Convolution Theorem and Applications

Introduction to Convolution

In mathematics, particularly in the field of signal processing and differential equations, convolution is a fundamental operation. It describes how the shape of one function is modified by another. Mathematically, the convolution of two functions, say $f(t)$ and $g(t)$, denoted by $(f * g)(t)$, is defined by the integral:

$(f * g)(t) = \int_{-\infty}^{\infty} f(\tau) g(t - \tau) d\tau$

This integral represents the weighted average of the first function $f(\tau)$ as it is shifted and flipped, with the weights given by the second function $g(t - \tau)$. The variable $\tau$ is a dummy integration variable. For functions defined on $[0, \infty)$, the integral is typically from $0$ to $t$:

$(f * g)(t) = \int_{0}^{t} f(\tau) g(t - \tau) d\tau$

Convolution is a commutative operation, meaning $f * g = g * f$. It is also associative and distributive.

The Convolution Theorem for Laplace Transforms

The Convolution Theorem is a powerful tool that relates the convolution of two functions in the time domain to the product of their Laplace transforms in the frequency (s-domain). This theorem significantly simplifies the process of finding the Laplace transform of a convolution and, more importantly, finding the inverse Laplace transform of a product of two functions.

The theorem states that if $F(s) = \mathcal{L}\{f(t)\}$ and $G(s) = \mathcal{L}\{g(t)\}$, then the Laplace transform of the convolution of $f(t)$ and $g(t)$ is the product of their individual Laplace transforms:

$\mathcal{L}\{(f * g)(t)\} = F(s)G(s)$

Conversely, and perhaps more practically for solving problems, the inverse Laplace transform of a product of two functions in the s-domain is the convolution of their inverse Laplace transforms:

$\mathcal{L}^{-1}\{F(s)G(s)\} = (f * g)(t) = \int_{0}^{t} f(\tau) g(t - \tau) d\tau$

This inverse relationship is crucial for solving many types of problems, especially linear ordinary differential equations with constant coefficients and initial conditions.

Proof of the Convolution Theorem (Optional but Insightful)

Let's consider the Laplace transform of the convolution $(f * g)(t)$:

$\mathcal{L}\{(f * g)(t)\} = \int_{0}^{\infty} e^{-st} \left( \int_{0}^{t} f(\tau) g(t - \tau) d\tau \right) dt$

This is a double integral. We can change the order of integration. The region of integration is defined by $0 \le \tau \le t$ and $0 \le t < \infty$. This region in the $\tau-t$ plane is a triangle with vertices at (0,0), (0,0), and (infinity, infinity). When we change the order of integration, the limits become $0 \le \tau < \infty$ and $\tau \le t < \infty$.

So, the integral becomes:

$\int_{0}^{\infty} f(\tau) \left( \int_{\tau}^{\infty} e^{-st} g(t - \tau) dt \right) d\tau$

Now, let's make a substitution in the inner integral. Let $u = t - \tau$. Then $t = u + \tau$, and $dt = du$. When $t = \tau$, $u = 0$. When $t \to \infty$, $u \to \infty$.

The inner integral becomes:

$\int_{0}^{\infty} e^{-s(u + \tau)} g(u) du = e^{-s\tau} \int_{0}^{\infty} e^{-su} g(u) du$

Recognize that $\int_{0}^{\infty} e^{-su} g(u) du$ is the Laplace transform of $g(u)$, which is $G(s)$.

Substituting this back into the outer integral:

$\int_{0}^{\infty} f(\tau) e^{-s\tau} G(s) d\tau$

Since $G(s)$ is independent of $\tau$, we can take it out of the integral:

$G(s) \int_{0}^{\infty} e^{-s\tau} f(\tau) d\tau$

The integral $\int_{0}^{\infty} e^{-s\tau} f(\tau) d\tau$ is the Laplace transform of $f(\tau)$, which is $F(s)$.

Therefore, we have:

$\mathcal{L}\{(f * g)(t)\} = G(s)F(s) = F(s)G(s)$

This completes the proof.

Applications of the Convolution Theorem

The Convolution Theorem is exceptionally useful in several areas:

1. Solving Linear Ordinary Differential Equations (ODEs): It provides a systematic way to find the particular solution of non-homogeneous linear ODEs.
2. Analyzing Systems: In control systems and signal processing, the output of a linear time-invariant (LTI) system is the convolution of the input signal with the system's impulse response. The theorem simplifies analysis in the frequency domain.
3. Calculating Inverse Laplace Transforms: When a function in the s-domain is a product of two simpler functions whose inverse transforms are known, the convolution theorem can be used to find the inverse transform.

Application 1: Solving ODEs using Convolution

Consider a second-order linear ODE with constant coefficients:

$ay'' + by' + cy = r(t)$

with initial conditions $y(0) = y_0$ and $y'(0) = y'_0$.

Taking the Laplace transform of both sides:

$a[s^2 Y(s) - sy(0) - y'(0)] + b[s Y(s) - y(0)] + c Y(s) = R(s)$

Substituting initial conditions:

$a[s^2 Y(s) - sy_0 - y'_0] + b[s Y(s) - y_0] + c Y(s) = R(s)$

Rearranging to solve for $Y(s)$:

$(as^2 + bs + c) Y(s) = R(s) + a(sy_0 + y'_0) + by_0$

$Y(s) = \frac{R(s)}{as^2 + bs + c} + \frac{a(sy_0 + y'_0) + by_0}{as^2 + bs + c}$

Let $H(s) = \frac{1}{as^2 + bs + c}$. This $H(s)$ is the transfer function of the system. Its inverse Laplace transform, $h(t) = \mathcal{L}^{-1}\{H(s)\}$, is the impulse response of the system (assuming zero initial conditions for the homogeneous part).

Then, the first term can be written as $Y_{particular}(s) = R(s) H(s)$. By the Convolution Theorem, the inverse Laplace transform of this term is:

$y_{particular}(t) = \mathcal{L}^{-1}\{R(s) H(s)\} = (r * h)(t) = \int_{0}^{t} r(\tau) h(t - \tau) d\tau$

The second term represents the response due to the initial conditions, which can be found using partial fraction decomposition or other standard inverse Laplace transform techniques.

The total solution is $y(t) = y_{homogeneous}(t) + y_{particular}(t)$. The convolution integral gives the particular solution directly.

Example ODE Problem:

Solve the ODE $y'' + 4y = \sin(3t)$ with $y(0) = 0$ and $y'(0) = 0$.

Take Laplace transform:

$s^2 Y(s) - sy(0) - y'(0) + 4 Y(s) = \mathcal{L}\{\sin(3t)\}$

$s^2 Y(s) + 4 Y(s) = \frac{3}{s^2 + 9}$

$(s^2 + 4) Y(s) = \frac{3}{s^2 + 9}$

$Y(s) = \frac{3}{(s^2 + 4)(s^2 + 9)}$

Here, $R(s) = \mathcal{L}\{\sin(3t)\} = \frac{3}{s^2 + 9}$ and $H(s) = \frac{1}{s^2 + 4}$.

We need to find the inverse Laplace transform of $Y(s) = R(s)H(s)$.

First, find the inverse Laplace transforms of $R(s)$ and $H(s)$:

$r(t) = \mathcal{L}^{-1}\{R(s)\} = \mathcal{L}^{-1}\left\{\frac{3}{s^2 + 9}\right\} = \sin(3t)$

$h(t) = \mathcal{L}^{-1}\{H(s)\} = \mathcal{L}^{-1}\left\{\frac{1}{s^2 + 4}\right\} = \frac{1}{2}\mathcal{L}^{-1}\left\{\frac{2}{s^2 + 2^2}\right\} = \frac{1}{2}\sin(2t)$

Now, use the Convolution Theorem to find $y(t) = \mathcal{L}^{-1}\{R(s)H(s)\} = (r * h)(t)$:

$y(t) = \int_{0}^{t} r(\tau) h(t - \tau) d\tau = \int_{0}^{t} \sin(3\tau) \left(\frac{1}{2}\sin(2(t - \tau))\right) d\tau$

$y(t) = \frac{1}{2} \int_{0}^{t} \sin(3\tau) \sin(2t - 2\tau) d\tau$

Use the trigonometric identity: $\sin A \sin B = \frac{1}{2}[\cos(A-B) - \cos(A+B)]$

Let $A = 3\tau$ and $B = 2t - 2\tau$.

$A - B = 3\tau - (2t - 2\tau) = 5\tau - 2t$ $A + B = 3\tau + (2t - 2\tau) = \tau + 2t$

So, $\sin(3\tau) \sin(2t - 2\tau) = \frac{1}{2}[\cos(5\tau - 2t) - \cos(\tau + 2t)]$

$y(t) = \frac{1}{2} \int_{0}^{t} \frac{1}{2}[\cos(5\tau - 2t) - \cos(\tau + 2t)] d\tau$

$y(t) = \frac{1}{4} \int_{0}^{t} [\cos(5\tau - 2t) - \cos(\tau + 2t)] d\tau$

Integrate with respect to $\tau$:

$y(t) = \frac{1}{4} \left[ \frac{\sin(5\tau - 2t)}{5} - \frac{\sin(\tau + 2t)}{1} \right]_{0}^{t}$

Evaluate at the limits:

At $\tau = t$: $\frac{1}{4} \left[ \frac{\sin(5t - 2t)}{5} - \frac{\sin(t + 2t)}{1} \right] = \frac{1}{4} \left[ \frac{\sin(3t)}{5} - \sin(3t) \right]$

At $\tau = 0$: $\frac{1}{4} \left[ \frac{\sin(-2t)}{5} - \frac{\sin(2t)}{1} \right] = \frac{1}{4} \left[ -\frac{\sin(2t)}{5} - \sin(2t) \right]$

Subtract the lower limit from the upper limit:

$y(t) = \frac{1}{4} \left[ \left(\frac{\sin(3t)}{5} - \sin(3t)\right) - \left(-\frac{\sin(2t)}{5} - \sin(2t)\right) \right]$

$y(t) = \frac{1}{4} \left[ \frac{\sin(3t) - 5\sin(3t)}{5} + \frac{\sin(2t) + 5\sin(2t)}{5} \right]$

$y(t) = \frac{1}{4} \left[ \frac{-4\sin(3t)}{5} + \frac{6\sin(2t)}{5} \right]$

$y(t) = \frac{1}{20} (-4\sin(3t) + 6\sin(2t))$

$y(t) = \frac{1}{10} (3\sin(2t) - 2\sin(3t))$

This is the particular solution. Since initial conditions were zero, this is the complete solution.

Application 2: Analyzing LTI Systems

A linear time-invariant (LTI) system can be characterized by its impulse response $h(t)$. If an input signal is $x(t)$, the output signal $y(t)$ is given by the convolution of the input and the impulse response:

$y(t) = (x * h)(t) = \int_{0}^{t} x(\tau) h(t - \tau) d\tau$

Using the Laplace transform and the Convolution Theorem:

$Y(s) = \mathcal{L}\{y(t)\} = \mathcal{L}\{x(t) * h(t)\} = X(s) H(s)$

Here, $X(s) = \mathcal{L}\{x(t)\}$ is the Laplace transform of the input signal, and $H(s) = \mathcal{L}\{h(t)\}$ is the transfer function of the system. The theorem states that the output in the s-domain is simply the product of the input's transform and the system's transfer function. This is a fundamental concept in system analysis, allowing engineers to analyze system behavior in the frequency domain, which is often simpler than time-domain analysis. For example, stability and frequency response can be easily determined from $H(s)$.

Example System Analysis:

Consider an LTI system with impulse response $h(t) = e^{-2t}u(t)$, where $u(t)$ is the unit step function. Let the input signal be $x(t) = e^{-t}u(t)$. Find the output $y(t)$.

First, find the Laplace transforms of $x(t)$ and $h(t)$:

$X(s) = \mathcal{L}\{e^{-t}u(t)\} = \frac{1}{s+1}$ for $s > -1$.

$H(s) = \mathcal{L}\{e^{-2t}u(t)\} = \frac{1}{s+2}$ for $s > -2$.

The output in the s-domain is $Y(s) = X(s)H(s)$:

$Y(s) = \frac{1}{s+1} \cdot \frac{1}{s+2} = \frac{1}{(s+1)(s+2)}$

To find $y(t)$, we need to find the inverse Laplace transform of $Y(s)$. We can use partial fraction decomposition for $Y(s)$:

$\frac{1}{(s+1)(s+2)} = \frac{A}{s+1} + \frac{B}{s+2}$

$1 = A(s+2) + B(s+1)$

If $s = -1$, then $1 = A(-1+2) + B(0) \implies 1 = A$.

If $s = -2$, then $1 = A(0) + B(-2+1) \implies 1 = -B \implies B = -1$.

So, $Y(s) = \frac{1}{s+1} - \frac{1}{s+2}$.

Now, take the inverse Laplace transform:

$y(t) = \mathcal{L}^{-1}\left\{\frac{1}{s+1}\right\} - \mathcal{L}^{-1}\left\{\frac{1}{s+2}\right\}$

$y(t) = e^{-t}u(t) - e^{-2t}u(t)$

$y(t) = (e^{-t} - e^{-2t})u(t)$

Alternatively, we could have used the convolution integral directly:

$y(t) = \int_{0}^{t} x(\tau) h(t - \tau) d\tau = \int_{0}^{t} e^{-\tau} e^{-2(t - \tau)} d\tau$

$y(t) = \int_{0}^{t} e^{-\tau} e^{-2t} e^{2\tau} d\tau = e^{-2t} \int_{0}^{t} e^{\tau} d\tau$

$y(t) = e^{-2t} [e^{\tau}]_{0}^{t} = e^{-2t} (e^t - e^0) = e^{-2t} (e^t - 1)$

$y(t) = e^{-t} - e^{-2t}$

This matches the result obtained using the s-domain approach.

Application 3: Calculating Inverse Laplace Transforms of Products

Sometimes, we encounter Laplace transforms of the form $F(s)G(s)$, where $f(t) = \mathcal{L}^{-1}\{F(s)\}$ and $g(t) = \mathcal{L}^{-1}\{G(s)\}$ are known functions. The Convolution Theorem provides a direct method to find $\mathcal{L}^{-1}\{F(s)G(s)\}$.

$\mathcal{L}^{-1}\{F(s)G(s)\} = (f * g)(t) = \int_{0}^{t} f(\tau) g(t - \tau) d\tau$

This is particularly useful when partial fraction decomposition is difficult or leads to complex terms.

Example Inverse Transform:

Find the inverse Laplace transform of $F(s) = \frac{1}{s^2-1}$ and $G(s) = \frac{1}{s-2}$.

We know $f(t) = \mathcal{L}^{-1}\left\{\frac{1}{s^2-1}\right\} = \sinh(t)$.

And $g(t) = \mathcal{L}^{-1}\left\{\frac{1}{s-2}\right\} = e^{2t}$.

Using the Convolution Theorem, $\mathcal{L}^{-1}\{F(s)G(s)\} = (f * g)(t)$:

$(f * g)(t) = \int_{0}^{t} \sinh(\tau) e^{2(t - \tau)} d\tau$

$(f * g)(t) = \int_{0}^{t} \sinh(\tau) e^{2t} e^{-2\tau} d\tau = e^{2t} \int_{0}^{t} \sinh(\tau) e^{-2\tau} d\tau$

Recall $\sinh(\tau) = \frac{e^{\tau} - e^{-\tau}}{2}$.

$(f * g)(t) = e^{2t} \int_{0}^{t} \frac{e^{\tau} - e^{-\tau}}{2} e^{-2\tau} d\tau$

$(f * g)(t) = \frac{e^{2t}}{2} \int_{0}^{t} (e^{\tau} e^{-2\tau} - e^{-\tau} e^{-2\tau}) d\tau$

$(f * g)(t) = \frac{e^{2t}}{2} \int_{0}^{t} (e^{-\tau} - e^{-3\tau}) d\tau$

$(f * g)(t) = \frac{e^{2t}}{2} \left[ -e^{-\tau} + \frac{1}{3}e^{-3\tau} \right]_{0}^{t}$

$(f * g)(t) = \frac{e^{2t}}{2} \left[ \left(-e^{-t} + \frac{1}{3}e^{-3t}\right) - \left(-e^{0} + \frac{1}{3}e^{0}\right) \right]$

$(f * g)(t) = \frac{e^{2t}}{2} \left[ -e^{-t} + \frac{1}{3}e^{-3t} - \left(-1 + \frac{1}{3}\right) \right]$

$(f * g)(t) = \frac{e^{2t}}{2} \left[ -e^{-t} + \frac{1}{3}e^{-3t} - \left(-\frac{2}{3}\right) \right]$

$(f * g)(t) = \frac{e^{2t}}{2} \left[ -e^{-t} + \frac{1}{3}e^{-3t} + \frac{2}{3} \right]$

Distribute $e^{2t}$:

$(f * g)(t) = \frac{1}{2} \left[ -e^{2t}e^{-t} + \frac{1}{3}e^{2t}e^{-3t} + \frac{2}{3}e^{2t} \right]$

$(f * g)(t) = \frac{1}{2} \left[ -e^{t} + \frac{1}{3}e^{-t} + \frac{2}{3}e^{2t} \right]$

This is the inverse Laplace transform of $\frac{1}{(s^2-1)(s-2)}$.

Summary of Convolution Theorem

The Convolution Theorem is a cornerstone of Laplace transform theory, bridging the gap between time-domain operations (convolution) and frequency-domain operations (multiplication).

Key takeaways:

• $\mathcal{L}\{(f * g)(t)\} = F(s)G(s)$
• $\mathcal{L}^{-1}\{F(s)G(s)\} = (f * g)(t) = \int_{0}^{t} f(\tau) g(t - \tau) d\tau$
• It simplifies solving linear ODEs with constant coefficients, especially for finding particular solutions.
• It is fundamental in the analysis of LTI systems, relating input, output, and impulse response in the Laplace domain.
• It provides an alternative method for finding inverse Laplace transforms of products of functions.

Mastering the Convolution Theorem and its applications is essential for advanced studies in differential equations, control theory, and signal processing.