⏱ Time Domain Amplitude vs Time
🌐 Frequency Domain (FFT) Magnitude Spectrum
📚 Concept Note — What you're observing:
Each smoothing filter acts as a low-pass filter in the frequency domain. The FFT panel shows how much high-frequency noise energy each filter suppresses. The Moving Average has a sinc-shaped frequency response with sidelobes; the Gaussian filter has no sidelobes (Gaussian is its own Fourier transform); the Savitzky-Golay preserves polynomial trends up to the chosen order, giving it better peak fidelity at the cost of less aggressive stopband attenuation. The Moving Median is a non-linear filter — unlike the others, it is not a convolution. It replaces each sample with the median of its window, making it exceptionally robust to impulse noise (spikes) while preserving sharp edges. However, it does not have a classical frequency response.
The SNR gain = 10·log₁₀(Powersignal/Powerresidual noise) — higher is better for noise suppression, but watch for peak distortion in the time domain!
Each smoothing filter acts as a low-pass filter in the frequency domain. The FFT panel shows how much high-frequency noise energy each filter suppresses. The Moving Average has a sinc-shaped frequency response with sidelobes; the Gaussian filter has no sidelobes (Gaussian is its own Fourier transform); the Savitzky-Golay preserves polynomial trends up to the chosen order, giving it better peak fidelity at the cost of less aggressive stopband attenuation. The Moving Median is a non-linear filter — unlike the others, it is not a convolution. It replaces each sample with the median of its window, making it exceptionally robust to impulse noise (spikes) while preserving sharp edges. However, it does not have a classical frequency response.
The SNR gain = 10·log₁₀(Powersignal/Powerresidual noise) — higher is better for noise suppression, but watch for peak distortion in the time domain!