STFT Understanding

[FCance]

Hello

I am analyzing the accelaration of a free vibration of a concrete column. This is the signal.

FCance_4-1678582785367.png

FCance_0-1678580888685.png

This gives the Power Spectrum.

FCance_1-1678580993551.png

and this the STFT

FCance_3-1678581316161.png

What's my problem?

Well, in the Power spectrum you can see a large component at about 3 Hz (Fundamental Frequency) and this sameone, in the STFT, you can barely see it 

My question is, why does this happen?

Another problem i have, related to the provious one, are the type of spectrum (Amplitude or Power) and values (RMS or Peak) that the STFT gives.

I read the LABview Help thousand of times but it doesn't say anyhing about that. I suposse that it gives Power Spectrum RMS. Am i right?

[altenbach]

We have absolutely no idea what parameters you are using. Maybe attach your VI and your data so we can analyze what's happening.

(Btw, it's LabVIEW, not LABview.)

[FCance]

Change the Menu to "Read Measure" and put "100" in Sensitivity which is that of the accelerometer.

Measure data: Samples=10000 , Sample Rate = 1000 Hz

[altenbach]

Sorry, I don't have a computer with the advanced signal processing toolkit in LabVIEW 2021. I only have 2020.

Can you do he following:

• Set the controls to reasonable values and make them the default. No need to tell me the settings.
• Use save for previous to save in LabVIEW 2020.
• Attach the down-converted code once more. It will be in a new folder.

Since the STFT uses a small sliding window, very low frequencies are masked, of course. You cannot tell the frequency of a banana.

[FCance]

I did what you asked, hope all is well.

[altenbach]

Sorry, I don't have the vibration toolkit so I cannot run the upstream code. Please add an indicator (right-click wire...create indicators), run the VI, and turn the indicator into a constant. Save for previous and attach once more.

altenbach_0-1678643959854.png

(I cannot tell what determines the loop rate. It is not reasonable to burn the CPU forever repeating the same code over and over with the same inputs.)

[altenbach]

You have 1000 points per second, so a 3Hz signal has a period of ~333 points. Your window is a 512 point Blackman which is much (much!) less than 333 points at half height, so your 3Hz signal ends up as banana with no full cycles remaining.

[FCance]

There, thanks for tips.

[FCance]

Tell me if im correct, what you are telling me is this

FCance_1-1678648791946.png

What would be the solution? Decrece the Sample rate? 

[altenbach]

---

@FCance wrote:

There, thanks for tips.

---

Sorry, this is again in LabVIEW 2021 so I cannot open it on my current computer (despite the file name).

One possibility would be to increase the window size, but then you lose time resolution. I am sure there are more adaptive solutions, but I haven't studied it. How about something based on wavelets? Do you have the advanced signal processing toolkit?