Professional Documents
Culture Documents
Frequency-Domain Analysis of Systems3
Frequency-Domain Analysis of Systems3
Mark Fowler
Note Set #19
C-T Systems: Frequency-Domain Analysis of Systems Reading Assignment: Section 5.2 of Kamen and Heck
Ch. 2 Convolution
C-T System Model Convolution Integral D-T Signal Model Convolution Sum
Recall that: Because h(t) and H() form a FT pair, one completely defines the other. h(t) and convolution completely describe the zero-state response of an LTI to an input i.e. h(t) completely describes the system.
Conv. Property from chapter 4!! Proof Step 1: Think of the input as a sum of complex sinusoids -Each component
= F ( )e jt
F ( )e jt
Step 2: We know how each component passes through an LTI -This is the idea of frequency response - H ( ) F ( )e jt is the out. component that is due to the input component Step 3: Exploit System Linearity (again Step 2 was the first time) -Total output is a sum of output components y (t ) =
1 2
[H ( ) F ( )]e jt d
Input-Output Relationship Characterized Two Ways 1. Time-Domain: y(t) = h(t)*f(t) 2. Freq-Domain: Y() = H()F() Given input f(t) and impulse response h(t), to analyze the system we could either: 1. Compute the convolution h(t)*f(t) or 2. Do the following: (a) Compute H() & compute F() (b) Compute the product Y() = H()F() (c) Compute the IFT: y(t) = F 1{ H()F()}
From the Freq-Domain view we can see how H() boosts or cuts the amounts of the various frequency components
Relationships between various modeling methods Recall: we are trying to find ways to model CT Linear Time-Invariant Systems in Zero-State
Example Scenario: You need to send a pulse signal into a computers interface circuit to initiate an event (e.g. next PTT slide) Q: What kind of signal should you use? Possibility: A rectangular pulse: AP (t ) A t
Q: Will this work? It depends on the interface circuitry already in the computer! Suppose the interface circuitry consists of an AC Coupled transistor amplifier as shown below
AC coupled
Input signal
Output signal
x(t )
y (t )
x(t ) X ( )
H()
y (t ) Y ( )
What is H()?? Use Sinusoidal Analysis to find it we did that once already for this circuit Use Phasors, Impedances, and Voltage Divider: v v Req V0 = V 1 i Req + jC
x(t )
y (t )
H ( ) =
jReq C 1 + jReq C
Nowwhat does the input pulse look like in the frequency domain? From FT table:
p (t ) sinc 2
X ( )
jReqC So the output FT looks like: Y ( ) = H ( ) X ( ) = sinc 2 1 + jReqC
Now how do we find y(t)?
y (t ) = F 1{ ( )} Y
YUCK!!! HARD!!!
Welldo we need to go back to the time domain? NO! Just look at Y() and see what it tells Think Parsevals theorem
The plots below show that very little energy gets through the system Systems Freq. Resp.
1 0.8 |H(f)| 0.6 0.4 0.2 -500 0 f (Hz) 500 1000 0 -1000 1 -500 0 f (Hz) 500 1000 0.1 0.08 |Y 1(f)| 0.06 0.04 0.02 0 -1000 01 -500 0 f (Hz) 500 1000
Input FT
0.1 0.08 |X 1(f)| 0.06 0.04 0.02 0 -1000 01
Output FT
The problem lies in that |H()| is small where |X()| is big (and vice versa) Pick an X() that does not do that!! Use a pulse that is Modulated Up to where |H()| allows it to pass
x2 (t ) = p (t ) cos(0t )
A modulated pulse
sinc shifted up
sinc shifted up
X 2 ( )
0
See actual plots on next page
Original Input FT
0.1 0.08 |X 1(f)| |H(f)| 0.06 0.04 0.02 0 -1000 0.1 0.08 |X 2(f)| |H(f)| 0.06 0.04 0.02 0 -1000 -500 0 f (Hz) 500 1000 -500 0 f (Hz) 500 1000 1 0.8 0.6 0.4 0.2 0 -1000 1 0.8 0.6 0.4 0.2 0 -1000
Original Output FT
-500
0 f (Hz)
500
1000
-500
0 f (Hz)
500
1000
0 -1000
-500
0 f (Hz)
500
1000
Alternate Input FT
Alt. Output FT
|X()|
Frequency (rad/sec)
X() (degrees)
Frequency (rad/sec)
Time (sec)
Comments on This Example We can use the FT to see at what frequencies there are undesired signals Then we can specify a desired system frequency response H() that will reduce (or attenuate) the undesired signal while keeping the desired signal
Note that it would be virtually impossible to try to directly specify a desired system impulse response that will do this
Once we have specified the desired H() we could try to find a circuit (i.e., a physical system) that will implement it (either exactly or approximately)
This is the design or system synthesis problem We havent yet learned how to do this!! Tools well learn later will help! However, if we have H() specified as a mathematical function we could possibly compute the inverse FT to get the impulse response h(t) then we could implement this digitally like we did earlier to simulate an RC circuit using D-T convolution.