Hi After removing high frequency white noise in my INS sensor, I want to simulate low frequency noise with Burg or Yuler. I used some functions in python to analyse the recorded data and to get parameters to use. But the the problem is the sequences generated diverge (and take very big values), is that normal ? I use the following equation : XN+4=XN+3*C3+XN+2*C2+XN+1*C1+W1 where W1 is white noise. I have another question. Why I have to use kalman filter to estimate the error ? Why not to integrate directly the value given by the simulation ?