Skip to content
Tutorials Center
My learning

Topics

Lesson 11 of 36 ยท LTI Systems and Convolution

The Convolution Sum

Compute the discrete-time convolution sum three ways, by the flip-and-slide method, by a table, and in Python, and learn the output length and edge effects.

The convolution sum is the formula that turns an input and an impulse response into an output:

y[n] = (x * h)[n] = sum over k of x[k] h[n - k].

It looks abstract, but you can calculate it by hand with a few reliable procedures, and you can code it in five lines. This lesson shows the flip-and-slide picture, the shifted-copies picture, a table method, and numerical checks.

Picture 1: flip, slide, multiply, add

To get one output sample y[n]:

  1. Flip h[k] to get h[-k].
  2. Slide it by n to get h[n - k].
  3. Multiply point by point with x[k].
  4. Add all the products.

Repeat for each n. The overlap between x[k] and the moving flipped h changes as n grows, which makes the output rise, plateau and fall.

flowchart LR
    A["h[k]"] -->|"flip"| B["h[-k]"]
    B -->|"slide by n"| C["h[n-k]"]
    C -->|"multiply with x[k], sum"| D["y[n]"]

Do this for every n, and the sequence of sums is the output signal.

Python Flip and slide, one output at a time

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.

Picture 2: shifted copies

The same sum, read column by column: every input sample x[k] launches a scaled, delayed copy of h. Adding the copies gives y. For x = [1, 2, 3] and h = [1, 1, 1, 1]:

k contribution n = 0..5
0 1 * h starting at 0 1 1 1 1 0 0
1 2 * h starting at 1 0 2 2 2 2 0
2 3 * h starting at 2 0 0 3 3 3 3
sum 1 3 6 6 5 3

The output is [1, 3, 6, 6, 5, 3]. This method is fast for hand calculations with short signals.

Output length

If x has Lx samples and h has Lh samples (both starting at n = 0), the output has Lx + Lh - 1 samples. The first output index is the sum of the starting indices and the last is the sum of the end indices. Convolving x on n = 2..5 with h on n = -1..1 gives y on n = 1..6.

Python Shifted copies and the table method

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.

Special inputs

  • Convolving with delta[n] gives the signal back: x * delta = x.
  • Convolving with delta[n - k0] delays: x * delta[n - k0] = x[n - k0].
  • Convolving with a step u[n] produces a running sum.
  • Two decaying exponentials a^n u[n] and b^n u[n] (with a != b) give (a^(n+1) - b^(n+1)) / (a - b) for n >= 0.
Python Convolution with delta, delayed delta and step

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.

Writing your own convolution

A direct implementation uses two loops and costs Lx * Lh multiplications. NumPy's convolve does the same in C. For long signals, FFT-based methods (scipy.signal.fftconvolve) are much faster.

Python Convolution from scratch against the library

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.

Common mistakes

  • Forgetting to flip h (that computes a correlation, not a convolution).
  • Getting the output length wrong; remember Lx + Lh - 1.
  • Using the wrong starting index when the signals do not start at zero.
  • Mixing up x * h (convolution) with x h (point-wise product).

Check your understanding

Quick check ยท 4 questions

of 4 answered You scored / 4 ยท perfect!

  1. 1. x has 5 samples and h has 3 samples. How many samples does x * h have?

    Output length is Lx + Lh - 1 = 5 + 3 - 1 = 7.
  2. 2. x[n] * delta[n - 3] equals:

    Convolving with a shifted impulse shifts the signal by the same amount.
  3. 3. What is [1, 1] * [1, 1]?

    Copies [1, 1] at 0 and [1, 1] at 1 overlap: 1, 1+1, 1.
  4. 4. Which step distinguishes convolution from correlation?

    Convolution reverses h; correlation does not.

Summary

  • y[n] = sum of x[k] h[n - k]: flip, slide, multiply, add.
  • Equivalent view: sum of scaled, delayed copies of h, one per input sample.
  • Output length is Lx + Lh - 1; the start index is the sum of the start indices.
  • Convolving with a delta returns the signal; with a delayed delta, a delayed signal.
  • np.convolve and scipy.signal.fftconvolve compute it numerically.

Next: The Convolution Integral