Synthetic parametric model

Description

On this page you will find a synthetic parametric model for which one can easily experiment with different system orders \( n \), values of the parameter \( \varepsilon \), as well as different poles and residues.

Also, the decay of the Hankel singular values can be changed indirectly through the parameter \( \varepsilon \).

Model

The parameter \(\varepsilon\) scales the real part of the system poles, that is, \(p_i=\varepsilon a_i+jb_i \, \), with \( j \) the imaginary unit. For a system in pole-residue form

\[ H(s,\varepsilon) = \sum_{i=1}^{n}\frac{r_i}{s-p_i} = \sum_{i=1}^{n}\frac{r_i}{s-(\varepsilon a_i+jb_i)} \]

we can write down the state-space realization

\[ H(s,\varepsilon) = \widehat{C}\Big(sI-\varepsilon \widehat{A}_\varepsilon - \widehat{A}_0\Big)^{-1}\widehat{B}+D\]

with system matrices defined as

\[\varepsilon \widehat{A}_\varepsilon + \widehat{A}_0 = \varepsilon \left[\begin{array}{ccc} a_1 & & \\ & \ddots & \\ & & a_n\end{array}\right] +\left[\begin{array}{ccc} jb_1 & & \\ & \ddots & \\ & & jb_n\end{array}\right] ,\]

\[\widehat{B} = [1,\ldots,1]^T,\quad \widehat{C} = [r_1,\ldots,r_n],\quad D = 0.\]

Notice that the system matrices have complex entries.

For simplicity, assume that \( n \) is even, \( n=2k \), and that all system poles are complex and ordered in complex conjugate pairs, i.e.

\[ p_1 = \varepsilon a_1+jb_1, p_2 = \varepsilon a_1-jb_1, \ldots, p_{n-1} = \varepsilon a_k+jb_k, p_n = \varepsilon a_k-jb_k, \]

and the residues also form complex conjugate pairs

\[ r_1 = c_1+jd_1, r_2 = c_1-jd_1, \ldots, r_{n-1} = c_k+jd_k, r_n = c_k-jd_k. \]

Then a realization with matrices having real entries is given by

\[ A_\varepsilon = \left[\begin{array}{ccc} A_{\varepsilon,1} & & \\ & \ddots & \\ & & A_{\varepsilon,k}\end{array}\right], \quad A_0 = \left[\begin{array}{ccc} A_{0,1} & & \\ & \ddots & \\ & & A_{0,k}\end{array}\right], \quad B = \left[\begin{array}{c} B_1 \\ \vdots \\ B_k\end{array}\right], \quad C = \left[\begin{array}{ccc} C_1 & \cdots & C_k\end{array}\right], \quad D = 0,\]

with \( A_{\varepsilon,i} = \left[\begin{array}{cc} a_i& 0 \\ 0 & a_i \end{array}\right] \), \( A_{0,i} = \left[\begin{array}{cc} 0& b_i \\ -b_i & 0 \end{array}\right] \), \( B_{i} = \left[\begin{array}{c} 2 \\ 0 \end{array}\right] \), \( C_{i} = \left[\begin{array}{cc} c_i& d_i\end{array}\right] \).

Numerical Values

We construct a system of order \(n = 100\). The numerical values for the different variables are

• \(a_i \) equally spaced in \( [-10^3, -10]\),

• \(b_i \) equally spaced in \([10, 10^3]\),

• \( c_i = 1\),

• \( d_i = 0\),

• \(\varepsilon\)\( \in [1/50,1]\).

In MATLAB, the system matrices are easily formed as follows:

 n = 100;
 a = -linspace(1e1,1e3,n/2).';   b = linspace(1e1,1e3,n/2).';
 c = ones(n/2,1);                d = zeros(n/2,1);
 aa(1:2:n-1,1) = a;              aa(2:2:n,1) = a;
 bb(1:2:n-1,1) = b;              bb(2:2:n-2,1) = 0;
 Ae = spdiags(aa,0,n,n);
 A0 = spdiags([0;bb],1,n,n) + spdiags(-bb,-1,n,n);
 B = 2*sparse(mod([1:n],2)).';
 C(1:2:n-1) = c.';               C(2:2:n) = d.';   C = sparse(C);

The above system matrices \(A_\varepsilon, A_0, B, C\) are also available in MatrixMarket format Synth_matrices.tar.gz.

Plots

We plot the frequency response \(H(s,\varepsilon) = C\big(sI-\varepsilon A_\varepsilon - A_0\big)^{-1}B\) and poles for parameter values \(\varepsilon \in [1/50, 1/20, 1/10, 1/5, 1/2, 1] \).

Error creating thumbnail: Unable to save thumbnail to destination

Frequency response of synthetic parametric system, for parameter values 1/50 (blue), 1/20 (green), 1/10 (red), 1/5 (teal), 1/2 (purple), 1 (yellow).

Error creating thumbnail: Unable to save thumbnail to destination

Poles of synthetic parametric system, for parameter values 1/50 (blue), 1/20 (green), 1/10 (red), 1/5 (teal), 1/2 (purple), 1 (yellow).

In MATLAB, the plots are generated using the following commands:

 r(1:2:n-1,1) = c+1j*d;     r(2:2:n,1) = c-1j*d;
 ep = [1/50; 1/20; 1/10; 1/5; 1/2; 1];                       % parameter epsilon
 jw = 1j*linspace(0,1.2e3,5000).';                           % frequency grid
 for j = 1:length(ep)
   p(:,j) = [ep(j)*a+1j*b; ep(j)*a-1j*b];                    % poles
   [jww,pp] = meshgrid(jw,p(:,j));
   Hjw(j,:) = (r.')*(1./(jww-pp));                           % freq. resp.
 end
 figure,  loglog(imag(jw),abs(Hjw),'LineWidth',2)
          axis tight,    xlim([6 1200])
          xlabel('frequency (rad/sec)')
          ylabel('magnitude')
          title('Frequency response for different \epsilon')
 figure,  plot(real(p),imag(p),'.')
          title('Poles for different \epsilon')

Other interesting plots result for small values of the parameter. For example, for \(\varepsilon = 1/100, 1/1000 \), the peaks in the frequency response become more pronounced, since the poles move closer to the imaginary axis.

Next, for \(\varepsilon \in [1/50, 1/20, 1/10, 1/5, 1/2, 1] \), we also plot the decay of the Hankel singular values. Notice that for small values of the parameter, the decay of the Hankel singular values is very slow.

Error creating thumbnail: Unable to save thumbnail to destination

Hankel singular values of synthetic parametric system, for parameter values 1/50 (blue), 1/20 (green), 1/10 (red), 1/5 (teal), 1/2 (purple), 1 (yellow).

Contact

Antonio Cosmin Ionita 14:38, 29 November 2011 (UTC)