Beats — Question 9

Section 17-7 · Physics pages 2768–2769

Question 9 Finite recording time broadens a spectral line A complex sinusoid e2πif0t is observed only for −T/2 ≤t ≤T/2. Its normalized Fourier amplitude at trial frequency f is A(f) = T −1 ∫T/2 −T/2 e2πi(f0−f)t dt. t p A finite recording window truncates the observed oscillation. (a) Evaluate the integral and find its nearest zeros to f0. (b) Two tones are 2.0 Hz apart. What recording duration puts one at the first spectral zero of the other? Relate this to the beat period without claiming it is an absolute limit on all frequency estimation methods.
Show solutionHide solution
Question 9 - Solution Finite recording time broadens a spectral line t p (a) Method and calculation For ∆= f0 −f̸ = 0, A(f) = eiπ∆T −e−iπ∆T 2πi∆T = sin(π∆T) π∆T . The continuous limit at ∆= 0 is one. The closest zeros satisfy |f −f0| = 1/T. (b) Interpretation and further calculation Setting 1/T = 2.0 Hz gives T = 0.50 s = 1/∆f. This equals one beat period. It is a useful resolution scale for the specified rectangular-window spectrum. Other windows, model-based fitting, noise levels and prior information can alter practical estimation performance; this calculation does not establish a universal hard bound. Physical check The product frequency offset times duration is dimensionless, and the zero spacing narrows as observation duration increases.

Question 9

Finite recording time broadens a spectral line

A complex sinusoid e2πif0te^{2\pi i f_0t} is observed only for −T/2≤t≤T/2-T/2\le t\le T/2. Its normalized Fourier amplitude at trial frequency ff is A(f)=T−1∫−T/2T/2e2πi(f0−f)tdtA(f)=T^{-1}\int_{-T/2}^{T/2}e^{2\pi i(f_0-f)t}\,dt.

Diagram shown in the original SVG worksheet.

  1. Evaluate the integral and find its nearest zeros to f0f_0.

  2. Two tones are 2.0Hz2.0\,\mathrm{Hz} apart. What recording duration puts one at the first spectral zero of the other? Relate this to the beat period without claiming it is an absolute limit on all frequency estimation methods.

Question 9 - Solution

Finite recording time broadens a spectral line

Diagram shown in the original SVG worksheet.

(a) Method and calculation

For Δ=f0−f≠0\Delta=f_0-f\ne0,A(f)=eiπΔT−e−iπΔT2πiΔT=sin⁡(πΔT)πΔT.A(f)=\frac{e^{i\pi\Delta T}-e^{-i\pi\Delta T}}{2\pi i\Delta T}=\frac{\sin(\pi\Delta T)}{\pi\Delta T}.The continuous limit at Δ=0\Delta=0 is one. The closest zeros satisfy |f−f0|=1/T|f-f_0|=1/T.

(b) Interpretation and further calculation

Setting 1/T=2.0Hz1/T=2.0\,\mathrm{Hz} givesT=0.50s=1/Δf.T=0.50\,\mathrm s=1/\Delta f.This equals one beat period. It is a useful resolution scale for the specified rectangular-window spectrum. Other windows, model-based fitting, noise levels and prior information can alter practical estimation performance; this calculation does not establish a universal hard bound.

Physical check

The product frequency offset times duration is dimensionless, and the zero spacing narrows as observation duration increases.