Vibrating String — Question 7

PDF ↗

Question 7

A unit fixed-end string with speed one is known to involve only modes 1,…,N1,\ldots,N: u(x,t)=∑n=1N[fncos(nπt)+gnnπsin(nπt)]sin⁡(nπx).u(x,t)=\sum_{n=1}^N\left[f_n\cos(n\pi t) +\frac{g_n}{n\pi}\sin(n\pi t)\right]\sin(n\pi x). A displacement sensor at x=s∈(0,1)x=s\in(0,1) records hs(t)=u(s,t)h_s(t)=u(s,t) for 0≤t≤20\le t\le 2.

Tasks

  1. Derive recovery formulas for fn,gnf_n,g_n from the trace whenever the nnth mode is visible.

  2. For a rational sensor position s=p/qs=p/q in lowest terms, identify exactly the invisible modes. Explain why continuous-time sampling cannot recover those modes.

  3. Use sensors at s=1/2s=1/2 and s=1/3s=1/3. Determine their common invisible modes and the largest NN for which the entire stated finite-mode state is uniquely recoverable.

  4. Prove that an irrational sensor sees every individual mode but can be arbitrarily poorly conditioned as NN grows. You may use the pigeonhole fact that for every integer Q≥1Q\ge 1 there are integers 1≤n≤Q,k1\le n\le Q,k with |ns−k|≤1/Q|ns-k|\le 1/Q.

Original worksheet page 1: question and worked solution for 9-8-007
Show solutionHide solution

Question 7 – Solution

Strategy. Time orthogonality separates frequencies, but it cannot restore a spatial mode whose node lies at the sensor.

Step 1: Project the temporal record. On [0,2][0,2], distinct integer-frequency sines and cosines are orthogonal and each squared norm is one. Consequently fn=∫02hs(t)cos⁡(nπt)dtsin⁡(nπs),gn=nπ∫02hs(t)sin⁡(nπt)dtsin⁡(nπs)\boxed{f_n=\frac{\int_0^2h_s(t)\cos(n\pi t)\,dt}{\sin(n\pi s)},\qquad g_n=\frac{n\pi\int_0^2h_s(t)\sin(n\pi t)\,dt}{\sin(n\pi s)}} whenever sin⁡(nπs)≠0\sin(n\pi s)\ne 0. These recover both initial displacement and velocity coefficients.

Step 2: Identify rational-position blindness. For s=p/qs=p/q in lowest terms, sin⁡(nπs)=0\sin(n\pi s)=0 exactly when qq divides nn. Every such mode contributes zero to hs(t)h_s(t) for every time and for every choice of its two coefficients. Thus even a perfect continuous trace cannot distinguish states that differ only in those modes.

Step 3: Combine the two sensors. The midpoint misses the even modes; the one-third sensor misses multiples of three. Their common invisible modes are exactly the multiples of six. For N≤5N\le 5, each mode has a nonzero factor at at least one sensor, so the formulas recover the entire state. For N≥6N\ge 6, arbitrary sixth-mode displacement or velocity can be added without changing either record. The largest fully identifiable cutoff is therefore N=5.\boxed{N=5.}

Step 4: Distinguish visibility from stability. If ss is irrational, no nsns is an integer, so each denominator is nonzero. But the supplied approximation gives |sin⁡(nπs)|=|sin⁡(π(ns−k))|≤π/Q|\sin(n\pi s)|=|\sin(\pi(ns-k))|\le\pi/Q. As Q→∞Q\to\infty, the selected indices cannot stay in a finite set: each such index has a fixed positive distance from the integers. Hence arbitrarily large modes have arbitrarily small sensor factors. An error ene_n in the cosine projection becomes en/sin⁡(nπs)e_n/\sin(n\pi s) in fnf_n; the sine projection has the additional factor nπn\pi for gng_n. Exact visibility at all modes does not provide a uniform noise bound.

Original worksheet page 2: question and worked solution for 9-8-007

Original worksheet layout. Use Enlarge or open the PDF for a closer view.