Professional Documents
Culture Documents
Mark Fowler
Note Set #30
C-T Systems: Laplace Transform and System Stability Reading Assignment: Section 8.1 of Kamen and Heck
1/30
Ch. 2 Convolution
C-T System Model Convolution Integral D-T System Model Convolution Sum
Y ( s ) = X ( s ) H ( s ) y (t ) = L -1{X ( s ) H ( s )}
Much insight can be gained by looking at H(s) and understanding how its structure will affect the form of y(t). Section 8.1: First well look at how H(s) can tell us about a systems stability Section 8.4: Then well see how the form of H(s) can tell us about how the system output should behave Section 8.5: Then well see how to design an H(s) to give the desired frequency response.
3/30
2. All poles are in the open left-half of the s-plane 3. Routh-Hurwitz test (section 8.2, well skip it)
4/30
This is not absolutely integrable as you integrate its absolute value from 0 to the value of the integral keeps growing without bound So the system having this impulse response is not BIBO stable it is unstable that means that there is a bounded input that will (eventually) drive the systems output to infinity.
Note: Real H/W will encounter problems long before the output goes to infinity!!!
x (t ) = sin(0t )u(t )
5/30
h( ) = sin(0 )u( )
x (t ) = h (t )
x(t )
x(t ) h() t
To get y(t) Integrate gives area of the humps as t grows you get more and more humps output grows without bound!!! Thus, the upper envelope of the output grows by the area of one of those humps each time we increase t by one-half the period of the sinusoid So the upper envelope grows linearly with time it grows without bound!!!
6/30
Doing numerical calculations like this are helpful but they dont PROVE that something happens the plot by itself does not show what happens after 0.5 seconds!! As far as we know anything could happen out there!! That is the value of mathematical analysis (based on sound models, of course!)
7/30
So we can use the absolute integrability check but it is not that easy to do in general. It also doesnt tell us much about why a system is stable. Well if the impulse response can be used to check for stability it should be no surprise that the transfer function can also be used!!!
h(t ) = h1 (t ) + ... + h p (t )
From our understanding of partial fraction expansion we know what each of these terms can look like: For Complex Pole Pair For Real Pole hi (t ) = ci e t cos(i t + i )u (t ) distinct pole hi (t ) = ci e p t u (t ) distinct pole
i
hi (t ) = ci e pit
This decays if pi < 0
We can capture the condition for decay with: the real-part of the pole is negative Although we wont prove it here it can be shown that this decay is fast enough to ensure absolute integrability and thus stability. For the repeated pole cases:
hi (t ) = ci t k e pi t u(t )
Weve got a race!!! The tk terms are going up and the exponentials are going down for poles whose real parts are negative WHO WINS???
So we have argued that a system is stable if all its poles have negative real parts
Re{pi } < 0, i
Stability condition (not proved here!) An Nth order system (with N poles pi = 1, 2, , N) is stable if and only if
x x
10/30
A system is said to be marginally stable if it has at least one distinct pole on the j axis but no repeated poles on j Marginally stable systems are not a good idea in most applications because they will still tend to get large outputs under certain conditions Comment: Just because the model satisfies BIBO stability, this does not mean the physical system will have no problems:
y (t ) < C2
Summary
In most applications we desire a stable system We can easily check for stability by looking to see where the systems poles are Also We wont show it here, but every stable system has an H(s) whose ROC includes the j axis therefore, the FT-based frequency response H() exists.
11/30
x(t )
y (t )
Recall:
So all we need to do is find where its poles are!!!! Use the Quadratic Formula on the denominator of H(s):
p1, 2 R 1 R = 2L 2 L LC
2
From this result we can find out how the component values impact the stability of this system. So well now systematically analyze this well look at four cases.
12/30
R 1 R = 2L 2 L LC
j s-plane
1 / LC 1 / LC
So for this case the poles are purely imaginary. They lie on the j axis Marginally Stable
1 / LC s 2 + (1 / LC )
h (t ) =
1 sin 1 / LC t LC
[(
)]
But we cant really build an oscillator this way because all real L & C have parasitic resistance (need to use special electronic circuits to get an osc.) All real passive circuits are stable
13/30
Case 2:
p1, 2
Note : If
1 R < LC 2L
R 1 R = 2L 2 L LC
the
~~ term is < 0
j
p1, 2
R = j0 2L
0 =
1 R LC 2 L
1 R LC 2 L
R / 2L
1 R LC 2 L
h (t ) = Ae( R / 2 L ) t cos(0t + )
h(t ) t
0< R<
2L LC
Case 3:
p1, 2
R 1 R = 2L 2 L LC
Note : If
1 R > , LC 2L
j
h(t ) = c1e 1t + c2 e 2t
h(t ) t
R>
2L LC
Case 4:
p1, 2
R 1 R = 2L 2 L LC
Note : If
2L R= , LC
h(t ) t
16/30
Note: For fixed L, C values we can drastically change the circuits behavior by adjusting R
Marginally stable stable
0
h(t) persistently oscillates h(t) decaying oscillation
2L LC
See the plot on the next page for pole positions as R is varied
17/30
j R=0
R=
2L LC
Repeated Roots
2L 0< R< LC
R=0
18/30
L
+
x(t)
y(t)
H ( s) =
2L LC
s 2 + (R L )s + 1 LC
LC
If R <
Now well explore the effect of changing the component values to change the location of the poles for the complex poles case Youve already seen the next 3 slides they are repeated here as a reminder!!!
19/30
From the Pole-Zero Plots we can Visualize the TF function on the s-plane:
The circuit were looking at doesnt have the zero at the origin!!
20/30
From our Visualization of the TF function on the s-plane we can see the Freq. Resp.:
21/30
Can also look at a pole-zero plot and see the effects on Freq. Resp. As the pole moves closer to the j axis it has a stronger effect on the frequency response H(). Poles close to the j axis will yield sharper and taller bumps in the frequency response.
By being able to visualize what |H(s)| will look like based on where the poles and zeros are, an engineer gains the ability to know what kind of transfer function is needed to achieve a desired frequency response then through accumulated knowledge of electronic circuits (requires experience accumulated AFTER graduation) the engineer can devise a circuit that will achieve the desired effect.
22/30
Effect of Changing o
23/30
j
2000 1000 j 0
0 1000
o = 500 a
-1000 -2000 -1000 600 400 h(t) 200 0 -200 1.5 1 |H( )| 0.5 0 -1 0 0.01 0.02 0.03 0.04 0.05 t (sec) 0.06 0.07 0.08 0.09 0.1
-0.8
-0.6
-0.4
-0.2
0 (rad/sec)
0.2
0.4
0.6
0.8 x 10
1
4
24/30
j
2000 1000 j 0
0 1000
o = 200 a
0.01
0.02
0.03
0.04
0.05 t (sec)
0.06
0.07
0.08
0.09
0.1
-0.8
-0.6
-0.4
-0.2
0 (rad/sec)
0.2
0.4
0.6
0.8 x 10
1
4
25/30
j
2000 1000 j 0
0 1000
o = 50 a
-1000 -2000 -1000 1000 500 h(t) 0 -500 -1000 10 0 0.01 0.02 0.03 0.04 0.05 t (sec) 0.06 0.07 0.08 0.09 0.1
|H( )| 5
0 -1
-0.8
-0.6
-0.4
-0.2
0 (rad/sec)
0.2
0.4
0.6
0.8 x 10
1
4
26/30
Effect of Changing o
27/30
j
2000 1000 j 0
0 1000
o = 200 a
0.01
0.02
0.03
0.04
0.05 t (sec)
0.06
0.07
0.08
0.09
0.1
-0.8
-0.6
-0.4
-0.2
0 (rad/sec)
0.2
0.4
0.6
0.8 x 10
1
4
28/30
j
2000 1000 j 0
0 1000
o = 200 a
-1000 -2000 -1000 1000 500 h(t) 0 -500 -1000 5 4 |H( )| 3 2 1 0 -1 -0.8 -0.6 -0.4 -0.2 0 (rad/sec) 0.2 0.4 0.6 0.8 x 10 1
4
0.01
0.02
0.03
0.04
0.05 t (sec)
0.06
0.07
0.08
0.09
0.1
29/30
j
2000 j 0
0 1000
o = 200 a
0.01
0.02
0.03
0.04
0.05 t (sec)
0.06
0.07
0.08
0.09
0.1
6 |H( )| 4 2 0 -1
-0.8
-0.6
-0.4
-0.2
0 (rad/sec)
0.2
0.4
0.6
0.8 x 10
1
4
30/30