Lissage des Données de la Fonction de Transfert
|
Lissage des Données de la Fonction de Transfert |
|
|
Introduction au Concept de Mesure > Lissage des Données de la Fonction de Transfert Smoothing is essentially another type of averaging that is available only for Transfer Function displays (Phase or Magnitude). This feature helps to reduce "jagginess" on transfer function traces and can make trends in the frequency response of the system under test easier to see. On a smoothed transfer function trace, each data point is averaged together with some number of the immediately adjacent points on either side.
Older versions of smart used a fixed size, rectangular moving average – basically a FIFO average turned sideways – for smoothing transfer function data but this was not ideal. A perennial problem with FFT-based analysis of audio data is that the frequency data points in a Discrete Fourier Transform (DFT) are spaced linearly, whereas people hear logarithmically. Getting good low frequency resolution from an FFT typically means having excess resolution at high frequencies, meaning that you would like to include more frequency data points in your smoothing average as you go up in frequency. For this application it would also be advantageous to give more weight to the data at the center of the smoothing window and less weight to points farther out at the edges, whereas a rectangular averaging window gives equal weight to all points in the average.
The smoothing function in Smaart 7 addresses both these issues, by utilizing a true logarithmic fractional octave smoothing window that adaptively increases in size as you ascend in frequency. The Smaart 7 smoothing window is bell shaped, rather than rectangular, giving relatively more weight in the average to the point being smoothed and the closest points on either side and less to the points that are farther away from the center. We had also observed that experienced users often tended to increase the amount of smoothing for the transfer function displays when looking more at phase data and decrease it to look at magnitude data, so in Smaart 7 we provide separate smoothing controls for phase and magnitude data in frequency response measurements.
See also: |