Skip to content
Tutorials Center
My learning

Topics

Lesson 19 of 36 · Fourier Transform

The Continuous-Time Fourier Transform

Learn the continuous-time Fourier transform: the analysis and synthesis integrals, magnitude and phase spectra, existence conditions, and numerical checks in Python.

A Fourier series describes a signal that repeats forever as a sum of harmonics. Most real signals do not repeat: a clap, a single heartbeat, a radar pulse. The Fourier transform extends the idea to any signal that dies out (or is at least well behaved) by letting the spacing between harmonics shrink to zero. The sum of discrete harmonics turns into an integral over a continuous range of frequencies.

In this lesson you meet the transform pair, learn how to read a spectrum, see when the integral exists, and check two classic results numerically. If you worked through the Fourier series chapter, the formulas will feel familiar; the main change is that coefficients become a function of frequency.

The transform pair

The Fourier transform of a signal x(t) is

X(f) = ∫ x(t) e^(-j2πft) dt, integrated over all time.

The inverse transform rebuilds the signal:

x(t) = ∫ X(f) e^(j2πft) df, integrated over all frequencies.

Here f is in hertz. Many textbooks use the angular frequency ω = 2πf, which moves a factor of 1/(2π) into the inverse formula. Both conventions describe the same thing; this course uses hertz because it is easier to relate to a spectrum analyzer.

From series to transform

Fourier series Periodic signal, discrete harmonics
Longer period Harmonics move closer together
Infinite period Harmonics fill a continuum
Fourier transform Spectrum X(f) for any signal

The transform is the series with the period pushed to infinity.

1 / 2

Magnitude and phase

X(f) is generally complex. Write it as |X(f)| e^(j∠X(f)). The magnitude tells you how strongly the frequency f is present; the phase tells you how the component is shifted in time. For a real signal the spectrum has conjugate symmetry: X(-f) is the complex conjugate of X(f). So the magnitude is even and the phase is odd, and you only need to inspect positive frequencies.

When does the transform exist?

A sufficient condition (the Dirichlet condition) is that x(t) is absolutely integrable, meaning the integral of |x(t)| is finite, plus finitely many discontinuities and extrema in any finite interval. Decaying exponentials, pulses and Gaussians qualify. Signals that do not decay, like a constant or a sine wave, are not absolutely integrable. Their transforms exist only with impulses (a constant maps to δ(f)); you meet those in transform pairs and Parseval's theorem.

A one-sided exponential

Take x(t) = e^(-at) u(t) with a > 0, where u(t) is the unit step. Then

X(f) = ∫ from 0 to ∞ of e^(-(a + j2πf)t) dt = 1 / (a + j2πf).

The magnitude is 1/√(a² + (2πf)²). It is largest at f = 0 and falls as f grows: a smooth, low-pass shape. The numerical version below approximates the integral with a Riemann sum on a very fine grid and compares it with the formula.

Python Numerical Fourier transform of an exponential

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

Program input · 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.

A rectangular pulse and the sinc function

A pulse of height 1 and width T centered at zero is the most important transform pair in the course. Its transform is

X(f) = T · sinc(fT), where sinc(u) = sin(πu) / (πu).

The spectrum is real (the pulse is even), crosses zero at multiples of 1/T, and decays slowly because the pulse has sharp edges. Narrow pulses in time produce wide spectra, and wide pulses produce narrow spectra. That inverse relationship is the heart of time-frequency trade-offs.

Python Pulse width and its spectrum

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

Program input · 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.

Note that np.sinc is the normalized sinc, sin(πu)/(πu), which matches the definition above.

Signal energy is finite for the signals in this lesson, so they are called energy signals. Parseval's theorem relates that energy in time to the energy in frequency; you meet it later in this chapter.

Common mistakes

  • Forgetting the 2π: if you use ω, the exponent is e^(-jωt) and the inverse transform carries 1/(2π).
  • Plotting only the real part. Always look at magnitude and phase.
  • Expecting a pure sine to have a finite-valued transform. It has impulses.

Check your understanding

Quick check · 4 questions

of 4 answered You scored / 4 · perfect!

  1. 1. What is the Fourier transform of x(t) = e^(-at) u(t) with a > 0?

    Integrating e^(-(a + j2πf)t) from 0 to infinity gives 1 / (a + j2πf). The other options either drop the complex part or flip its sign.
  2. 2. A real signal has X(3) = 2 + 5j. What is X(-3)?

    Real signals have conjugate-symmetric spectra, so X(-f) is the complex conjugate of X(f).
  3. 3. What happens to the main lobe of a pulse spectrum when you make the pulse wider in time?

    Time and frequency widths are inversely related. The first zero sits at f = 1/T, so a larger T moves it closer to zero.
  4. 4. Which signal is NOT absolutely integrable, so its transform needs impulses?

    A constant never decays, so the integral of |x(t)| diverges. Its transform is an impulse at f = 0.

Summary

  • The Fourier transform X(f) = ∫ x(t) e^(-j2πft) dt describes the frequency content of non-periodic signals.
  • The inverse transform rebuilds x(t) from X(f).
  • Real signals have even magnitude and odd phase spectra.
  • e^(-at) u(t) maps to 1 / (a + j2πf); a pulse of width T maps to T sinc(fT).
  • Narrow in time means wide in frequency, and the reverse.

Next: Properties of the Fourier Transform