A car goes over a bump, a landing gear takes an aircraft’s weight, an aircraft noses up and settles: one equation, three numbers. 5.2 Constant Coefficients & the Characteristic Equation solved in general. Here you apply it to the system that made it famous, and the question becomes which numbers a designer can choose, and what each buys. We name the dimensionless number that sets the shape of the motion, measure it from a recording, follow the energy, and test the folk wisdom that critical damping settles fastest.
Transcript
- A mass on a spring and a damper. The spring pulls back: minus k x. The damper, from memory, resists in proportion to speed: minus c x prime. Newton's second law then gives m x double prime plus c x prime plus k x equals zero: last lesson's equation, with a equal to m.
- Omega zero, root k over m, is the undamped frequency. The damping ratio zeta is c over the critical coefficient, two root m k; it has no units. Example: two kilograms, k two hundred, c eight: omega zero is ten, critical damping forty, zeta point two.
- Now vary zeta. At zero the roots sit on the imaginary axis. Raising it slides the pair along the circle of radius omega zero until they meet at minus omega zero: critical damping. Then they split along the real axis, product still omega zero squared. At zeta two: -2.6795 and -37.3205.
- Released from rest, three motions. Zeta point two rings at omega d, 9.797959, so the period is 0.641275 seconds, inside an envelope that shrinks by a factor e every half second. Critical damping never crosses zero. And zeta two is slower than critical: its slow root, -2.6795, rules the late motion.
- Tap and record. Every peak is the same fraction of the last; the log of that ratio is the log decrement, delta. The simulated peaks are 0.641275 seconds apart with ratio 0.277329, so delta is 1.282550. The shortcut, delta over two pi, gives 0.204124, two percent high; the exact inverse recovers 0.200000.
- Energy is kinetic plus spring. Differentiate, use Newton's law, and the spring terms cancel: d E d t equals minus c x prime squared, the damper's force times velocity. Only the total falls. At t equals point one the slope is -0.962939, matching minus c x prime squared.
- Energy shrinks by e to the minus two delta each period, since x and x prime both shrink by e to the minus delta. The rule four pi zeta: at zeta point zero one it is 6.4 percent high; at point two it claims 2.513274, more than all the energy, against the true 0.923088.
- Settling time is the last moment x is outside a band, here two percent. Zeta point nine undershoots by only point one five percent and settles at 4.6996; critical damping needs 5.8339. Zeta point seven undershoots by 4.6 percent, poking out, so it settles only at 5.9788: slower than critical.
- So is critical damping the fastest? Only among non-ringing returns. The best design has its first undershoot just touching the band: for two percent, zeta 0.7797 settles at 3.6025, against 5.8339 at critical. A hair lower and the lobe pokes out: 5.0684. To be inside and stay, ring slightly.
- Dry friction: a constant one point three newtons. Each half-cycle reflects the turning point about its centre, so the amplitude drops by a fixed point zero two six: linearly. It sticks at 11.9381 seconds, at point zero one two. The log ratio grows from 0.0534 to 1.674: no single delta.
- The same numbers sit inside the integrators. With no damping, one Euler step multiplies the energy by one plus h squared, like a damper with negative c: from point five to 10479.5778 in a thousand steps, exactly the formula. Symplectic Euler stays between 0.4762 and 0.5263.
- Damping wins stability back only for small steps: Euler's per-step growth is below one exactly when h is under two zeta over omega zero, here point zero four. RK4 allows about two point six to two point nine over omega zero, until heavy damping turns stiff: at zeta ten, 0.1396.
- To recap: a spring and a damper give m x double prime plus c x prime plus k x equals zero; omega zero and zeta set the motion. Energy drains at c x prime squared. Critical damping is fastest only among non-ringing returns. Next: pushing the mass.
From Newton’s law to the equation
A mass on a frictionless track is tied to a wall by a spring and a damper. Let be its displacement from the spring’s relaxed position.
- Spring. Hooke’s law: , proportional to the stretch and pointing back to rest; is in N/m.
- Damper. A viscous damper opposes the velocity in proportion to it, , with in N s/m (from memory: oil forced through a small gap at low speed resists in proportion to its speed).
Newton’s second law, , gives
the equation of the last lesson with , and third coefficient . For a hanging mass, measure from the stretched rest position and gravity cancels. With , every solution decays exactly when and (the decay criterion of the last lesson). Both forces are assumed linear.
Natural frequency and damping ratio
With no damper the roots are , so the mass oscillates at the natural frequency rad/s. Divide by and write , which defines the damping ratio
Why this grouping? has no units: in time units of the equation is , one family at different clock speeds. The discriminant is , so
| roots | motion | |
|---|---|---|
| undamped: constant-amplitude oscillation | ||
| , | underdamped: decaying oscillation, envelope | |
| (double) | critical: , no oscillation | |
| two negative reals, product | overdamped: two decaying exponentials |
The damped frequency is below only at second order (). As grows the pair slides along the circle to meet the real axis at , then splits into two reals whose product is still .
Worked example. A kg block, N/m, N s/m: rad/s, , . The roots are , the period is s and the envelope shrinks by every s. Released from rest, the last lesson’s formulas (, ) give . Block A prints these numbers; the first sim preset is this system.
Try the presets: no damper gives a closed ellipse in the (x, x’) plane, light damping a spiral, and at critical damping () the spiral is gone and the curve never crosses . At () the roots are and and the motion is slower than at critical damping. The mechanism is Vieta: the roots multiply to , so when the damper pushes one root out to the other must come in to , and that slow root governs the late motion.
Measuring damping: the logarithmic decrement
In a lab you tap the structure and record the ring-down. The cosine factor of repeats every , so for every , , and the same holds for . The maxima of are where , so they are exactly apart and each is a fixed fraction of the last. The logarithmic decrement is
The inverse solves for ; for light damping .
For the worked example , the peak ratio is , the shortcut gives (two percent high) and the exact inverse . Block A integrates with RK45, finds the maxima as zero crossings of , and prints log ratios of three times and gaps of s. In practice use , which divides reading error by .
Energy
Let . Differentiate and substitute :
The mechanism: is the damper’s force times the velocity, the power it removes, while the spring only trades energy between the two terms of . So never grows, and with it is constant. Block B checks it on the example: against at ; the sim prints beside the running integral of .
Since and both shrink by over one period, at every , so the fraction lost per cycle is exactly
The approximation uses (error about ) and . The first is the one that fails: the condition is , i.e. , not just ” small”. Block B’s simulated energy loss between successive maxima equals the exact column to six digits:
| exact | |||
|---|---|---|---|
The worked example loses per cycle; the rule would claim more than all of it.
How fast is “settled”?
Two things get called fast: the decay rate of the slowest mode and the settling time , the last moment is outside a band of times the release distance.
Decay rate. For the envelope rate is ; for the slow root’s rate is . Only at is it : critical damping has the fastest asymptotic decay, and the fastest among the non-ringing ones.
Settling time. Released from rest (), and . The velocity next vanishes at , where : the first undershoot. If it lies inside the band, the motion enters the band once and stays; if it pokes out, jumps to when that lobe re-enters. Block C computes by root-finding (times in units of ):
| lowest | , 5% band | , 2% band | |
|---|---|---|---|
Above the slow rate falls, so the times only grow ( and at ). Below it the picture is not smooth: at the undershoot () is inside both bands and (2%) beats critical damping’s . At the undershoot () is inside the 5% band, so is about times faster than critical, but it pokes out of the 2% band, and then is slower than critical.
The best case is the where the undershoot just touches the band: , i.e. with . For the 2% band and , against at critical damping. At the lobe pokes out and ; at it is . The other bands:
| band | at | at | ratio | |
|---|---|---|---|---|
| 5% | ||||
| 2% | ||||
| 0.1% |
So critical damping has the fastest asymptotic decay and non-ringing return, but for “inside 2% or 5% and stay there” after a release from rest the best design rings slightly, at or , and creeps toward as the band shrinks. The sim presets “zeta 0.78” and “critical” () read s and s, i.e. and .
The same ω₀ and ζ inside the integrators
For , , one Euler step , multiplies the energy by , as a damper with would. With and steps block E gives from , matching . Symplectic Euler instead stays inside over the run: the story of 4.7 Energy & Symplectic Integrators.
Damping wins stability back only for small : for , , below exactly when . For , , block E prints at and at (bound ). For RK4 it finds the largest stable :
The first four sit between and , but a heavily overdamped system is stiff (4.5 Stability Regions & Stiffness): at the fast root is and the limit is .
Where the viscous model stops
The derivation assumed and . Block D simulates one other force.
Coulomb (dry) friction: a constant-size force N opposing the motion, with , and no viscous damper. Moving left, oscillates about the centre ; moving right, about . A half-cycle reflects the turning point about its centre, so the amplitude falls by each half-cycle, linearly. From the turning points are . The motion also stops: once the spring cannot pull the mass loose (static and sliding friction taken equal), and it sticks after half-cycles, at s, at , not at zero. The log ratio of successive maxima is for the first pair and for the last: there is no single .
So plot against : constant means a linear viscous system and a usable ; growing with means dry friction. A spring that stiffens with stretch (Duffing, ) breaks it differently: its effective stiffness grows with the swing, so the period depends on the amplitude and no single exists (from memory; I did not simulate it here).
Lab
Blocks: A the worked example, B energy, C settling, D friction, E integrators.
Exercise: critical damping and the decrement from data
Implement critical_damping, peak_indices (local maxima of a sampled signal), log_decrement (the average of over the peaks) and zeta_from_delta (the exact inverse: the tests include , where is almost five percent off).
Problems
A kg platform sits on springs of total stiffness N/m. (a) Find and the critical damping. (b) With N s/m, find , and . (c) Released from rest, how far past rest does it first swing, as a fraction of the release distance?
Show worked solution
(a) rad/s and N s/m.
(b) , rad/s, s.
(c) The first undershoot is with exponent , so the fraction is , about .
A kg machine is tapped. Its peaks fall from mm to mm over cycles of s each. Estimate , and , and judge .
Show worked solution
and . The measured period is the damped one: , so rad/s. Then N/m and N s/m (critical: ). The shortcut gives , only high at this tiny .
A kg mass on an N/m spring has N s/m and is released from rest at m. (a) Find and . (b) Find the starting energy, the fraction lost per cycle, the energy after exactly cycles and the energy dissipated by then.
Show worked solution
(a) rad/s, , so ; and s.
(b) J. With , the loss per cycle is . After three cycles J, so the damper has dissipated J.
Key Takeaways
- A Hooke spring and a viscous damper give ; sets the time scale and the shape, with roots .
- Underdamped motion rings at inside ; overdamped rates multiply to , so more damping slows the slow one.
- Successive maxima shrink by a fixed ratio: and .
- ; the loss per cycle is exactly , and needs .
- Critical damping has the fastest asymptotic decay, but for a 2% or 5% band a design at or settles about times faster.
- and also govern the solver (Euler needs ; heavy damping is stiff), and Coulomb friction decays linearly and sticks.