How to remove noise from the acquired signal?

[JP187]

Hi,

I am measuring the output of 4 strain gauges. NITB-4330 front block mounted on NIPXIe-1073 chasis has been used as DAQ device. I am getting some weird signals. I want to know is there noise present in the output? What would be the possible source of noise? How can i distinguish between noise and actual signal?  And how can i remove noise? Does noise has anything to do with sample rate? I have attached output of strain gauges for two different sample rate. I am new to LabVIEW and concept of noise.

Thanks in advance,

JP

[altenbach]

All real data has noise. What counts is the signal/noise. Your noise covers a relatively small range but everything looks noisy of you zoom in close enough :D. How does it compare to real signals?

Nose can have any frequency, so the sampling rate can fold it differently unless you do proper analog filtering according to niquist to keep the alias frequencies of higher frequencies out. What are your sampling frequencies? How is the wiring (e.g. don't zip tie your signal cables to the power cable. 🐵

[JP187]

thanks altenbach for reply. I couldn't reply earlier, as i wasn't in the lab.

Can you please tell me how can i distinguish between original signal content and noise content in the signal. I have used 100 Hz and 1000 Hz as sampling frequency. I have taken spectral analysis of the output signal and instead of frequency, it was showing time on x-axis. I had applied filter (lowpass butterworth, 5th order, with cut off frequency of 0.2 Hz) on the output signals. I know, 0.2 Hz cut off frequency is very lower, but with that only i am getting better signal. I have setup this experiment on breadboard using jumper wires.

Please refer the images in the attachment. Note that, now i am measuring output of 8 strain gauges instead of 4.

Regards,

JP

[GerdW]

Hi JP,

I have taken spectral analysis of the output signal and instead of frequency, it was showing time on x-axis.

- It is showing "Time" as x axis label as this is the default value. You can always change this label to anything else…

- When your "Waveform Graph" shows the FFT result then you have 60db SNR: this is quite ok.

Do you shake your sensor chips or do you just want to measure a static offset of your strain gauges?

(Did you made those sensor chips on your own? I once worked in a similar environment…)

[JP187]

Hi GerdW,

No, I didn't make those sensor chips. I am not apyling any input to them, i just want to measure static offset.

Regards,

JP

[Henrik_Volkers]

You capture some raw data.

You look at the data (time domain, frequency domain*)

*) the conversion (usually called FFT ) has some side effects (window, leakage,..) you need to know, generally any signal processing has effects you need to keep in mind!  (filters will have some lead in time and other artifacts). Articels/lectures/books on signal processing (numerical and analog) help.

A first test: Shortcut the input at your DUT. Theory tell you should read zero. In real world you read the digitizer/emc noise and you can monitor drifts (temp, ADC ref drift, etc)

By adding/varying shields and GND topology you get experience on your lab emc condition.

Maybe try shielded twisted pair cables (LANcable is a nice source)  and a 'white metal' cooky box. (linear Jim Williams style for insiders 😉 )

You didn't tell what the bridge is sensitive to, however a next step could be a bridge simulation (a bridge with really stable resistors) that can give hints to supply voltage noise.

Your MEMS sensors on the picture looked nice but the window allows light influence!

You should monitor the temperature too. All devices are temperature sensitive (OK, the best only some to fractions of ppm).

Finally: A low pass filter always help 😄 😄

the jump and ringing on your filtered signals look like filter artifacts ...