Unit 5, lesson 29

Sinusoidal Steady State and Frequency Response

How complex gain gives the amplitude ratio and phase lag of the sinusoidal steady state, how the frequency response tracks them across driving frequencies, and how a decaying transient meets the initial data.

  • Video 29
  • Frequency Response and Circuits

A damped mass on a spring is pushed by a force proportional to cos⁡2t\cos 2t. After a while the mass moves at the frequency of the force, but with its own amplitude, and its peaks arrive later than the peaks of the force. How large is the motion, and how late are its peaks? This lesson finds both from one complex number, follows them across all driving frequencies and then adds the decaying free motion that the initial data require.

Key terms

Complex gain

Definition Let tt be time and let primes denote derivatives with respect to tt. Consider

mx′′+cx′+kx=F0cos⁡ωtmx'' + cx' + kx = F_0\cos\omega t

with real constants m>0m > 0, c≥0c \ge 0, k>0k > 0, a force amplitude F0>0F_0 > 0 and a driving angular frequency ω≥0\omega \ge 0, and let p(r)=mr2+cr+kp(r) = mr^2 + cr + k be the characteristic polynomial. When p(iω)≠0p(i\omega) \ne 0, the complex gain at ω\omega is

H(iω)=1p(iω)=1k−mω2+icω.\begin{aligned} H(i\omega) &= \frac{1}{p(i\omega)} \\[0.4em] &= \frac{1}{k - m\omega^2 + ic\omega}. \end{aligned}

The condition p(iω)≠0p(i\omega) \ne 0 holds for every real ω\omega when c>0c > 0, and for every ω≠k/m\omega \ne \sqrt{k/m} when c=0c = 0. The reason it works is that (eiωt)′=iωeiωt(e^{i\omega t})' = i\omega e^{i\omega t} and (eiωt)′′=−ω2eiωt(e^{i\omega t})'' = -\omega^2 e^{i\omega t}, so a trial ZeiωtZe^{i\omega t} with a complex constant ZZ turns the left side into p(iω)Zeiωtp(i\omega)Ze^{i\omega t}. Matching F0eiωtF_0e^{i\omega t} gives Z=H(iω)F0Z = H(i\omega)F_0. Since the coefficients are real, the real part of a complex solution solves the equation with the real part of the force, and F0cos⁡ωt=Re⁡(F0eiωt)F_0\cos\omega t = \operatorname{Re}(F_0e^{i\omega t}). The imaginary unit is only a tool for the calculation. The force and the displacement are real.

In plain words Differentiating eiωte^{i\omega t} only multiplies it by iωi\omega, so for a sinusoidal input the whole differential equation acts as multiplication by one complex number. Dividing by that number gives the gain, which carries both the change in size and the delay.

Example For x′′+2x′+5x=F0cos⁡ωtx'' + 2x' + 5x = F_0\cos\omega t the characteristic polynomial is p(r)=r2+2r+5p(r) = r^2 + 2r + 5, so

H(iω)=15−ω2+2iω.H(i\omega) = \frac{1}{5 - \omega^2 + 2i\omega}.

At ω=2\omega = 2 the denominator is 1+4i1 + 4i, and multiplying the top and the bottom by the conjugate 1−4i1 - 4i gives

H(2i)=1−4i17.H(2i) = \frac{1 - 4i}{17}.

Sinusoidal steady state

Definition For the drive F0cos⁡ωtF_0\cos\omega t with p(iω)≠0p(i\omega) \ne 0, the sinusoidal steady state is the particular solution xs(t)=Re⁡(H(iω)F0eiωt)x_s(t) = \operatorname{Re}(H(i\omega)F_0e^{i\omega t}). Writing the gain in polar form as H(iω)=∣H(iω)∣e−iϕH(i\omega) = \lvert H(i\omega) \rvert e^{-i\phi}, it is

xs(t)=Acos⁡(ωt−ϕ),A=F0 ∣H(iω)∣,\begin{gathered} x_s(t) = A\cos(\omega t - \phi), \\[0.4em] A = F_0\,\lvert H(i\omega) \rvert, \end{gathered}

a sinusoid at the driving frequency. It is the only solution of the form acos⁡ωt+bsin⁡ωta\cos\omega t + b\sin\omega t. When c>0c > 0, every solution approaches it as t→∞t \to \infty, because the difference of two solutions solves the homogeneous equation and decays. That is why it is called the steady state. Without damping the same formula still gives a periodic particular solution, but other solutions do not approach it.

In plain words Once the starting conditions have been forgotten, the mass copies the force at the same frequency, scaled in size and shifted later in time.

Example For x′′+2x′+5x=13cos⁡2tx'' + 2x' + 5x = 13\cos 2t the gain above gives

xs=Re⁡(13(1−4i)17e2it)=1317cos⁡2t+5217sin⁡2t.\begin{aligned} x_s &= \operatorname{Re}\left(\frac{13(1 - 4i)}{17}e^{2it}\right) \\[0.4em] &= \tfrac{13}{17}\cos 2t + \tfrac{52}{17}\sin 2t. \end{aligned}

A real substitution confirms it. For x=acos⁡2t+bsin⁡2tx = a\cos 2t + b\sin 2t we have x′′=−4xx'' = -4x, so the left side is (a+4b)cos⁡2t+(b−4a)sin⁡2t(a + 4b)\cos 2t + (b - 4a)\sin 2t. With a=1317a = \tfrac{13}{17} and b=5217b = \tfrac{52}{17} this is 13cos⁡2t+0sin⁡2t13\cos 2t + 0\sin 2t, exactly the force.

Frequency response

Definition The frequency response of the oscillator is the complex gain viewed as a function of the driving angular frequency, ω↦H(iω)\omega \mapsto H(i\omega) for ω≥0\omega \ge 0. It is usually shown as two real curves against ω\omega, the amplitude ratio ∣H(iω)∣\lvert H(i\omega) \rvert and the phase lag ϕ(ω)\phi(\omega). Because the equation is linear, a drive that is a finite sum of sinusoids at different frequencies has as its steady state the sum of the separate steady states, each scaled and delayed according to its own frequency. With mm, cc and kk positive, the amplitude ratio starts at 1/k1/k when ω=0\omega = 0, behaves like 1/(mω2)1/(m\omega^2) for large ω\omega, and has a peak at a positive frequency only when c2<2mkc^2 < 2mk.

In plain words One formula answers the question for every driving frequency at once. Slow pushes are followed almost as if the spring were stretched by a steady force, and very fast pushes barely move the mass and run nearly half a cycle behind.

Example For x′′+2x′+5x=F0cos⁡ωtx'' + 2x' + 5x = F_0\cos\omega t the table lists the amplitude ratio and the lag, in radians, at a few frequencies.

ω\omega∣H(iω)∣\lvert H(i\omega) \rvertϕ\phi
000.2000.20000
110.2240.2240.4640.464
3\sqrt{3}0.2500.2501.0471.047
220.2430.2431.3261.326
5\sqrt{5}0.2240.2241.5711.571
330.1390.1392.1592.159
10100.0100.0102.9342.934

The ratio reaches its largest value 14\tfrac{1}{4} at ω=3\omega = \sqrt{3}, which is possible because c2=4<10=2mkc^2 = 4 < 10 = 2mk. The lag passes π/2\pi/2 at the natural angular frequency 5\sqrt{5} and approaches π\pi as ω\omega grows.

Amplitude ratio

Definition The amplitude ratio at the angular frequency ω\omega is the modulus of the complex gain,

∣H(iω)∣=1∣p(iω)∣,\lvert H(i\omega) \rvert = \frac{1}{\lvert p(i\omega) \rvert},

where ∣p(iω)∣2=(k−mω2)2+c2ω2\lvert p(i\omega) \rvert^2 = (k - m\omega^2)^2 + c^2\omega^2. It is the amplitude of the steady displacement divided by the amplitude of the force, in meters per newton. Geometrically, ∣p(iω)∣\lvert p(i\omega) \rvert is the length of the vector from 00 to the point (k−mω2,cω)(k - m\omega^2, c\omega) of the complex plane, and the ratio is its reciprocal.

In plain words Multiply the force amplitude by this number to get the size of the steady motion.

Example For x′′+2x′+5x=13cos⁡2tx'' + 2x' + 5x = 13\cos 2t the denominator 1+4i1 + 4i has length 17\sqrt{17}, so the amplitude ratio is 1/171/\sqrt{17} and the steady amplitude is 13/17≈3.1513/\sqrt{17} \approx 3.15. The coefficients of the steady state agree, since

(1317)2+(5217)2=1317.\sqrt{\left(\tfrac{13}{17}\right)^2 + \left(\tfrac{52}{17}\right)^2} = \frac{13}{\sqrt{17}}.

Phase lag

Definition Write atan2⁡(y,x)\operatorname{atan2}(y, x) for the angle in (−π,π](-\pi, \pi] of the point (x,y)≠(0,0)(x, y) \ne (0, 0). The phase lag at ω\omega is the angle of the point (k−mω2,cω)(k - m\omega^2, c\omega) that represents p(iω)p(i\omega),

ϕ=atan2⁡(cω, k−mω2),\phi = \operatorname{atan2}\left(c\omega,\ k - m\omega^2\right),

so that p(iω)=∣p(iω)∣eiϕp(i\omega) = \lvert p(i\omega) \rvert e^{i\phi} and H(iω)=∣H(iω)∣e−iϕH(i\omega) = \lvert H(i\omega) \rvert e^{-i\phi}. For c>0c > 0 and ω>0\omega > 0 the second coordinate cωc\omega is positive, so 0<ϕ<π0 < \phi < \pi, and the steady state Acos⁡(ωt−ϕ)A\cos(\omega t - \phi) reaches each peak ϕ/ω\phi/\omega seconds after the most recent peak of the force. The lag is less than π/2\pi/2 when ω<k/m\omega < \sqrt{k/m}, equal to π/2\pi/2 when ω=k/m\omega = \sqrt{k/m}, and greater than π/2\pi/2 beyond that. The one-argument formula arctan⁡(cω/(k−mω2))\arctan(c\omega/(k - m\omega^2)) gives the same angle only when k−mω2>0k - m\omega^2 > 0. Above the natural frequency it returns a negative angle in the wrong quadrant, and at the natural frequency it is undefined.

In plain words The lag says how far behind the force the motion runs, measured as an angle of the cycle. Slow driving gives almost no lag, driving at the natural frequency gives a quarter cycle, and fast driving approaches half a cycle.

Example At ω=2\omega = 2 for x′′+2x′+5xx'' + 2x' + 5x, the point is (1,4)(1, 4), in the first quadrant, so ϕ=arctan⁡4≈1.326\phi = \arctan 4 \approx 1.326 radians, about 7676 degrees, and the response peaks trail the force peaks by ϕ/2≈0.663\phi/2 \approx 0.663 seconds. With A=13/17A = 13/\sqrt{17}, cos⁡ϕ=1/17\cos\phi = 1/\sqrt{17} and sin⁡ϕ=4/17\sin\phi = 4/\sqrt{17}, expanding Acos⁡(2t−ϕ)A\cos(2t - \phi) returns 1317cos⁡2t+5217sin⁡2t\tfrac{13}{17}\cos 2t + \tfrac{52}{17}\sin 2t, and the positive sine coefficient is the mark of a lag. At ω=3\omega = 3 the point is (−4,6)(-4, 6), in the second quadrant, so ϕ=π−arctan⁡32≈2.159\phi = \pi - \arctan\tfrac{3}{2} \approx 2.159. The careless value arctan⁡(6/(−4))≈−0.983\arctan(6/(-4)) \approx -0.983 would wrongly suggest that the motion runs ahead of the force.

Transient and steady-state response

Definition Let c>0c > 0 and consider the drive F0cos⁡ωtF_0\cos\omega t with initial values x(0)x(0) and x′(0)x'(0). The solution splits as

x=xs+xtr,x = x_s + x_{\text{tr}},

where xsx_s is the sinusoidal steady state and the transient xtrx_{\text{tr}} is the solution of the homogeneous equation mx′′+cx′+kx=0mx'' + cx' + kx = 0 with xtr(0)=x(0)−xs(0)x_{\text{tr}}(0) = x(0) - x_s(0) and xtr′(0)=x′(0)−xs′(0)x_{\text{tr}}'(0) = x'(0) - x_s'(0). Because c>0c > 0, both characteristic roots have negative real part, so the transient decays to zero and x−xs→0x - x_s \to 0 as t→∞t \to \infty. The steady state depends only on the drive, and the transient carries the initial data.

In plain words The steady state is what the drive asks for, and the transient is the correction that makes the motion start where it actually starts. With damping, the correction fades and only the response to the drive remains.

Example Solve x′′+2x′+5x=13cos⁡2tx'' + 2x' + 5x = 13\cos 2t from rest, x(0)=0x(0) = 0 and x′(0)=0x'(0) = 0. The steady state starts at xs(0)=1317x_s(0) = \tfrac{13}{17} with velocity xs′(0)=10417x_s'(0) = \tfrac{104}{17}, so it cannot be the whole answer. The transient has the form e−t(Acos⁡2t+Bsin⁡2t)e^{-t}(A\cos 2t + B\sin 2t), with A=−1317A = -\tfrac{13}{17} and initial velocity −A+2B-A + 2B, which must equal −10417-\tfrac{104}{17}. This gives B=−11734B = -\tfrac{117}{34}, so

xtr=−e−t(1317cos⁡2t+11734sin⁡2t)\begin{aligned} x_{\text{tr}} = -e^{-t}\bigl(&\tfrac{13}{17}\cos 2t \\ &+ \tfrac{117}{34}\sin 2t\bigr) \end{aligned}

and x=xs+xtrx = x_s + x_{\text{tr}}. At t=0t = 0 the two parts cancel in both position and velocity. By t=5t = 5 the transient is smaller than 0.0240.024 in size, while the steady amplitude stays 13/17≈3.1513/\sqrt{17} \approx 3.15.

Check yourself

  1. Find the steady state of x′′+2x′+5x=10cos⁡tx'' + 2x' + 5x = 10\cos t, where primes are derivatives in tt, together with its amplitude and phase lag.
Show the answer

Here p(i)=5−1+2i=4+2ip(i) = 5 - 1 + 2i = 4 + 2i, so

10H(i)=104+2i=10(4−2i)20=2−i.\begin{aligned} 10H(i) &= \frac{10}{4 + 2i} \\[0.4em] &= \frac{10(4 - 2i)}{20} = 2 - i. \end{aligned}

The steady state is Re⁡((2−i)eit)=2cos⁡t+sin⁡t\operatorname{Re}((2 - i)e^{it}) = 2\cos t + \sin t. Its amplitude is ∣2−i∣=5≈2.24\lvert 2 - i \rvert = \sqrt{5} \approx 2.24, and its lag is atan2⁡(2,4)=arctan⁡12≈0.464\operatorname{atan2}(2, 4) = \arctan\tfrac{1}{2} \approx 0.464 radians, so xs=5cos⁡(t−ϕ)x_s = \sqrt{5}\cos(t - \phi). A real check uses x′′=−xx'' = -x for this function, so the left side is 4(2cos⁡t+sin⁡t)+2(−2sin⁡t+cos⁡t)=10cos⁡t4(2\cos t + \sin t) + 2(-2\sin t + \cos t) = 10\cos t.

  1. For the same oscillator driven at ω=4\omega = 4, a classmate computes the lag as arctan⁡(8/(−11))≈−0.629\arctan(8/(-11)) \approx -0.629 and concludes that the motion runs ahead of the force. Find the correct lag and the amplitude ratio.
Show the answer

The denominator is p(4i)=5−16+8i=−11+8ip(4i) = 5 - 16 + 8i = -11 + 8i, a point in the second quadrant. So

ϕ=atan2⁡(8,−11)=π−arctan⁡811≈2.513\begin{aligned} \phi &= \operatorname{atan2}(8, -11) \\[0.4em] &= \pi - \arctan\tfrac{8}{11} \approx 2.513 \end{aligned}

radians, about 144144 degrees, a lag between π/2\pi/2 and π\pi as expected above the natural frequency 5\sqrt{5}. The one-argument arctangent sees only the ratio 8/(−11)8/(-11), which is also the ratio for the point (11,−8)(11, -8) in the fourth quadrant, so it picks the wrong quadrant. The amplitude ratio is 1/∣−11+8i∣=1/185≈0.07351/\lvert -11 + 8i \rvert = 1/\sqrt{185} \approx 0.0735.

  1. Solve x′′+2x′+5x=10cos⁡tx'' + 2x' + 5x = 10\cos t with x(0)=0x(0) = 0 and x′(0)=0x'(0) = 0, and name the transient and the steady state. What would change without the damping term?
Show the answer

From question 1 the steady state is xs=2cos⁡t+sin⁡tx_s = 2\cos t + \sin t, with xs(0)=2x_s(0) = 2 and xs′(0)=1x_s'(0) = 1. The transient e−t(Acos⁡2t+Bsin⁡2t)e^{-t}(A\cos 2t + B\sin 2t) must start at −2-2 with velocity −1-1, so A=−2A = -2 and −A+2B=−1-A + 2B = -1, which gives B=−32B = -\tfrac{3}{2}. So the transient is

xtr=−e−t(2cos⁡2t+32sin⁡2t),\begin{aligned} x_{\text{tr}} = -e^{-t}\bigl(&2\cos 2t \\ &+ \tfrac{3}{2}\sin 2t\bigr), \end{aligned}

which decays like e−te^{-t}, and the solution is x=2cos⁡t+sin⁡t+xtrx = 2\cos t + \sin t + x_{\text{tr}}. Without damping the equation x′′+5x=10cos⁡tx'' + 5x = 10\cos t has free solutions cos⁡5 t\cos\sqrt{5}\,t and sin⁡5 t\sin\sqrt{5}\,t that never decay. The solution from rest is then 52(cos⁡t−cos⁡5 t)\tfrac{5}{2}(\cos t - \cos\sqrt{5}\,t), which keeps both frequencies forever and never settles onto a single sinusoid.

How this lesson connects