ELEC/TELE/PHTN 4123

Electrical/Telecommunications Design Proficiency

Lab2 Control/Electronics System Design
z3371039 Lian HOU
z3371069 Zheyu WENG
Requirements Analysis
The main task of this lab is to design a design an control system which can
control the output of power generation plant and minimize the transient
effects in both load and non load scenarios. The power generation plant
includes an generator and an motor which represents the turbine. The
turbine(motor) will be driven by the voltage provided by control system
designed in this lab. The specific requirement need to be included:

1. The desired voltage should be set from using Matlab via DAQ
2. Model the plants in order to design control system.
3. The system should work in load and non load situation. small value resistor will
be used in load situation.
4. Overshoot is smaller than 5 percent
5. Controller should have zero stead state error.
6. The speed and output voltage measurements should be displayed in Matlab.
Components
1 Bread Board
1 Motor Generator Set
Resistors
Capacitors
4 LM741 op-amp
High level design
reference controll current The verificatio
input er amplifier plant n
Desired voltage is from DAQ analogue pin and set up in Matlab. With DAQ,
analogue port can output exactly the same value we input to the computer.
So it can be used as reference voltage to the control system.
A negative feedback control system will be implemented in this design. The
output of generator plant connects back to control system. Before starting the
controller designing, the transfer function of generation plant is required to
model first. And base on the transfer function, controller needs to calibrate
the whole system to match the requirements above, such as minimizing
steady state error and overshoot.
The motor driver current amplifier is given. The whole controller system is
purely voltage-voltage type. Which means that the input and output of system
is demonstrated by voltage. The system cannot guarantee the current is high
enough to drive the motor. So the motor driver current amplifier is necessary
to be used between controller and plant.
The generator plant includes a motor and a generator. The motor represents
turbine and the speed is controlled by the voltage of input. The transfer bar
will put generation in motion.
In the display and verification part, the speed of generation and the output
voltage provided by generator should be displayed in Matlab via DAQ. The
method of speed measurement can be achieved by using a light and a photo
transistor. Both transistor output and output of generator connect to DAQ
analogue input.
Low Level Design
MODELLING
Function Derivation
Before the controller is implemented, the transfer function of system(including
generator and motor) needs to be estimated first.

Graph 1
Ra : resistance of the armature winding
La : inductance of the armature winding
Graph 1 show the equivalent circuit of motor. the assumption is that
permanent magnet which can provide fixed field is in the motor. So current
from power supply only flow into armature circuit.
Theoretical deviation will show below by some assumption.

LI a ( s )s+ R a I a ( s ) +V b ( s ) =Ea ( s ) ......(1)

T (s )
I a ( s) = m (Kt is armature constant)
Kt
V b ( s )=K bs m (s ) (Kb is voltage constant)
R
( a+ Las )T m ( s )
+ K bs m ( s )=Ea ( s ) .......(2)
Kt

motor J generator

Fiction force is ignored in transfer bar.

T m ( s ) =J mssm ( s )
Substituted equation(2) and assumed that the armature inductance La is
small compared to the armature resistance, Ra.
m ( s ) K1
=
E a ( s ) s(s+ a1)
K1 and a1 is a constant number combined by different value
The generator has the same structure as the motor.

E out ( s ) s(s+ a2)

=
m ( s ) K2
The final result will be like

E out(s ) s+a 2
=
E (s) s+ a1
So the assumption of transfer function is first order equation.
Low Pass Filter
When the input is constant 5 volt DC from DAQ analog output through motor
deriver. We can see the output is noisy. A digital filter will be implemented in
order to reduce noise.
Now the transfer function will be estimated by using step response graph.

Graph 2. Unfiltered output Graph 3. Output under different cut off

frequency
In previous step, the transfer function we assume is first order. The easiest
approach is finding time constant directly after filtering. But the problem is
how to design low pass filter, when we try to remove noise, genuine
information may cut off as well. The filter used is for noise removal but the
slight delay or shift influence will happen in low frequencies. The graph 3 here
shows step response after filtering by two different cut off frequencies low
pass filters. Time constant between two step responses is different and it is
impossible to determine which one is more close to the real step response.
Therefore, using low pass filter cannot be chosen.

Least square method

The second option is using linear least square method to estimate the time
constant. Least square method is the most common method for data fitting.
least square means the sum of square error will be minimized. In this
t
modeling task, after assumption of equation V =V 0(1e )

This method can be used to estimate parameter . V 0 is final value of

voltage, it get from step response output value.
Linearization should be done before using least square method
t
V
=(1e )
V0
t
V
1 =e
V0
V
(
log 1
V0 )=
t

V
cause there are lots of noise in the data, it is possible that is higher than
V0
V
1. So 1 can be negative, which is not acceptable.
V0

So, non linear least square method will be used.

 t
The assumption of equation still is V =V 0(1e ) ;

Both and V 0 will be treat as parameters

the basic idea of non linear least square method is that:

y is output voltage of step response in this model.

t
V =V 0(1e ) ;

and V0 is parameter, And they will be given 1 at first time. t is

variable.

t t
V V 1
=1e =V 0te ( 2 )
V 0

delta beta is the gap between real result and estimate result.

Substituted in equation below

According to least square method formula,

V V
c = [ V 0 ] ; X =[ ];
V 0

Update the value of and V0

Do 10 times iteration.
Graph 4. Modeling transfer function
t
(
V =V 0 1e ; ) V 0 = 4.0969; = 0.1726;
To find transfer function from unit step response, Laplace Transform table will
be used.

The final result, the open loop transfer function is.

19.598
KG( s)=
s+22.22

Controller Design (Design of PID controller)

Plant K*G(s)
Motor Motor And
Controller Generator
Driver
C(s)

Feedback
Graph 5. Block Diagram H(s)

19.598
With the transfer function of Plane known, KG ( s )= , the PID
s+22.22
controller has the transfer function know, C(s) = Kp + Kd*s +Ki/s. According to
block diagram, the open loop transfer function with controller is
19.598(Kp+ Kds+ Ki /s)
KC ( s)G( s)= . ----
s+22.22
Since there is a unity feedback H(s) = 1 ----
KC ( s )G( s)
The overall transfer function is T ( s )= . ----
1+ KC ( s )G ( s )H (s)

By substituting in and into equation , the closed-loop transfer function
can be found.

19.598(Kd s 2 + Kps+ Ki)

19.598 Kd +1
T ( s )=
19.598 Kp+22.22 19.598 Ki
s2 + s+
19.598 Kd+1 19.598 Kd+1
Apparently the closed-loop transfer function is a second-order system.
Applying the property of second-order system.
19.598 Kp+ 22.22
2 n= ----
19.598 Kd +1
19.598 Ki
n 2= ----
19.598 Kd+1
According to the requirements:
1. Maintain max 5% overshoot. %OS = 5%
Damping Ratio

2. Minimize the rise time.

Since the Rise time has a relationship with damping ratio as shown below. By
modeling the relationship between Overshoot and Damping Ratio with Matlab.

Graph 6. Rise time vs Damping Ratio Graph 7. Damping Ration

vs %OS
It's shown that Rise time increases with Damping Ratio, and in ordering to
reduce Rise time(as well as Damping Ratio), the Percentage OS needs to be as
big as possible. Therefore the %OS value will have to be taken as big as
possible.
Therefore, substituting %OS =5%, the damping ratio = 0.69.
There is no requirement of settling time, so it's assumed that the settling time
cannot be worse than the plant transient response.
4 4
For the plant transient response, Ts= = =0.18 s . Therefore, say the
a 22.22
closed-loop transient response has a settling time Ts0.18s.
4
For second-order system, T s= 0.18 , therefore n 22.22 . As
n
=0.69 known, n 32.2 , say n=33 .

Referring to the requirements of the PID controller:

1. minimise the rise time (Can be achieved by varying Kp)
2. maximum overshoot of 5% (Can be achieved by varying Kp)
3. the controller should have zero stead state error (Can be achieved by
 varying Ki)
Considering the plant can be modeled as a first-order system, the
complication of the controller circuit and most importantly the requirements
can be achieved by PI controller only. So During the design, a PI controller will
be implemented first to see if it meets all the requirements in this case Kd =
0. Substituting n=33, =0.69, Kd=0 to equation and .
The solution is:

Ki = 55.57
Kp = 1.1899
Kd = 0

Substituting Ki and Kp into the Closed-loop transfer function T(s).

19.598(Kd s 2 + Kps+ Ki)

19.598 Kd +1
T ( s )=
19.598 Kp+22.22 19.598 Ki
s2 + s+
19.598 Kd+1 19.598 Kd+1

23.320( s+ 2.383)
T ( s )= 2
s +45.54 s+ 1089.061

Routh Table
S2 1 1089.061
S1 45.54 0
S0
[ 1
45.54
1089.061
0 ] [ 1
45.54
1089.061
0 ]
45.54 45.54
= 1089.039 =0
According to the Routh table, there is no sign change, therefore the system is
stable.

Circuit Design
Op-amp
2

Feedba
ck
Output
Vref

Op-amp Op-amp
1 4
Op-amp
3
Graph 8. Circuit Diagram
1 R2
As Ki= and Kp= . With Ki and Kp known, taking one value of
R 3C R1
the components in each equation, the other value could be calculated.
The op-amp 1 is an error amplifier, the non-inverting node of op-amp 1 is
connected to the reference voltage connected to the output of DAQ, the
inverting node of op-amp 1 is connected to the feedback. The output will be
Gain*(Vref-Vfeedback). In our case, Gain =1.

Parameter Rise time Overshoot Settling time Steady-state error Stability[9]

Decrease Increase Small change Decrease Degrade

Decrease Increase Increase Eliminate Degrade

Graph 9. Effect of Ki Kp
The op-amp 2 and 3 is a proportional controller and an integral controller
respectively. The proportional controller decreases rising time and integral
controller eliminates steady state error.
Since the gain of op-amp 2 and 3 are negative, therefore a summing and
inverting op-amp 4 is added in the circuit.
Since the value of Ki and Kd is known.
1 R2
Ki= Kp=
R 3C R1
The final values are C = 10uF, R3 = 1.8K, R1 = 10K, R2 = 12K.

Analog low pass filter

In order to reduce the noise of output, analog low pass filter will be
implemented after the generator.
It's not a good idea to put a LPF in the circuit to observe the transfer function,
however it's necessary to put one behind the output of the generator. The
unfiltered output is really noisy and unable to observe its performance and the
final value. For modulation of transient response, tiny difference in the
transient response will affect the final transfer and design of PI controller.
However the real output of the system is a DC constant voltage, the final
value will be less affected.
By setting the cut-off frequency at 5Hz, wc = 1/RC = 2*5*pi
Picking C = 220uF. Thus R is 144.686,150 is picked.

Decoupling capacitor
A decoupling capacitor is the capacitor connect between power supply lines
and the ground. Decoupling capacitor not only can kill the noise from power
supply lines, it also can cancel noise from other circuit elements which is
shunted to the capacitor. Controller based on negative feedback still is
sensitive to the noise. decoupling capacitor is necessary to add in the circuit.
There are three 220uF decoupling capacitors, through V++ and G, V-- and G,
V++ and V-- respectively.

Load and non load scenarios

Graph 9. Load and no-load
This is a simple example to show the different between load and non load
scenarios. because of voltage drop from Rout, Rload Voltage cannot achieve
5V anymore. In this task, the output voltage of generation plan in load
scenarios is smaller than it in non load scenarios. so DC gain of transfer
function will be different. It would influence controller design. The previous
steps are all base on non load scenario. because it is impossible to design two
circuits to be confront to two different scenarios. The components value would
be chosen or changed according to the loading test after non loading test.

Speed Measurement
The method of speed measurement is using photo transistor and a light. The
light is connect to 5V power supply instead of theatrical value 9V from
datasheet, because it is impossible to connect a 9V from power supply in this
lab. A disc with an empty hole is on the transfer bar that between motor and
generator. An light and an photo transistor are on different sides of the disc.
Light will go through the hole on the disc once a rotation and transistor will
receive input. The graph below shows components inside the plants and the
circuit should be partially developed. DAQ analogue input pin would get 5V
signal once photo transistor receive light.

Graph 10. Sample output of Photo Transistor Graph11. FFT response of

output
The graph shows the output of transistor and each pulse means a rotation.
Therefore, Frequency analysis is the simplest way to find the speed or
frequency of data. DC should be remove otherwise the amplitude of DC will be
the biggest after Fourier Transform. Since transistor output is not an pure sin
wave, it have to include harmonics. but the biggest amplitude one or first
peak would be the target frequency. Using fft and max command to find the
peak of frequency which is speed of generator.

Implementation and Test Plan

Before building the circuit, the value of components is determined. However
the transfer function from modeling is based on some assumptions. And the
parameter of PI controller calculation is based on transfer function. Therefore,
checking the actual circuit can meet the specific requirement is important. If it
does not satisfy to the requirement, then the adjust the components value.
The table below should be fulfilled.

Cause the parameter of inverter and summer's component is constant, so

they will not be changed.
Table 1: Nominal and Actual Component Values and Design Parameters of PI
controller
Component Nominal value Actual value
R1 10K
R2 12K
R3 1.8K
C1 10uf

Table 2: Nominal and Actual Component Values and Design Parameters of low
pass filter

Component Nominal value Actual value

R11 144.686
C12 220uF

Table 3: transient response of different desired voltage

desired With load Steady overshoot Response Output
voltage state time volatge
Error
2V Yes
2V No
3V Yes
3V No
4V Yes
4V No
1. Both current amplifier and PI controller set power supply Vcc(12), Vee(-
12),ground.
2 Use Matlab to display the output voltage of generator.
3. If response of output is not expected, test isolated Op-amp and
demonstrate its functionality.
4. Check input and output of proportional amplifier
5. Check input and output of integrator
6. Check input and output of summer.
7. Check input and output of inverter.
8. Check k the capacitors polarity.
9. If the circuit is work, change the value of Kp and Ki which is based on
R1,R2,R3,C1, if steady state error or overshoot does not meet the
requirement.

table 4: impact of changing Kp and Ki.

Paramet Rise Overshoo Settling Steady-state Stability[

9]
er time t time error

Small
Decrease Increase Decrease Degrade
change

Decrease Increase Increase Eliminate Degrade