low frequency harmonics in the force signals

[Oleks]

Hello all,

I am trying to find out what causes noise in my signal, whether it is a faulty hardware or my lack of knowledge.

The system that I am working consist of two stepper motors(parker HV232) that provide harmonic heave+pitch motion with phase difference of 90 degrees (fish tail flapping kind of motion). I have a 3-axial load cell attached to the shaft of the pitch motor, and it measures  lift, drag, and torque on a plate that oscillates in the water tunnel(oscillation is provided by those two steppers). Tunnel provide incoming flow.

Force signals are captured through SG-23 + PXI-6221 daq board, I set gain on the last one to +/-200mV for better resolution.

When motors are turned off, the noise level on the static signal is below resolution threshold, and nothing strange is present neither in signal nor in its spectra.

The problems begin when I initialize motion on the steppers, FFT of the signal shows low frequency spikes that are exact multiplies of the oscillation frequency, I attached a pdf file with example.

We opt out from amplifier/filter that was suggested by manufacture and I am suspecting that what I observing might be due to that, but not 100% sure. The VI that I am using applies a band pass filter to the data in the 0.2Hz - 6Hz, but somehow I still can see noise beyond 6 Hz.  

I was wandering maybe someone had experience with similar issues and can suggest a solution on how to get rid of those harmonics.

Thank you in advance

[kikiduu · Accepted solution]

I am not an expert in tunnel dynamics, but from what you are saying it looks like they may not be noise but components generated due to rotation of a mechanical assembly. Normally people refer  to such harmonics as 'order components' ,meaning they are a multiple of the rotating frequency.

So may be this is the forward

1. Verify from an expert (one who understands the mechanical system that you are using) that at the point where you are measuring force you won't get any order components.

2. Understand what exactly is the measurand. What exactly do you want to do with the measured signal (RMS, FFT etc?. Then you may be able to figureout whether the order components can affect your measurement.

---

Oleks wrote:

We opt out from amplifier/filter that was suggested by manufacture and I am suspecting that what I observing might be due to that, but not 100% sure. The VI that I am using applies a band pass filter to the data in the 0.2Hz - 6Hz, but somehow I still can see noise beyond 6 Hz.  

---

During filtering, the attenuation in stop-band determines how fast the signals will die-down in the stop band.

Message Edited by kikiduu on 01-29-2009 04:54 AM

[Oleks]

Thank you very much Kikiduu,

That seems what is happening. I would expect to see a spike in the spectra that corresponds to the oscillation frequency, but its multiples were something that I did not foresee. The signal eventually will be integrated to get power and efficiency and for that reason I would like it to be as much cleaner as possible.

By the way, are software base filters (VI that are available in Labview), capable to handle order components, or should I think about a hardware solution?

Thanks again!

[WCR]

If you clean up the signal with a software lowpass filter that removes, or at least attenuates, the harmonics above the fundamental, the cleaned up signal will have less power in it.  Is that what you want?  If the signal is "real" then maybe you shouldn't clean it up.

If you remove the higher harmonics (thinking in the frequency domain), it is equivalent to reducing the time domain signal to a sinusoid at the fundamental frequency - the sinusoid that best fits the data.  Even a "perfect" (noise-free) square or triangle wave will have the higher harmonics. Whether you really want to reduce your signal to a sinusoid is up to you, but you can easily do it by selecting a filter such as Butterworth, Chebyshev, etc. from Functions -> Signal Processing -> Filters.  Watch out for Labview's Bessel filter; its frequency cutoff is not what it should be (except when the filter order =1, a trivial case).

Bill

[Now_With_Underscores]

Hey Oleks,

Just want to mention that National Instruments sells a toolkit that provides order analysis.  Obviously, for your setup we're trying to remove these orders from your system, but just thought I'd mention that we do have this toolkit to pin down order harmonics.  Have a great afternoon!

[Oleks]

That is interesting,

On the link that you provided, the toolkit shown as a part of  "NI Sound and Vibration Measurement Suite", is it available separately from it?

And once again, I would like to thank everybody for insights, at least I know where to dig now, which is a big deal for me!

[Oleks]

Edit: Apparetnly we have the NI Sound and Vibration Measurement Suite  I am going to check what that toolkit can do.

[Now_With_Underscores]

The order analysis toolkit was a seperate purchase from the sound and vibration toolkit up to version 2.0.  The order analysis toolkit can now only be purchased as part of the sound and vibration measurement suite.