I get it, about poles and zeroes, when we talk in terms of the transfer function.

But, if the transfer function is the ratio of the output to the input, then if the input signal is zero, then the gain should go to infinity, right?

Am I missing something?

  • 3
    $\begingroup$ The transfer function is defined for a non zero input. $\endgroup$
    – AnVij
    Nov 26 '17 at 7:53
  • $\begingroup$ The denominator D(z) and numerator N(z) of transfer function are written in form of polynomials. We solve for $D(z)=0$ and $N(z)=0$. Poles are solution for $D(z)$, zeros are solution for $N(z)$. $\endgroup$
    – learner
    Nov 26 '17 at 11:10
  • $\begingroup$ @learner I know that, but I wanted to know in terms of input and output signals. $\endgroup$ Nov 26 '17 at 11:14
  • 2
    $\begingroup$ The transfer function of an LTI system clearly does not depend on the input signal. So your statement "if the input signal is zero, then the gain should go to infinity" does not make sense. Why should the gain go to infinity? Why wouldn't the output be zero for zero input? $\endgroup$
    – Matt L.
    Nov 26 '17 at 17:11
  • $\begingroup$ Transfer function is actually $Y(s) = G(s) U(s)$. provided that $u(t)=0$ for all $t$ you cannot divide.Hence it just means output is also zero $\endgroup$
    – percusse
    Nov 26 '17 at 19:49

A transfer function is defined as the ratio of the transform of output to the transform of input where all initial conditions are zero.

  • $\begingroup$ What if the input of the system is zero, at the initial condition? $\endgroup$ Dec 18 '19 at 17:49

The short answer, from Matlab

>> 0/0

ans = NaN

The gain is undefined if the input is zero because the output is also zero.

The long answer with some digressions:

If you look at a Control / Linear Systems Textbook, one should see that a system is defined as linear if it is BOTH zero state linear AND zero input linear. In most Signal Processing Textbooks, only the zero state linearity is stated which seems to be a contradiction but really isn't. Trying to Control a NonCausal system would seem to be kind-of futile, so $t=0$ is where things begin (mathematically expressed by the one sided Laplace Transforms ) but one has to account for initial conditions of physical objects like capacitor voltages and inductor currents so you can have output at $t=0$ with input $u(t)=0$, but the gain is still causal input to output.

In Signal Processing, you can record a signal and process forward and/or/both in-reverse time, so the most general expressions have lower integrals $t=-\infty$. The entire history of components such as capacitor voltages and inductor currents are realized as zero state conditions at $t=-\infty$ and the full input $u(t)$ history from $t=-\infty$


A transfer function is a property of a circuit or a black-box network with reference to its input and output ports. Causality requires that there is no output signal before there is an input signal.

See also: http://www.rfcurrent.com/poles-and-zeroes


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.