The usage of the vfreq function in ADE Explorer calculator.

Did you actually run a harmonic balance (hb) analysis? The expression won't work if you did (and you would have to have simulated a lot of harmonics to get a graph as you showed). That looks more like a PSD/DFT plot from a transient simulation. 

As for returning to the calculator window once you've clicked on the node in the schematic - presumably it's just hidden behind the schematic window? For me it pops to the front once I've clicked on a net but it can dynamically change again as you move your cursor.

Andrew

Hi ，Andrew Beckett ！

Your speculation is entirely correct: the calculator window happens to be covered by the schematic window; the waveform in the example plot is obtained through Fourier Transform. 

My goal is to transform the waveform shown in the picture  below into the frequency domain.

Which function should I choose  or what should I do ？

Best regards，

Qunshan Liang

Qunshan,

Presumably you're going to want to do a discrete Fourier transform via the dft() function. I'm not sure what you are aiming for here, but the bottom waveform isn't periodic, so you'd need to apply windowing - and given that it appears to be a single period of the top waveform, the distortion caused by the window will be fairly significant.

Do you understand how to use Fourier transforms and the effect of various parameters?

Andrew

Hi ，Andrew 

I have previously learned about Fourier Transform and I know that it can be divided into four cases based on the continuity/discreteness and periodicity/non-periodicity of the time-domain signal. What I want to convert to the frequency domain is a continuous non-periodic signal, which means I need to apply Fourier Transform to it. The formula used for this purpose is as follows:

But I don't know how to operate the calculator in AED Explorer. Do you know how to operate it?

Best regards，

Qunshan Liang

Dear liangqunshan,

liangqunshan said:

I have previously learned about Fourier Transform and I know that it can be divided into four cases based on the continuity/discreteness and periodicity/non-periodicity of the time-domain signal. What I want to convert to the frequency domain is a continuous non-periodic signal, which means I need to apply Fourier Transform to it. The formula used for this purpose is as follows:

The expressions you included in your Forum post are for continuous time signals. In your case, you do not have a continuous time signal, but one that is sampled. Hence, you cannot use the expression in your Forum post to examine the frequency components of the two signals you show (signal /VD and signal /M3/D). The analog in the sampled-time domain to examine the frequency components of a sampled signal is the discrete Fourier transform or DFT or its closely related power spectral density (PSD).

The ViVA Calculator (which you showed in your screenshot of your first Forum post entry) provide access to the dft() function as well as to the psd() function. As an alternative, ViVA offers the Spectrum Assistant which is accessible from ViVA's menu as Window->Assistants->Spectrum. The Virtuoso Visualization and Analysis XL User Guide provides information on both of these functions. There are also many support articles and Troubleshooting articles on the Cadence On-line support portal regarding their use. In the last entry to the Forum post at URL:

community.cadence.com/.../spectrummeasurement-function

there is a link to a document that details an example of using the Spectrum Assistant for some simple waveforms and some notes on some subtleties to its use.

However, to obtain accurate results using the DFT of PSD requires some knowledge about sampling. Without this background, I think it will be very difficult for you to use these functions in a manner that will provide accurate results.

I also think, as Andrew mentioned, it would be helpful to understand what your main objective is in studying the Fourier components of either of the two waveforms you displayed. Are you interested in their bandwidths or their spectral peaks? In both cases, your transient data needs to include multiple base periods of the waveforms in order to provide any reasonably accurate spectral information. Your lower signal /M3/D is not periodic in the time interval you show and the upper signal /VD possibly shows about 1 1/2 periods (my guess).

Shawn

Dear ShawnLogan ，thank you for reply ！

After studying, I have learned to use the dft function, and I have converted the previous signal (VD) into the frequency domain, obtaining the following results.

The purpose of transforming the time-domain waveform to the frequency domain is to observe the magnitude in the high-frequency region to determine if there will be significant electromagnetic noise. It appears that the amplitude in the high-frequency part is relatively small.

Is the result I obtained correct ?

Best regards，

Qunshan Liang

Dear liangqunshan,

I have spent some time trying to duplicate your result to better illustrate the impact of modifying your dft() analysis to both improve its frequency resolution and accuracy. As I mentioned in my earlier posts, the use of the dft() function to provide an accurate estimate of frequency content requires some study of the mechanics of sampling. I hope this provides some insights to help you. To avoid the spam filter for this Forum, please access the note by copying and appending the following three pieces of text to form the following URL.

/scl/fi/taf1c46eogo6ald55397j/

/fft_summary_liangqunshan_study_100623v1p0.pdf?

rlkey=99s64wigemmgmn6btmzs5fwym&dl=0

Shawn

Dear liangqunshan,

I apologize, but I noticed there is a problem with the link I provided as jt reports the file was deleted. Please use the following link:

Copy and concatenate the following two text lines to the dropbox prefix to form the correct link (shown above)

/scl/fi/b8jnac46sz5gzh4p9udy1/

fft_summary_liangqunshan_study_100723v1p1.pdf?rlkey=qokye4uhscm64p8fy48umz061&dl=0

Shawn

Dear ShawnLogan ，thank you for help ！

It seems that my country prohibits the use of Dropbox as I am unable to access its official website. Regarding the sampling theorem, I have studied it before and reviewed it recently. I understand that the sampling frequency should be at least twice the highest frequency component in the signal to avoid overlapping in the spectrum. Since I am not aware of the highest frequency component in the signal, so I will increase the number of sampling points as much as possible.

What I intend to do is to perform a discrete Fourier transform on the waveform of a non-periodic function over a certain period of time to see if it exhibits high magnitude in the high-frequency region. However,I am uncertain about the correctness of such operation and whether there will be significant deviations in the results.

Best regards，

Qunshan Liang

ear liangqushan,

liangqunshan said:

It seems that my country prohibits the use of Dropbox as I am unable to access its official website.

I did not realize that. I have placed the note on OneDrive at the location:

fft_summary_liangqunshan_study_100723v1p1.pdf

but also attached a lower resolution version of it in case you are not able to access that site either.

liangqunshan said:

What I intend to do is to perform a discrete Fourier transform on the waveform of a non-periodic function over a certain period of time to see if it exhibits high magnitude in the high-frequency region.

When you take the discrete Fourier transform of a signal, you are innately assuming it is periodic. Specifically, if you take an N point DFT of a signal, the signal is assumed to be periodic over those N samples and the resultant DFT will represent that "N-periodic" signal. That does not mean you cannot take its DFT, but you need to understand the implications of it. For example, if the starting and ending time points are not the same, the DFT will show spectral leakage that can distort the resultant DFT. There are what are called "windowing" functions to minimize this effect, but you need to understand this to use the DFT to get an accurate estimate of the frequency contents of the signal.

Shawn

fft_summary_liangqunshan_study_100723v1p1.pdf