Prev: Any way to get around SVD's reordering of basis vectors?
Next: Developing DFIG detailed Model in Simulink
From: Pawel on 25 Oct 2009 21:55 I'm dealing with automobile simulated by two-degree-of-freedom system. After forming continuous-time state-space model, determining the system's highest frequency (~1.5Hz) I set sampling frequency (~6Hz) and determined corresponding discrete-time model. Next I generated pulse response with pulse input. Now I'm suppose to add random noise signal with zero-mean and variance - 3% of the maximum absolute amplitude of pulse response. Since it is a system with high daming ratio maximum absolute amplitude of oscillating pulse response is 1.2500e-004 which gives variance=3.7501e-006. To generate noise signal I'm using randn that requires standard deviation and here is the thing: noise = sqrt(variance)*randn(m,n); But sqrt(variance) is much higher than variance since variance<<1 and ratio noise/signal becomes huge. Later I'm about to use this pulse response (Markov parameters) to form Hankel matrices and apply Eigensystem Realization Algorithm. But I'm affraid that this anomalously big noise will mess things up. Any idea what to do with it? |