0
$\begingroup$

The CTFT of a signal as a function of $ \omega $ and $ f $ is identical:

$$ X(j\omega) = \int_{-\infty}^{\infty} x(t) \, e^{-j \omega t} \operatorname{dt} \;\;\;\;\bigg|\;\;\;\; X(j f) = \int_{-\infty}^{\infty} x(t) \, e^{-j 2 \pi f t} \operatorname{dt} $$

But the Inverse CTFT has a scaling factor of $ 1 / (2 \pi) $ if $ f(t) $ is transformed from $ X(\omega) $:

$$ f(t) = {1 \over {2\pi}} \int_{-\infty}^{\infty} X(j\omega) \, e^{j \omega t} \operatorname{d\omega} \;\;\;\;\bigg|\;\;\;\; f(t) = \int_{-\infty}^{\infty} X(j f) \, e^{j 2 \pi f t} \operatorname{df} $$

1. What is the conceptual implication of having the scaling factor?

Say if if I have a signal that is a pure sinusoid ( $ 1V $ in voltage amplitude and $ 60 \operatorname{Hz} $ in frequency):

$$ x(t) = \cos(2\pi \cdot 60t) $$

The spectrums between $ X(j\omega) $ and $ X(jf) $ is practically identical

enter image description here

If I convert the $ X(\omega)$ and $ X(f) $ spectrum back into its time domain function $ f(t) $, I yield different results. I can reconstruct $ x(t) $ from $ X(jf) $ but there is this (attenuating) scaling factor of $ 1 / (2\pi) $ from $ X(j\omega) $:

$$ \begin{align} \text{from } X(\omega) \implies f(t) &= \frac{1}{2\pi} \big[ 0.5e^{j2\pi60t} + 0.5e^{-j2\pi60t} \big] = \frac{1}{2\pi} \big[ 0.5\cos(2\pi60t) + 0.5\cos(2\pi60t) \big] \\ &= \frac{1}{2\pi} \cdot \cos(2 \pi 60t) \neq x(t) \\ &\implies 0.16V \text{ in voltage amplitude and frequency of } 60 \operatorname{Hz} \end{align} $$

whereas,

$$ \begin{align} \text{from } X(f) \implies f(t) &= 0.5e^{j2\pi60t} + 0.5e^{-j2\pi60t} = 0.5\cos(2\pi60t) + 0.5\cos(2\pi60t) \\ &= \cos(2 \pi 60t) = x(t) \end{align} $$

2. Does that imply that I have to normalized whichever value I read on a $ X(\omega) $ vs $ \omega $ plot by multiplying it by $ 2\pi $ or else the reconstructed voltage amplitude would be lower than what it really is?

$\endgroup$
1
  • $\begingroup$ Reason for the downvote? $\endgroup$
    – KMC
    Oct 29, 2021 at 2:47

1 Answer 1

1
$\begingroup$

Your $X(\omega)$ is incorrect. It should use

$$e^{j2\times\pi\times 60 t} \leftrightarrow 2\pi \delta(\omega-2\times\pi\times 60).$$ See this table.

See this table for the same thing in $f$.

$\endgroup$
5
  • $\begingroup$ $ X(\omega) $ has two spectral lines each with amplitude 0.5. How do I write it in terms of $ 2\pi \delta(\omega-2\pi 60 $? $$ X(\omega) = 0.5 \cdot 2\pi \delta(\omega -2\pi 60) \text{???} $$ But that would give me $$ 0.5 \cdot 2\pi \cdot 2 = 2\pi V $$ $ \approx $ six times the voltage magnitude of the original signal ($ 1V$) $\endgroup$
    – KMC
    Oct 28, 2021 at 18:31
  • $\begingroup$ @KMC Sorry! My bad. I updated my answer to use the correct Fourier transform pair. Yours will have a positive and negative frequency exponential with a coefficient of 0.5. So each will generate a term of $\pi\delta(\omega-2\times\pi\times 60)$. $\endgroup$
    – Peter K.
    Oct 28, 2021 at 19:06
  • $\begingroup$ I'm still not getting this ... so $$ x(t) = 0.5[\pi \delta(\omega-2 \pi 60)t] + 0.5[\pi \delta(-\omega-2\pi60)t] $$ does not look like $$ x(t) = \cos(2\pi 60t) $$ $\endgroup$
    – KMC
    Oct 29, 2021 at 1:53
  • $\begingroup$ @KMC But it is precisely that. That's what the Fourier transform tables I've posted say. $$0.5[2\pi \delta(\omega-2 \pi 60)] + 0.5[2\pi \delta(-\omega-2\pi60)] \leftrightarrow \cos(2\pi 60t)$$. $\endgroup$
    – Peter K.
    Oct 29, 2021 at 2:32
  • $\begingroup$ I always thought (and I was taught that) the height of the spectral line in $ X(\omega) $ IS the value of the amplitude of the time-domain sinusoid of a corresponding $ \omega $, but it is wrong. So if I want to correlated the scalar value of the transformed function, I really need to normalize height by dividing the value by $ 2\pi $. If I plot my spectral lines in $ X(\omega) $ I have two impulse each with height $ 0.5 \cdot 2\pi $. Whereas in $ X(f) $ their impulse height are exactly $ 0.5 $ which sum up to $ 1V $ for the sinusoid in the time domain. $\endgroup$
    – KMC
    Oct 29, 2021 at 2:45

Your Answer

By clicking “Post Your Answer”, you agree to our terms of service and acknowledge that you have read and understand our privacy policy and code of conduct.

Not the answer you're looking for? Browse other questions tagged or ask your own question.