0
$\begingroup$

I have some mems accelerometer (ADXL355 and IIS2ICLX) and i want to get a accuracy of 0,005 ° incline measuring by using the axis of it.

I am measuring with arduino over I2C and have a Samplerate about 20 Hz and 30 Hz.

With FFT i got a spectrum, that shows me a 1/f-noise and white noise. Lowpassfilter can reduce white noise for sure, but what about the 1/f noise ?

I am using the moving average, which works not bad, but are there better filters ?

And if there are better filters, could you explain why these are better?

Thank you for your help.

$\endgroup$
2
  • $\begingroup$ Can you add a photo of your spectrum? $\endgroup$
    – Baddioes
    Dec 13, 2023 at 18:19
  • $\begingroup$ It's pretty hard to get the toothpaste back into the tube. $\endgroup$ Dec 14, 2023 at 20:27

1 Answer 1

2
$\begingroup$

The best ways of dealing with noise is to prevent it in the first place.

  1. Identify the root cause and eliminate it as much as is possible
  2. Look at the transmission between the noise source and the sensor/detector. Add isolation/blocking
  3. Block noise at the detection end: blocking, shielding, damping, isolation, clean up power supplies, etc.

Once you have optimized on the physical side, you can start looking at algorithmic noise reduction. That tends to be complicated but in general: the more a priori knowledge about the noise and the signal can incorporate the better it will work. That typically starts with a careful characterization of both and signal and noise property.

I am using the moving average, which works not bad, but are there better filters ?

Moving average is simple, but it's just a low-pass filter and not a great one at that as it has high side lobes. A better filter would be a bandpass filter that reduces all frequency where the SNR isn't good enough.

There are many other more advanced filters (Wiener, Kalman, etc.) that are dynamic, signal dependent and time variant.

$\endgroup$
1
  • $\begingroup$ Thanks for your comment. The main problem is, that my signal should be dc signal (stataic seismic mass in an accelerometer, which causes increasing the capacity). And some sources says, that the root cause is an 1/f-noise and the main reaeson for that noise are 1) thermic fluctuations (Model ofVoss and Clarke), 2) superposition of two Partial waves – scattered electron and non-scattered electron (model of handel) last 3) superposition of generation recombination noise (model of McWorther) So it seems the root cause cannot be compensated... ? $\endgroup$
    – Rayjay
    Dec 14, 2023 at 21:01

Your Answer

By clicking “Post Your Answer”, you agree to our terms of service and acknowledge you have read our privacy policy.

Not the answer you're looking for? Browse other questions tagged or ask your own question.