How to obtain IFFT for a set of experimental data in a text file?

I have the data taken from spectrum analyzer in a text file. (This is a frequency domain signal) I want to get the time domain signal of this signal using IFFT. Can anyone explain me the procedure to get the time domain signal? I have attached the text file with this.

[Star Strider]

It depends what the three columns represent. If you have both amplitude and phase, you can likely reconstruct the complex spectrum from a one-sided Fourier ttransform, and then compute its complex conjugate, concatenate them, and invert it. The frequency axis must be continuous from zero to the Nyquist frequency.

If the data do not meet those conditions, then it is not possible to recover the time domain signal.

[ridma kaveendra]

Actually I only need the 2nd and 3rd coulumn data. Second column has the frequency data (from 1.5GHz to 2.5GHz) and the third column has the amplitute. I have attached the plot of those two columns with this. will these data be enough to get the time domain signal?

[Star Strider]

No.

You need the phase as well in order to reconstruct the complex Fourier transform.

[ridma kaveendra]

i dont understant, this is a 2GHz sine wave signal generated from a signal generator and it is viewed using a spectrum analyzer. only one sinusoidal wave so what is the relation to phase with this? look below and you can see the original sine wave which i am talking about. the data in above excel.txt are the frequancy domain data of that signal. can you explain the relation of phase between these two

[Star Strider]

The Fourier transform produces a complex frequency-domain representation of the time-domain signal. That result can be considered as a complex quantity, or the magnitude and phase, however you need both components to reconstruct the signal. That is simply how the transforms work.

[ridma kaveendra]

Assuming the phase is zero, can you show me how to obtain the time domain signal sir?

[Walter Roberson]

The code I posted does that.

[ridma kaveendra]

That code output is not a sine wave, does that mean phase can not be zero?

[Walter Roberson]

Padding with zeros is a convolution that often ends up creating a sync . Padding with zeros was needed because of the missing frequencies.

Answer 1

[Walter Roberson]

You can zero-pad the 900 entries before x = 1500000000 . You can then take the complex conjugate of the augmented y, and flipud() it, append the two, and prefix with a zero.

Now you would have a two-sided complex conjugate that you can use ifft() on to get the time domain signal.

The 1666666 sampling interval affects the labeling of the the time domain signals, but not their relative value.

If your frequency bins were not all 1666666 or 1666667 apart, then you would need to use other approaches.

[ridma kaveendra]

I got the conjuage of y and append the two. I dont understand the next steps? also when I looked at the matrices conj(y) is equal to y. Is that correct? I have attched the modified code.

[Walter Roberson]

y_extended = [0; y; flipud(conj(y))];
yt = ifft(y_extended);

figure()
plot(yt)

When I look at that, and zoom in on points 550 to 650, I find that I doubt that the original data represents what you are saying it represents.

[ridma kaveendra]

This is what i am doing. I am generating a signal and connect it to oscilloscope(time domain signal) and spectrum analyzer(frequency domain signal). below is the signal taken from oscilloscope. What I need to do is to obtain this waveform by using inverse fft on spectrum analyzer waveform. The attched file is the spectrum analyzer data(frequency(from 0 to 3GHz) and amplitude).

[Walter Roberson]

The file does not contain any data below 1.5Ghz (though I need to recheck the units for that number)

[ridma kaveendra]

this has values from zero hertz

[Walter Roberson]

At line 602 the file suddenly shifts from 3 column to 1 column. We do not know how to deal with that?

[ridma kaveendra]

sorry that is my mistake. here is the new file. Still output is not a sine wave as above image