Ear Acoustics Explorer
Model the ear canal as an acoustic transmission line ending in a lumped middle-ear network. See how fluid, pressure and ossicle problems change absorbance, the tympanogram and the reflected waveform, then size the probe: sample rate, FFT length, SNR and stimulus band.
1 · Condition
Middle ear
Stiffness
2 · Ear canal & probe
3 · Measurement design
Result
| Measure | Model | Normative 90 % range | Source |
|---|
Wideband absorbance
A = 1 − |R|² at the probe. Shaded: normal-adult 90 % range. Dashed: the normal ear for the same age group and canal.
Tympanogram
Admittance magnitude at the probe while the ear-canal pressure sweeps. Shaded: normal peak-pressure region (−100 to +50 daPa).
Incident vs reflected
The stimulus, and the wave that comes back from the eardrum to the probe (inverse FFT of the stimulus spectrum × R(f)). The probe mic actually hears their sum.
Acoustic reflectometry
An unsealed tip at the canal entrance hears incident + reflected sound. Reflection from the eardrum cancels the incident wave near the quarter-wave frequency, c/(4L). A stiffer, fluid-backed eardrum reflects more and deepens the dip. Shaded: the 1.8–4.4 kHz sweep used by commercial reflectometers.
Measurement design
What the electronics have to deliver for this canal and condition. All values are engineering estimates.
Per-bin SNR across the band
Network element values for this condition
| Block | Element | SI | CGS |
|---|
A tube, a lumped middle ear, and the numbers a probe has to measure
Every acoustic middle-ear test does the same thing: put a known sound into a sealed (or nearly sealed) ear canal and work out how much of it the eardrum takes in and how much it sends back. Tympanometers do it at one tone while the canal pressure sweeps. Wideband acoustic immittance (WAI) systems do it across 0.2–8 kHz. Acoustic reflectometers, including the smartphone-plus-paper-funnel approach, listen to the interference between incident and reflected sound near the canal's quarter-wave frequency. This page models all three with one physics chain.
1. The ear canal as a transmission line
Below about 8 kHz the canal carries plane waves, so it behaves as a uniform tube of area A = πd²/4 between the probe tip and the eardrum (residual length L = canal length − insertion depth). Air properties are taken at body temperature (37 °C: c = √(γRT) = 353 m/s, ρ = 1.14 kg/m³).
With losses switched on, the propagation constant and Z0 pick up first-order viscous and thermal boundary-layer terms (Γ = (jω/c)·√((1+εv)(1+εt)), εv = (1−j)δv/a, εt = (γ−1)(1−j)δv/(a√Pr), δv = √(2μ/ρω)). That adds a little absorbance in the canal itself, so the probe reads slightly more than the eardrum.
2. The middle ear as a lumped network
The eardrum load uses the topology of Kringlebotn's (1988) network for the human middle ear, in a Zwislocki-style analog where series elements share volume velocity:
Starting values are Kringlebotn's, read from S. Lansbergen's MIT-licensed MATLAB implementation of that network. The code here is a new JavaScript implementation of the same topology. The element values were then refitted by least squares, pulled toward the original values, so that the model's normal adult hits:
- the normal-adult mean absorbance of Karuppannan et al. (2024) at ten frequencies, on a dense 1/6-octave grid with a smoothness penalty (so no hidden resonances);
- the 226 Hz norms of Margolis & Heller (1987) and an admittance phase inside Wiley et al.'s (1987) 56–76°;
- the full-effusion mean absorbance of the same 2024 study, with a flat 226 Hz tympanogram.
The fit moved no element more than 2.5× from its original value. The table below is the evidence:
| Frequency | Model, normal | Published mean ± SD | Model, full effusion | Published effusion mean ± SD |
|---|
| 226 Hz measure | Model, normal adult | Margolis & Heller (1987) mean, 90 % range |
|---|
3. Static pressure and the tympanogram
A tympanometer sweeps the canal from +200 to −400 daPa. The pressure difference across the eardrum, Δp = p_canal − p_middle-ear, stretches it. That is modelled with an empirical stiffening law applied to the eardrum compliances Cs and Cr. It follows the same idea as Vanhuyse et al. (1975): reactance is a symmetric, peaked function of pressure.
w and the exponent were set so that the normal adult reproduces the Margolis & Heller means for Ytm, ECV and TW together. The residual eardrum admittance at +200 daPa makes ECV read about 20 % high, as real tympanometers do (Shanks & Shohet 2009 report 24–39 %). Pressurizing only the eardrum branches also gives the behavior seen in wideband tympanometry: as |Δp| grows, low-frequency absorbance collapses and the absorbance ridge moves up toward 3–5 kHz.
Classification follows Jerger (1970) and Jerger et al. (1972), with numeric limits from the Margolis & Heller 90 % ranges:
- B: no usable peak (Ytm < 0.1 mmho), split by volume. Large volume (adult > 2.5 cm³, child > 1.0 cm³) means a perforation or open tube. A small volume (< 0.3 cm³) means an occlusion. A normal volume is consistent with effusion.
- C: peak below −100 daPa.
- As / Ad: normal pressure, with Ytm below the 5th or above the 95th percentile.
- A: everything inside the ranges.
At 1000 Hz the canal and eardrum are near resonance, so tympanograms can be notched. The tool shows them but does not type them.
4. Conditions: what each preset changes
- Normal adult: the calibrated baseline: 2.5 cm × 7.5 mm canal, 0.5 cm insertion, 6 cm³ middle-ear air (published averages run 4–9 cm³; Molvaer et al. 1978; Stepp & Voss 2005).
- Normal child: 2.0 cm × 6.5 mm canal, 3 cm³ of air (the mastoid is still developing), eardrum 1.4× stiffer and w = 17 daPa. That puts Ytm, ECV and TW inside the preschool ranges.
- Negative middle-ear pressure: p_ME = −250 daPa. The peak moves to −250 daPa (Type C). At ambient the eardrum is pre-stretched, so absorbance below ~2 kHz falls.
- Serous effusion, partial: 40 % fill with −150 daPa. Fluid lowers the air compliance (stiffer) and touches the eardrum (mass + loss), the two mechanisms identified by Ravicz, Rosowski & Merchant (2004). The 226 Hz peak survives because a little air is enough at low frequency, while absorbance at 0.5–1.5 kHz drops hard. That is why WAI beats 226 Hz admittance for effusion (Beers et al. 2010).
- Mucoid effusion, full: 100 % fill with viscous fluid, 2.7 % of the air left as bubbles. The result is a flat tympanogram with a normal canal volume, and absorbance of 0.1–0.4 everywhere, close to the measured effusion means (Ellison et al. 2012; Karuppannan et al. 2024).
- Otosclerosis: stapes and incudostapedial joint 40× stiffer, plus the eardrum–malleus coupling 2.5× stiffer. Review note: in this network part of the eardrum motion bypasses the ossicles, so stapes fixation alone lowers Ytm only ~30 % (still Type A, which is true of many otosclerotic ears). The extra coupling stiffness stands in for advanced fixation and produces the classic As pattern, with low absorbance below ~1.25 kHz (Nakajima et al. 2012).
- Ossicular discontinuity: the load beyond the incudostapedial joint drops to zero. The eardrum and malleus resonate sharply near 0.5 kHz: a narrow absorbance peak (a reflectance notch at 0.5–0.8 kHz in patients; Nakajima et al. 2012) and a deep Ad tympanogram.
- TM perforation / PE tube: a 2 mm hole (acoustic mass ρ(t + 1.7a)/(πa²) plus orifice wall loss) couples the canal to the middle-ear air, per Voss et al. (2001). Static pressure equalizes through the hole, so the tympanogram is flat, and the probe sees canal plus middle-ear volume.
5. Acoustic reflectometry and the spectral gradient
Reflectometry was introduced by Teele & Teele (1984). An unsealed probe at the canal entrance sends a 1.8–4.4 kHz sweep and records incident plus reflected sound. Near the quarter-wave frequency the two cancel and the response shows a dip. A mobile eardrum absorbs energy and gives a broad, shallow dip. Fluid behind the eardrum reflects more and gives a narrow, deep one.
The commercial device reduces the dip to a "spectral gradient" angle between the slopes on either side, with cut-offs near 95° and 75° in its own units (Combs et al., US 5,699,809). The smartphone version (Chan et al. 2019, the research behind WavelyDx) plays 150 ms, 1.8–4.4 kHz FMCW chirps through a paper funnel with a 7 mm tip, records at 48 kHz, finds the dip in 2.3–3.8 kHz, and classifies the shape around it. It reported an AUC of 0.898 versus 0.776 for a commercial reflectometer. The angle alone, computed with ±200 Hz reference points, gave an AUC of only 0.687.
Here the response is modelled as
The angle uses the same ±200 Hz reference points. Treat it as an approximation: the axis scaling is ours, and the model's funnel is an idealized continuation of the canal.
6. Measurement design
Three results usually surprise people:
- Time separation is not available. A 2 cm residual canal gives a 113 µs round trip. Separating incident from reflected sound in time would need about 18 kHz of bandwidth, so real WAI systems use a source calibrated in long tubes (Thévenin method; Allen 1986; Keefe et al. 1992) instead of time gating.
- The quarter-wave null at the probe (about 4.4 kHz for a 2 cm residual canal) puts the probe mic up to 20+ dB below the incident level when the eardrum reflects strongly. That costs ADC and mic dynamic range right where the effusion signature lives.
- Chirps buy SNR. A click puts its energy into a fraction of a millisecond, a chirp spreads the same peak level over tens of milliseconds. Note the ceiling: broadband stimuli above roughly 65–70 dB SPL can trigger the stapedius reflex, which changes the very impedance being measured. Body noise (breathing, blood flow, swallowing) usually dominates the mic noise below ~500 Hz.
Limits
- This is a teaching and scoping model. It is not a medical device, it is not validated for any clinical use, and it must not be used for diagnosis or screening.
- The canal is a uniform, rigid-walled cylinder. Real canals taper and bend, have compliant cartilaginous walls (much more so in infants), and change area near the eardrum. Above ~6–8 kHz the uniform-tube result is only indicative.
- The middle ear is one lumped network with one set of values calibrated to group means. Real ears vary widely: normative absorbance SDs are 0.04–0.16. The network cannot reproduce eardrum modal behaviour or the ear-to-ear spread.
- The pressure law is empirical and symmetric. Real tympanograms are slightly asymmetric, show hysteresis with sweep direction and rate, and overestimate middle-ear pressure by 10–70 daPa. The model has none of that.
- The normative absorbance band comes from one adult study on one instrument (Interacoustics Titan, ambient pressure, mean ± 1.645 SD). Systems differ, especially above 4 kHz, and children need their own norms (Beers et al. 2010). The same adult band is shown for the child presets.
- The perforation preset gives higher mid-frequency absorbance than the measured means (about 0.9 vs 0.5–0.6) and is sensitive to hole size. The otosclerosis preset's tympanogram is wide, while clinical otosclerotic tympanograms are often narrow ("peaky").
- Fluid is a series mass and loss on the eardrum plus lost air volume. The serous/mucoid difference (a 2/3 loss ratio) is a modelling choice; fluid volume matters far more than viscosity (Ravicz et al. 2004).
- The SNR budget counts mic self-noise only. Add body noise, leaks, transducer non-linearity, source calibration error and probe-fit variation before you size a real channel.
References
- Feeney MP, Hunter LL, Kei J, et al. Consensus statement: Eriksholm workshop on wideband absorbance measures of the middle ear. Ear Hear 2013;34 Suppl 1:78S–79S.
- Karuppannan A, Barman A, Mamatha NM. Wideband absorbance pattern and its diagnostic value in adults with middle ear effusions and tympanic membrane perforation. J Int Adv Otol 2024;20(2):158–163 (normal and effusion absorbance, Table 2).
- Margolis RH, Heller JW. Screening tympanometry: criteria for medical referral. Audiology 1987;26:197–208. Ranges as tabulated in Shanks J, Shohet J, Tympanometry in clinical practice, in Katz J (ed.), Handbook of Clinical Audiology, 6th ed., ch. 9, Table 9.1 (2009).
- American Speech-Language-Hearing Association. Tympanometry (ASHA Working Group on Aural Acoustic-Immittance Measurements), 1988: definitions of TW, the 226 Hz volume equivalence.
- Jerger J. Clinical experience with impedance audiometry. Arch Otolaryngol 1970;92:311–324. Jerger J, Jerger S, Mauldin L. Studies in impedance audiometry. I. Normal and sensorineural ears. Arch Otolaryngol 1972;96:513–523 (As and Ad subtypes).
- Molvaer OI, Vallersnes FM, Kringlebotn M. The size of the middle ear and the mastoid air cell. Acta Otolaryngol 1978;85:24–32.
- Wiley TL, Oviatt DL, Block MG. Acoustic-immittance measures in normal ears. J Speech Hear Res 1987;30:161–170 (226 Hz phase 56–76°).
- Kringlebotn M. Network model for the human middle ear. Scand Audiol 1988;17(2):75–85.
- Lansbergen S. middle_ear_model: lumped-element dynamic middle-ear model with pressure variation (MATLAB). MIT License. github.com/Lansbergen/middle_ear_model. Topology and starting element values; no code copied.
- Zwislocki J. Analysis of the middle-ear function. Part I: Input impedance. J Acoust Soc Am 1962;34:1514–1523.
- Vanhuyse VJ, Creten WL, Van Camp KJ. On the W-notching of tympanograms. Scand Audiol 1975;4:45–50.
- Voss SE, Rosowski JJ, Merchant SN, Peake WT. Middle-ear function with tympanic-membrane perforations. II. A simple model. J Acoust Soc Am 2001;110:1445–1452.
- Stepp CE, Voss SE. Acoustics of the human middle-ear air space. J Acoust Soc Am 2005;118:861–871.
- Ravicz ME, Rosowski JJ, Merchant SN. Mechanisms of hearing loss resulting from middle-ear fluid. Hear Res 2004;195:103–130.
- Ellison JC, Gorga M, Cohn E, Fitzpatrick D, Sanford CA, Keefe DH. Wideband acoustic transfer functions predict middle-ear effusion. Laryngoscope 2012;122:887–894.
- Beers AN, Shahnaz N, Westerberg BD, Kozak FK. Wideband reflectance in normal Caucasian and Chinese school-aged children and in children with otitis media with effusion. Ear Hear 2010;31:221–233.
- Merchant GR, Al-Salim S, Tempero RM, Fitzpatrick D, Neely ST. Improving the differential diagnosis of otitis media with effusion using wideband acoustic immittance. Ear Hear 2021;42:1183–1194.
- Rosowski JJ, Nakajima HH, Hamade MA, et al. Ear-canal reflectance, umbo velocity, and tympanometry in normal-hearing adults. Ear Hear 2012;33:19–34. Nakajima HH, et al. Comparison of ear-canal reflectance and umbo velocity in patients with conductive hearing loss. Ear Hear 2012;33:35–43.
- Voss SE, Allen JB. Measurement of acoustic impedance and reflectance in the human ear canal. J Acoust Soc Am 1994;95:372–384. Keefe DH, Ling R, Bulen JC. Method to measure acoustic impedance and reflection coefficient. J Acoust Soc Am 1992;91:470–485. Allen JB. Measurement of eardrum acoustic impedance. In Peripheral Auditory Mechanisms, Springer, 1986.
- Siegel JH. Ear-canal standing waves and high-frequency sound calibration using otoacoustic emission probes. J Acoust Soc Am 1994;95:2589–2597.
- Teele DW, Teele J. Detection of middle ear effusion by acoustic reflectometry. J Pediatr 1984;104:832–838. Combs JT, et al. Device and process for generating and measuring the shape of an acoustic reflectance curve of an ear. US Patent 5,699,809 (1997).
- Chan J, Raju S, Nandakumar R, Bly R, Gollakota S. Detecting middle ear fluid using smartphones. Sci Transl Med 2019;11:eaav1102. Product description: wavelydx.com.
- Keefe DH. Acoustical wave propagation in cylindrical ducts: transmission line parameter approximations for isothermal and nonisothermal boundary conditions. J Acoust Soc Am 1984;75:58–62.
Scoping an acoustic sensing channel for a medical device?
Kyros designs the probe electronics, transducer drive, low-noise capture and firmware for acoustic measurements, and builds the verification evidence that a regulated product needs.
