Filtering Noise & Its subharmonics

[Haristhota]

I am trying to acquire data using capacitive sensor from a rotating object (Spindle).

Spindle is rotating at 1000RPM and i am getting 16.667 Hz(1000/60) sine wave (due to spindle eccentricity) as a major component and its sub harmonics in the signal.

For my analysis, i have to completely eliminate this frequency component and its sub harmonics.

If i apply 8th order band stop Butterworth filtering, this frequency component is getting suppressed but not getting eliminated. Sub harmonics are remains same.

What would be the better approach to solve this problem? Do i need to implement adaptive filtering algorithms? Or any simple way is there? Please suggest.

[crossrulz]

A Band Stop filter will just reduce frequencies in the stop band.  It will not affect the harmonics.

I think what you really need is a low pass filter.

[johnsold]

What is the nature of the desired signal? Its frequencies, waveform, amplitude, ...? How does the amplitude of the spindle signal compare to that of the desired signal? How many subharmonics are relevant and what are their amplitudes?  Does the spindle speed change during a single measurement? What kind of analysis will you be doing on the desired signal?  What is your sampling rate? How much data do you analyze at one time?

It really does not make sense to talk about filters or other signal processing until you have defined what you need to do.

Lynn

[Haristhota]

The desired signal is the signal without rotational frequency and it's harmonics. The spindle speed is constant so that i get constant rotation frequency i.e. speed rpm/60.

I attached the screen shots of my signal. I need to eliminate 2000RPM/60 frequency i.e. 33.33 Hz (Major sinusoidal component) and its subharmonics from the signal for further analysis.

The data is initially collected at 10KSPS. it's DC offset is removed and  down sampled to 4 KSPS. It is bandstop filtered from 32-34 Hz and FFT is obtained.

[Henrik_Volkers]

Please post a vi with just the data. copy the graph with the raw data into a new vi , make current values default, save the vi....

If you want to detekt changes in the waveform, maybe you can try to store a wfrm as a zero offset and use it calc the difference?

Another way is to add a 180° shifted (half periode delayed)  or substract a one periode delayed signal.

I would try to sync the samplerate to the RPM ...  or do a resampling ...

And finally it's still not clear to me what you want to measure...  do you know about the sound and vibration tools?  Order analyses?

[Haristhota]

Please find the VI. I know about NV toolkit of LabVIEW. 

Adding 180 shift, that i didn't understand clearly.

I would like to measure the rotational errors of a machine tool spindle. Rotational errors are different from the eccentricity of the spindle. 

Due to eccentricity in mounting, i am getting that sinusoidal component in the signal that i would like to eliminate, so i implemented a bandstop filter.

But i also would like to eliminate the harmonics due to rotational frequency component for accurate measurements.

[Henrik_Volkers]

here is something to play 🙂

I shifted the signal for one periode (actually about 3 periodes to have a better samples to periodes match 😉 )  and substracted the signal.  every content in periode will be removed.

and removed the offset and applied a STFT.

have fun

and again: What information are you looking for??  Would be of great help to help you 😉

[Henrik_Volkers]

ups file missing

and keep in mind, that is a 'sledge hammer' ....   but since you didn't told us what  you are ooking for 😄

[Haristhota]

Hi Henrik,

This is working well and better than IIR bandstop filtering that i implemented. Signal is very clean.

Is sub-harmonics also getting eliminated here?

Could you please enlight me on this shifting technique that you have implemented in that VI?

What exactly is happening? What is the name of this technique? any mathematical modelling available for this?

and why my bandpass filtering is giving initial high amplitude sinusoids?

I am trying to find out the spindle rotational errors, using a capacitance probes and a precision artefact. Please check below link.

http://www.lionprecision.com/sea/sea.html

My setup is pretty basic and for coarse lathe spindles.

Vielen Dank

[Henrik_Volkers]

Have a look at the vi ....    My method take the data from one (three) periode(s) and substract it from the next (three) periode(s), so ALL identical parts that equal one (three) periode(s) of the signal are canceled out. (harmonics are multiples of the base periode, so yes, they are canceld out)

However by looking at your link, I have the impression that these signals are already a measure of spindle quality because it's already tilting 😉

But I'm not in the health monitoring buisness .... 

May I ask what do you know about signal processing?

BTW the main frequency is 30.07 Hz  ~ 1984 RPM