# How to calculate the $H_2$ and $H_\infty$ norms of TransferFunctionModels or StateSpaceModels?

Lets assume we have a system model:

tf = TransferFunctionModel[{{
(s^3 - 2.841 s^2 + 2.875 s - 1.004)/
(s^3 - 2.417 s^2 + 2.003 s - 0.5488)}}, s];
ss = StateSpaceModel[tf] My question is:

How to calculate $H_2$ and $H_\infty$ norms of this system or transfer function? Is there a built-in function for it?

Some notes:

• In MATLAB norm(ss,2) or norm(ss,inf) does that.
• Mathematical calculations can be done as in this link.

From the LyapunovSolve examples one can compute the $H_2$ norm of an asymptotically stable continuous-time system as below:

{a, b, c} = {{{-1, 2, 3}, {0, -2, 2}, {0, 0, -3}},
{{1, 1}, {1, 1}, {1, 1}},
{{1, 1, 1}, {1, 1, 1}}};

x = LyapunovSolve[a, -b.Transpose[b]];
h2norm = Sqrt[Tr[c.x.Transpose[c]]] // N


6.12917

The $H_2$ norm can be computed as $\sqrt{\text{Trace}\left(b q b^{\dagger}\right)}$ or $\sqrt{\text{Trace}\left(c p c^{\dagger}\right)}$. Here $b$ and $c$ are the input and output matrices of $ss$, and $q$ and $p$ are the observability and controllability gramians.

However if you tried ObservabilityGramian[ss] there is an error because this particular system is unstable and hence its $H_2$ norm is $\infty$.

The $H_{\infty }$ norm can be obtained from the peak value of the the SingularValuePlot (or magnitude plot in BodePlot).

SingularValuePlot[tf]


This turns out to be $\sim 5.2 db$ for this system. (Unfortunately, no built-in functions yet for the norm computations.)

• what kind of transpose/inverse is it? – Zero Aug 22 '16 at 20:09
• ConjugateTranspose – Suba Thomas Aug 22 '16 at 20:18