3
\$\begingroup\$

How can we represent a time shifted unit impulse function in frequency domain ? Fourier transform of unit impulse function is 1. So I think it will be an infinite parallel line in frequency domain. What about for the time shifted version ?

\$\endgroup\$
1
  • 1
    \$\begingroup\$ \$f(\omega)= cos(\omega T)-j\:sin(\omega T)\$, where the impulse is shifted to \$t=T\$ \$\endgroup\$
    – Chu
    Jan 26 '18 at 15:44
3
\$\begingroup\$

Apply the Time Shifting property of Fourier Transform.

If $$ x(t) \implies X(j\omega) $$ then $$ x(t-a) \implies e^{-j\omega a} X(j\omega)$$ For impulse, $$ \delta (t) \implies1$$ Means, it contains all frequencies and their magnitudes are unity.

For the time shifted version, $$\delta(t-a) \implies e^{-j\omega a}$$

This is a complex number with magnitude = 1. Hence the magnitudes of all the frequency components are still unity in the frequency domain. But their phase angles are different, $$ \phi = -\omega a $$ In the earlier case, the phase \$ \phi \$ was zero for all the frequency components.

\$\endgroup\$
1
\$\begingroup\$

The magnitude of the FFT is 1 for all frequencies, but don't forget that the result is a complex function — there's a phase angle component, too. For the non-shifted pulse, those angles are all zero, but when you time-shift the pulse, the phase angles for different frequencies become different.

\$\endgroup\$
1
\$\begingroup\$

We expect the same magniitude but phase of spectrum for an impulse will integrate down from 0 to -2π at 1/f=T for a given time delay T. It will then be recursive from 0 to -2π repeating at intervals of f=1/T in the spectrum.

We expect the same phase negative integral ramp for a delayed Step Response.

I found (that) understanding this feature in class to be useful 10 yrs later, when I was investigating an EMI problem. (explained below picture)

(real world example)

What if the impulse is just a pulse of some amplitude and some pulse width?

enter image description here

  • It is a step with an delayed inverted step then both repeated
  • Can you see how the spectrum of the pulse drops below -60dB or null here?
  • That is your time delay or width of the pulse. You will not find any sine wave energy at that frequency = f=1/T or any harmonic of that 2/T,3/T.
    • Since my rise time was finite the recursive humps decay with rising f.
    • In the simulation above I used 10 Hz rep. rate and hand drew a pulse on this webpage and displayed the log amplitude and phase of the FFT.
  • in my real world problem, I discovered someone installed an air deionizer in the ceiling producing 30kV DC on the ceiling outside the cleanroom with electrode pins insulated a 2 cm metal cup with alternating impulses to deionize the airborne dust.
    • unfortunately it produced continuous spectrum to up the antenna wavelength of the arc pin gap and then repeated which then interfered with my Servo Writer.
      We had mysterious errors in a delicate magnetic recording machine called a SERVO WRITER that writes impulse currents on a rotating disk later used inside a hard disk drive for servo positioning. When I examined the fault, I recognized stray RF noise impulses in the narrow band frequency (~10MHz) but occurring at a low repetition rate like < 10 pulses per second. I discovered the source of interference outside the Class 100 Clean room from an air deionizer installed in the hallway. It was installed outside the cleanroom at a Burroughs disk drive factory in the mid-80's to reduce the airborne dust contamination outside the room in the hopes that translated to less dust being transported past the air shower and inside the clean room.

Next question ?

\$\endgroup\$

Your Answer

By clicking “Post Your Answer”, you agree to our terms of service, privacy policy and cookie policy

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