Filter to Extract and Smooth DC Level

[aaronb]

I am trying to filter and smooth the peak DC level of 100kHz sampled data. I can't seem to find a suitable filter that filters and smooths the data without rounding off the leading edge of the signal and/or creating new artifacts like undershoot, overshoot, extended rise time, etc. I have attached a VI with sample data. I have included some of the more promising filters I have tried. In the end I am looking for as flat of a DC level as I can achieve with minimal rise time at the 100ms mark without adding additional artifacts.   

[GerdW]

Hi Aaron,

can you attach your "SmartSmoothingFilter" subVI too?

Btw. when comparing your original data with any of the filtered ones all of them are quite nice… 😄

[aaronb]

Sorry forgot about that one. Here are both files...

[GerdW]

Hi Aaron,

you should use the original and the Median-Savitzky-Golay-w-Butterworth:

Use the original filter for the parts of your signal with big changes and the Median for parts with low changes.

Then again: keep in mind the comparison of the filtered result with your original data - it has be senseful in the end…

[aaronb]

Yea I liked that one as well however I am not thrilled with the ~75ms hit I get on the rise time. If I could figure out a way to reliability grab the unfiltered data at the beginning power on rise and then just apply the filter to the data after that would probably work.   

[GerdW]

Hi Aaron,

calculate the running stdev of your data. When your "data rise to noise" this stdev will get bigger. Use this to decide for unfiltered or filtered output…