This week I spent some time with Alex who updated the adaptive code to save the filter's coeffs all the time, stop when I open the loop, and reload the latest coeffs. when I start it again.
The point was to minimize the adaptation rate. Unfortunately, seems it is making some filters go wild, so it is not in use yet.
After taking some more measurements with the adaptive filter running, I have noticed a new peak in the signal around 22-23 Hz. My first assumption was that this is caused due to internal resonance of MC1 (which is driven when the adaptive code is running, and not when it's not). Therefore, I drove MC1 without the adaptive filter looking for the same peak... which wasn't there.
This sent me back to the adaptive code... Seems there is a matrix in the simulink file of the adaptive filter which doesn't have an MEDM screen. I am now working on making this screen. Once I am done with that, and make sure there is correlation between simulink and MEDM, I'll keep on chasing the peak in the code. |