. Limits are set per third-octave band, a slice of the spectrum one third of an octave wide, named by its center frequency. The striking feature is the cliff at 20 to 25 kHz: Health Canada allows 75 dB in the 20 kHz band but 110 dB from 25 kHz upward, and its table notes that subjective effects can occur immediately, so the limits do not depend on exposure time (Health Canada, Safety Code 24, 1991).
These numbers are old and thinly supported, and their authors’ successors say so. In its 2024 review, ICNIRP concluded that the 1984 IRPA limits rest on limited evidence and could be either too high or too low, but that formal revision must wait for more research (ICNIRP statement on airborne ultrasound, 2024).
| Metric | Value | Source |
|---|---|---|
| Canadian limit, 16 kHz and 20 kHz third-octave bands | 75 dB | Health Canada Safety Code 24, 1991 |
| Canadian limit, 25, 31.5, 40 and 50 kHz third-octave bands | 110 dB | Health Canada Safety Code 24, 1991 |
| Canadian total sound pressure level limit for other body parts when ear protection is worn | 137 dB | Health Canada Safety Code 24, 1991 |
| IRPA 1984 general-public limit, 20 kHz band | 70 dB (5 dB below the occupational 75 dB) | ICNIRP statement 2024 |
| IRPA 1984 general-public limit, 25 kHz band and above | 10 dB below the occupational 110 dB (100 dB) | ICNIRP statement 2024 |
| IRPA allowance for occupational exposure under 1 hour | +9 dB (ICNIRP says it lacks justification) | ICNIRP statement 2024 |
| Ultrasonic levels not demonstrated to cause hearing loss | Below 120 dB | Health Canada Safety Code 24, 1991 |
| Level above which acute harmful effects occur in humans | Approximately 155 dB | Health Canada Safety Code 24, 1991 |
OSHA’s guidance, in place since 2004, allowed an extra 30 dB on maximum permissible levels; it has since been dropped (Acoustics Today, Fall 2020, Exposure to Ultrasound in Air). Most recent available data: Health Canada Safety Code 24, 1991, and IRPA interim guidelines, 1984, as reviewed by ICNIRP in 2024.
7. Measured Exposures and Human Studies Near Ultrasonic Beams
A published laboratory audit of a commercial parametric array is not reassuring. Measuring on axis at 3.5 m in an anechoic chamber (a room with no echoes; see our anechoic chamber testing statistics), Dolder and colleagues found 118 dB in the 40 kHz band, above every guideline they compared. Turning the gain down cut the audible A-weighted output by 10 dB but the 40 kHz pump signal by only 1 dB, so a quieter-sounding beam is not necessarily a lower ultrasound exposure (Dolder et al., ICA 2019).
The human data are small and mostly null. A 2019 study exposed one ear of nine young listeners to 40 kHz ultrasound at up to 120 dB and found no significant functional change compared with nine controls (Carcagno, Di Battista and Plack, Acta Acustica united with Acustica 2019), and a nine-subject physiological study at 0.3 m and 1.0 m from a 40 kHz parametric speaker found no significant difference from a conventional speaker. Hearing sensitivity is the wild card: ICNIRP notes that 40% of children tested had thresholds below 100 dB at 26 kHz.
| Metric | Value | Source |
|---|---|---|
| Ultrasound in the 40 kHz band at 3.5 m from a commercial parametric array, high gain vs low gain | 118 dB vs 117 dB | Dolder et al., ICA 2019 |
| Exceedance of the ICNIRP general-public guideline at 40 kHz | At least 17 dB (18 dB at high gain) | Dolder et al., ICA 2019 |
| Exceedance of the ICNIRP occupational and OSHA-based values at 40 kHz, high gain | 8 dB and 3 dB | Dolder et al., ICA 2019 |
| Ultrasound from haptic feedback arrays at about 40 kHz, at 60 cm and at 20 cm (for comparison) | 125 dB and 155 dB | Battista 2019 and Lieber et al. 2019, as tabulated in Acoustics Today, Fall 2020 |
| Sources measured in a public-exposure survey; level of a hand dryer at 40 kHz | 16 sources; 84 dB | Fletcher et al., JASA 2018 |
| Exposure study design: 40 kHz amplitude-modulated tone, levels and duration | 105, 110, 115 and 120 dB, 10 minutes per level; 9 exposed vs 9 controls; no significant changes | Carcagno et al., Acta Acustica 2019 |
| Physiological study near a 40 kHz parametric speaker: subjects and distances | 9 subjects at 0.3 m and 1.0 m; no significant difference vs a conventional speaker | Lee, Shimomura and Katsuura 2012 |
| Ultrasonic hearing thresholds cited by ICNIRP | 88 dB or higher at 24 kHz; 3 of 32 ears below 100 dB at 28 kHz; 40% of children below 100 dB at 26 kHz | ICNIRP statement 2024 |
The Dolder test covers one generic unit under laboratory conditions; it does not establish what a listener at a museum exhibit receives. The AGNIR report put the wider gap plainly: human exposures to ultrasound have not been well quantified except for those from medical devices. Most recent available data: Dolder et al., 2019, and Carcagno et al., 2019.
Summary: Parametric Acoustic Arrays by the Numbers
| Metric | Value | Source |
|---|---|---|
| Crossref citations of Westervelt’s 1963 parametric array paper | 812 | Crossref, JASA 1963 |
| First parametric loudspeaker paper (“the audio spotlight”) | 1983 | Yoneyama et al., JASA 1983 |
| Range of carrier frequencies in the devices reviewed here | 27 kHz to 90 kHz | Aoki et al. 1991; Ahn et al., Sensors 2019 |
| Carrier measured on a 0.6 m by 0.6 m panel | 64 kHz | Zhong, Kirby and Qiu, JASA 2020 |
| Beamwidth of a low-distortion parametric array | 3 degrees | Pompei, JASA 1999 |
| Null-to-null beamwidth of an 8-channel, 40 kHz parametric array | 40 degrees | Shi, Kajikawa and Gan, APSIPA 2014 |
| Beamwidth change, hot and humid vs cold and dry | Can double | Shi, Kajikawa and Gan, APSIPA 2014 |
| Loss in audible output per halving of audio frequency | About 12 dB | Zhong, Kirby and Qiu, JASA 2020 |
| Ultrasound-to-audio conversion efficiency | Generally below 0.15% | Li et al., JASA 2023 |
| Audible level from a 125 dB primary wave at 0.45 m | About 62 dB | Ahn et al., Sensors 2019 |
| THD with pure AM and narrow-band transducers vs after preprocessing | Up to 50% vs 5% | Pompei, JASA 1999 |
| Plain double-sideband AM distortion | Above -20 dB | San Martin et al., Applied Sciences 2020 |
| Audible band of a 2026 metamaterial parametric speaker | 500 Hz to 10 kHz | Kim et al., Nature Communications 2026 |
| Canadian limit, 25 to 50 kHz bands | 110 dB | Health Canada Safety Code 24, 1991 |
| IRPA general-public limit, 20 kHz band | 70 dB | ICNIRP statement 2024 |
| Measured ultrasound in the 40 kHz band at 3.5 m from a commercial unit; exceedance of the ICNIRP public guideline | 118 dB; at least 17 dB | Dolder et al., ICA 2019 |
| Highest exposure level in a controlled 40 kHz hearing study (no significant changes) | 120 dB | Carcagno et al., Acta Acustica 2019 |
Methodology and Sources
We ran 15 targeted searches and read the primary papers, standards and agency documents directly, through publisher pages, PubMed Central, institutional repositories and the Crossref and OpenAlex metadata services. Every figure above was read in a source during research for this update. Manufacturer figures are labeled as vendor claims and kept to a minority of each section.
- Westervelt, “Parametric Acoustic Array”, JASA 1963 and Yoneyama et al., “The audio spotlight”, JASA 1983: founding papers; citation counts from Crossref metadata (Bennett and Blackstock, JASA 1975, also from Crossref).
- Shi, Kajikawa and Gan, “An overview of directivity control methods of the parametric array loudspeaker”, APSIPA Transactions on Signal and Information Processing 2014
- Pompei, “Beams of sound: A low-distortion airborne parametric array”, JASA 1999; Pompei’s 2002 MIT thesis abstract for the directivity comparison.
- Li, Zhong, Jing and Lu, “A note on the audio sound power generated by a parametric array loudspeaker”, JASA 2023
- Zhong, Kirby and Qiu, “A non-paraxial model for the audio sound behind a non-baffled parametric array loudspeaker”, JASA 2020
- San Martin, Tello, Valencia and Marzo, “Experimental Evaluation of Distortion in Amplitude Modulation Techniques for Parametric Loudspeakers”, Applied Sciences 2020
- Ahn et al., “A Critical Step to Using a Parametric Array Loudspeaker in Mobile Devices”, Sensors 2019
- Kim, Oh, Moon and Rho, “Dual-domain metamaterials co-integrated with a compact ultrasonic transducer for highly directional audio generation”, Nature Communications 2026
- Skinner, Groves and Hinders, “Demonstration of a length limited parametric array”, Applied Acoustics 2019
- Aoki, Kamakura and Kumamoto, Electronics and Communications in Japan 1991, and Lee and Gan, IEEE Transactions on Audio, Speech and Language Processing 2006 (abstract and bibliographic metadata via Crossref and OpenAlex).
- Lee, Shimomura and Katsuura, “Physiological responses at short distances from a parametric speaker”, Journal of Physiological Anthropology 2012
- Dolder, Fletcher, Lloyd Jones, Lineton and Leighton, “Exposure levels for parametric arrays in light of guideline ambiguities”, Proceedings of ICA 2019
- Fletcher et al., “Public exposure to ultrasound and very high-frequency sound in air”, JASA 2018
- Carcagno, Di Battista and Plack, Acta Acustica united with Acustica 2019
- Health Canada, Safety Code 24: Guidelines for the Safe Use of Ultrasound, Part II, 1991
- ICNIRP, Validity of the 1984 Interim Guidelines on Airborne Ultrasound and Gaps in the Current Knowledge, Health Physics 2024
- UK Advisory Group on Non-ionising Radiation, RCE-14, Health Effects of Exposure to Ultrasound and Infrasound, 2010
- Acoustics Today, Fall 2020, exposure table reproduced from Leighton et al.
- Holosonics AS-24iX product page and company history, 2026: vendor claims only.
- Data watch: The Holosonics page gives both a 5 degree beam and a 1.5 m by 1.5 m coverage at 2 m; a 5 degree full angle would cover only about 0.17 m at that distance (2 x 2 m x tan 2.5 degrees), so the two numbers likely use different edges or definitions, and Skinner et al. cite a separate Holosonics claim of about 2 m of beam width at 4 m. The Ahn et al. abstract labels the 83 kHz beam as a difference-frequency wave while the results section treats 83 and 90 kHz as primaries; we report them as carriers. The Carcagno et al. abstract mentions a 50 second total duration that conflicts with its 10 minutes per level, so we report only the per-level figure. The IRPA public value of 100 dB for 25 kHz and above follows ICNIRP’s “10 dB below occupational” wording and matches the 100 dB that Dolder et al. tabulate for the 25, 31.5 and 40 kHz bands. The Zhong et al. 315 Hz front and back levels come from their model and measurement figures for one panel and may not generalize. No source we could read gave air-absorption coefficients at 40 to 64 kHz, so this roundup leaves them out. Most exposure data date from 1991 to 2019; ICNIRP said in 2024 that its guidelines cannot yet be formally revised, so new limits would change sections 6 and 7.
Last updated: October 27, 2026. We update this roundup quarterly, and the next refresh is expected when ICNIRP publishes its next airborne ultrasound guidance or new JASA measurements of commercial parametric loudspeakers appear.