1
$\begingroup$

I am on chapter 6 of "Understanding digital signal processing", 3rd Edition by Richard G. Lyons. But I am stuck on question 6.15. Here is the network in question. enter image description here

I wrote down the difference equation as $$y(n)=x(n)+y(n-1)-Qy(n-1),$$ where $x(n)=D$(i.e. constant for all $n$), $Q$ is positive and less than $1$ and network is at rest at $n=0$ (i.e $y(n-1)=0$ when $n=0.$) In order to answer the question, I need to develop the series equation for $y(n)$ and find its closed form.

I have developed the series expression for $$y(n)=D\sum^n_{k=0}(-Q)^k {{(n+1)}!\over{(k+1)!(n-k)!}}.$$

And the terms without simplifying the factorials are $$y(n)=D{(n+1)!\over n!} - DQ{(n+1)!\over 2(n-1)!} +DQ^2{(n+1)!\over 6(n-2)!} -DQ^3{(n+1)!\over 24(n-3)!} . .. +(-Q)^nD$$.

My issue is that I cannot find where I have gone wrong because this series for $y(n)$ does not seem to have a common ratio which makes it hard for me to write it in the closed form as in $$\sum^{N-1}_{n=p}r^n={r^p-r^N\over{1-r}}.$$

My question is, is my developed $y(n)$ suitable? If it is suitable, does it have a closed form?

Thanks

$\endgroup$
  • $\begingroup$ i dunno why you don't combine $1$ and $-Q$ into $1-Q$. $\endgroup$ – robert bristow-johnson Sep 14 '17 at 0:25
  • $\begingroup$ or $Q-1$. however you like it. i don't understand why $Q$ should be by itself in any later equations. $\endgroup$ – robert bristow-johnson Sep 14 '17 at 0:28
  • $\begingroup$ is $x(n)=D=$ constant $\ne 0$ for all $n$ or just $n \ge 0$ ? $\endgroup$ – robert bristow-johnson Sep 14 '17 at 0:30
  • $\begingroup$ @robertbristow-johnson sorry I have tried to improve the question. $\endgroup$ – Chika Sep 14 '17 at 7:02
  • $\begingroup$ @robertbristow-johnson yes $(x)$=D=constant and the question considers only $n$ greater or equals zero $\endgroup$ – Chika Sep 14 '17 at 7:06
0
$\begingroup$

$$y[n] = x[n] + (1-Q) y[n-1]$$

where $y[-1] = 0$ and otherwise $n \ge 0$.

this is a difference equation with both a "natural response" (when the input is always zero, but you might have initial conditions) and a "forced response". a.k.a. the "transient response" and the "steady-state response" the math guys might call these the "homogeneous solution" and the "particular solution".

this is LTI so then the exponential function is an eigenfunction for the natural response.

for the natural solution, try $$ y_\text{n}[n] = y_\text{n}[0] a^n $$

then $$\begin{align} y[n] &= x[n] + (1-Q) y[n-1] \\ y_\text{n}[0] a^n &= 0 + (1-Q) y_\text{n}[0] a^{n-1} \\ a^n &= (1-Q) a^{n-1} \\ &= (1-Q) a^n a^{-1} \\ 1 &= (1-Q) a^{-1} \\ \end{align}$$

so i guess $a = 1-Q$.

for steady state, you posit a solution that is like the input plus all of its finite differences. since the input is constant and all finite differences of a discrete constant function are zero, then the steady state output is a constant. you sum the scaled steady state with the scaled transient response for the complete response.

$$\begin{align} y[n] &= x[n] + (1-Q) y[n-1] \\ C + y_\text{n}[0] a^n &= D + (1-Q)(C + y_\text{n}[0] a^{n-1}) \\ C + y_\text{n}[0] (1-Q)^n &= D + (1-Q)(C + y_\text{n}[0] (1-Q)^{n-1}) \\ &= D + (1-Q)C + y_\text{n}[0] (1-Q)^n) \\ C &= D + (1-Q)C \\ \end{align}$$

i guess that means that $QC = D$ and $C = \tfrac{D}{Q}$.

now you add your natural response to your forced response.

$$ y[n] = y_\text{n}[n] + C $$

lastly, for your initial condition

$$\begin{align} y[n] &= x[n] + (1-Q) y[n-1] \\ y[0] &= x[0] + (1-Q) y[-1] \\ C + y_\text{n}[0] (1-Q)^0 &= D + (1-Q)\cdot 0 \\ C + y_\text{n}[0] &= D \\ \end{align}$$

i guess that means that $ y_\text{n}[0] = D-C = D - \tfrac{D}{Q} $.

$$\begin{align} y[n] &= y_\text{n}[n] + C \\ &= y_\text{n}[0]a^n + C \\ &= (D - \tfrac{D}{Q})(1-Q)^n + \tfrac{D}{Q} \\ \end{align}$$

for $n \ge 0$.

$\endgroup$
  • $\begingroup$ As $n$ becomes large, the closed form becomes $D/Q$ as suggested by the book. I am not quite sure how I could have got this using the series expression as suggested by the book. It will take me a while (newbie here) but I am gonna try to understand what you have done which seem quite advanced for me! Any suggestion of what I can type into youtube? $\endgroup$ – Chika Sep 14 '17 at 21:19
  • $\begingroup$ you would have to recognize the series as that of an exponential function (and the constant term would be a little different). $\endgroup$ – robert bristow-johnson Sep 14 '17 at 21:54
  • $\begingroup$ you say you're a newbie. have you had a class in differential equations? although this is a finite difference equation (not having derivatives), the same concepts are used. but any of these LTI difference equations will have exponentials in their solutions. $\endgroup$ – robert bristow-johnson Sep 14 '17 at 21:56
  • $\begingroup$ I did some classes years ago on differential equations, I will find and review, thanks a lot! $\endgroup$ – Chika Sep 14 '17 at 22:51

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.