Skip to content
Tutorials Center
My learning

Topics

Lesson 13 of 36 ยท LTI Systems and Convolution

Convolution Properties and Difference Equations

Use commutative, associative and distributive laws to combine LTI systems in series and parallel, then connect convolution to difference equations and recursive filters.

Convolution is an operation with algebraic rules, and those rules explain how LTI systems combine. Once you know them, a block diagram with several stages collapses to a single impulse response. This lesson also connects convolution to the other way systems are specified in practice: difference equations (discrete time) and, by analogy, differential equations.

The algebra of convolution

For signals x, h, g and a constant a:

  • Commutative: x * h = h * x. You can flip either signal.
  • Associative: (x * h) * g = x * (h * g).
  • Distributive: x * (h + g) = x * h + x * g.
  • Linear in each argument: (a x) * h = a (x * h).
  • Identity: x * delta = x.
  • Shift: x[n - a] * h[n - b] = y[n - a - b].

Consequences for block diagrams:

flowchart LR
    A["x"] --> H1["h1"] --> H2["h2"] --> B["y"]
    C["x"] --> HS["h1 * h2"] --> D["y"]

Systems in series (cascade) have the impulse response h1 * h2, and by commutativity the order does not matter for LTI blocks.

  • Cascade: h = h1 * h2.
  • Parallel (outputs added): h = h1 + h2.
  • Feedback is not a simple convolution, but it leads to the transfer-function algebra of later chapters.
Python Verifying the convolution laws

Press Run (or Ctrl + Enter) to see the result here.

Program input ยท what you'd type while the program runs

Type the answers before you press Run, one per line. Each time the program reads a line (with input()), it gets the next line from this box.

Cascade and parallel in practice

Two echoes in series produce echoes of echoes: the combined impulse response is the convolution of the two.

Python Series and parallel systems

Press Run (or Ctrl + Enter) to see the result here.

Program input ยท what you'd type while the program runs

Type the answers before you press Run, one per line. Each time the program reads a line (with input()), it gets the next line from this box.

Stability and causality of combinations

If both h1 and h2 are causal, so is h1 * h2. If both are BIBO stable, so is the cascade (since sum |h1 * h2| <= sum |h1| * sum |h2|). Stability of a parallel combination also follows from the two parts.

Difference equations

Many discrete LTI systems are described by a constant-coefficient linear difference equation:

y[n] = b0 x[n] + b1 x[n-1] + ... - a1 y[n-1] - a2 y[n-2] - ...

  • If only inputs appear on the right (all a are zero), the system is non-recursive and its impulse response is just [b0, b1, ...]: an FIR filter.
  • If past outputs appear (recursive), the impulse response typically never ends: an IIR filter.

Example: y[n] = 0.8 y[n-1] + x[n] has impulse response h[n] = 0.8^n u[n]. Try the recursion on a delta and compare with the formula.

Python Recursion produces an exponential impulse response

Press Run (or Ctrl + Enter) to see the result here.

Program input ยท what you'd type while the program runs

Type the answers before you press Run, one per line. Each time the program reads a line (with input()), it gets the next line from this box.

In lfilter(b, a, x) the denominator is a = [1, -0.8] because the equation is written y[n] - 0.8 y[n-1] = x[n]. Remember the sign flip when converting.

Initial conditions and rest

A difference equation alone does not give a unique system: you also need initial conditions. A system is initially at rest if the output is zero before the input starts. LTI behavior (including zero output for zero input) holds only under that assumption; a nonzero initial state adds a zero-input response on top of the zero-state response x * h.

Python Zero-state and zero-input responses

Press Run (or Ctrl + Enter) to see the result here.

Program input ยท what you'd type while the program runs

Type the answers before you press Run, one per line. Each time the program reads a line (with input()), it gets the next line from this box.

Continuous-time counterpart

The continuous analog is a linear constant-coefficient differential equation, for instance RC dy/dt + y = x(t) for the RC circuit. With zero initial conditions it is LTI, and its impulse response is h(t) = (1/RC) e^(-t/RC) u(t). The Laplace transform (later chapter) turns such equations into algebra.

Check your understanding

Quick check ยท 5 questions

of 5 answered You scored / 5 ยท perfect!

  1. 1. Two LTI systems with impulse responses h1 and h2 are connected in series. The overall impulse response is:

    A cascade of LTI systems has the convolution of the individual impulse responses.
  2. 2. Which law lets you swap the order of two LTI systems in a cascade?

    x * h1 * h2 equals x * h2 * h1 because convolution is commutative.
  3. 3. What is the impulse response of y[n] = 0.8 y[n-1] + x[n] (at rest)?

    Feeding a delta, each step multiplies the previous output by 0.8, giving 1, 0.8, 0.64, ...
  4. 4. In scipy's lfilter, what is the a vector for y[n] - 0.5 y[n-1] = x[n]?

    a holds the coefficients of y on the left side: 1 for y[n] and -0.5 for y[n-1].
  5. 5. A nonrecursive difference equation produces:

    With only input terms on the right, the impulse response has finite length.

Summary

  • Convolution is commutative, associative and distributive, with delta as identity.
  • Cascade means h1 * h2; parallel means h1 + h2.
  • Difference equations describe the system recursively; at rest they define a unique LTI system.
  • No past outputs gives FIR; feedback of outputs gives IIR.
  • Initial conditions add a zero-input response to the zero-state response.

Next: Introduction to Fourier Series