I was looking over some DSP and came across the following signal: $x[n] = 1/n$. So I wondered whether it had a z-transform but I soon realized that it fails to meet two conditions described here:

...(1) the signal starts at a finite time and (2) it is asymptotically exponentially bounded...

So I modified it to meet those requirements: $x[n] = \frac{1}{n^2} u[n] $

(I assume $n>0$ is a condition or perhaps $u[n]$ should be changed to $u[n-1]$.)

Considering that signal, I believe it satisfies both requirements however I don't see anything like it in a table of common z-transforms, but it can be rewritten as:

$ x[n] = \frac{1}{1^2}*\delta[n-1] + \frac{1}{2^2}*\delta[n-2] + \frac{1}{3^2}*\delta[n-3]\ +\ ...$

Which I believe transforms to $ X(z) = \frac{1}{z^1} + \frac{1}{(2z)^2} + \frac{1}{3^2z^3} +\ ... $

Does this transformation make sense? Is it the best approach? And if it's correct, is there a more compact way to express it?

  • $\begingroup$ The Z-transform of $\frac{1}{n^2} u[n-1]$ is explained in this answer. $\endgroup$
    – Matt L.
    Jul 20, 2016 at 14:48

1 Answer 1


What you've got is definitely the z-transform of that signal (by definition). For a more compact expression, notice that the polylogarithm is exactly the infinite sum you've shown (just replace $ z \rightarrow 1/z$ in the Mathworld expression). Unfortunately I don't think there's any nicer way to write it; sometimes there just isn't.


Your Answer

By clicking “Post Your Answer”, you agree to our terms of service and acknowledge you have read our privacy policy.

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