Correct time scaling for IFFT of VNA S-parameter data

2 replies · 4K views · last reply
Nora
Messages
2
Reaction score
0
TL;DR
time (distance) axis of Inverse Fast Fourier Transform (IFFT) from Vector Network Analyzer (VNA) S-parameter data.
Hi
I would like to transform the S-parameter responce, collected from a Vector Network Analyzer (VNA), in time domain by using the Inverse Fast Fourier Transform (IFFT) . I use MATLAB IFFT function to do this and the response looks correct, the problem is that I do not manage to the time scaling correct. The VNA has a build in time domain toolbox so I know how it should look like, see the figure below.
Line graph titled Return-Loss, time domain, showing LINMAG1 against TIME1.SEC from -1 to 4 ×10^-9 s, with a main peak near 0.9 ×10^-9 s reaching about 0.083 and a second peak near 1.95 ×10^-9 s at about 0.046.

In the figure below I have plotted the MATLAB transformed data with the index on x-axis (also zoomed in between N=0 to 20, N_max is 1600)
Line plot titled Return-Loss, time domain, with response on the y-axis from 0 to 0.08 and N on the x-axis from 0 to 20. The curve peaks near 0.073 at N=5 and again near 0.048 at N=9, then decays toward 0 by N=20.

At first I assumed that the increment on x-axis would be 1/bw/N_max where bw is the band width (i.e. f_max-f_min) and N_max is the number of samples. This is obviously wrong with a factor of 1e-3.
I would much appreciate if someone would like to help to get this correct! I have attached the S-matrix data in the txt file. First column is the frequency in GHz, second is real part of the response and third is imaginary part of the response.
My thanks in advance
/Thomas
 
Attachments
  • Line graph titled Return-Loss, time domain, plotting LINMAG1 against TIME1.SEC from -0.4 to 1.2 s, with one blue trace labelled PNA transfomed data showing a main peak near 0.27 s and a smaller peak near 0.58 s.
    Return-Loss, time domain, VNA.jpg
    29.5 KB · Views: 300
  • data.txt
    data.txt
    50 KB · Views: 321
The time step is given by
$$
\Delta t = \frac{1}{N \Delta f}
$$
From the data file, it appears that ##\Delta f = 0.00125\ \mathrm{GHz}## (the frequencies appear to be rounded values).