Easy Z-Transform doesn't match in Wolfram/Matlab

157 Views Asked by At

I have this function:

$$g(k)=2*(\frac14)^{k-1}+(\frac18)^k$$

I calculate the Z-transform in this way, applying the delay property: $Z[x(k-1)]=z^{-1}X(z)$

$$G(z)=\frac{8}{4z-1}+\frac{8z}{8z-1}$$

The problem is that both Wolfram and Matlab solved it in this way:

$$G(z)=\frac{32z}{4z-1}+\frac{8z}{8z-1}$$

They are considering: $$(\frac14)^{k-1}=(\frac14)^{k}*(\frac14)^{-1}$$

How can it be possible? The delay property is not applicable here, or they are wrong?

1

There are 1 best solutions below

0
On BEST ANSWER

Z-transform can be two-sided (bilateral), namely $$\sum_{k=-\infty}^\infty g(k) z^{-k} \tag1$$ In this case, delay property holds as you stated: $$\sum_{k=-\infty}^\infty g(k-1) z^{-k} =\sum_{k=-\infty}^\infty g(k) z^{-k-1} = z^{-1}\sum_{k=-\infty}^\infty g(k) z^{-k} $$

However, Maple et al calculate one-sided Z-transform: $$\sum_{k=0}^\infty g(k) z^{-k} \tag2$$ For this transform the delay property is different, since there is no $g(-1)$ in (2). We have the following: $$ z^{-1}\sum_{k=0}^\infty g(k) z^{-k} = \sum_{k=0}^\infty g(k) z^{-k-1} = \sum_{k=1}^\infty g(k-1) z^{-k} $$ Now, if we are dealing with $g(k) = 2/4^{k-1}$, then $g(k-1)=4g(k)$, hence $$ z^{-1}\sum_{k=0}^\infty g(k) z^{-k} = 4 \sum_{k=1}^\infty g(k) z^{-k} = 4 \sum_{k=0}^\infty g(k) z^{-k} - 4g(0) $$ Solve: $$\sum_{k=0}^\infty g(k) z^{-k} = \frac{4g(0)}{4-z^{-1}} = \frac{32z}{4z-1}$$ in agreement with Maple et al.


By the way, it is easier not to use any kind of delay property and calculate directly using the sum of geometric series: $$\sum_{k=0}^\infty ar^k = \frac{a}{1-r}$$