11

I am trying to figure out what the fourier transform of a constant signal is and for some reason i am coming to the conclusion that the answer is 1. Or better yet a step function.

Micheal C.
  • 123
  • 1
  • 1
  • 5
  • can you use $\LaTeX$ and math notation to spell out what you mean by a "constant signal"? do you mean $$ x(t) = C \qquad \forall t $$ ? – robert bristow-johnson May 04 '17 at 02:48
  • I'm not that good with latex but yes that is what i mean. – Micheal C. May 04 '17 at 03:11
  • 8
    the continuous Fourier Transform of a constant is not 1 (a constant), but is a dirac delta function. $$ \mathscr{F} { C } = C \cdot \delta(f) $$ and that is not 1. – robert bristow-johnson May 04 '17 at 03:36
  • @robertbristow-johnson that's the right and best answer, why not put it below even though it is short? – Dan Boschen May 04 '17 at 21:27
  • 2
    this SE thing actually imposes a length limitation (i forgot how many characters). that's one of the annoying things about SE. – robert bristow-johnson May 05 '17 at 00:46
  • 2
    Huh?${}{}{}{}{}$ – Dilip Sarwate May 05 '17 at 21:46
  • 1
    @robertbristow-johnson Comments need to have at least 15 characters in them, but as my Huh? above illustrates, one can still get around the requirement by surrounding pairs of {} with dollar signs. – Dilip Sarwate May 05 '17 at 21:49
  • 1
    @DilipSarwate, answers must be at least 30 characters. i s'pose my commment was longer than 30 characters. but it was still to trivial to deserve an answer. {} {} what really bothers me is, when i am on the meta site, that SE's bot copy edits my response and, due to terseness or lacking capitals, says that the quality of my answer is not good enough. that's when i would like to kick that bot's metal ass. – robert bristow-johnson May 06 '17 at 01:09

3 Answers3

16

I'll complete a bit the answer given in a comment above.

Intuitively first, to which frequency corresponds a signal constant in time, for exemple $x(t) = 1$ $\forall t$ ? Such a signal shows no variation in time and hence contains only a component with frequency 0 (this is a DC signal). This means that its Fourier transform must be 0 everywhere, except in $f=0$. Mathematically, $$X(f) = \delta(f).$$ Now, can we prove this? Yes, simply take the inverse Fourier transform of $\delta(f)$ and use the properties of the Dirac delta $\delta(f)$ $$x(t) = \int_{-\infty}^\infty \delta(f)e^{j2\pi ft} \mathrm{d}f = \int_{-\infty}^\infty \delta(f) \mathrm{d}f = 1.$$

anpar
  • 957
  • 1
  • 6
  • 15
1

Fourier transforms (they are legion) somehow reflect the amplitude of (complex) sines in data. A flat signal "should" only have non-zero amplitudes on the $0$th frequency, and $0$ amplitude on the others. But what are we calling a flat signal? I will restrict to two common acceptions.

  1. In the continuous time, the signal spreads from $-\infty$ to $\infty$, and a continuous-time Fourier transform naturally transforms this infinite spread into an infinite amplitude at the $0$th frequency, theoretically turned into a distribution, denoted by the Dirac $\delta$ function, as answered by @anpar
  2. In a spatially bounded interval (like a constant-valued image), either continuous or discrete, assuming periodicity to maintain some flatness (using Fourier series or the discrete Fourier transform), you obtain a finite constant at $0$ frequency, and zero elsewhere.

This finite constant depends on how you normalize your Fourier transform.

Finally, on a single-sample signal, the DFT or FFT indeed gives you a constant "Fourier" transform:

fft(1)

ans = 1

Laurent Duval
  • 31,850
  • 3
  • 33
  • 101
1

When you take the Fourier Transform of a constant signal, the result will be a delta function centered at zero frequency. The delta function has a value of the amplitude of the constant signal.

To see why this happens, let's recall the formula for the Fourier Transform of a continuous-time signal $x \left( t \right)$:

$$ X \left( f \right) = \int x \left( t \right) e^{-j 2 \pi f t } dt $$

where $X \left( f \right)$ is the Fourier Transform of $x \left( t \right)$, and $f$ is the frequency variable.

If $x \left( t \right)$ is a constant signal, then $x \left( t \right) = C$ for all $t$, where $C$ is a constant. Substituting this into the above formula, we get:

$$ X \left( f \right) = \int C e^{-j 2 \pi f t} dt \implies X \left( f \right) = C \int e^{-j 2 \pi f t} dt$$

The integral in the above expression is equal to zero except when $f = 0$, in which case it is equal to the integration interval. Assuming that the integration interval is from $- \infty$ to $\infty$, we get:

$$ X \left( f \right) = C \int e^{-j 2 \pi f t} dt \implies X \left( f \right) = C \int e^{0} dt \implies X \left( f \right) = C \left[ t \right]^{\infty}_{- \infty} \implies X \left( f \right) = C \left[ \infty - \left( - \infty \right) \right] \implies X \left( f \right) = C \cdot \infty$$

Therefore, the Fourier Transform of a constant signal $x \left( t \right)$ is a delta function centered at zero frequency with a value of $C$, the amplitude of the constant signal.

In summary, when you take the Fourier Transform of a constant signal, the frequency will be zero, and the result will be a delta function centered at zero frequency with a value equal to the amplitude of the constant signal.

lennon310
  • 3,590
  • 19
  • 24
  • 27
jar123
  • 11
  • 1