Removing a trend

removing a trend

Data samples often contain trends that can perturb the computation of an FFT. The simplest example is a sine wave with a non-zero mean value. This mean value will generate a low frequency peak that can hide the spectrum of the curve.

For this reason, it is possible to compute the trend of the curve and get rid of it before FFT computation. The trend can be estimated by:

• a constant value (constant detrend)

• a straight line (linear detrend)

• a second order polynomial (quadratic detrend)

The example below shows an oscillating signal (red curve) before and after a quadratic detrend

quadratic detrend detrend operation.

Figure 12: Original signal (red) and quadratic detrend curve (green)

Figure 13: Signal after detrend operation

Simcenter Amesim computes the least squares fit of the chosen trend and subtracts it from the original data sample before computing the FFT.