system identification command line- impulseest

Hi; studying the link below http://www.mathworks.com/help/ident/gs/identify-linear-models-using-the-command-line.html i faced something unclear. The help says: "Estimating the Empirical Step Response

To estimate the step response from the data, first estimate a non-parametric impulse response model (FIR filter) from data and then plot its step response. % model estimation |*_Mimp = impulseest(Ze1,60)_;*||| " why did it use 60||; why it did not use 20 or 650 or ...;

Answer 1

[Rajiv Singh]

The "60" refers to the length of the FIR filter. An FIR model typically has larger number of coefficients than a rational transfer function model. Hence the "large number" 60. The exact value of the filter length is not known in advance. You have to try out various values and evaluate the resulting quality of the model using tools such as COMPARE, RESID, and evaluating confidence bounds on impulse response and step response plots.

Note that the estimation is "regularized" unless you choose regularization kernel to be "none" (see IMPULSEESTOPTIONS). This means that choosing too large a value for the filter length would not adversely affect its quality too much. So go with a reasonably large number of coefficients (such as 1/10 of data length).

Answer 2

[mohsen]

thank you.

my estimation data is 20000 samples in about 0.3 second and i could not enter 2000 in impulseest. it showed me an error

when i was sampling, i did not know how to compute the sampling rate and i used the maximum sampling rate that could be supported by ma daq card.

i dont know whether it leads to something wrong whith my system identification. but i computed the arx and amx models.

i also could not compute the process models with any order. it shows me an error.

could you please make me some advices