Introduction to Fourier Series

Fundamentals of Fourier Series What Are Periodic Functions? A periodic function is one that repeats its pattern indefinitely. If $f(t)$ is periodic, then there exists a fixed time interval $T > 0$ (called the period) such that $f(t+T) = f(t)$ for all values of $t$. The period is the time it takes for the function to complete one full cycle. For example, a sine wave with period $T=2\pi$ completes one full oscillation every $2\pi$ units of time. From the period, we can define the fundamental frequency (also called angular frequency): $$\omega0 = \frac{2\pi}{T}$$ This tells us how many radians per unit time the function cycles through. Think of it as measuring "how fast" the oscillation occurs. The Key Idea: Decomposing Periodic Functions Here's the remarkable insight: any reasonable periodic function can be written as a sum of sines and cosines. This is the core idea behind Fourier series. Why is this useful? Because sines and cosines are simple building blocks. If we can decompose a complicated periodic signal into sines and cosines, we can: Understand which frequency components are present Design filters to remove unwanted frequencies Predict behavior and solve equations more easily The Fourier series representation of a periodic function $f(t)$ with period $T$ is: $$f(t) = a0 + \sum{n=1}^{\infty}\bigl[an\cos(n\omega0 t) + bn\sin(n\omega0 t)\bigr]$$ Here's what each piece means: $a0$ is a constant term (the average value of the function) $an$ are coefficients that tell us the "strength" of each cosine component $bn$ are coefficients that tell us the "strength" of each sine component $n=1$ corresponds to the fundamental harmonic — the component oscillating at the fundamental frequency $\omega0$ $n=2,3,\ldots$ correspond to higher harmonics — components oscillating at integer multiples of the fundamental frequency ($2\omega0, 3\omega0$, etc.) Why This Works: Orthogonality The reason we can use sines and cosines as building blocks is that they form an orthogonal set of functions. Orthogonality means that any two different basis functions are "independent" of each other in a mathematical sense. Specifically, over one period $[0,T]$: The integral of the product of two different sine or cosine functions equals zero The integral of the product of a sine and cosine (with different frequencies) equals zero This is crucial because it means each coefficient $an$ and $bn$ captures only the contribution of that specific sinusoidal component — it doesn't get "mixed" with other components. The Projection Interpretation: Computing each Fourier coefficient is analogous to projecting a vector onto a coordinate axis in linear algebra. Just as a vector's component along the $x$-axis tells you how much of the vector points in the $x$-direction, the coefficient $an$ tells you how much of the function $f(t)$ aligns with the cosine component at frequency $n\omega0$. Computing Fourier Coefficients Now that we understand what a Fourier series is, we need to know how to find the coefficients $a0$, $an$, and $bn$ for a given function. Finding $a0$ (The Average Value) The coefficient $a0$ is the simplest — it equals the average value of $f(t)$ over one complete period: $$a0 = \frac{1}{T}\int{0}^{T} f(t)\,dt$$ This makes intuitive sense: it's the "DC offset" or mean level around which the function oscillates. Finding $an$ (Cosine Coefficients) For each integer $n \geq 1$, the cosine coefficient is computed using: $$an = \frac{2}{T}\int{0}^{T} f(t)\cos(n\omega0 t)\,dt$$ Notice the factor of $2/T$ (not $1/T$). This scaling ensures that the magnitude of $an$ correctly represents the strength of that harmonic component. To compute this integral, you multiply the original function by $\cos(n\omega0 t)$ (essentially asking "how much does the function oscillate in sync with this cosine?") and integrate over one period. Finding $bn$ (Sine Coefficients) Similarly, for each integer $n \geq 1$, the sine coefficient is: $$bn = \frac{2}{T}\int{0}^{T} f(t)\sin(n\omega0 t)\,dt$$ The formula is nearly identical to $an$, except we use sine instead of cosine. Key point: If your function $f(t)$ has special symmetry properties, some coefficients may be zero. For example, if $f(t)$ is an even function (symmetric about the vertical axis), all $bn = 0$ and only cosine terms appear. Convergence and Quality of Approximation Does the Fourier Series Actually Converge? A natural question is: does the infinite sum actually equal the original function? The answer depends on the smoothness of $f(t)$. Dirichlet's Convergence Theorem provides conditions: If $f(t)$ is piecewise-continuous, bounded, and has only a finite number of discontinuities in one period, then the Fourier series converges to $f(t)$ at every point where $f(t)$ is smooth (not jumping). Most practical signals satisfy these conditions, so you can generally trust that the Fourier series representation works. At Jump Discontinuities What happens at a point where $f(t)$ suddenly jumps? The Fourier series does something clever: it converges to the average of the left and right limits: $$\text{Fourier series at jump} = \frac{f(t^-) + f(t^+)}{2}$$ This is called the Cesàro value of the discontinuity. So even though the Fourier series can't exactly match a jump, it settles for the value halfway between the two sides. The Gibbs Phenomenon: Overshooting Near Discontinuities Here's something unexpected: near a jump discontinuity, the partial sums (truncated Fourier series) exhibit oscillations and overshoot the function. This is the Gibbs phenomenon. The image above shows this beautifully. As you include more and more terms (higher rows), the approximation improves, but notice how it always overshoots slightly near the discontinuity and then undershoots. This overshoot is not an error in our calculation—it's an inherent feature of trying to approximate a discontinuous function with smooth sine and cosine curves. The overshoot amounts to approximately 9% above and below the jump, regardless of how many terms you include. However, the region where this overshoot occurs shrinks as you add more terms, so the error becomes more localized. Truncating the Series for Practical Approximation In practice, you almost never keep all infinitely many terms. Instead, you truncate the series by keeping only the first few terms: $$f(t) \approx a0 + \sum{n=1}^{N}\bigl[an\cos(n\omega0 t) + bn\sin(n\omega0 t)\bigr]$$ The question is: how many terms do you need? The answer depends on how much accuracy you need. The good news: the first few terms often capture most of the behavior. For smooth functions, the coefficients decay rapidly, so truncating at $N=5$ or $N=10$ might give you a very good approximation. For functions with sharp corners or discontinuities, you may need more terms to accurately represent the jumps. This is why Fourier series are powerful in applications — a complicated signal often needs only a handful of sinusoidal components to reconstruct. <extrainfo> Applications of Fourier Series Signal Processing and Filtering In audio and communications, periodic signals are naturally decomposed into frequencies using Fourier series. By examining the coefficients $an$ and $bn$, engineers can see which frequencies are strong and which are weak. Filters can then be designed to suppress unwanted frequencies (like noise at 60 Hz from electrical power lines) while preserving desired signal frequencies. Vibrations and Modal Analysis Mechanical systems (bridges, buildings, machinery) vibrate at natural frequencies. Fourier series allows engineers to decompose complex vibrations into individual harmonic modes, making it easier to identify resonances and design dampers to prevent failure. Heat Conduction and Differential Equations When solving the heat equation for periodic boundary conditions, the solution naturally involves a Fourier series. This is one of the classic applications in mathematical physics. Connection to the Fourier Transform The Fourier series works for periodic functions. The Fourier transform is a natural extension to non-periodic signals—it shows that any signal can be viewed as a superposition of sinusoids, just at a continuum of frequencies rather than discrete harmonics. This is the foundation of modern signal processing, audio compression, wireless communications, and many other technologies. </extrainfo>