0
$\begingroup$

I am trying to model the pdf of the peak and rms of the time domain response $y(t)$ in a reverberation chamber. The system is stimulated by pulsed linear chirp $x(t)$ for 200 ns that goes linearly from 4-5 GHz and its Fourier transform is denoted by $X(f)$. The frequency response of the system, $H(f)$ is a random variable and is measured 50 times for 50 different tuner steps over a frequency range of 1 to 6 GHz. We know the distribution of $H(f)$ over a narrow frequency band (~500 MHz) is given by Lomax distribution. Now we know that we can use the following equation to calculate the time-domain response $y(t)$. $$ y(t) = \mathscr{F}^{-1}\big\{X(f)*H(f) \big\} $$

Where $\mathscr{F}^{-1}$ is the inverse Fourier transform. On the other hand, we have measured $y(t)$ in the time domain as well. As you might imagine, $y(t)$ is a random variable. We are interested in the distribution of the peak and rms of this variable. Here is a plot of 200 measurements in the time-domain.

enter image description here

Now you might be wondering why we did the convolution at first, the answer is that since we know the distribution of $H(f)$, we would like to link that to the distribution of the peak and rms of y(t) so we understand how the spectrum of the pulsed chirp and characteristics of $H(f)$ (modal density, frequency spacing, etc.) influence the distribution of the peak and rms of $y(t)$

$\endgroup$
5
  • $\begingroup$ I'm still trying to decode this. Are you tracking the peak with a sliding maximum and the rms with a sliding average on the square of the signal? And then you want a "distribution" of values of both? Like a histogram of peak values and another histogram of rms values? $\endgroup$ Apr 27, 2023 at 5:04
  • $\begingroup$ so, $H(f)$ is the Fourier transform of $y(t)$, right? $\endgroup$ Apr 27, 2023 at 8:47
  • $\begingroup$ I added plenty of details. In hindsight, my original post wasn't descriptive enough. $\endgroup$
    – Reza Afra
    Apr 28, 2023 at 8:17
  • $\begingroup$ robert bristow-johnson I know how to get the histogram. That's trivial. I would like to gain insight into the best approach to model peaks and rms of y(t). A random variable is fully specified by its CDF, so ultimately we would like to find an analytical expression for that. $\endgroup$
    – Reza Afra
    Apr 28, 2023 at 8:19
  • $\begingroup$ If $$ y(t) = \mathscr{F}^{-1}\big\{X(f) \circledast H(f) \big\} $$ then $$ y(t) = x(t) \cdot h(t) $$ $\endgroup$ Apr 28, 2023 at 14:01

0

Your Answer

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

Browse other questions tagged or ask your own question.