A tabla and a knocking engine ring for the same mathematical reason: each is a vibrating circle, so its natural frequencies are set by the zeros of Bessel functions. A plain drumhead's overtones fall at 1.59, 2.14 and 2.30 times its fundamental, which sounds unpitched; the tabla's black spot (syahi) loads the skin so the overtones land near 2, 3, 4 and 5 times, which is what C. V. Raman showed in the 1920s and 1930s. Engine knock rings the hot gas in the cylinder at its own circular modes, about 6.4 kHz for a KTM 390's 89 mm bore and 7.9 kHz for a Pulsar NS200's 72 mm bore.
Calculated in this post from the Bessel zeros (Wikipedia: Vibrations of a circular membrane), Draper's relation for a cylinder (Broatch et al., 2016), the KTM 390 Adventure X bore of 89 mm (moto-data.net) and the Pulsar NS200 bore of 72 mm (Bajaj owner's manual), with these example inputs: burned gas at 2,500 K, γ = 1.28 and M = 0.0285 kg/mol (my assumed values for hot combustion products).
- Bessel zeros set the pitch of any round vibrating thing, whether a drumhead or the gas in a cylinder.
- 1 : 1.59 : 2.14 is a plain drum's overtone pattern; the loaded tabla pulls it to near 1 : 2 : 3 : 4 : 5.
- f = 1.841c/πB, Draper's 1938 equation, gives the first knock mode of a bore B.
- 6.4 kHz for a KTM 390's 89 mm bore and 7.9 kHz for a Pulsar NS200's 72 mm bore, at an assumed 2,500 K.
Calculated in this post from Draper's relation (Broatch et al., 2016) and the KTM 390 and NS200 bores (moto-data.net; Bajaj owner's manual), with these example inputs: burned gas at 2,500 K, γ = 1.28 and M = 0.0285 kg/mol.
- 5–20 kHz is the band knock sensors and researchers filter for.
That frequency band is why knock sensors listen between roughly 5 and 20 kHz. This is a tier-4 collision: two unrelated things paired to find the physics they truly share.
The case is built with numbers, from the wave equation on a circular membrane to the pressure waves in a combustion chamber, and the verdict at the end says how far the link really goes.
Why does a plain drum sound unpitched?
Because the overtones of a uniform circular skin are not whole-number multiples of its fundamental. A stretched membrane obeys the two-dimensional wave equation, and with its rim clamped, the allowed shapes are Bessel functions Jm(kr) times cos(mθ), where m counts nodal diameters.
The rim must not move, so Jm(ka) = 0 at radius a, which gives:
Here τ is the tension per unit length, σ the mass per unit area, and jmn the n-th zero of Jm. The lowest zeros are 2.405, 3.832, 5.136, 5.520, 6.380 and 7.016, so the mode ratios are 1, 1.59, 2.14, 2.30, 2.65 and 2.92.
No integer pattern, so the ear hears a thud with no clear note. That is fine for a bass drum keeping time; it would be a disaster accompanying a raga.
What does the tabla's black spot do?
The syahi, a paste of iron filings, soot and starch built up in layers at the centre of the dayan (the right-hand drum), makes the membrane heavier in the middle. Mass per unit area σ now depends on radius, which changes the wave equation to:
With the right σ(r), the modes shift until several of them land on near-integer ratios. Raman and S. Kumar reported harmonic overtones in Indian drums in 1920, and Raman's 1934 paper described the loaded, damped drumhead in detail. Later work solved the loaded equation numerically and recovered Raman's result.
In 1919, in Calcutta, C. V. Raman experimented on a fine large mridanga, using sand on its head to make the vibrations visible. When a drumhead is struck while a finger damps it at a chosen point, the sand gathers along the lines that stay still, drawing each vibration mode on the skin. Mr Sivakali Kumar supplied the drums. What the sand showed was unlike any Western drum: the first nine normal modes of the loaded head grouped themselves into just five tones, harmonically related, with the third, fourth and fifth harmonics each made by two or three modes ringing together. Raman and Kumar reported it in a letter to Nature in 1920, titled Musical Drums with Harmonic Overtones, and Raman returned to it in a full paper in 1934. The tabla's black spot had been doing advanced mode-tuning for centuries before anyone wrote down the equation.
PPST Bulletin: Raman's work on musical instruments (reproducing his 1922 and 1934 texts)The authors, working at St. Stephen's College, Delhi, solved the wave equation for a membrane whose density falls off from the centre like the syahi, using a Runge–Kutta method mode by mode. With a continuous loading they found overtone ratios of 2.00, 2.98, 2.99, 4.00, 4.00, 5.01, 5.01 and 5.02 against the measured tabla pattern of 2, 3, 3, 4, 4, 5, 5, 5, with the lowest mode at 1.07 instead of 1. They note the residual errors of 0.01 to 0.02 are below what the ear distinguishes for a 240 Hz Sa. The limit: it is a student project with an idealised loading and constant tension, not a measurement of a real drum.
Read the paper (free)What is engine knock, physically?
Knock is the unburned end of the fuel-air charge igniting by itself before the flame front reaches it. The spark lights a flame that spreads across the chamber in a few milliseconds, compressing and heating the gas ahead of it.
If that end gas gets hot enough for long enough, it auto-ignites almost at once, and the sudden pressure rise launches pressure waves that bounce around the cylinder. Those waves are what you hear as ping, and what wears and cracks pistons.
A 2026 simulation study of a hydrogen engine graded knock intensity as the amplitude of these oscillations:
- about 1 MPa for mild deflagration knock;
- about 3 MPa for detonation knock;
- over 7 MPa for "superknock".
Using CONVERGE simulations of a four-cylinder direct-injection hydrogen engine (73.5 mm bore, 90.44 mm stroke, 11.6:1 compression), the authors separated deflagration knock, detonation knock and superknock. They detected knock by band-pass filtering the cylinder pressure at 5–20 kHz and used a 0.2 MPa threshold; knock intensities were about 1.08 MPa (deflagration), 3.11 MPa (detonation), 7.05 MPa (superknock) and 9.34 MPa for pre-ignition knock. The limit: it is a numerical model of a hydrogen engine, not a petrol motorcycle, though the filtering band is the standard one.
Read the paper (free)Why does knock ring at a particular frequency?
Because the cylinder is a short, round, rigid-walled drum of hot gas, and the gas has its own circular modes. This time the wall does not move, so the boundary condition is that the gas cannot flow into it: the radial pressure gradient is zero at the bore.
That swaps the zeros of Jm for the zeros of its derivative J′m. Working it step by step:
- Pressure field. Near top dead centre the chamber is a flat disc, so take \( p(r,\theta,t) = J_m(kr)\cos(m\theta)\,e^{i\omega t} \).
- Rigid wall. \( \partial p/\partial r = 0 \) at \( r = B/2 \), so \( J'_m(kB/2) = 0 \), giving \( kB/2 = \alpha_{mn} \), the zeros 1.841, 3.054, 3.832, 4.201 and 5.318.
- Frequency. With \( \omega = ck \), \( f = \omega/2\pi \):
This is Draper's equation, published by C. S. Draper in 1938, and it is still the starting point for knock sensing. The speed of sound c depends on the temperature of the burned gas, from \( c = \sqrt{\gamma R T / M} \).
With γ = 1.28, M = 0.0285 kg/mol and T = 2,500 K (assumed values for hot combustion products):
Calculated in this post from the ideal-gas speed of sound (Wikipedia: Speed of sound), with these example inputs (my assumed values for hot combustion products): γ = 1.28, M = 0.0285 kg/mol and a burned-gas temperature of 2,500 K.
- KTM 390 Adventure, 89 mm bore, first mode. \( f_{10} = 1.841 \times 966 / (\pi \times 0.089) = 6{,}360 \) Hz.
- Same engine, next modes. \( f_{20} = 3.054/1.841 \times 6{,}360 = 10{,}550 \) Hz; \( f_{01} = 3.832/1.841 \times 6{,}360 = 13{,}240 \) Hz.
- Pulsar NS200, 72 mm bore. \( f_{10} = 1{,}778 / (\pi \times 0.072) = 7{,}860 \) Hz. Smaller bore, higher pitch, just as a smaller tabla rings higher.
- Mode ratios. 1, 1.66, 2.08, 2.28, 2.89: as inharmonic as a plain drum. Knock is an unpitched ping.
See the data as a table
| Plain membrane (zeros of Jm) (× lowest mode) | Tabla, measured pattern (× lowest mode) | Cylinder gas (zeros of J′m) (× lowest mode) | |
|---|---|---|---|
| Mode 1 | 1 | 1 | 1 |
| Mode 2 | 1.593 | 2 | 1.659 |
| Mode 3 | 2.136 | 3 | 2.081 |
| Mode 4 | 2.295 | 3 | 2.282 |
| Mode 5 | 2.653 | 4 | 2.888 |
| Mode 6 | 2.917 | 4 | 2.896 |
Source: Bessel-function zeros (computed); tabla pattern from Malu and Siddharthan (2000) · The tabla's loading pulls its modes onto whole numbers. The engine's are set by a rigid round wall and stay inharmonic.
Use this to check what you just read: put in your own engine's bore and see where its knock should ring.
Knock frequency for your bore
How sensitive are these numbers?
Mostly to temperature, because frequency goes with the square root of it. Drop the burned gas from 2,500 K to 2,000 K and c falls from 966 to 864 m/s, taking the KTM's first mode from 6.36 to 5.69 kHz; at 2,800 K it rises to 6.73 kHz.
Calculated in this post from the ideal-gas speed of sound (Wikipedia: Speed of sound) and Draper's relation (Broatch et al., 2016), with these example inputs (my assumed values for hot combustion products): γ = 1.28, M = 0.0285 kg/mol and temperatures of 2,000, 2,500 and 2,800 K.
As the piston moves down after a knock event, the gas expands and cools, so the ringing pitch slides downward over a millisecond or two, like a tabla stroke bending with palm pressure on the bayan.
Geometry and bore
Engine geometry matters less but still counts: a domed head or deep piston bowl is not a flat disc, and engineers often calibrate the Bessel constants against measured pressure traces. The bore is the big lever. Size scales pitch inversely in both instruments.
| Change | KTM 390 first mode | Shift |
|---|---|---|
| Baseline, 2,500 K | 6.36 kHz | none |
| Cooler gas, 2,000 K | 5.69 kHz | −11% |
| Hotter gas, 2,800 K | 6.73 kHz | +6% |
| Smaller bore, 72 mm (NS200) | 7.86 kHz | +24% |

How does a knock sensor hear it?
A knock sensor is a small piezoelectric accelerometer bolted to the cylinder block, and it listens for those 5 to 15 kHz vibrations transmitted through the metal. Bosch Motorsport's KS4-R2, for example, is specified for 3 to 30 kHz with its own resonance above 30 kHz, so its response is reasonably flat across the knock band.
- Open a window. The engine computer listens only just after top dead centre.
- Filter. It band-pass filters the signal around the expected modes.
- Compare. It compares that energy with the background noise.
- Act. If it detects knock, it retards the ignition timing on the next cycles, lowering peak temperature and pressure, then creeps the timing back.
The design problem is the tabla player's problem in reverse.
The tabla maker adds mass to move the modes where the ear wants them; the engine calibrator finds where the modes already are and tunes the filter to them.
Both are reading the same Bessel table.
An engine without an electronic knock-control loop cannot retard its timing in response, so it has to rely on conservative ignition timing and the right fuel grade from the start. That is part of why the fuel debate on older bikes, covered in whether premium petrol is worth it and what E20 does to motorcycle engines, is really a knock debate.
The basics of compression and combustion are in how engine power and torque work.

| Property | Tabla (dayan) | Knocking cylinder |
|---|---|---|
| What vibrates | Goatskin membrane under tension | Hot burned gas in a rigid bore |
| Boundary condition | Edge clamped: Jm(ka) = 0 | Wall rigid: J′m(kB/2) = 0 |
| Wave speed | √(τ/σ), set by tuning tension | √(γRT/M), about 966 m/s at 2,500 K |
| Typical lowest frequency | Tuned to the singer's Sa, around 240 Hz | About 6–8 kHz for motorcycle bores |
| Overtones | Loaded to near 2, 3, 4, 5 times | 1.66, 2.08, 2.28 times: inharmonic |
| What excites it | A finger strike | Auto-ignition of the end gas |
| Who listens | The audience | A piezo knock sensor, 5–20 kHz band |
So what does a tabla actually teach you about knock?
Three things, each with a number:
- Size sets pitch. Round things ring at frequencies set by Bessel zeros and their size: halve the diameter and the pitch doubles, which is why the 72 mm NS200 bore knocks higher (7.9 kHz) than the 89 mm KTM bore (6.4 kHz).
- The boundary matters. A clamped skin uses the zeros of Jm, a rigid wall uses the zeros of J′m, so the two are cousins, not twins.
- Harmony is arranged. A ring is only musical if someone arranges the overtones. Nobody arranges the engine's, which is why knock pings instead of sings.
Where the analogy fails is the medium: the tabla is a solid membrane whose stiffness and mass you can engineer, while knock rings in a gas whose sound speed changes by the millisecond as it cools.
A tabla membrane and the gas in a knocking cylinder both ring in circular modes set by Bessel functions, and both scale inversely with diameter. The tabla uses the zeros of Jm (clamped edge) and the syahi loads them onto near-harmonic ratios of 2, 3, 4 and 5; the cylinder uses the zeros of J′m (rigid wall), giving Draper's f = 1.841c/πB, about 6.4 kHz for an 89 mm bore at 2,500 K. Same mathematics, different boundary and medium, and only one of them is tuned.
For more of the machines behind these posts, see Yantra.
What's new
- Simulation study sorts hydrogen-engine knock into deflagration, detonation and superknock ACS Omega, Sep 2026
- Semi-supervised machine learning transcribes tabla strokes with confidence-guided weighting arXiv, Sep 2026
- Rhythm-aware language model for weakly supervised tabla transcription arXiv, Aug 2026
What if next?
- Why your chain and a pearl necklace wear out the same way. From rings of sound to rings of metal and silk.
- A cricket ball's swing and your helmet's buffeting are one phenomenon. Another case of one equation, two worlds.
- Your suspension is an idli. Damping, the reason a tabla stroke and a pothole both die away.
Questions people ask
What frequency is engine knock?
Typically about 5 to 20 kHz, with the strongest first mode set by Draper's equation, f = 1.841c/πB. For motorcycle bores of 70 to 90 mm and burned gas near 2,500 K, that is roughly 6 to 8 kHz.
Why does the tabla have a black spot?
The syahi adds mass to the centre of the drumhead, which shifts its vibration modes from the inharmonic ratios of a uniform skin to near whole-number multiples. That gives the dayan a clear, tunable pitch, as C. V. Raman showed.
Can you hear engine knock?
Often, as a metallic pinging or rattling under load, especially when accelerating hard at low rpm in a high gear. Light knock can be inaudible over engine noise, which is why modern engines use sensors.
Do motorcycles have knock sensors?
It depends on the model, so check the service manual. A bike without an electronic knock-control loop cannot pull its timing back when knock starts, so it relies on conservative ignition timing and the fuel grade the maker specifies.
Sources
- Malu, S. S. and Siddharthan, A.: Acoustics of the Indian Drum, arXiv (2000) (source for the loaded-membrane model, overtone ratios and the 240 Hz Sa)
- Wu, J., Zhao, Z. and Zheng, Z.: Classification and Mechanism Analysis of Knock Modes in Hydrogen Engines, ACS Omega 11(39) (2026) (source for knock intensities and the 5–20 kHz band)
- Raman, C. V.: The Indian musical drums, Proc. Indian Acad. Sci. 1, 179–188 (1934) (source for Raman's loaded, damped drumhead)
- Raman, C. V. and Kumar, S.: Musical Drums with Harmonic Overtones, Nature 104, 500 (1920) (the original letter)
- PPST Bulletin: Raman's work on musical instruments (source for the 1919 Calcutta mridanga experiments, sand figures and the nine-modes-into-five-tones result)
- Bosch Motorsport: Knock Sensor KS4-R2 (source for the 3–30 kHz range)
- Wikipedia: Vibrations of a circular membrane (source for the clamped-membrane solution)
- moto-data.net: KTM 390 Adventure X (2025) specifications (source for the 89 mm bore)
- Bajaj Auto: Pulsar NS200 owner's manual (source for the 72 mm bore)
- Confidence-Guided Markov Weighting for Semi-Supervised Tabla Stroke Transcription, arXiv (2026)
