Freq. to Time conversion of S-parameter with IFFT function (negative freq. and zero padding?)

Dear all,

I want to convert Freq. Data(S-parameter) to time domain with IFFT function.

Here is my code. But, converted time signal is not concided with my thought.

I saw that negative freq. generation and zero padding are necessary from some documents.

How can I fix my code?

Thank you~!

--------------------

clear all;

close all;

% Define time and frequency axis variables

fs = 35*10^9; % samples/s, total BW : 75~110GHz -> 35GHz

N = 801; % number of points of measured S-parameter : 801

dt = 1 / fs; % time step

t = (0:N-1)*dt; % time axis

df = 1 / N / dt; % Hz, frequency step

f = (0:N-1)*df+75*10^9;

St=load('w-band_air-10mm_headerx.txt');

S21=St(:,4) + i*St(:,5);

S21t=S21';

figure(1)

plot(f,20*log10(abs(S21t))) % plot freq. domain signal

title('Frequency Domain Signal (S21, dB)')

xlabel('Freq. (Hz)')

ylabel('Mag. (dB)')

S21_TD = ifft(ifftshift(S21t));

figure(2)

plot(t,S21_TD)

title('Time Domain Signal (S21)')

xlabel('Time (s)')

[Jeffrey Clark]

@dhlee, the frequency-amplitude data in w-band_air-10mm_headerx.txt columns 4 and 5 seems to have a bias of one, and more curiously has negative values regardless of the bias? Are you sure it isn't time-amplitude which looks like this plot(t,S21t):

[dhlee]

Frequency-amplitude data was measured value from test equipment(Vector network analyzer), so It could have a bias of one.

My expected IFFT signal looks like below.

[Jeffrey Clark]

@dhlee, have you worked this out for yourself? In w-band_air-10mm_headerx.txt there are 9 columns of data. The first is the frequency being scanned, then 4x2 columns of complex data. You use (4,5) as the value of readings for the presumed 75-110Ghz data. As I pointed out, the data in (4,5) seems odd for what should be frequency amplitudes; whereas (2,3) and (8,9) seem more like expected data.

If (4,5) is really where the frequency response should be taken, it appears to be encoded onto the 75-110Ghz chirp (the cosine picture I posted a while back). One would generally need to know the initial phase and rate of the chirp to remove its amplitude and phase from the (4,5) data, or fit a LS cosine function to the data to estimate its start phase and rate of change. In my case I eyballed it and extracted it from (4,5) to give a resonable response.

It is also curious that you didn't use the frequencies in (1) in your program. So how about working from the attached untitled.m; I'm in the process of adding the negative frequency and zero padding but don't know if it will help.