Filtering random EM noise from a large 3-phase motor?

[r_h_a]

Hi All,

First, an overview of the system: we have a DAQ system running Labview, two PCI-6225 cards, a PCI-6224, and a PCI-6254 card.  All connections run through SCB-68 terminal blocks.  All power supplies for the instrumentation and DAQ cabinet run through isolation transformers.  Additionally, most lines are shielded twisted pair lines and are run from the test table to the DAQ terminal blocks through a grounded conduit. The signals we are having issues with are measured at 1kS/s per channel on the PCI-6225 cards.

Under normal operation, the signals are clean.  Where we are having problems is with a set of test we need to run where we use a large 3-phase motor used to drive a rotating machinery testbed engine.  The table is bolted to the same large metal table as the test instrumentation.  When the motor comes on, our nice clean data becomes an angry fuzzy caterpillar.  The noise does not appear to have a specific frequency and in the raw data it shows up as single-point, relatively high amplitude blips.  I would say the overall amplitude is 10s of mVs. 

A hardware low-pass filter removed a large amount of this noise, but we do not have enough hardware filters for every channel.  I attempted to use a software filter (Butterworth with similar inputs to the hardware filters) and it smooths out to blips and lowers their overall amplitude, but is not near as effective as the hardware filter.

Any ideas?

Thanks in advance,

Hunter

[John_P1]

Hi Hunter,

What was the design of the hardware filter that effectively reduced the noise, and how do you have your SW butterworth filter configured?  One of the drawbacks of using a software-defined filter is that it does NOT provide anti-aliasing.  That is, if you have noise components at a higher frequency than Nyquist (in your case only 500 Hz), then this noise would be aliased back down into your signal of interest.

*You could characterize the noise a bit better by sampling on a single channel of the 6225 at a much higher rate, then look at the data in the Frequency domain by taking an FFT in LabVIEW.  Make sure to acquire enough points to have adequate resolution when you take your FFT (see The Fundamentals of FFT-Based Signal Analysis and Measurement in LabVIEW and LabWindows/CVI).

So, increasing your Nyquist frequency (by increasing your sample rate) could give some benefit depending on the characteristics of the noise you are seeing (you would have a higher alias-free bandwidth which you would be able to filter in software).  If there is noise present at certain frequencies higher than your nyquist frequency, we should be able to choose a sample rate that will alias this noise to a frequency that we can still filter out in software.

Also, if your twisted-pair wiring is not connected differentially I suggest doing so if you have enough channels available on the 6225 (you can connect up to 40 differential pairs).  For more information about connecting your signals you can check out the Field Wiring and Noise Considerations article.

Best Regards,

John

[AnalogKid2DigitalMan]

Does the large metal table have a good connection to earth ground? Having one may reduce the amplitude of the noise.

-AK2DM

[Henrik_Volkers]

Big Motor => big currents!  And if a converter is used to control the motor speed .... they are a horrible source of NOISE in not seriously filtered in the converter ($$$). 

Avoid ground loops, provide low!! impedance grounding from the motor directly back to the converter , adding Chokes sometimes help (http://ts.nist.gov/MeasurementServices/Calibrations/upload/72C-2.pdf)

As mentioned some sort of input filter is needed (nyquist). 

If there are still spikes the median filter can help.

[amswak · Accepted solution]

Hi All,

I just wanted to give some closure to this topic.  After trial and error and many phone calls to VFD companies, we narrowed the source of the noise to the inverter.  The fix (and over 98% of the noise is now gone) was a sin wave filter on the output (going to the motor) and a reator on the input (coming in from facility power).  If you dont want to spent the big bucks on a sin wave filter, just putting a reactor on the inverter output reduced the noise almost 80%.

Best regards and thanks for the help!

Hunter