Quality Factor

Strike a wine glass with a spoon and it sings — a clear tone that hangs in the air for half a second, a thousand oscillations after your spoon has left. Strike a lump of clay and you get a dull thud: the energy you delivered is gone before the clay completes even one vibration. Both objects are damped oscillators obeying the same equation; what separates the singer from the thud is a single dimensionless number, the quality factor $Q$ . It is the oscillator's ratio of energy stored to energy leaked per cycle, and it is arguably the most important figure of merit in all of oscillation physics. What makes $Q$ remarkable is its reach. The damper on a slamming door has $Q \approx 0.5$ — it cannot oscillate at all. A tuning fork manages a few thousand. The quartz crystal in a wristwatch holds $10^4$ to $10^5$ . A cesium fountain clock resolves a resonance with $Q \sim 10^{10}$ , a superconducting accelerator cavity stores microwave photons at $Q \sim 10^{11}$ , and the strontium transition in a modern optical lattice clock has a line quality above $10^{16}$ . Seventeen orders of magnitude, one number, one equation of motion. This article builds $Q$ from its energy definition, derives its equivalence to the frequency ratio $\omega_0/2\beta$ , and turns it into a practical ruler: how to read $Q$ off a ringdown, what it says about the damping regimes, and why the same number that counts surviving cycles also sets the bandwidth and the amplification of a driven resonance — the triple identity that makes $Q$ the universal currency of resonant physics.

Key takeaways

$Q$ is the dimensionless figure of merit of any oscillator: $Q = 2\pi \times$ (energy stored)/(energy lost per cycle) — equivalently, energy stored divided by energy lost per radian of oscillation.
For the standard damped oscillator $\ddot x + 2\beta\dot x + \omega_0^2 x = 0$ , the energy decays as $E = E_0 e^{-2\beta t}$ , and the definition evaluates to $Q = \omega_0/2\beta = \omega_0\tau$ , with $\tau = 1/2\beta$ the energy decay time.
Each cycle a fraction $2\pi/Q$ of the stored energy leaks away; the amplitude envelope e-folds after $Q/\pi$ cycles and falls to $e^{-\pi} \approx 4.3\%$ after exactly $Q$ cycles.
Ringdown measurement: the logarithmic decrement $\delta = \ln(A_n/A_{n+1})$ between successive peaks gives $Q = \pi/\delta$ — no clock calibration, no force sensor, just two amplitude readings.
The damping regimes are a statement about $Q$ : overdamped for $Q < \tfrac12$ , critically damped at $Q = \tfrac12$ , oscillatory (underdamped) for $Q > \tfrac12$ .
The same $Q$ controls the driven response: fractional bandwidth $\Delta\omega/\omega_0 = 1/Q$ and on-resonance amplification $A(\omega_0) = Q \times$ static deflection — one number, three measurements.
Key idea: $Q = \omega_0/2\beta$ counts how many radians an oscillator swings before its energy decays by $1/e$ — and the very same number sets how sharply and how strongly it responds when driven.

Where this lives today

The frontier of $Q$ engineering today is a fight against thermal noise, and the weapon is the fluctuation–dissipation theorem: any mode that can lose energy to a heat bath must also receive random kicks from it, in exact proportion. Low $Q$ therefore means loud thermal jitter smeared over a wide band; high $Q$ concentrates the same jitter into a needle at the resonant frequency, where it can be avoided or filtered. This is the entire logic of LIGO's suspensions — fused-silica fibres with $Q \sim 10^9$ — and of the current hunt for better mirror coatings, where the amorphous oxide layers' comparatively low mechanical $Q$ (loss angles near $10^{-4}$ , i.e. $Q \sim 10^4$ ) now sets the detectors' mid-band noise floor. Crystalline AlGaAs coatings with tenfold higher $Q$ are one of the headline upgrades planned for the 2026-era detector improvements and the next-generation Cosmic Explorer designs. At the same time, superconducting radio-frequency cavities with $Q \gtrsim 10^{11}$ — built to accelerate particle beams — have been repurposed as quantum hardware: a microwave photon that coasts for ten seconds is an excellent quantum memory, and the same cavities serve as haloscopes listening for dark-matter axions, where the detectable signal power scales directly with $Q$ . And the clock ladder keeps climbing: strontium and ytterbium optical lattice clocks now interrogate lines with $Q$ above $10^{16}$ , so stable that two clocks a centimetre apart in height measurably disagree — gravitational time dilation read off a tabletop, with the quality factor of equation (2) doing the heavy lifting.

The physical picture

Every free oscillator is having an argument with friction. The spring (or the glass's elasticity, or the pendulum's gravity) wants to keep the energy sloshing between kinetic and potential form forever; damping siphons off a little of that energy every swing. $Q$ is the score of the argument: the number of radians of oscillation completed before friction has claimed a $1/e$ fraction of the energy. High $Q$ means the sloshing wins for a long time — the oscillator coasts. Low $Q$ means friction wins within a cycle or two.

Figure

The ringdown — a high-Q oscillator coasts while a low-Q oscillator dies within two swings

Read the two panels as one experiment repeated with two different dampers. Both oscillators start with the same energy and the same rhythm; only the leak rate differs. The $Q = 30$ oscillator loses $2\pi/Q \approx 21\%$ of its energy per cycle and takes $Q/\pi \approx 9.5$ cycles for its envelope to fall by $1/e$ ; the $Q = 6$ oscillator bleeds more than half its energy every swing and is effectively silent after four. The rule of thumb hiding in the mathematics is worth memorizing: an oscillator rings for about $Q/\pi$ cycles — after exactly $Q$ full cycles the amplitude is down to $e^{-\pi} \approx 4.3\%$ , gone for any practical purpose. There is a second, subtler picture to hold: the amplitude and the energy do not decay together. Energy goes as amplitude squared, so it dies exactly twice as fast — the source of a famous factor-of-2 trap:

Figure

Amplitude decays at rate beta, energy at rate two-beta — the energy is always one step ahead

The red curve is the energy, the blue curve the amplitude envelope. Because $E \propto A^2$ , squaring $e^{-\beta t}$ gives $e^{-2\beta t}$ : the energy reaches $1/e$ of its starting value at the energy decay time $\tau = 1/2\beta$ , while the amplitude needs $2\tau$ to do the same. Both decay times are honest ways to describe the same oscillator, and $Q$ is built from the energy one: $Q = \omega_0\tau$ . Keep this diagram in mind whenever a ringdown measurement asks you to convert between an amplitude trace (what an oscilloscope shows) and an energy budget (what $Q$ is defined by).

One number, three questions

Everything starts from the free, damped oscillator — the equation whose driven cousin is analyzed in the Resonance article. With no drive, the mass is left alone with its spring and its friction:

\ddot x + 2\beta\dot x + \omega_0^2 x = 0 \qquad (1)

where: $x$ — displacement from equilibrium; $\omega_0 = \sqrt{k/m}$ — natural angular frequency; $\beta$ — damping constant (dimensions of inverse time), defined so the amplitude envelope of free oscillations decays as $e^{-\beta t}$ . Both $\omega_0$ and $\beta$ are rates, so their ratio is the only dimensionless number the oscillator possesses — every dimensionless property of its motion must be a function of it. The physically primary definition of $Q$ is a bookkeeping statement about energy, made before any equation is solved. Take the energy currently stored in the oscillation, divide by the energy the damping destroys in one full cycle, and scale by $2\pi$ :

Q \;=\; 2\pi\,\frac{E_{\text{stored}}}{E_{\text{lost per cycle}}} \;=\; \frac{E_{\text{stored}}}{E_{\text{lost per radian}}} \qquad (2)

where: $E_{\text{stored}}$ — instantaneous mechanical energy; $E_{\text{lost per cycle}}$ — energy dissipated during one period $T$ . The second form explains the otherwise mysterious $2\pi$ : a cycle contains $2\pi$ radians of phase, so $Q$ is simply the energy stored measured in units of the energy lost per radian of oscillation. Said inversely: each cycle costs the oscillator the fraction $2\pi/Q$ of whatever energy it currently holds — a $Q = 1000$ tuning fork loses just $0.63\%$ per cycle, which is why it can afford to ring three hundred cycles before its envelope e-folds. For the linear oscillator of equation (1), this energy definition collapses to a ratio of the two rates, and that is the identity everything else in this article flows from.

Derivation AFrom the Energy Ledger to the Frequency Ratio

We are deriving the central identity $Q = \omega_0/2\beta$ — that the energy-bookkeeping definition of equation (2), which mentions no equation of motion at all, evaluates for the linear damped oscillator to the simple ratio of natural frequency to twice the damping constant. The strategy is direct: take the underdamped solution of equation (1), compute how the stored energy decays in time, extract the loss in a single period, and feed both numbers into the definition. The one step that needs care is the weak-damping assumption $Q \gg 1$ : only then does the energy change slowly on the scale of a cycle, so that "energy lost per cycle" is a well-defined small quantity and the exponential can be linearized. Watch also the factor of 2 between amplitude decay and energy decay — it is exactly what converts the naive guess $\omega_0/\beta$ into the correct $\omega_0/2\beta$ .

Step 1. Write down the underdamped solution of (1): $x(t) = A_0\,e^{-\beta t}\cos(\omega_1 t + \phi)$ with $\omega_1 = \sqrt{\omega_0^2 - \beta^2}$ . The envelope decays at the rate $\beta$ . Why we can treat the envelope as constant within a cycle: for weak damping $\beta \ll \omega_0$ , the envelope changes by only $\beta T \approx 2\pi\beta/\omega_0 = \pi/Q \ll 1$ per period.

Step 2. Compute the cycle-averaged stored energy. Averaging $E = \tfrac12 m\dot x^2 + \tfrac12 m\omega_0^2 x^2$ over a cycle with the envelope frozen gives $\bar E(t) = \tfrac12 m\omega_0^2 A_0^2\,e^{-2\beta t}$ . Why the exponent doubled: energy is quadratic in the displacement and velocity, and squaring $e^{-\beta t}$ gives $e^{-2\beta t}$ — the energy decay rate is $2\beta$ , not $\beta$ .

Step 3. Extract the loss per cycle. Over one period $T = 2\pi/\omega_0$ (using $\omega_1 \approx \omega_0$ for weak damping), the energy falls by

E_{\text{lost per cycle}} = \bar E\,\big(1 - e^{-2\beta T}\big) \approx \bar E \cdot 2\beta\,T = \bar E\,\frac{4\pi\beta}{\omega_0}

Why linearizing is legitimate: the exponent $2\beta T = 2\pi/Q$ is small precisely when $Q$ is large, and the correction terms are of relative order $1/Q$ — invisible until $Q$ approaches unity, where the whole cycle-counting picture is dissolving anyway.

Step 4. Feed both into the definition (2):

Q = 2\pi\,\frac{\bar E}{\bar E \cdot 4\pi\beta/\omega_0} = \frac{\omega_0}{2\beta} \qquad (3)

where: $Q$ — the quality factor, now expressed purely in terms of the oscillator's two rates. The stored energy cancels out completely: $Q$ does not care how hard the oscillator was struck, only how leaky it is. That is what makes it a property of the system rather than of any particular motion.

Step 5. Repackage using the energy decay time. Since $\bar E = E_0 e^{-2\beta t}$ e-folds in the time $\tau = 1/2\beta$ , equation (3) is equivalently

Q = \omega_0\,\tau \qquad (4)

where: $\tau$ — the time for the stored energy to fall by $1/e$ . This is the most compact reading of all: $Q$ is the number of radians of oscillation completed during one energy decay time. A cavity with $Q = 10^{11}$ at $\omega_0 = 2\pi \times 1.3\ \mathrm{GHz}$ holds its microwave energy for $\tau = Q/\omega_0 \approx 12$ seconds — twelve seconds of coasting for a photon that would cross the cavity in a nanosecond. Verified: dimensions check ( $\omega_0\tau$ is rad/s times s, dimensionless); the result matches Morin Ch. 4 and Taylor Ch. 5 (who write $Q = \omega_0/\gamma$ with $\gamma = 2\beta$ ); and in the limit $\beta \to 0$ , $Q \to \infty$ recovers the undamped oscillator that never loses energy.

Equations (2), (3), and (4) are one number wearing three outfits — an energy ratio, a frequency ratio, and a radian count. Which outfit you reach for depends on what your instrument measures: a calorimetric or power measurement hands you (2), a fitted decay constant hands you (3), and a "how long does it ring" stopwatch measurement hands you (4).

The ringdown ruler

The fastest way to measure $Q$ in a laboratory needs no force sensor and no calibrated clock — just the decaying oscillation itself. Strike the oscillator, record the successive peak amplitudes, and count. The mathematics of the envelope converts the count directly into $Q$ . Since the amplitude decays as $e^{-\beta t}$ and each cycle lasts $T = 2\pi/\omega_0$ , after $N$ full cycles the amplitude has fallen to

\frac{A_N}{A_0} = e^{-\beta N T} = e^{-\pi N/Q} \qquad (5)

where: $A_N$ — peak amplitude after $N$ cycles; the exponent uses $\beta T = 2\pi\beta/\omega_0 = \pi/Q$ from equation (3). Two landmarks follow instantly. Setting the exponent to $1$ : the amplitude e-folds after $N = Q/\pi$ cycles. Setting $N = Q$ : after $Q$ cycles the amplitude has fallen by $e^{-\pi} \approx 0.043$ — to $4.3\%$ , essentially gone. So " $Q$ counts the cycles an oscillator survives" is literally true if you take "survives" to mean "retains more than a twentieth of its swing."

Go deeper · Derivation B · 2 minReading Q Off a Decaying Trace

Derivation BReading Q Off a Decaying Trace

The goal is the practical inverse problem: given an experimental ringdown trace — a sequence of measured peak amplitudes $A_0, A_1, A_2, \dots$ — extract $Q$ with nothing but a logarithm. The tool is the logarithmic decrement $\delta$ , the constant difference between the logs of successive peaks, and we will show $\delta = \pi/Q$ , so that $Q = \pi/\delta$ . The strategy is to evaluate the envelope at times exactly one period apart and use the weak-damping period. The point to watch is which period appears: the oscillator rings at the damped frequency $\omega_1$ , not $\omega_0$ , but for any $Q$ large enough to make ringdown counting sensible the difference is of order $1/Q^2$ and is swallowed by measurement noise long before it matters.

Step 1. Take the ratio of successive peaks. Peaks occur one period apart, so the envelope gives $A_{n+1}/A_n = e^{-\beta T}$ , the same ratio for every pair. Define the logarithmic decrement

\delta \equiv \ln\frac{A_n}{A_{n+1}} = \beta\,T \qquad (6)

where: $\delta$ — the log-decrement, dimensionless. Why the constancy is itself a diagnostic: if $\ln(A_n/A_{n+1})$ drifts as the amplitude falls, the damping is not linear — air drag at large amplitude, or dry friction at small — and no single $Q$ describes the system.

Step 2. Convert to $Q$ . With $T = 2\pi/\omega_1 \approx 2\pi/\omega_0$ and $\beta = \omega_0/2Q$ from equation (3), $\delta = \beta T = \pi/Q$ , hence

Q = \frac{\pi}{\delta}

Why this beats fitting an exponential: the decrement is a ratio of two amplitudes read off the same trace, so calibration errors in the sensor gain cancel exactly; and the time base never enters, so the clock need not be calibrated either.

Step 3. Use widely separated peaks for precision. Comparing peak $0$ with peak $N$ instead of adjacent peaks gives $\delta = \tfrac{1}{N}\ln(A_0/A_N)$ , dividing the reading error by $N$ . Why one should not push $N$ too far: once $A_N$ sinks toward the noise floor, the logarithm amplifies the noise; the optimum is to stop while the trace is still clean, typically around one to two e-foldings. *Verified: for the worked example below, $A_0 = 10$ cm and $A_{20} = 6$ cm give $\delta = \ln(10/6)/20 = 0.0255$ and $Q = \pi/0.0255 = 123$ ; the internal cross-check $2\delta = 0.051$ correctly equals the fractional energy loss per cycle $2\pi/Q = 0.051$ .*

The grand tour — seventeen orders of magnitude

$Q$ earns its status as the figure of merit because it spans real systems the way almost no other physical number does. Here is the ladder, every ring-down time recomputed from $Q$ and $f_0$ via $\tau_{\text{amp}} = 2Q/\omega_0$ (the amplitude $1/e$ time):

Figure

The Q ladder — real oscillators spanning seventeen decades of quality factor

System	$f_0$	$Q$	Amplitude ring-down $2Q/\omega_0$
hydraulic door closer	$\sim 1$ Hz	$\approx 0.5$	none — overdamped, no oscillation
car suspension (good shocks)	$1.3$ Hz	$\approx 1$	$0.24$ s — one bounce and done
Earth's gravest normal mode $_0S_2$	$0.309$ mHz	$\approx 510$	$6.1$ days of planetary humming after a great earthquake
tuning fork (A440)	$440$ Hz	$\approx 5\times10^3$	$\approx 3.6$ s of audible ring
quartz watch crystal	$32{,}768$ Hz	$10^4$ – $10^5$	$\approx 0.5$ s at $Q = 5\times10^4$
LIGO fused-silica suspension fibre	$\sim 500$ Hz (violin mode)	$\sim 10^9$	$\approx 7$ days — it essentially never stops
cesium fountain clock transition	$9.19$ GHz	$\sim 10^{10}$	measured as a $\sim 1$ Hz linewidth
superconducting niobium cavity	$1.3$ GHz	$10^{10}$ – $3\times10^{11}$	stored microwaves coast for seconds
strontium optical lattice clock	$429$ THz	$\sim 1.7\times10^{16}$	a $26$ mHz line — parts in $10^{16}$

Check any row with equation (4). The niobium cavity: $\tau = Q/\omega_0 = 10^{11}/(2\pi \times 1.3\times10^9\ \mathrm{Hz}) \approx 12\ \mathrm{s}$ for the energy — light that would cross the cavity in a nanosecond instead sloshes back and forth ten billion times. The Earth row: $Q \approx 510$ means $Q/\pi \approx 160$ cycles to e-fold, and at a $54$ -minute period that is $160 \times 54\ \mathrm{min} \approx 6$ days — which is precisely why seismometers still recorded the planet ringing nearly a week after the 2004 Sumatra earthquake.

Q marks the damping regimes

Because $Q$ is the only dimensionless knob in equation (1), the qualitative character of the motion — does it oscillate at all? — must be decided by $Q$ alone. The dividing line falls at $Q = \tfrac12$ .

Go deeper · Derivation C · 2 minWhere the Oscillation Dies

Derivation CWhere the Oscillation Dies

We are recasting the classification of damping regimes — underdamped, critically damped, overdamped — as a pure statement about $Q$ , and deriving the damped frequency formula $\omega_1 = \omega_0\sqrt{1 - 1/4Q^2}$ . The route is the standard characteristic-root analysis of equation (1), but with every appearance of $\beta$ traded for $\omega_0/2Q$ so that the regime boundary emerges as a pure number. The payoff to watch for: the discriminant changes sign exactly at $Q = \tfrac12$ , and the frequency correction enters only at order $1/Q^2$ — which is why a $Q = 10$ oscillator already rings within $0.13\%$ of its natural frequency.

Step 1. Insert the exponential ansatz $x = e^{rt}$ into (1) to get the characteristic equation $r^2 + 2\beta r + \omega_0^2 = 0$ , with roots

r = -\beta \pm \sqrt{\beta^2 - \omega_0^2}

Why the sign of the discriminant is everything: real roots mean pure exponential decay (no oscillation); complex roots mean decay times a sinusoid.

Step 2. Substitute $\beta = \omega_0/2Q$ and factor out $\omega_0$ :

\frac{r}{\omega_0} = -\frac{1}{2Q} \pm \sqrt{\frac{1}{4Q^2} - 1}

where: the bracket is now built entirely from the dimensionless $Q$ . The discriminant $1/4Q^2 - 1$ is positive for $Q < \tfrac12$ (two real decaying roots: overdamped), zero at $Q = \tfrac12$ (degenerate root: critically damped), and negative for $Q > \tfrac12$ (complex pair: underdamped, genuine oscillation).

Step 3. In the underdamped regime, write the imaginary part of the root as the ringing frequency:

\omega_1 = \omega_0\sqrt{1 - \frac{1}{4Q^2}} \qquad (7)

where: $\omega_1$ — the observed frequency of free damped oscillation. Why the shift is tiny in practice: at $Q = 10$ the correction factor is $\sqrt{1 - 0.0025} = 0.99875$ , a $0.13\%$ slowing; at $Q = 5\times10^4$ (a watch crystal) it is one part in $10^{10}$ — damping steals amplitude long before it perceptibly changes the pitch.

Step 4. Check a concrete overdamped case, $Q = 0.4$ . The dimensionless roots are $-1/(0.8) \pm \sqrt{1/0.64 - 1} = -1.25 \pm 0.75$ , i.e. $r = -0.5\,\omega_0$ and $r = -2.0\,\omega_0$ : two pure decays, one slow and one fast, no oscillation anywhere. Why the slow root dominates the approach to equilibrium: after a short time only $e^{-0.5\omega_0 t}$ survives — an overdamped system is sluggish, which is exactly why critical damping ( $Q = \tfrac12$ ), not heavy damping, gives the fastest return without overshoot. *Verified: at $Q = \tfrac12$ the two roots merge at $r = -\omega_0$ , reproducing the critical-damping double root; as $Q \to \infty$ , $r \to \pm i\omega_0$ , recovering undamped SHM; matches the damping-regime classification in Taylor Ch. 5 and Morin Ch. 4.*

The classification deserves a moment of appreciation. "Overdamped versus underdamped" sounds like it should depend on the mass, the spring, and the damper separately — three parameters. It does not. One dimensionless number decides, and the door-closer engineer and the seismometer designer are just choosing opposite sides of $Q = \tfrac12$ .

One number, three measurements

Here is the deepest fact about $Q$ , and the reason it appears on every resonator datasheet ever printed. Three completely different experiments — a stopwatch experiment, a frequency-sweep experiment, and an amplitude experiment — yield the same number:

Q \;=\; \underbrace{\frac{\omega_0}{2\beta}}_{\text{ring-down}} \;=\; \underbrace{\frac{\omega_0}{\Delta\omega}}_{\text{bandwidth}} \;=\; \underbrace{\frac{A(\omega_0)}{A(0)}}_{\text{amplification}} \qquad (8)

where: $\Delta\omega$ — the full width at half maximum of the driven power-absorption curve; $A(\omega_0)/A(0)$ — the on-resonance amplitude divided by the static (zero-frequency) deflection. The first leg is the one this article has derived: the free-decay ratio of equation (3). The other two belong to the driven oscillator, and their stories are told by this article's siblings — but the fact that all three coincide is the point to internalize here. The decay leg and the bandwidth leg are Fourier twins: an oscillation that persists for $\tau = Q/\omega_0$ seconds cannot have a frequency defined more sharply than $\Delta\omega \sim 1/\tau = \omega_0/Q$ . Long ringing is narrow bandwidth — the same information read in the time domain or the frequency domain. The full half-power algebra that turns this scaling into the exact statement $\Delta\omega = 2\beta$ is worked out in Resonance Curve and Bandwidth. The amplification leg — that driving on resonance multiplies the static response by exactly $Q$ , and that the energy definition (2) applied at the peak of the driven response reproduces the same number — is derived via energy balance in the Resonance article (its equations 5 and 8). One number, purchasable in three currencies:

you measured	you compute	leg of (8)
ring-down envelope, log-decrement $\delta$	$Q = \pi/\delta$	decay
frequency sweep, peak at $\omega_0$ , FWHM $\Delta\omega$	$Q = \omega_0/\Delta\omega$	bandwidth
response amplitude at resonance vs. at DC	$Q = A(\omega_0)/A(0)$	amplification

Any two of the three measured on the same device constitute a consistency test of the linear-damping model itself. Real resonator engineering leans on this constantly: if the ringdown $Q$ and the bandwidth $Q$ of your MEMS device disagree, something in the physics — nonlinearity, multiple loss channels, a second mode hiding under the peak — is waiting to be found.

Limiting cases

Case	Result	Meaning
weak damping, $Q \gg 1$	$\omega_1 \approx \omega_0\big(1 - \tfrac{1}{8Q^2}\big)$ , loses $2\pi/Q$ of $E$ per cycle	the cycle-counting picture is exact in spirit: rings $Q/\pi$ cycles per e-fold, pitch shift negligible
moderate, $Q = 1$	rings $\approx 0.3$ cycles to $1/e$ ; $\omega_1 = 0.87\,\omega_0$	barely an oscillator — one overshoot, then done; the three definitions of $Q$ begin to visibly disagree
critical damping, $Q = \tfrac12$	double root $r = -\omega_0$ ; fastest non-overshooting return	the engineer's optimum for pointers, door closers, car suspensions
overdamped, $Q < \tfrac12$	two real decay rates $r_\pm = \omega_0\big(-\tfrac{1}{2Q} \pm \sqrt{\tfrac{1}{4Q^2}-1}\big)$	no cycles to count — $Q$ survives only as a formal parameter
undamped ideal, $Q \to \infty$	$E$ conserved, $\Delta\omega \to 0$	simple harmonic motion recovered; infinitely sharp resonance, eternal ring

Common trapThe amplitude and the energy decay at the same rate

They never do — energy decays exactly twice as fast, because $E \propto A^2$ . The amplitude envelope goes as $e^{-\beta t}$ (e-folds at $t = 1/\beta$ ), the energy as $e^{-2\beta t}$ (e-folds at $\tau = 1/2\beta$ ). Mix them up and every $Q$ you compute is wrong by a factor of 2. The confusion is compounded by a genuine convention split in the literature: many texts write the equation of motion as $\ddot x + \gamma\dot x + \omega_0^2 x = 0$ , so their $\gamma$ equals our $2\beta$ and their formula reads $Q = \omega_0/\gamma$ — the same physics, different letter. Landau uses $\lambda$ for our $\beta$ . Before using any $Q$ formula from any book, check which decay the damping symbol describes: if the amplitude decays as $e^{-(\text{symbol})t/2}$ , the symbol is $2\beta$ ; if as $e^{-(\text{symbol})t}$ , it is $\beta$ . The one convention-proof anchor is the energy definition (2) — $2\pi$ times energy stored over energy lost per cycle — which contains no symbols to misread at all.

Worked examples

1 · Q of a lab pendulum from a ringdown (confidence builder)

A pendulum of period $T = 2.0$ s is released with amplitude $10.0$ cm. Twenty full cycles later its amplitude is $6.0$ cm. Find $Q$ , the damping constant $\beta$ , the fraction of energy lost per cycle, and how long the pendulum takes to ring down to $1/e$ of its starting amplitude.

Step 1. Log-decrement over $N = 20$ cycles, from equation (6): $\delta = \tfrac{1}{20}\ln(10.0/6.0) = \tfrac{0.5108}{20} = 0.02554$ .

Step 2. Quality factor: $Q = \pi/\delta = 3.1416/0.02554 = \boxed{123}$ .

Step 3. Damping constant: $\beta = \delta/T = 0.02554/2.0 = 0.0128\ \mathrm{s^{-1}}$ . Cross-check against equation (3): $\omega_0 = 2\pi/T = 3.1416\ \mathrm{rad/s}$ , and $\omega_0/2\beta = 3.1416/0.02554 = 123$ . ✓

Step 4. Energy loss per cycle: $2\pi/Q = 2\pi/123 = \boxed{5.1\%}$ . Sanity check it directly: energy ratio per cycle $= (A_{n+1}/A_n)^2 = e^{-2\delta} = e^{-0.0511} = 0.950$ , a $5.0\%$ loss. ✓ (The tiny difference is the linearization of the exponential.)

Step 5. Ring-down: the amplitude e-folds after $Q/\pi = 39$ cycles, i.e. $39 \times 2.0\ \mathrm{s} = \boxed{78\ \mathrm{s}}$ . And after $Q = 123$ cycles (about four minutes) the amplitude will be $10.0 \times e^{-\pi} = 0.43$ cm — the $e^{-\pi}$ landmark in action.

2 · The watch crystal's energy budget (exam level)

The quartz tuning fork in a wristwatch oscillates at $f_0 = 32{,}768$ Hz. When the drive circuit is switched off, its vibration amplitude falls to $60\%$ of its initial value in $0.100$ s. (a) Find $\beta$ , $Q$ , and the energy decay time $\tau$ . (b) What fraction of its energy does the crystal lose per cycle? (c) If the crystal stores $E = 1.0$ nJ while running, what average power must the battery supply to sustain the oscillation, and what does this say about battery life?

Step 1. The envelope obeys $A(t) = A_0 e^{-\beta t}$ , so $e^{-\beta(0.100)} = 0.60$ , giving $\beta = \ln(1/0.60)/0.100 = 0.5108/0.100 = 5.11\ \mathrm{s^{-1}}$ .

Step 2. Quality factor from equation (3): $\omega_0 = 2\pi \times 32{,}768 = 2.059\times10^5\ \mathrm{rad/s}$ , so $Q = \omega_0/2\beta = 2.059\times10^5/10.22 = \boxed{2.0\times10^4}$ . Note the scale of the measurement: in that $0.100$ s the crystal completed $3{,}277$ full cycles while losing only $40\%$ of its swing — cycle-counting at its finest.

Step 3. Energy decay time from equation (4): $\tau = 1/2\beta = 1/10.22 = 0.098\ \mathrm{s}$ , and consistently $Q = \omega_0\tau = 2.059\times10^5 \times 0.098 = 2.0\times10^4$ . ✓

Step 4. Fractional energy loss per cycle: $2\pi/Q = 6.283/(2.0\times10^4) = \boxed{3.1\times10^{-4}}$ — about $0.03\%$ per tick of the crystal.

Step 5. Sustaining power. In steady oscillation the drive must replace dissipation, and the dissipation rate is $P = E/\tau = 2\beta E = 10.22 \times 1.0\times10^{-9}\ \mathrm{J} = \boxed{10\ \mathrm{nW}}$ . Cross-check per cycle: $P/f_0 = 1.02\times10^{-8}/32{,}768 = 3.1\times10^{-13}\ \mathrm{J} = (2\pi/Q)\,E$ . ✓

Step 6. Interpret. A coin cell stores roughly $1000\ \mathrm{J}$ ; at $10$ nW the resonator itself could run for thousands of years — in practice the counting electronics, not the crystal, drain the battery. High $Q$ is why a watch runs for years on a battery the size of a shirt button: the oscillator only ever asks for $2\pi/Q$ of its energy back per cycle.

Origins · 1 minLord Rayleigh and how this was discovered

Lord Rayleigh

1842–1919 · Cambridge and Terling, Essex

John William Strutt, third Baron Rayleigh, built the mathematical house that $Q$ lives in. His Theory of Sound (1877–78) treated free and forced vibration with damping systematically, and his "dissipation function" — a quadratic form encoding frictional losses alongside the kinetic and potential energies — is precisely the bookkeeping that makes "energy lost per cycle" a computable object. Rayleigh understood viscosity and vibration as two faces of one subject: he analyzed why tuning forks ring long and why resonators select their frequencies sharply, connecting persistence in time to selectivity in pitch decades before radio engineers needed the same mathematics. Characteristically, his greatest discovery came from refusing to let a small number slide: a $0.1\%$ discrepancy in the density of nitrogen led him to argon and the 1904 Nobel Prize. The letter " $Q$ " itself came later — coined around 1920 by the radio engineer Kenneth S. Johnson at Western Electric, who needed a symbol for a coil's "quality" — but the quantity it names is Rayleigh's ratio of energy stored to energy dissipated, unchanged from the Theory of Sound to the LIGO coating lab.

History

1877

— Rayleigh's Theory of Sound systematizes damped and forced vibration, including the dissipation function that makes per-cycle energy loss calculable.

1920

— Kenneth S. Johnson at Western Electric coins the symbol

Q

for the "quality" of radio coils; the reciprocal-bandwidth meaning becomes standard in circuit engineering.

1927

— Marrison and Horton build the first quartz-crystal clock at Bell Labs; the crystal's

Q \sim 10^5

makes it more stable than any pendulum.

1955

— Essen and Parry run the first cesium atomic clock at NPL; atomic transitions push effective line

Q

past

10^8

, and timekeeping leaves mechanics behind.

2015

— LIGO detects gravitational waves using test masses hung on fused-silica fibres with

Q \sim 10^9

, engineered so thermal noise concentrates into needle-thin frequency lines away from the signal band.

Going further

References. Morin, Introduction to Classical Mechanics, Ch. 4 (Oscillations) · Goldstein, Classical Mechanics, 3rd ed., Ch. 3 / oscillation sections · Landau & Lifshitz, Mechanics (Vol. 1), §25–26 · Taylor, Classical Mechanics, Ch. 5

Provenance. Definitions and formulas are standard textbook results: damped-oscillator solution and $Q=\omega_0/\gamma$ (Morin Ch. 4, Taylor Ch. 5); energy-loss-per-cycle definition and resonance width $\Delta\omega=\gamma$ (Landau & Lifshitz Mechanics §25–26); driven steady-state amplitude and Lorentzian lineshape (Goldstein Ch. 3). A verifier should recheck: (1) the factor-of-2 between amplitude decay $\gamma/2$ and energy decay $\gamma$ ; (2) the Lorentzian half-width $\gamma/2$ giving FWHM $=\gamma$ ; (3) the resonant amplification ratio $A(\omega_0)/A(0)=Q$ ; (4) $\omega_d=\omega_0\sqrt{1-1/4Q^2}$ with critical damping at $Q=1/2$ . Numerical example values (tuning fork $Q\sim10^4$ , quartz $\sim10^5$ , car suspension $\sim1$ ) are order-of-magnitude figures quoted in standard references, not precise measured constants. Worked-example numbers on this page ( $Q = 123$ from $\delta = 0.0255$ ; $\beta = 5.11\ \mathrm{s^{-1}}$ , $Q = 2.0\times10^4$ , $\tau = 98$ ms, $10$ nW for the watch crystal; cavity coast time $12$ s at $Q = 10^{11}$ ; Earth $_0S_2$ ring-down $6.1$ days at $Q = 510$ ) are illustrative arithmetic recomputed from the boxed formulas; this page uses the $\beta$ convention with $\gamma = 2\beta$ .

Builds on: Damped Harmonic Oscillator · Energy in Simple Harmonic Motion · Underdamped, Critically Damped, and Overdamped Regimes · Complex Exponential Representation of Oscillations Leads to: Resonance · Driven Oscillations

Why is there a factor of 2π in the definition of Q? It looks arbitrary.

It is there to make the three faces of $Q$ agree. Without it, the energy definition would give $E/\Delta E_{\text{cycle}} = Q/2\pi$ , which equals $\omega_0/4\pi\beta$ — an awkward number that matches neither the bandwidth ratio $\omega_0/\Delta\omega$ nor the amplification ratio $A(\omega_0)/A(0)$ . Inserting $2\pi$ aligns all three at the value $\omega_0/2\beta$ . The cleanest way to see it is the per-radian reading in equation (2): a cycle contains $2\pi$ radians of phase, so $2\pi E/\Delta E_{\text{cycle}}$ is exactly the energy stored divided by the energy lost per radian. Radians are the natural angle unit of oscillation ( $\omega_0$ itself is in radians per second), so measured per radian, the definition carries no arbitrary constant at all.

My textbook writes Q = ω₀/γ but this article says Q = ω₀/2β. Which is correct?

Both — they describe the same physics in different notation, and this genuinely causes wrong answers on exams. Texts that write the equation of motion as $\ddot x + \gamma\dot x + \omega_0^2 x = 0$ have an amplitude envelope $e^{-\gamma t/2}$ , so their $\gamma$ is our $2\beta$ and their $Q = \omega_0/\gamma$ is numerically identical to our $Q = \omega_0/2\beta$ . Landau writes $2\lambda\dot x$ , making $\lambda = \beta$ . The test that never fails: find how the amplitude decays in the book's own solution. If it decays as $e^{-st}$ for some symbol $s$ , then $Q = \omega_0/2s$ , always. Or bypass symbols entirely and use the energy definition (2), which is convention-proof.

Does Q change as the oscillation dies down?

Not for the linear oscillator of equation (1) — and that constancy is exactly what "exponential decay" means. Each cycle removes the same fraction $2\pi/Q$ of whatever energy remains, whether the swing is ten centimetres or ten microns; the log-decrement between successive peaks is the same all the way down the trace. Real systems can break this. Air drag grows quadratically with speed at large amplitude (loss fraction rises with amplitude), while dry sliding friction stays constant as the motion shrinks (loss fraction grows as amplitude falls, killing the motion in finite time with a linearly — not exponentially — shrinking envelope). Measuring how the apparent $Q$ varies with amplitude is a standard diagnostic for identifying which friction mechanism you actually have.

How can an atomic clock have a Q of 10¹⁰ when there is no spring and nothing mechanical is oscillating?

Because $Q$ was never really about springs — it is a property of any resonance, defined by equation (2) or by the ratio $\omega_0/\Delta\omega$ . A cesium atom driven near its hyperfine transition responds like a resonator with center frequency $9{,}192{,}631{,}770$ Hz; the interrogation scheme gives that resonance an effective width near $1$ Hz, hence $Q \approx 9\times10^9$ . The correspondence is not loose analogy: the classical damped oscillator and the decaying quantum state produce the same Lorentzian lineshape, with the state's lifetime playing the role of $\tau$ in $Q = \omega_0\tau$ . That is why one language covers wine glasses and strontium atoms — long-lived means narrow, in quantum mechanics as in mechanics.

If high Q is good, why are car suspensions and door closers deliberately built with Q near or below 1/2?

"Quality" is a figure of merit for a task, and the tasks differ. A clock or a filter wants to preserve and select one frequency: high $Q$ means low loss and a narrow line — pure win. A suspension's job is the opposite: absorb a road shock and return to equilibrium fast, without ringing. For that job the ideal is critical damping, $Q = \tfrac12$ — the fastest non-overshooting return derived in the regimes derivation. A high- $Q$ car would bounce down the road for $Q/\pi$ cycles after every pothole, and a high- $Q$ door would slam back and forth. Same number, same equations; whether you maximize or minimize it depends entirely on whether you want the system to remember or to forget.

Is Q meaningful for an overdamped system, where nothing oscillates and there are no cycles to count?

Formally you can still compute $Q = \omega_0/2\beta < \tfrac12$ , and it usefully labels how far into the overdamped regime you are — the two decay rates $r_\pm = \omega_0(-\tfrac{1}{2Q} \pm \sqrt{\tfrac{1}{4Q^2}-1})$ depend only on it. But every interpretation built in this article quietly dies: there are no cycles, so "energy lost per cycle" has no referent; there is no ring-down to log-decrement; and there is no resonant peak whose width you could measure (Resonance shows the displacement peak vanishes even earlier, at $Q = 1/\sqrt2$ ). Engineers therefore describe overdamped systems with the damping ratio $\zeta = 1/2Q$ instead, which stays conveniently near 1 at the critical boundary.

The article says Q equals the number of radians before the energy decays by 1/e. Radians of what, exactly?

Radians of oscillation phase. Free oscillation advances its phase $\phi = \omega_1 t \approx \omega_0 t$ steadily, at $2\pi$ radians per cycle — think of the reference point sweeping around the circle in the phasor picture. The energy meanwhile decays as $e^{-2\beta t}$ , e-folding at $\tau = 1/2\beta$ . The phase accumulated in that time is $\omega_0\tau = \omega_0/2\beta = Q$ radians — equation (4). Dividing by $2\pi$ radians per cycle converts to the equivalent cycle counts: $Q/2\pi$ cycles per energy e-fold, or $Q/\pi$ cycles per amplitude e-fold (the amplitude takes twice as long). All the memorable landmarks — $e^{-\pi}$ after $Q$ cycles, $0.63\%$ loss per cycle at $Q=1000$ — are this one radian statement re-expressed.

Two of my measurements disagree: the ringdown gives Q = 5000 but the bandwidth sweep gives Q = 3500. Which one is wrong?

Possibly neither — the disagreement itself is data. The triple identity (8) is a theorem about the linear, single-mode oscillator, so when two legs disagree, one of its assumptions is failing. Common culprits, in rough order: a second resonant mode hiding under the main peak broadens the swept curve without touching the ringdown; drive-amplitude nonlinearity (the swept measurement pushes the oscillator harder than the free decay) adds amplitude-dependent loss; and instrument bandwidth or sweep speed can smear the measured linewidth. A genuinely amplitude-dependent damping shows up as a curved log-decrement plot, which the ringdown itself will reveal. Precision resonator labs measure $Q$ both ways on purpose, precisely because agreement certifies the linear model and disagreement points at the interesting physics.

Quality Factor

Where this lives today

The physical picture

One number, three questions

The ringdown ruler

The grand tour — seventeen orders of magnitude

Q marks the damping regimes

One number, three measurements

Limiting cases

Worked examples

1 · Q of a lab pendulum from a ringdown (confidence builder)

2 · The watch crystal's energy budget (exam level)

Going further

Before this

Where this leads