Vadim,
I think if your model (ARMA) is already defined (in terms of known
coefficients), using idsim and filter should be equivalent (although filter
may work faster). However, I may point out the following:
1. In the latest version of the Sys Id toolbox (ver 5.0 in Matlab R12),
idsim/poly2th have been replaced by a single function "sim". sim provides
more options for simulations. Please see the help on idmodel/sim in version
5.0 (also pasted below). In particular, noise-corrupted simulations can be
performed, and special input types can be used via "idinput" functionality.
2. You can still use idsim in the new version of Sys Id, which works exactly
like "sim". There is no need of poly2th, as theta format has been obsoleted.
hth,
Rajiv Singh
help idmodel/sim
SIM Simulates a given dynamic system.
Y = SIM(MODEL,UE)
MODEL: contains the parameters of the model in any of the IDMODEL
formats, IDSS, IDPOLY, IDARX, or IDGREY.
UE: the input-noise data UE = [U E]. Here U is the input data, that
could be given as an IDDATA object (with the signal defined as input)
or as a matrix U = [U1 U2 ..Un] with the column vector Uk as the k:th
input. Similarly, E is either an IDDATA object or a matrix of noise
inputs (as many columns as there are output channels). If E is omitted
a noise-free simulation is obtained.
Example: U = iddata([],idinput(200),'Ts',0.1);
E = iddata([],randn(200)); Y = SIM(MODEL,[U E]);
The noise contribution is scaled by the variance information con-
tained in MODEL (that is, MODEL.NoiseVariance). This means that E should
be white noise with unit covariance matrix to obtain the correct noise
contribution according to the model.
Y: The simulated output. If U is an IDDATA object, Y is also
delivered as an IDDATA object, otherwise as a matrix, whose k:th
column is the k:th output channel.
If UE is a multiple experiment IDDATA object, so will Y be.
If MODEL is continuous time, it is first sampled according to the
information in the input U which then must be an IDDATA object,
('Ts' and 'InterSample' properties).
With [Y,YSD] = SIM(MODEL,UE) the estimated standard deviation of the
simulated output, YSD, is also computed. YSD is of the same format as Y.
Y = SIM(MODEL,UE,INIT) gives access to the initial state:
INIT = 'm' (default) uses the model's initial state.
INIT = 'z' gives zero initial conditions.
INIT = X0 (column vector). Uses X0 as the initial state.
See also IDINPUT, IDMODEL for input generation and model creation
and COMPARE, PREDICT for model evaluation.
--
=
Instructions for joining and leaving this list and remarks about
the problem of INAPPROPRIATE MESSAGES are available at
http://jse.stat.ncsu.edu/
=
