You are on page 1of 9

Preliminary Theory (Expt 2)

Linear State Variable Feedback Controller


1.1 Plant:
Let the transfer function of a SISO system be G(s).

The plant was chosen whose transfer function is:


1000
𝐺𝐺 (𝑠𝑠) = 𝑠𝑠 3 +10𝑠𝑠 2 +100𝑠𝑠+1000 (1.1)

1.2 Implementation of Plant


The plant is first modelled in the Controllable Canonical Form (CCF) state space representation:
Dividing both the numerator and denominator of G(s) by s3, we have
1000
𝑠𝑠 3
𝐺𝐺(𝑠𝑠) = 10 100 1000
1+ + 2 +
𝑠𝑠 𝑠𝑠 𝑠𝑠 3

By Mason’s gain formula, the transfer function of a system as obtained from its signal flow graph is given
by,
1
𝐺𝐺 (𝑠𝑠) = ∑𝑃𝑃 𝛥𝛥
𝛥𝛥 𝑖𝑖 𝑖𝑖
Where, Δ=the determinant of the graph
Pi=path gain of the ith forward path between input and output
Δi= the cofactor value of Δ for the ith forward path, with the loops touching the ith forward path
removed.

By analogy with Mason’s gain formula [N. Nise], the transfer function can be represented in the form of
a signal flow graph as shown in figure 1.1:

Transforming this signal flow graph into a Simulink block diagram, we get the following system:

Fig 1.1
Fig. 1.2

Now the plant is modelled in the Controllable Canonical Form (CCF) state space representation:
In the graph of Fig. 1.1 let us define a set of state variables x1, x2, and x3 by assigning one state after each
integrator. Then the variable just before the integrator will be the derivative of the corresponding state.
The state and output equations can then be written from the graph in Fig. 1.1 as:

𝑥𝑥1̇ (𝑡𝑡) = 𝑥𝑥2 (t)


𝑥𝑥̇ 2 (𝑡𝑡) = 𝑥𝑥3 (t)
𝑥𝑥̇ 3 (𝑡𝑡) = −1000 𝑥𝑥1 (t) −100 𝑥𝑥2 (t) −10 𝑥𝑥3 (t)+u(t)
y(t) = 1000 𝑥𝑥1 (t)

Thus, we can now write,


𝑥𝑥̇1 0 1 0 𝑥𝑥1 0
𝑥𝑥̇
� 2� = � 0 0 𝑥𝑥
1 � � 2 � + �0� 𝑢𝑢 (1.4)
𝑥𝑥̇ 3 −1000 −100 −10 𝑥𝑥3 1
𝑥𝑥1
𝑦𝑦 = [1000 0 0] �𝑥𝑥2 � (1.5)
𝑥𝑥3
Then the state space model will be

𝑋𝑋̇(𝑡𝑡) = 𝐴𝐴𝐴𝐴(𝑡𝑡) + 𝐵𝐵𝐵𝐵(𝑡𝑡) (1.6)


𝑦𝑦(𝑡𝑡) = 𝐶𝐶𝑋𝑋(𝑡𝑡) (1.7)

𝑥𝑥1 (t) 0 1 0 0
Where, 𝑋𝑋(𝑡𝑡) = �𝑥𝑥2 (t)�;𝐴𝐴 = � 0 0 1 �; 𝐵𝐵 = �0�; 𝐶𝐶 = [1000 0 0];
𝑥𝑥3 (t) −1000 −100 −10 1

(1.4) is in Controllable Canonical Form. Now x1, x2,and x3 are used as inputs to the LSVF controller.
1.3 Step response of uncompensated open loop plant
The poles of the uncompensated plant are at [-10, 10j, -10j]. Hence, in open loop, the system is
marginally stable. The time domain response of the plant to a unit step input is shown in Fig. 1.3.

Fig. 1.3
1.4 LSVF Controller: In LSVF control, state variables x1, x2,and x3, are fed back to the input via gain
blocks k1,k2,and k3 respectively. The block diagram of an LSVF controller is shown in Fig. 1.4:

The control law is: Fig 1.4

𝑢𝑢(𝑡𝑡) = −𝐾𝐾𝐾𝐾(𝑡𝑡) + 𝑟𝑟(𝑡𝑡) = [−(𝑘𝑘1 𝑥𝑥1 (𝑡𝑡) + 𝑘𝑘2 𝑥𝑥2 (𝑡𝑡) + 𝑘𝑘3 𝑥𝑥3 (𝑡𝑡)) + 𝑟𝑟(𝑡𝑡)] (1.8)

Replacing (1.8) in (1.4) the closed loop system is obtained as:


𝑥𝑥1̇ (𝑡𝑡) = 𝑥𝑥2 (t)
𝑥𝑥̇ 2 (𝑡𝑡) = 𝑥𝑥3 (t)
𝑥𝑥̇ 3 (𝑡𝑡) = −(1000 + 𝑘𝑘1 )𝑥𝑥1 (t) – (100 + 𝑘𝑘2 ) 𝑥𝑥2 (t) – (10 + 𝑘𝑘3 ) 𝑥𝑥3 (t)+u(t) (1.9)
y(t) = 1000 𝑥𝑥1 (t)

The transfer function of the LSVF controlled plant then becomes:


1000
𝐺𝐺𝑐𝑐𝑐𝑐 (𝑠𝑠) = (1.10)
𝑠𝑠 3 +(10+𝑘𝑘 3 )𝑠𝑠 2 +(100+𝑘𝑘 2 )𝑠𝑠+(1000+𝑘𝑘 1 )

By varying k1, k2 and k3, the poles of the compensated system can be placed at any desired location.The
Simulink diagram representation of the LSVF controlled closed loop system is shown in Fig. 1.5.

Fig. 1.5
The LSVF controlled plant is to be implemented using an Op-Amp based circuit. To avoid saturation in
the Op-Amps, individual gain blocks can have a limited range of gain values. Thus, although theoretically
k1, k2 and k3 can assume any value, in practice their range is limited to within ±6 for the schematic shown
in Fig. 1.4 (using a step input of 1 volt and a DC supply of ±15V to the Op-Amps). This does not allow us
to get a wide range of plant outputs using the LSVF controller.

However, if we combine the gain blocks used for obtaining the CCF state variables with the LSVF gain
blocks as shown in Fig. 1.5, we can increase the ranges of the LSVF gains that can be implemented. In
this case, k1, k2 and k3 can assume any value within ±60, ±600, and ±6000 respectively. Thus, a wide
variety of plant responses can be obtained with the help of the LSVF controller. Hence, implementation
of the LSVF controlled plant is based on the schematic shown in Fig. 1.5.
Fig 1.5

1.5 Step Response of LSVF controlled plant


The step response of the LSVF controlled plant was obtained graphically using Simulink for 3 sets of LSVF
gains. The 3 sets of LSVF gains were chosen such that 3 different types of step responses could be
observed – underdamped, overdamped, and critically damped.

1.5.1 Underdamped Step Response


For k1=10, k2=10, and k3=10, using (2.13), the transfer function of the LSVF controlled plant becomes:
1000
𝐺𝐺𝑐𝑐𝑐𝑐 (𝑠𝑠) =
𝑠𝑠3 + 20𝑠𝑠2 + 110𝑠𝑠 + 1010
Gcl(s) has poles at -17, -1.49+7.56j, -1.49-7.56j. Thus, the step response of this system is underdamped.
The simulated time domain response of the compensated plant to a unit step input is shown in Fig. 1.6.

Fig 1.6

1.5.2 Critically Damped Step Response


For k1=0, k2=200, and k3=20, using (1.13), the transfer function of the LSVF controlled plant becomes:
1000
𝐺𝐺𝑐𝑐𝑐𝑐 (𝑠𝑠) =
𝑠𝑠3 + 30𝑠𝑠2 + 300𝑠𝑠 + 1000
Gcl(s) has 3 poles at -10. Thus, the step response of this system is critically damped. The simulated time
domain response of the plant to a unit step input is shown in Fig. 1.7.
`

Fig 1.7
1.5.3 Overdamped Step Response

For k1=0, k2=520, and k3=52, using (1.13), the transfer function of the LSVF controlled plant becomes:

1000
𝐺𝐺𝑐𝑐𝑐𝑐 (𝑠𝑠) =
𝑠𝑠3 + 62𝑠𝑠2 + 620𝑠𝑠 + 1000

Gcl(s) has poles at -2, -10, and -50. Thus, the step response of this system is overdamped. The time
domain response of the compensated plant to a unit step input is shown in Fig. 1.8.

Fig 1.8

_________________________________________

You might also like