Eliminate the periodic noise from the signals.

[Shrinivas]

Hi,

I am acquiring the signal from displacement sensor with resolution 0.1 micron. Analog( 10 to -10 V) output of the sensor connected to DAQ-6215.

When I measure the signal, I am getting the periodic noise , when I connect it in the RSE mode as attached here.

When I change it to differential connection, I am getting random noise (please refer the attachment).

I know that, I can use filter to remove this noise. But I am scared that my actual signal may get affected with the filter as my measured signals will be similar to the noise.

Please let me know how can I go about it.

Thank you 

[AdamTrojak]

Can you attache powerspectrum of your signal? What kind of measurements will you take? Will displacement be periodical as well?

[Bob_Schor]

Several questions:

• Have you looked at your signal with an oscilloscope?  Do you see noise?
• It is very difficult to interpret the plots you showed without knowing the time base, the sampling rate, and the Y-Axis scale.  The Differential plot seemed to show a signal that changed slowly (i.e. around 1 Hz), but without knowing the sampling rate, this is hard to interpret.
• Is there a possibility that you are seeing "signal"?
• There are numerous sources of "noise", including picking up 60Hz A/C "from the air".  Good grounding, proper shielding, and hooking up your acquisition hardware properly can help.  Ultimately, there is also "measurement noise", which can be removed by filtering.  Knowing the frequency you expect of the "signal" and the (observed) frequency of the "noise" helps in designing your filter.

Bob Schor

[johnsold]

First, those signals are very small, about 1.5 to 3.5 mV pp on a 1 V base. What is the magnitude of your signal for a 100 nm displacement change?

Second, the time scales on the two graphs are very different, making comparisons difficult.

"Guesstimating" from the time scale on the differential graph, it appears that the sampling rate may be a few samples per second. This is almost certainly too low to correctly identify any interfering periodic signal. The Nyquist criterion requires that sampling occur at more than twoce the frequency of the signal being acquired - including the interference. Most likely you are picking up power line frequency interference (50 or 60 Hz) and aliasing by undersampling.

Do not even think about any filtering or other signal processing to reduce the interference until you know its source and how it is getting in.  If it is aliased, it may be very difficult to remove.

Tell us more so that we can help.

Lynn

[mikeporter]

First, from looking at the images I have no idea what is noise and what is signal.

Second, as Lynn said, you need to locate the noise source and fix it through proper shielding, termination, isolation, etc.

Third you need a hardware antialiasing filter that is set at 2x your sample rate - whatever that may be. Assuming you have selected your sample rate correctly it won't effect your signal, but it will keep out of band energy from being aliased into your data. Once out of band energy is aliased into your data, its pretty much game over. You're hosed.

The filter doesn't to be anything too fancy, a simple RC filter will do the job nicely. Depending on the daq hardware you are using there may already be areas on the board that are preconfigured for building such a filter. (For example the SCB-68A, kudos NI, very nicely done)

Again, to highlight one of Lynn's points, don't worry about signal processing or any other sort of software gymnastics until you have your data sampled cleanly and without any aliased energy.

Mike...

[billko]

---

@mikeporter wrote:
First, from looking at the images I have no idea what is noise and what is signal.

Second, as Lynn said, you need to locate the noise source and fix it through proper shielding, termination, isolation, etc.

Third you need a hardware antialiasing filter that is set at 2x your sample rate - whatever that may be. Assuming you have selected your sample rate correctly it won't effect your signal, but it will keep out of band energy from being aliased into your data. Once out of band energy is aliased into your data, its pretty much game over. You're hosed.

The filter doesn't to be anything too fancy, a simple RC filter will do the job nicely. Depending on the daq hardware you are using there may already be areas on the board that are preconfigured for building such a filter. (For example the SCB-68A, kudos NI, very nicely done)

Again, to highlight one of Lynn's points, don't worry about signal processing or any other sort of software gymnastics until you have your data sampled cleanly and without any aliased energy.

Mike...

---

Cleaning up the signal physically first will not only make it easier for you to deal with in software, it will also extend the lifetime of your measuring equipment by not exposing it to all that spiky noise.