Meaning
Digital signal processing routines smooth noisy experimental data by fitting local low-degree polynomials to adjacent data points using least-squares approximation. Implementing Savitzky Golay filtering is common when preparing electrochemical battery cycler data for differential capacity or differential voltage analysis. The method reduces high-frequency noise while preserving the shape and height of physical peaks.
Algorithm Operation
Moving window calculations fit a polynomial of a predetermined degree to a selected subset of data points centered around each target value. This local fit generates a smoothed output point that replaces the raw center value, and then the window shifts forward by one data step. By choosing an odd-numbered window size and a polynomial order, the user balances noise attenuation against signal distortion.
Noise Reduction
Experimental measurements of cell voltage and current always contain random high-frequency fluctuations from electrical interference or temperature drift. Simple moving averages would smear out the steep transitions and close peaks that indicate phase changes in the electrode. This specialized smoothing technique resolves these features by maintaining higher moments of the underlying distribution.
Signal Integrity
Traditional low-pass filters tend to shift peak positions and flatten peak heights, which introduces error into thermodynamic calculations. The preservation of original peak locations ensures that tracking routines for state of health estimate electrode degradation accurately.
