1
$\begingroup$

Let $h$ be the impulse response of the gammachirp filter:

$$ h(t) = \exp(c_1 + c_2 t + c_3 \ln t) \cos(c_4 + c_5 t + c_6 \ln t) $$

The output $y$ of the filter is given by the convolution of the input $x$ and $h$:

$$ y = x \ast h $$

What is the state-space representation of this filter?

$\endgroup$
2
$\begingroup$

A state-space representation can either indicate a LTI, LTV or nonlinear system. As @fibonatic said, your system surely does not have a LTI state-space representation. Assume that the system is described in s-s form as: $$\begin{align}\dot{x}&=f(x,t)+u\\y&=x\end{align}$$ Then we can write $$\begin{align}\dot{h}&=f(h,t)+\delta(t)\\ &=\left\{\left(c_2+\frac{c_3}t\right)-\left(c_5+\frac{c_6}t\right)\tan(c_4 + c_5 t + c_6 \ln t)\right\}h(t)\\&\stackrel{\Delta}= g(t)h(t)\end{align}$$ This might give us the idea that we can write $f(x,t)$ as $$f\left(x,t\right)=g(t)x-\delta(t)$$ but as you see, this $f$ will be dependent on the input which contradicts our initial assumption that $f(x,t)$ is independent of $u$. So there can't be any affine state space representation of this filter.

$\endgroup$
  • $\begingroup$ The system doesn't have a finite-dimensional state-space representation. But infinite-dimensional realizations exist. $\endgroup$ – Pait May 11 '17 at 12:08
  • $\begingroup$ @Pait Correct. Another fact is, the system doesn't have a real state space representation but complex ones can be found. $\endgroup$ – polfosol May 11 '17 at 12:18
  • $\begingroup$ You mean a finite-dimension complex representation? Interesting - can you elaborate or point to a source? Thanks! $\endgroup$ – Pait May 11 '17 at 13:57
2
$\begingroup$

The impulse response of a LTI system with a finite dimensional state space can be computed with,

$$ h(t) = \left\{\begin{array}{l l} C\, e^{A\, t} B + D\, \delta(t) & \text{if } t \geq 0 \\ 0 & \text{else} \end{array} \right. $$

which should always consist out of a finite sum of exponentials (potentially complex conjugate pairs and/or multiplied by polynomials when $A$ has Jordan blocks of larger than size one).

When $c_3$ and $c_6$ are not zero, then the impulse response you provided can not be written as a finite sum of such terms. Therefore there is no state space model, with a state space of finite size, which behaves exactly like such filter.

$\endgroup$
  • $\begingroup$ There is no finite linear time invariant state space representation. Thus your justification is not enough. $\endgroup$ – Bilal Jafar Karaki Aug 6 '17 at 1:12

Your Answer

By clicking “Post Your Answer”, you agree to our terms of service, privacy policy and cookie policy

Not the answer you're looking for? Browse other questions tagged or ask your own question.