0
$\begingroup$

Let $a \in (-1,1)$ and $x\in {\mathbb R}$ and consider a following sequence of rational functions: \begin{eqnarray} R^{(0)}[x]&:=& \frac{x (a x-1)}{a x-x^2+x-1}\\ R^{(1)}[x]&:=& -\frac{(x-1) x^3 \left(\left(a^2+a-1\right) x^2-2 a x+1\right)}{(a x-1) \left(a x-x^2+x-1\right)^3}\\ R^{(2)}[x]&:=&(x-1) x^4 \frac{\left( -1+ (5 a-3) x+ \left(-10 a^2+5 a+10\right) x^2+ \left(10 a^3+a^2-30 a-1\right) x^3+ \left(-5 a^4-5 a^3+30 a^2+9 a-4\right) x^4+ \left(a^5+2 a^4-12 a^3-9 a^2+2 a+2\right) x^5+ \left(2 a^4+2 a^3-a^2+a-1\right)x^6 \right) }{(a x-1)^3 \left(a x-x^2+x-1\right)^5}\\ \vdots \end{eqnarray} In general we have: \begin{eqnarray} R^{(n)}[x]:=\sum\limits_{p=1}^n \frac{x^{p+1} (1-x)}{(1-a x)^p (1-(1+a) x+x^2)}\cdot (x \frac{d}{d x})^p R^{(n-p)}[x] \end{eqnarray} for $n=1,2,\cdots$. Those rational functions are Z-transforms of certain wave-forms in the time domain. In other words we have: \begin{equation} r_k^{(m)} := \left. \frac{1}{k!} \frac{d^k}{d x^k} R^{(m)}[x] \right|_{x=0} \end{equation} for $m=0,1,\cdots$ and $k\in {\mathbb N}$. Here $m$ is the order of the waveform.

Now, by decomposing the rational functions into partial fractions we have derived the following results: \begin{eqnarray} &&r^{(0)}_k = \frac{\sqrt{\Delta} \left(a^2+3 a+1\right) \sin ((k+1) \phi )}{a+3}+\left(a^2+a-1\right) \cos ((k+1) \phi )\\ \hline\\ &&r^{(1)}_k= \frac{a^k}{1-a}+\\ &&\mbox{$-\frac{\left(a^2-2\right) \cos ((k+1) \phi )}{a-1}-\frac{\left(-a^4-4 a^3+a^2+10 a-8\right) \sin ((k+1) \phi )}{(a-1) (a+3) \sqrt{\Delta}}$}+\\ &&\mbox{$\binom{k+1}{1} \left(\frac{\left(-2 a^4-11 a^3-13 a^2+9 a+10\right) \sin ((k+2) \phi )}{(a+3) \sqrt{\Delta}}+\frac{\left(2 a^4+7 a^3-a^2-11 a+2\right) \cos ((k+2) \phi )}{(a-1) (a+3)}\right)$}+\\ &&\mbox{$-\binom{k+2}{2} \left(\frac{\left(a^4+5 a^3+6 a^2-a-2\right) \cos ((k+3) \phi )}{a+3}-\frac{a \left(a^3+3 a^2-3\right) \sin ((k+3) \phi )}{\sqrt{\Delta}}\right)$}\\ \hline\\ &&r^{(2)}_k= \frac{(3 a-1) a^{k+1}}{(a-1)^2}+\frac{\left(3 a^2-3 a+1\right) (k+1) a^{k-1}}{(a-1)^2}+\frac{(k+1) (k+2) a^{k-1}}{2 (a-1)}+\\ &&\mbox{$-\frac{(3 a-1) \cos ((k+1) \phi )}{(a-1)^2}-\frac{\left(3 a^5+29 a^4+92 a^3+80 a^2-89 a-113\right) \sin ((k+1) \phi )}{(a-1)^2 (a+3)^3 \sqrt{-a^2-2 a+3}}+$}\\ &&\mbox{$\binom{k+1}{1} \left(-\frac{\sqrt{\Delta} \left(2 a^6+12 a^5+9 a^4-44 a^3-26 a^2+69 a-29\right) \sin ((k+2) \phi )}{(a-1)^3 (a+3)^3}-\frac{\left(2 a^6+16 a^5+33 a^4-22 a^3-94 a^2+31 a+97\right) \cos ((k+2) \phi )}{(a-1)^2 (a+3)^3}\right)+$}\\ &&\mbox{$-\binom{k+2}{2} \left(-\frac{\sqrt{\Delta} \left(7 a^6+54 a^5+113 a^4-34 a^3-249 a^2-29 a+107\right) \sin ((k+3) \phi )}{(a-1)^2 (a+3)^3}-\frac{\left(7 a^6+40 a^5+33 a^4-114 a^3-87 a^2+117 a+11\right) \cos ((k+3) \phi )}{(a-1)^2 (a+3)^2}\right)+$}\\ &&\mbox{$\binom{k+3}{3} \left(-\frac{\sqrt{\Delta} \left(8 a^6+43 a^5+39 a^4-84 a^3-77 a^2+52 a+14\right) \sin ((k+4) \phi )}{(a-1)^2 (a+3)^2}-\frac{\left(8 a^6+59 a^5+125 a^4+10 a^3-175 a^2-62 a+34\right) \cos ((k+4) \phi )}{(a-1) (a+3)^2}\right)+$}\\ &&\mbox{$-\binom{k+4}{4} \left( -\frac{3 \sqrt{\Delta} \left(a^6+7 a^5+15 a^4+6 a^3-11 a^2-6 a+1\right) \sin ((k+5) \phi )}{(a-1) (a+3)^2} -\frac{3 \left(a^6+5 a^5+5 a^4-6 a^3-7 a^2+2 a+1\right) \cos ((k+5)\phi)}{(a-1) (a+3)}\right)\quad$}\\ \end{eqnarray} where $\phi:=\arccos[(1+a)/2]$ and $\Delta:=3-2 a-a^2$.

To convince the reader that there are no typos I include a piece of Mathematica code which verifies the above results. We have:

Clear[RR]; Clear[RR];
RR[x_] := {(
   x (-1 + a x))/(-1 + x + a x - 
    x^2), -(((-1 + x) x^3 (1 - 2 a x - x^2 + a x^2 + a^2 x^2))/((-1 + 
       a x) (-1 + x + a x - x^2)^3))};
rr[k_] := {-(1 - a - a^2) Cos[(1 + k) phi] + (
    Sqrt[3 - 2 a - a^2] (1 + 3 a + a^2) Sin[(1 + k) phi])/(3 + a), 
   a^k/(1 - a) - ((-2 + a^2) Cos[(1 + k) phi])/(-1 + 
     a) - ((-8 + 10 a + a^2 - 4 a^3 - a^4) Sin[(1 + k) phi])/((-1 + 
       a) (3 + a) Sqrt[
     3 - 2 a - 
      a^2]) + (1 + 
       k) (((2 - 11 a - a^2 + 7 a^3 + 2 a^4) Cos[(2 + k) phi])/((-1 + 
          a) (3 + a)) + ((10 + 9 a - 13 a^2 - 11 a^3 - 
          2 a^4) Sin[(2 + k) phi])/((3 + a) Sqrt[3 - 2 a - a^2])) - 
    1/2 (1 + k) (2 + 
       k) (((-2 - a + 6 a^2 + 5 a^3 + a^4) Cos[(3 + k) phi])/(
       3 + a) - (a (-3 + 3 a^2 + a^3) Sin[(3 + k) phi])/Sqrt[
       3 - 2 a - a^2])};



l1 = Simplify[
   Table[D[Together[RR[x][[1 + n]]], {x, k}]/k!, {n, 0, 1}, {k, 0, 
      10}] /. x :> 0];
l2 = Simplify[
   FunctionExpand[
    Table[rr[k][[1 + n]] /. phi :> ArcCos[(1 + a)/2] , {n, 0, 1}, {k, 
      0, 10}]]];
MatrixForm[{l1, l2}]
l1 - l2

After running the code above we get the following output:

enter image description here

and the left hand side matches the right hand side as it should be.

Now, if we take higher orders ($m\ge 2$) then both the rational functions and the corresponding wave-forms in the time domain become to big to be written down not to mention to be dealt with.

My question is therefore is there any other way of inverting the Z-transforms in question ? In other words do we always have to resort to partial fraction decomposition if we want to compute the $k$th derivative of a rational function at the origin, or is there some other way of deriving the result?

$\endgroup$
4
  • $\begingroup$ @ThomasAndrews: I beg your pardon? The expression $r^{(m)}_k$ is the $k$th derivative of the $m$th order rational function at $x=0$. Therefore the quantities $r^{(m)}_k$ do not depend on $x$ but instead they depend on the time $k$ where $k=0,1,2,3,\cdots$. $\endgroup$
    – Przemo
    Commented Feb 2, 2018 at 18:00
  • $\begingroup$ @ThomasAndrews: Allright, I think I know what you mean. You probably thought that the quantity $R^{(1)}[x]$ is a derivative of $R^{(0)}[x]$. It is not. The rational functions are generated according to some recursive algorithm which, for the sake of simplicity, I do not include here. $\endgroup$
    – Przemo
    Commented Feb 2, 2018 at 18:02
  • $\begingroup$ Whoops, sorry, didn't read carefully. $f^{(n)}(x)$ is usually the $n$th derivative or $f$, and since the subject was repeated derivatives, I jumped to a conclusion in terms of notation. $\endgroup$ Commented Feb 2, 2018 at 18:05
  • $\begingroup$ @ThomasAndrews: The rational functions in question are being generated by repeatedly applying some sort of differentiation operator -- see my update above -- yet this is not simple differentiation.The rational functions along with their inverse Z-transforms can be seen as higher order corrections in some sort of perturbational theory calculation. Each of the waveforms have constant frequencies yet the combination of all of them has a time-varying frequency. It is this time-varying frequency that I am trying to figure out in this question. $\endgroup$
    – Przemo
    Commented Feb 5, 2018 at 11:38

0

You must log in to answer this question.

Browse other questions tagged .