Fourier Series and Fourier Transform With easy examples.

The Fourier series expresses any periodic signal as a sum of sines and cosines (harmonics) at multiples of the fundamental frequency. The Fourier transform extends the idea to non-periodic signals, representing them as a continuous spectrum over all frequencies. Together they translate signals from the time domain, where we see waveforms, to the frequency domain, where we see which frequencies are present and how strong each is.

This matters because most real-world processing is frequency-selective: equalizers boost bass and cut treble, radios pick one station's frequency, and noise is often concentrated in frequency bands we can filter out. Designing any of this in the time domain is guesswork; in the frequency domain it is arithmetic. Fourier analysis is the lens that makes frequencies visible.

Key Concepts

• Fundamental frequency: omega0 = 2 pi / T for a signal with period T; all harmonics are integer multiples n x omega0.
• Fourier coefficients: a0/2 is the DC (average) value; an and bn are the cosine and sine amplitudes of the nth harmonic, found by integrating the signal against cos(n omega0 t) and sin(n omega0 t) over one period.
• Exponential form: x(t) = sum of cn x e^(j n omega0 t); cn are complex amplitudes carrying both magnitude and phase.
• Fourier transform pair: X(omega) = integral of x(t) e^(-j omega t) dt; x(t) = (1/2 pi) integral of X(omega) e^(j omega t) d(omega).
• Spectrum: |X(omega)| shows how much of each frequency is present; periodic signals have line spectra (discrete lines), aperiodic signals have continuous spectra.
• Gibbs phenomenon: near a sharp jump, partial sums overshoot by about 9 percent — adding more harmonics never removes it.

Formulas

• Fourier series: x(t) = a0/2 + sum from n=1 to infinity of [an cos(n omega0 t) + bn sin(n omega0 t)].
• an = (2/T) integral over one period of x(t) cos(n omega0 t) dt; bn = (2/T) integral of x(t) sin(n omega0 t) dt.
• Square wave (amplitude A, 50 percent duty): x(t) = (4A/pi)[sin(omega0 t) + (1/3) sin(3 omega0 t) + (1/5) sin(5 omega0 t) + …] — only odd harmonics appear.
• Key properties: time shift x(t - t0) becomes X(omega) e^(-j omega t0); differentiation dx/dt becomes j omega X(omega); convolution in time becomes multiplication in frequency.

Worked Example

A symmetric square wave swings between +1 V and -1 V (amplitude A = 1). Find the amplitudes of its first two nonzero harmonics.

1. The waveform has zero average, so a0 = 0, and half-wave symmetry kills even harmonics.
2. Using the square-wave series with A = 1: fundamental (n = 1) amplitude = 4 x 1 / pi = 1.273 V.
3. Next nonzero harmonic is n = 3: amplitude = 4 / (3 pi) = 0.424 V.
4. The following one (n = 5) would be 4 / (5 pi) = 0.255 V, and so on, falling as 1/n.

Just the first two sine terms, 1.273 sin(omega0 t) + 0.424 sin(3 omega0 t), already capture the squarish shape — and the slow 1/n decay explains why square waves need wide bandwidth: significant energy lives in high harmonics.

Common Misconceptions

• "Fourier series works for any signal." Only periodic ones. Aperiodic signals need the Fourier transform (which you can view as the limit when the period goes to infinity).
• "More harmonics fix the Gibbs overshoot." They squeeze the overshoot closer to the jump, but its ~9 percent height never disappears — it is a mathematical property of the jump, not a computation error.
• "Negative frequencies are physically real." They are a mathematical convenience of the complex exponential form; a real signal's spectrum is symmetric, so positive frequencies tell the whole story.

Real-World Applications

• Audio and music: spectrum analyzers, equalizers and MP3 compression all operate on Fourier spectra.
• Communications: modulation shifts spectra to carrier frequencies; OFDM (used in Wi-Fi and 5G) packs many orthogonal subcarriers together.
• Power systems: harmonic analysis of distorted currents keeps the grid within standards.
• Medical imaging: MRI machines literally sample the Fourier transform of the body and invert it to form images.
• Vibration analysis: bearing faults in motors show up as new peaks in the vibration spectrum.

Fourier series decomposes periodic signals into harmonic sines and cosines; the Fourier transform does the same for aperiodic signals with a continuous spectrum. Time-domain convolution becomes frequency-domain multiplication, which is why filters, modulation and spectral analysis all live in the frequency domain.