Revised software uploaded on January 9, 2013
Accelerometers signals for pyrotechnic events often have spurious offsets.
Numerous potential causes include:
1. Excitation of the accelerometer’s piezoelectric crystal natural frequency
2. Signal conditioner overload
3. Base strain sensitivity
4. Instrumentation noise
5. Electromagnetic interference
These causes are often referred to loosely as “saturation effects.”
Data is precious, so there is a need to salvage the data using editing.
* * *
The Matlab scripts in mean_filter_saturation_rank.zip use the mean filtering method to remove spurious offsets from a pyrotechnic shock signal.
The main script is: mean_filter_saturation_rank.m
The remaining scripts are supporting functions.
The scripts optimize the mean filtering to meet several goals including:
1. Render the positive and negative SRS curves approximately equal to one another.
2. Achieve a user-specified initial slope for each SRS polarity curve, typically 6 to 12 dB/octave.
3. Achieve a net velocity of approximately zero.
4. Minimize any changes to the “good” portion of the raw data SRS curves.
Trial-and-error is used to determine the optimum number of passes and window size in the mean filter.
Engineering judgment is required to assess the results. This method is rather aggressive and may remove some “good energy” but hopefully only a negligible amount.
This is a long-term “work-in-progress.” More later…
– Tom Irvine