Line Resolution and Frequency

By the Nyquist criterion, to avoid aliasing of higher frequencies into lower frequency bins, you'll need to set your sample rate at more than 2x the maximum frequency in your measured signal. If your accelerometer is sensitive to frequencies of 6000 Hz, you'll need your sample rate to be at least 12,000 Hz to avoid aliasing. That said, your <0.2 Hz results might not come from aliasing problems, but its a good place to start.

Hope that helps

TinTin

Thans a lot for your helps. According to specification of this accelarometer, it can not measure the frequency bellow the 0.2 Hz.But I got the spectrum bellow the 0.2 Hz in FFT analysis. Whether the spectrum bellow 0.2Hz is considerable or not ?.

If the accelerometer can't measure below 0.2 Hz (a cycle with a period of 5 seconds) - then any results you get in bins below 0.2 Hz are not worth considering, because if the results don't come from your sensor, they either come from aliasing effects or noise sources. My pleasure to help

Regards,

TinTin

Thanks a lot for giving your valuable response. What I observe in my measurement that at any perticular time the vibration is 30 mm/s but the amplitude(ASD) below the 0.2 Hz is greater than 28 mm/s and above the 0.2 Hz is below the 1 mm/s. According to FFT the sum of all amplitude corresponds to all frequency gives the amplitude in time domain.So can I ignore the spectrum bellow the 0.2 Hz?.

For the first question please elaborate .

For the second question whenever u are calculating FFt what actually the sofware does is it calculates either the asd (amplitude spectral density ) or PSD(Power spectral density ) which will be always in that range .

If u want to check the amplitude at certain frequency band as in if u need " g " Vs " Hz " u need to calculate manually for the specific frequency band .

How I can select the line resolution of FFT analyser?. I use the FFT to see the frequency spectrum of Vibration signal. I observe that depending upon the line resolution of FFT, the amplitude of frequency spectrum at any perticular frequency is also change. I use the DEWETRON data logger. The data accquring rate of my data is 10kHz.

Tin Tin is right about Nyquist sampling rates but what equipment are you using for the data capture? A good scope should sample at 250MHz but if you are using LabVIEW and some hardware you might have to set the sample rate yourself - if you can get it to 12kHz.

I use DEWETRON data logger. My data accquiring rate is 10k but the sampling rate of A/D card is 100k.

Hi,

you need either an analog sufficient filtering in the instrument (your accelerometer) or in-between instrument and datalogger.

So I would suggest to check the accelerometer specification: what output type: analog voltage, analog frequency digitised or what else?

Then decide how much error (x% of full scale) you will tolerate at any frequency within your band of measurement (0.2Hz to 6KHz).

Then search in the output versus frequency curve (either from instrument or from filter after the instrument) at which frequency it has a response of this x% of FS.

This is the lowest permitted sampling frequency.

Without knowing these details I would put a 6KHz second order analog low pass filter after the accmeter, sample with 100KHz and use a next stage of digital low pass filter set to 6KHz or below depending on wanted bandwidth.

RHABE

I thing the out put is voltage type because the sensitivity of this sensor is 500 mV/g. As the sensitivity below the 0.2 Hz is less for this sensor so why i get the very high amplitude peak below the 0.2 Hz compare the amplitude spectrum between 0.2 to 4200Hz?

That was a comprehensive treatment of the problem Rhabe; technically thorough yet explained with simplicity - GA to you!