FFT sidebands

[Langelier]

Hello everyone.

I have been programming a UI that uses FFT and that can modify its settings. I have adapted the controls and everything works. The problem is that, even with a good resolution and averaging, I have much more sidebands in the FFT calculated by LabVIEW than in the one that I have with an oscilloscope. Is there a way to remove these sidebands? No filtering or averaging has been done. Is that the reason? 

I am sure that the measured signal should be a neat sine wave, so the multiple sidebands should not appear as high as they do. I put an example of the spectrum that I can obtain with a high resolution:

[johnsold]

Assuming that your signal does not have those spurious signals, then you need to give us much more information.

Consider these possibilities: Aliasing. Window effects. Spectral leakage. The signal really is like that but your scope does not show it.

What is your sample rate? How many samples are you processing at a time? What is the signal frequency supposed to be? What is the signal waveform? What interfering signals are present? What is the amplitude of the desired signal and any noise or interference?

Please post your VI and include some typical data seved as default.

Lynn

[Langelier]

My sample rate is 15 kHz and 15 000 samples were processed at a time. In the graph that is shown, the frequency wa supposed to be 12 Hz and the waveform was a sine wave sent to a speaker. The other peaks are normal to see because of the nature of the shaking element, but not the sidebands necessarily. As I said, with a scope, I did not have the sidebands. I cannot tell what the amplitude should be, because the gain was set arbitrarily in the device that measures the vibration. 

I post the VI with this message. 

[johnsold]

Your VI does not have any data. Please run it so that some data is in Waveform In control, then Make Current Values Default. Save.

I do not have the SVFA toolkit, but I can use other spectral analysis tools to look at the data.

Lynn

[ejkaiser]

My first guess would be clipping. Have you checked the time domain signal? You may be overdriving the analog signal chain or the AD converter.

Cheers

Edgar

[Langelier]

The measured signal is in the low frequencies, below 20 Hz, so could I really overdrive the ADC or the analog signal chain? I understand that at high frequencies, the sample rate will affect the measured signal, but at such low frequencies I thought that it would not. 

[Langelier]

I now cannot send you any data, because the test setup has been dismounted so we have the room for other tests. As soon as I can, I will. 

[johnsold]

What Edgar was asking related to the peak amplitude of the signal, not the frequency. If you graph the time domain signal, do the peaks show any sign of clipping or flattening?

Lynn