Skip to content

Continuous and Discrete Signals, Fourier Series and Transforms

Signal classes, elementary signals, Fourier series, Dirichlet conditions, Gibbs effect, Parseval and Fourier transform pairs for CSIR NET geophysics.

  • 21 min read
  • Updated
On this page · 15 sections
  1. Prerequisites
  2. The Hook: A Tide Gauge Record
  3. What a Signal Is
  4. Elementary Signals
  5. Discrete-Time Signals Are Not Just Sampled Curves
  6. Fourier Series
  7. From Series to Transform
  8. Where These Ideas Show Up in Geophysics
  9. Memory Tricks
  10. Don’t Confuse
  11. Common Mistakes
  12. Practice
  13. Quick Revision Table
  14. Key Takeaways
  15. What’s Next

Prerequisites

The Hook: A Tide Gauge Record

How CSIR NET asks it. In the Geophysics section of Part C, expect short derivations and recognition questions: is a sequence periodic, is a signal an energy or a power signal, which harmonics a waveform contains, the value a Fourier series converges to at a jump, Parseval sums, and transform pairs such as boxcar to sinc.


What a Signal Is

A signal is any quantity that varies with an independent variable, usually time or distance: ground velocity at a geophone, the total magnetic field along a flight line, gravity along a profile, temperature in a borehole. Mathematically it is a function \( x(t) \) or a sequence \( x[n] \).

Classification

BasisClassesTest
Independent variableContinuous-time \( x(t) \) / discrete-time \( x[n] \)Is it defined for every \( t \) or only at integers \( n \)?
AmplitudeAnalog (any value) / digital (quantised levels)A digital seismogram is both discrete-time and quantised
RepetitionPeriodic / aperiodic\( x(t + T) = x(t) \) for all \( t \); smallest such \( T > 0 \) is the fundamental period
SymmetryEven / odd\( x(-t) = x(t) \) or \( x(-t) = -x(t) \)
PredictabilityDeterministic / random (stochastic)A sweep is deterministic; microseismic noise is random
SizeEnergy / power signalFinite energy, or finite non-zero average power
SupportCausal / non-causal\( x(t) = 0 \) for \( t < 0 \)

Even–odd decomposition. Every signal splits uniquely as \( x = x_e + x_o \) with

$$x_e(t) = \tfrac{1}{2}\left[x(t) + x(-t)\right], \qquad x_o(t) = \tfrac{1}{2}\left[x(t) - x(-t)\right].$$

Energy and power

$$E = \int_{-\infty}^{\infty} \lvert x(t)\rvert^2\,dt, \qquad P = \lim_{T\to\infty}\frac{1}{2T}\int_{-T}^{T}\lvert x(t)\rvert^2\,dt .$$

For sequences replace the integral by a sum over \( n \). Then:

  • Energy signal: \( 0 < E < \infty \), so \( P = 0 \). Transients: a seismic wavelet, \( e^{-at}u(t) \) with \( a > 0 \).
  • Power signal: \( 0 < P < \infty \), so \( E = \infty \). Periodic signals, the unit step, stationary noise.
  • Some signals are neither: \( x(t) = t \) has infinite energy and infinite power.

For \( x(t) = e^{-at}u(t) \), \( E = \int_0^\infty e^{-2at}dt = 1/(2a) \). A sinusoid of amplitude \( A \) has average power \( A^2/2 \); a constant \( C \) has power \( C^2 \).


Elementary Signals

SignalContinuous formDiscrete formWhy it matters
Unit impulse\( \delta(t) \): zero for \( t \ne 0 \), \( \int\delta\,dt = 1 \)\( \delta[n] = 1 \) at \( n = 0 \), else 0Impulse response defines an LTI system
Unit step\( u(t) = 1 \) for \( t > 0 \)\( u[n] = 1 \) for \( n \ge 0 \)Switching on; causal signals
Boxcar (rect)1 for \( \lvert t\rvert < T/2 \)1 for \( 0 \le n \le N-1 \)Every finite record is a boxcar times the signal
Sinusoid\( A\cos(\omega t + \phi) \)\( A\cos(\Omega n + \phi) \)Eigenfunction of LTI systems (real form)
Complex exponential\( e^{i\omega t} \)\( e^{i\Omega n} \)Building block of Fourier analysis
Sinc\( \mathrm{sinc}(t) = \sin(\pi t)/(\pi t) \)—Spectrum of a boxcar; ideal interpolator
Ramp\( r(t) = t\,u(t) \)\( n\,u[n] \)Integral of the step

Key relations: \( u(t) = \int_{-\infty}^{t}\delta(\tau)\,d\tau \), \( \delta[n] = u[n] - u[n-1] \), and the sifting property

$$\int_{-\infty}^{\infty} x(t)\,\delta(t - t_0)\,dt = x(t_0), \qquad x[n] = \sum_{k} x[k]\,\delta[n - k].$$

The second form says any sequence is a sum of scaled, shifted impulses. That one line is the root of convolution (next page).

Operations on the independent variable

OperationResultNote
Time shift \( x(t - t_0) \)Delays by \( t_0 \) if \( t_0 > 0 \)
Time reversal \( x(-t) \)Mirror about \( t = 0 \)Used in correlation and convolution
Time scaling \( x(at) \)Compresses if \( \lvert a\rvert > 1 \)In discrete time \( x[2n] \) drops samples (decimation)
Combined \( x(at - b) \)Shift by \( b \) first, then scale by \( a \)Equivalent to \( x(a(t - b/a)) \)

Discrete-Time Signals Are Not Just Sampled Curves

Sampling a continuous signal at interval \( \Delta t \) gives \( x[n] = x(n\Delta t) \). Two facts about discrete sinusoids catch candidates every year.

1. A discrete sinusoid is periodic only if its frequency is a rational multiple of \( 2\pi \). For \( x[n] = \cos(\Omega_0 n) \) to repeat after \( N \) samples we need \( \Omega_0 N = 2\pi m \) for integers \( m, N \):

$$\frac{\Omega_0}{2\pi} = \frac{m}{N}\ \text{(rational)}, \qquad N_{\min} = \frac{2\pi m}{\Omega_0}\ \text{with the smallest integer } m \text{ that makes } N \text{ an integer.}$$
  • \( \cos(0.3\pi n) \): \( \Omega_0/2\pi = 3/20 \), so \( N = 20 \).
  • \( \cos(n) \): \( 1/(2\pi) \) is irrational, so the sequence is never periodic, although \( \cos t \) is.

2. Discrete frequencies repeat every \( 2\pi \). \( e^{i(\Omega_0 + 2\pi)n} = e^{i\Omega_0 n} \) for integer \( n \). Only \( -\pi < \Omega \le \pi \) is distinct, and the highest oscillation rate is \( \Omega = \pi \), the alternating sequence \( (-1)^n \). With \( \Omega = \omega\Delta t \), this is the frequency \( 1/(2\Delta t) \), the Nyquist frequency. Higher continuous frequencies fold back: this is aliasing, treated fully in Sampling Theorem, FFT, Z-Transform and Convolution.

Sum of periodic signals. \( x_1 + x_2 \) with periods \( T_1, T_2 \) is periodic only if \( T_1/T_2 \) is rational; the period is the least common multiple. \( \sin(4\pi t) + \cos(6\pi t) \) has periods \( 1/2 \) and \( 1/3 \) s, so the sum repeats every 1 s. \( \sin t + \sin(\sqrt{2}\,t) \) never repeats.


Fourier Series

A periodic signal with period \( T \) and fundamental angular frequency \( \omega_0 = 2\pi/T \) can be written as a sum of harmonics.

Trigonometric form

$$x(t) = a_0 + \sum_{n=1}^{\infty}\left[a_n\cos(n\omega_0 t) + b_n\sin(n\omega_0 t)\right]$$$$a_0 = \frac{1}{T}\int_T x\,dt, \qquad a_n = \frac{2}{T}\int_T x(t)\cos(n\omega_0 t)\,dt, \qquad b_n = \frac{2}{T}\int_T x(t)\sin(n\omega_0 t)\,dt .$$

\( a_0 \) is the mean (the “DC” value). The coefficients come from the orthogonality of sines and cosines over one period.

Complex exponential form

$$x(t) = \sum_{n=-\infty}^{\infty} c_n e^{in\omega_0 t}, \qquad c_n = \frac{1}{T}\int_T x(t)\,e^{-in\omega_0 t}\,dt .$$

Links: \( c_0 = a_0 \), \( c_n = \tfrac{1}{2}(a_n - i b_n) \) for \( n > 0 \), and for real \( x \), \( c_{-n} = c_n^{*} \). The amplitude spectrum \( \lvert c_n\rvert \) is even; the phase spectrum \( \arg c_n \) is odd. A periodic signal has a line spectrum: energy only at \( n\omega_0 \).

Dirichlet conditions (sufficient for convergence)

Over one period, \( x(t) \) must

  1. be absolutely integrable, \( \int_T\lvert x\rvert\,dt < \infty \);
  2. have a finite number of maxima and minima;
  3. have a finite number of finite discontinuities.

Then the series converges to \( x(t) \) where \( x \) is continuous, and to the midpoint \( \tfrac{1}{2}\left[x(t^-) + x(t^+)\right] \) at a jump.

Symmetry shortcuts

SymmetryConditionSeries contains
Even\( x(-t) = x(t) \)\( a_0 \) and cosines only (\( b_n = 0 \)); \( c_n \) real
Odd\( x(-t) = -x(t) \)sines only (\( a_0 = a_n = 0 \)); \( c_n \) imaginary
Half-wave\( x(t + T/2) = -x(t) \)odd harmonics only, zero mean
Even + half-wavebothodd cosines only
Odd + half-wavebothodd sines only (the square wave)

Three standard series

Square wave (odd, amplitude \( \pm A \), period \( T \)):

$$x(t) = \frac{4A}{\pi}\left[\sin\omega_0 t + \frac{\sin 3\omega_0 t}{3} + \frac{\sin 5\omega_0 t}{5} + \cdots\right]$$

Sawtooth \( x(t) = t \) on \( (-\pi, \pi) \), repeated:

$$x(t) = 2\left[\sin t - \frac{\sin 2t}{2} + \frac{\sin 3t}{3} - \cdots\right] = 2\sum_{n=1}^{\infty}\frac{(-1)^{n+1}}{n}\sin nt$$

Triangle wave \( x(t) = \lvert t\rvert \) on \( (-\pi, \pi) \), repeated:

$$x(t) = \frac{\pi}{2} - \frac{4}{\pi}\left[\cos t + \frac{\cos 3t}{9} + \frac{\cos 5t}{25} + \cdots\right]$$

How fast the coefficients fall

Smoothness of \( x(t) \)Coefficients decay asExample
Jump discontinuity in \( x \)\( 1/n \)square wave, sawtooth
\( x \) continuous, jump in \( x' \)\( 1/n^2 \)triangle wave
\( x, x' \) continuous, jump in \( x'' \)\( 1/n^3 \)parabolic wave
Infinitely smoothfaster than any powera pure sinusoid (one term)

A sharp edge in time needs a lot of high-frequency energy. The same rule explains why a seismic wavelet with an abrupt onset has a broad spectrum.

The Gibbs phenomenon

Truncate a Fourier series of a function with a jump and the partial sum overshoots next to the jump. As more terms are added the overshoot moves closer to the jump and gets narrower, but its height does not shrink: it tends to about 9% of the jump (about 0.0895 of the jump height on each side). The partial sums converge pointwise but not uniformly near the discontinuity. In processing, the same ringing appears when a spectrum is cut off sharply (a boxcar filter), which is why practical filters taper their edges (see Filters, Deconvolution and Fractal Analysis).

Parseval’s theorem for series

$$\frac{1}{T}\int_T\lvert x(t)\rvert^2\,dt = \sum_{n=-\infty}^{\infty}\lvert c_n\rvert^2 = a_0^2 + \frac{1}{2}\sum_{n=1}^{\infty}\left(a_n^2 + b_n^2\right).$$

Average power is the sum of the powers of the harmonics. Applied to the square wave (power \( A^2 \)):

$$A^2 = \frac{1}{2}\sum_{n\ \text{odd}}\left(\frac{4A}{n\pi}\right)^2 \;\Rightarrow\; \sum_{n\ \text{odd}}\frac{1}{n^2} = \frac{\pi^2}{8}.$$

The fundamental alone carries \( 8/\pi^2 \approx 81\% \) of the square wave’s power.


From Series to Transform

Let the period \( T \to \infty \). The line spacing \( \omega_0 = 2\pi/T \) shrinks to zero, the lines merge into a continuous spectrum, and the series becomes the Fourier transform. In the frequency-in-hertz convention used by most geophysics texts:

$$X(f) = \int_{-\infty}^{\infty} x(t)\,e^{-i2\pi f t}\,dt, \qquad x(t) = \int_{-\infty}^{\infty} X(f)\,e^{i2\pi f t}\,df .$$

With angular frequency, \( X(\omega) = \int x\,e^{-i\omega t}dt \) and the inverse carries \( 1/2\pi \). Sign conventions differ between books; the physics does not.

Transform pairs to know cold

\( x(t) \)\( X(f) \)
\( \delta(t) \)\( 1 \) (flat, white)
\( 1 \)\( \delta(f) \)
\( \mathrm{rect}(t/T) \) (width \( T \))\( T\,\mathrm{sinc}(fT) \), first zeros at \( f = \pm 1/T \)
\( e^{-at}u(t),\ a > 0 \)\( 1/(a + i2\pi f) \)
\( e^{-a\lvert t\rvert} \)\( 2a/\left[a^2 + (2\pi f)^2\right] \)
Gaussian \( e^{-\pi t^2} \)\( e^{-\pi f^2} \) (its own transform)
\( \cos(2\pi f_0 t) \)\( \tfrac{1}{2}\left[\delta(f - f_0) + \delta(f + f_0)\right] \)
Comb \( \sum_n\delta(t - n\Delta t) \)\( \frac{1}{\Delta t}\sum_k\delta(f - k/\Delta t) \)

Properties

PropertyTimeFrequency
Linearity\( a x + b y \)\( aX + bY \)
Shift\( x(t - t_0) \)\( X(f)\,e^{-i2\pi f t_0} \): amplitude unchanged, linear phase
Modulation\( x(t)\,e^{i2\pi f_0 t} \)\( X(f - f_0) \)
Scaling\( x(at) \)\( \frac{1}{\lvert a\rvert}X(f/a) \)
Reversal\( x(-t) \)\( X(-f) \) (\( = X^{*}(f) \) for real \( x \))
Differentiation\( dx/dt \)\( i2\pi f\,X(f) \)
Convolution\( x * y \)\( X\,Y \)
Multiplication\( x\,y \)\( X * Y \)
Correlation\( x \star y \)\( X^{*}Y \) (convention-dependent)
Parseval–Rayleigh\( \int\lvert x\rvert^2dt \)\( \int\lvert X\rvert^2df \)

Two consequences used constantly in geophysics:

  • Time–bandwidth trade-off. Scaling says that compressing a pulse spreads its spectrum. A boxcar of width \( T \) has a main lobe of half-width \( 1/T \): a short wavelet means wide bandwidth, which means good resolution.
  • Delay is a phase ramp. Shifting a trace changes only its phase, by \( -2\pi f t_0 \). A zero-phase wavelet is symmetric about \( t = 0 \); a pure delay adds linear phase.
graph LR
    A["Periodic, continuous
x(t)"] -->|"Fourier series"| B["Discrete line spectrum
c_n at n/T"] C["Aperiodic, continuous
x(t)"] -->|"Fourier transform"| D["Continuous spectrum
X(f)"] E["Aperiodic, discrete
x[n]"] -->|"DTFT"| F["Continuous, periodic
in f with period 1/Δt"] G["Periodic, discrete
N samples"] -->|"DFS / DFT"| H["Discrete, periodic
N lines"]

The diagram carries one rule: discrete in one domain means periodic in the other. Sampling in time makes the spectrum periodic (hence aliasing); sampling the spectrum makes the time signal periodic (hence the wrap-around of circular convolution).

Fourier analysis of discrete sequences

  • DTFT: \( X(\Omega) = \sum_n x[n]e^{-i\Omega n} \), continuous and \( 2\pi \)-periodic in \( \Omega \).
  • Discrete Fourier series / DFT: for \( N \) samples, \( X[k] = \sum_{n=0}^{N-1}x[n]e^{-i2\pi kn/N} \), \( k = 0, \dots, N-1 \). There are only N distinct coefficients, because \( e^{-i2\pi(k+N)n/N} = e^{-i2\pi kn/N} \). The DFT samples the DTFT at \( N \) equally spaced frequencies.

How the DFT is computed fast, and what its frequency spacing means, is on the sampling and FFT page.


Where These Ideas Show Up in Geophysics

SignalVariableTypical treatment
Seismogramtimewavelet, reflectivity, spectra, filtering
Gravity or magnetic profiledistancewavenumber \( k = 2\pi/\lambda \) spectra; upward continuation multiplies by \( e^{-\lvert k\rvert h} \)
Tide-gauge or geomagnetic observatory recordtimeharmonic (Fourier) analysis for periodic constituents
Well logdepthspectral and fractal (scaling) analysis
Magnetotelluric fieldstimeimpedance computed in the frequency domain

Spatial signals obey the same mathematics with \( t \to x \) and \( f \to 1/\lambda \). The gravity and magnetic field theory behind continuation is a separate Part C unit; the basics of the fields are in Gravity, Geomagnetism, Heat Flow and Isostasy.


Memory Tricks

Don’t Confuse

PairFirstSecond
Energy vs power signalfinite energy, zero average power (transients)infinite energy, finite power (periodic, stationary noise)
\( \cos t \) vs \( \cos n \)periodic, period \( 2\pi \)not periodic (no integer period)
Fourier series vs transformperiodic signal, discrete linesaperiodic signal, continuous spectrum
DTFT vs DFTcontinuous in \( \Omega \), infinite sequence\( N \) samples of the DTFT for an \( N \)-point record
Odd function vs odd harmonics\( x(-t) = -x(t) \): sines onlyhalf-wave symmetry: \( n = 1, 3, 5, \dots \) only
Zero phase vs linear phasesymmetric about \( t = 0 \), phase \( = 0 \)symmetric about \( t_0 \), phase \( = -2\pi f t_0 \)
Amplitude vs power spectrum\( \lvert X(f)\rvert \)\( \lvert X(f)\rvert^2 \) (or its expected value for random signals)

Common Mistakes


Practice

All questions below are original practice questions written for this page. They are not from any CSIR NET paper.

Level 1: Foundation

  1. What is the fundamental period of \( x[n] = \cos(0.35\pi n) \)?
  2. Is \( x(t) = e^{-2t}u(t) \) an energy or a power signal? Give its energy.
  3. To what value does the Fourier series of a function converge at a finite jump?
Answers
  1. 40 samples. \( \Omega_0/2\pi = 0.175 = 7/40 \), so \( N = 40 \) (with \( m = 7 \)).
  2. Energy signal, \( E = \int_0^\infty e^{-4t}dt = 1/4 \). Its average power is zero.
  3. The midpoint \( \tfrac{1}{2}[x(t^-) + x(t^+)] \), provided the Dirichlet conditions hold.

Level 2: Part B Style

Practice Q1. The fundamental period of \( x(t) = \sin(4\pi t) + \cos(6\pi t) \) is:

(a) 1/6 s (b) 1/3 s (c) 1/2 s (d) 1 s

Answer: (d) 1 s

Periods are \( 2\pi/4\pi = 1/2 \) s and \( 2\pi/6\pi = 1/3 \) s. The sum repeats at their least common multiple, 1 s (two cycles of the first, three of the second). (a) multiplies the periods instead of taking the LCM; (b) and (c) are the individual periods.

Practice Q2. A square wave alternates between \( +3 \) and \( -3 \) with odd symmetry. The amplitude of its third harmonic is closest to:

(a) 0.42 (b) 1.27 (c) 1.91 (d) 3.82

Answer: (b) 1.27

\( b_3 = \dfrac{4A}{3\pi} = \dfrac{12}{3\pi} = \dfrac{4}{\pi} \approx 1.27 \). (d) is the fundamental, \( 4A/\pi \approx 3.82 \). (c) is \( 2A/\pi \), the fundamental with 2 in place of 4. (a) is \( 4A/(9\pi) \), using the \( 1/n^2 \) decay of a triangle wave.

Practice Q3. Whose Fourier coefficients decrease as \( 1/n^2 \)?

(a) A square wave (b) A sawtooth wave (c) A triangle wave (d) A periodic train of Dirac impulses

Answer: (c)

The triangle wave is continuous with a discontinuous slope, so its coefficients fall as \( 1/n^2 \). Square and sawtooth waves have jumps (\( 1/n \)). An impulse train has coefficients that do not decay at all: every \( c_n = 1/T \).

Practice Q4. The average power of \( x(t) = 3 + 2\cos\omega_0 t + 4\sin 3\omega_0 t \) is:

(a) 10 (b) 14.5 (c) 19 (d) 29

Answer: (c) 19

\( P = 3^2 + \tfrac{1}{2}(2^2 + 4^2) = 9 + 10 = 19 \). (a) drops the DC term. (b) halves the DC term too: \( 4.5 + 10 \). (d) forgets to halve the sinusoids: \( 9 + 4 + 16 \).

Level 3: Part C Style

Practice Q5. A periodic signal of period 2 s is defined by \( x(t) = t \) for \( -1 < t < 1 \). Consider its Fourier series \( \sum b_n\sin(n\pi t) \).

Which of the following is correct?

(a) \( b_2 = +1/\pi \) and the series equals 1 at \( t = 1 \) (b) \( b_2 = -1/\pi \) and the series equals 0 at \( t = 1 \) (c) \( b_2 = -2/\pi \) and the series equals 0 at \( t = 1 \) (d) \( b_2 = 0 \) because the signal is odd

Answer: (b)

With half-period \( L = 1 \): \( b_n = \dfrac{2}{T}\displaystyle\int_{-1}^{1} t\sin(n\pi t)\,dt = \int_{-1}^{1} t\sin(n\pi t)\,dt = \dfrac{2(-1)^{n+1}}{n\pi} \).

So \( b_1 = 2/\pi \approx 0.637 \), \( b_2 = -2/(2\pi) = -1/\pi \approx -0.318 \), \( b_3 = 2/(3\pi) \).

At \( t = 1 \) the periodic extension jumps from \( +1 \) to \( -1 \), so the series converges to the midpoint, 0 (every \( \sin(n\pi) = 0 \) confirms it).

(a) has the sign wrong and ignores the jump. (c) drops the \( 1/n \). (d) confuses “odd function” (which removes cosines) with “no even harmonics” (which needs half-wave symmetry; the sawtooth does not have it).

Practice Q6. Consider these statements about Fourier analysis:

  1. A real periodic signal with half-wave symmetry, \( x(t + T/2) = -x(t) \), has no even harmonics and no DC term.
  2. Adding more terms to the Fourier series of a square wave reduces the Gibbs overshoot to zero.
  3. The DFT of an \( N \)-point sequence has exactly \( N \) distinct coefficients.
  4. Delaying a signal by \( t_0 \) changes its amplitude spectrum by the factor \( \lvert e^{-i2\pi f t_0}\rvert \), which reduces high frequencies.

Which are correct?

(a) 1 and 3 only (b) 1, 2 and 3 only (c) 2 and 4 only (d) 1, 3 and 4 only

Answer: (a) 1 and 3 only
  1. True. For even \( n \), the two half-periods cancel in the coefficient integral; \( n = 0 \) is even, so the mean is zero.
  2. False. The overshoot narrows but tends to about 9% of the jump.
  3. True. \( X[k + N] = X[k] \); only \( k = 0, \dots, N-1 \) are distinct.
  4. False. \( \lvert e^{-i2\pi f t_0}\rvert = 1 \) for every \( f \): a delay changes only the phase.

Practice Q7. A boxcar time window of length \( T = 0.25 \) s is applied to a long monochromatic 40 Hz signal. Which statement about the spectrum of the windowed signal is correct?

(a) It is a single line at 40 Hz, because windowing does not change frequency content (b) It is a sinc centred on \( \pm 40 \) Hz whose main lobe extends from 36 to 44 Hz (c) It is a sinc centred on \( \pm 40 \) Hz whose main lobe extends from 38 to 42 Hz (d) It is a Gaussian centred on 40 Hz with standard deviation 4 Hz

Answer: (b)

Windowing multiplies in time, so it convolves in frequency: the delta at 40 Hz is replaced by the window’s transform, \( T\,\mathrm{sinc}(fT) \), centred at 40 Hz. The first zeros of the sinc are at \( \pm 1/T = \pm 4 \) Hz from the centre, so the main lobe spans 36–44 Hz. (c) uses \( \pm 1/(2T) \). (a) ignores the multiplication theorem. (d) would need a Gaussian window. The side lobes of the sinc are the spectral leakage discussed on the sampling page.


Quick Revision Table

IdeaFact or formula
Energy / power\( E = \int\lvert x\rvert^2dt \); \( P \) = time average of \( \lvert x\rvert^2 \); energy signals have \( P = 0 \)
Even/odd parts\( x_e = [x(t) + x(-t)]/2 \), \( x_o = [x(t) - x(-t)]/2 \)
Discrete periodicity\( \cos\Omega_0 n \) periodic iff \( \Omega_0/2\pi \) rational
Discrete frequency range\( -\pi < \Omega \le \pi \); \( \Omega = \omega\Delta t \)
Fourier coefficients\( c_n = \frac{1}{T}\int_T x\,e^{-in\omega_0 t}dt \); \( c_{-n} = c_n^{*} \) for real \( x \)
Dirichletabsolutely integrable, finite extrema, finite jumps; midpoint at jumps
Symmetryeven: cosines; odd: sines; half-wave: odd harmonics only
Decayjump \( 1/n \); kink \( 1/n^2 \)
Gibbsovershoot about 9% of the jump, does not vanish
Parseval (series)\( P = a_0^2 + \frac{1}{2}\sum(a_n^2 + b_n^2) = \sum\lvert c_n\rvert^2 \)
Square wave\( \frac{4A}{\pi}\sum_{\text{odd }n}\frac{\sin n\omega_0 t}{n} \); \( \sum_{\text{odd}}1/n^2 = \pi^2/8 \)
Boxcar ↔ sincwidth \( T \) ↔ first zeros at \( \pm 1/T \)
Shift\( x(t - t_0) \leftrightarrow X(f)e^{-i2\pi f t_0} \)
Duality rulediscrete in one domain ↔ periodic in the other

Key Takeaways

What’s Next