How to get (frequency response) from this [difference equation]?

AI Thread Summary
The discussion focuses on deriving the frequency response H(f) from a given difference equation of a bandpass FIR filter. The user expresses difficulty in handling the double sums in the equation for y(k) and attempts to identify patterns through expansion. They apply the Z-transform to express the transfer function H(z) but struggle to relate the resulting expression to the desired frequency response when evaluated on the unit circle. The conversation highlights the need for assistance in connecting the derived Z-transform results to the expected frequency response formula. Overall, the thread emphasizes the complexities involved in analyzing FIR filters through difference equations and Z-transforms.
courteous
Messages
37
Reaction score
0

Homework Statement


Given this difference equation y(k) (of a bandpass FIR filter) ...
y(k) = \frac{1}{K^2} \sum_{m = k-K+1}^k \; \sum_{n = m-K+1}^m x(n) - \frac{1}{L^2} \sum_{m = k-L+1}^k \; \sum_{n = m-L+1}^m x(n)

... how does one derive this frequency response H(f)?
H(f) = \frac{1}{K^2} \left( \frac{\sin{\pi f K}}{\sin{\pi f}} \right)^2 e^{-2j\pi f(K-1)} - \frac{1}{L^2} \left( \frac{\sin{\pi f L}}{\sin{\pi f}} \right)^2 e^{-2j\pi f(L-1)}

Homework Equations


Usually, I'd do Z-transform on (disregarding general case of IIR filters) y(n) = \sum_{k=0}^K b_l x(n-k) to get the transfer function H(z) = \frac{Y(z)}{X(z)} = \sum_{k=0}^K b_l z^{-k}X(z) and then evaluate on unit circle (i.e. change z with e^{jw}).

The Attempt at a Solution


Don't know how to handle those double-sums in y(k). :shy:

Since the issue is "symmetrical", I've tried solving the 1st half of y(k) (the lowpass filter with a higher cutoff frequency, K < L). I've tried expanding the sum, hoping a pattern would emerge:
\sum_{m = k-K+1}^k \; \sum_{n = m-K+1}^m x(n) = \\<br /> \left[x(k-2K+2) + x(k-2K+3) + \dots + x(k-K+1) \right] +\\<br /> \left[x(k-2K+3) + x(k-2K+4) + \dots + x(k-K+2) \right] +\\<br /> \cdots +\\<br /> \left[x(k-K+1) + x(k-K+2) + \dots + x(k) \right]

So, there's some kind of "triangular/diagonal pattern" since x(k-2K+2) and x(k) (the very first and last x-element) are the only one to appear only once. Now what? Is this even the right approach? :confused:
 
Last edited:
Physics news on Phys.org
I'd appreciate (any) help. Continuing from 1st post:
\cdots = \left[ x(k-2K+2) + 2x(k-2K+3) + \cdots + (K-1)x(k-K+1) + \cdots + x(k) \right]

Applying Z-transform, we get
X(z) \left[ z^{-2K+2} + 2z^{-2K+3} + \cdots + (K-1)z^{-K+1} + \cdots + 1 \right]

So, the transfer function H(z) of the whole y(k) (from the 1st post) can be written as
H(z)=\frac{Y(z)}{X(z)}= \frac{1}{K^2}\left[ z^{-2K+2} + 2z^{-2K+3} + \cdots + (K-1)z^{-K+1} + \cdots + 1 \right] - \frac{1}{L^2}\left[ z^{-2L+2} + 2z^{-2L+3} + \cdots + (L-1)z^{-L+1} + \cdots + 1 \right]

If the above is correct, then how can this
\left[ z^{-2K+2} + 2z^{-2K+3} + \cdots + (K-1)z^{-K+1} + \cdots + 1 \right]
be equal to this, when evaluated on the unit circle z=e^{j2\pi f}?
\left( \frac{\sin{\pi f K}}{\sin{\pi f}} \right)^2 e^{-2j\pi f(K-1)}
 
Last edited:
I haven't made any progress on this ... anybody help me out, please?
 

Similar threads

Replies
0
Views
141
Replies
1
Views
2K
Replies
2
Views
2K
Replies
3
Views
3K
Replies
2
Views
2K
Replies
10
Views
2K
Back
Top