Professional Documents
Culture Documents
Signals and Systems
Signals and Systems
Know that how to derive the gain, as a function of frequency, of the amplifiers covered
particularly for low-pass and high-pass filters (41,44)
Be able to express and interpret dB Be able to draw bode plots and analyse op amp circuits using Bode plots (41-46) Appreciate that the analogy between electrical circuits and electromechanical and electrothermal systems enables the application of Bode plots to transducers (sensors and actuators) (62-77) To be able to draw the Bode plot of transducer with a low-pass filter characteristic by use of its time constant in response to a step input.(7) Be able to include the response of a sensor with a low-pass filter characteristic into a system Bode plot (60-61)
78
What were going to cover in the remainder of the course Amplifier circuits for conditioning the output of sensors Sensors for measuring
especially for interfacing to a computer mechanical displacements temperature pressure
Further relating the temporal transfer function of a sensor/amplifier to its frequency response
impulse response convolution
79
80
Basic principles
measure a physical parameter that changes with the mesurand (i.e the thing to be measured)
e.g. temperature dependent resistors
microphones (movement to emf) tape heads (magnetic field to emf) gramophone pickups (stylus movement to emf)
Chemical generation
81
boost the signal to a level significantly above that of noise pick up - normally ~ 1V to drive an actuator to drive a device such as ADC to a significant fraction of full-scale
82
Modern cars employ engine management computers to optimise engine performance. The outputs from a variety of sensors are converted into voltages and read by a computer that calculates optimum operating conditions. The interface to the computer depends on the sensor principle And frequency-characteristics
low-pass, differentiator etc voltage, current source or resistor
Airflow Sensor Inlet air temperature Knock sensor Exhaust oxygen sensor Accelerator Position Sensor Oil and coolant temperature Sensors Voltage Sensor for engine load Exhaust Pressure Sensor Engine Speed Sensor
83
O2
V nal~0.45 g s en s o r s i
84
Volts Amps W
85
In most electrical sensors an electrical parameter varies relation in proportion to the parameter measured
output voltage output current resistance capacitance inductance
Signal conditioning
Prior to reading into a computer or for transmission this parameter is converted to a voltage and filtered For example- a record player (gramophone) pickup
This is done using op amplifies as previously discussed The process is called signal conditioning
the output is and amplified from a peak value of few mV to 1V the pickup acts as a differentiator - to give a flat frequency response the signal is also passed through an op amp integrator.
86
resolution or accuracy with which an analogue input voltage is converted into an output number.
For an n-bit ADC the input voltage is represented by a number between 2n-1 representing full-scale where n is the number of bits typical resolutions include 8, 12, 16 bits which represent voltages with 255, 4096 or 65,532 levels respectively For optimum use of the resolution, the output of the signal conditioning should match the input voltage range to that of the ADC
0 representing zero volts
The sampling frequency at which the input voltage is converted into numbers
e.g 0 to +5V
This should be at least twice the highest frequency present in the signal
87
ADCs are used to convert an analogue input voltage into a number that can be interpreted as a physical parameter by a computer.
0111 0110 0100 0101 0011 0010 0001 0000 1111 1110 1100 1010 1101 1011 1001 0000 0110 0111 0011 1100 1001 1011
88
For example: i.e. Read a number into an engine management computer that tells the computer what the angle of the throttle is.
Rotary potentiometer
ADC
a throttle angle
89
voltage is a fraction of supply voltage is typically a few V small gain required may require buffering with an amplifier to prevent sensitivity to load impedance and long cables
i.e. to make into voltage source
Vs
Vs
R1 Vo = Vs R2 + R1 x = Vs l
Ro =
R2 R1
R2 R1 R2 + R1
Vo ' =
Rf Ri
Vo
l x
Ro ' ~ 0
Potentiometer
Ri
Rf
90
Measurement techniques
Resistance varies in relation to a parameter use an amplifier to buffer against lower load impedances
thermisters (temperature) potentiometers (position)
VS
R1 + sensor RS
Vo
R f Rs Vo = 1 + R + R VS Ri 1 S
Vo = Rf Rs VS Ri R1 + RS
Ri
Rf
Vo =
Rf Ri
VS
91
Many sensors vary their electrical characteristic by a small fraction over the range over which they are used A differential amplifier can help to interface this to an ADC (or other device) to give a large fractional change
Wheatstone bridge thermister - resistance varies by 4%/C temperature IC varies by 1mA/K (i.e. 290-300 K = 3% change in i)
R3 R3 Rs R1
+ -
R1 Vo
+
R2 RS R2
x1
Vo' = Vo -
R2 Vs R1 + R2
92
For example:an electronic medical thermometer Employs a thermister to measure temperature in range 35-40 C.
< 0.1 C sensitivity required 8-bit ADC Thermister RS=10kW at 35 C Temperature coefficient
v0 vmax
0.5 0.4 0.3 0.2 0.1 0 35 36 37 38 39 temperature 40 128 255
VS
Vmax=0.549V
Vo
10 k sensor
Vref
ADC RS
Vo = Rs VS R1 + RS
93
Variation is 0.049/0.549=9%
ADC output
v0 vmax
0.22
36 37 38 39 temperature
VS
235 40
Thermister voltage can be applied directly to ADC without amplification But the poor use of the ADC dynamic range means that the required resolution (0.1 C ) is not achieved...
voltage across thermister is high ADC has high input impedance ADC is close to thermister
10 k
Vo
ADC sensor R S
Vo = Rs VS R1 + RS
94
ADC output
Or in graphical format
0.549
255
A solution is to subtract the bias voltage with a differential amplifier (slide 93) The voltage range is unchanged, but the voltage at the ADC input at the minimum temperature of 35 C is now zero volts Using a differential amplifier with a gain of 10 with a Vref of 0.5V is now 0.49V, making good use of the ADC dynamic range VS 10k 10k Vo VS/2 RS 10k 0.5VS + x10
Vo
Vref
Vo'
ADC
95
255
0.023 10
0.02
5
0 35.5
96
ADC output
v0 vmax
ADC output
v0 vmax
Recap on sensors
Use in a potential divider to convert to voltage A Wheatstone Bridge and differential amplifier increases fractional change in voltage But one should take care
removes voltage offset so that voltage is zero at a specific temperature a potential divider is effected by
variations in supply voltage load impedance
potentiometer
97
Voltage sources
Ideal sources
Output voltage is independent of load current All real sources have a finite output impedance
Vo=Vs-iRi
a.k.a. internal resistance when the load draws a current i a voltage iRi is dropped across Ri To be a voltage source require RL>>Ro
Examples
VS
Current sources
Ideal sources
and hence for the current to be independent of load (i.e. a current source) we require RL <<Ro
for example, the virtual earth of a current-to-voltage convertor
io Ro RL
99
For example
output voltages of mV (electrostatic) mV (moving coil) gains of 100s-1000s to reach levels of 1 volt used for connecting equipment output ~0.45 V
Rwires +
Vo
Vo = -
Rf Ri
VS
Rf
Rf VS Vo = 1 + Ri
R f Ri Vo = 1 + Ri Ro + Rwires + Ri VS
For example
Rf=100kW, Ri=100W =>G=1001, amplifier microphone mV output to 1V line level 741 with G.BW product ~106?
100
A photodiode
electrons are produced by incoming photons these charge the photodiode capacitance the capacitor discharges through the parallel combination of the photodiode shunt resistance and amplifier input resistance
RC time constant low-pass filter with knee frequency 1/2pRC
Rf i Cd Rd Rwires + Vo=-iRf
101
For example
a digital fibre-optic receiver that produces TTL (5 volt) pulses in response to light pulses a light pulse generates a current of 1 mA for a logical one photodiode has a capacitance of 10 pF drives a remote transresistance amplifier through a 10m length 5 of cable with 100W resistance R = = 5M
f
10 - 6 what is Rf ? 1 f knee = = 159MHz what is the knee frequency ? -11 2p .10 .100 How might the knee frequency be increased ?
Rf
1 mA 10 pF
Rwires=100 Vo=-iRf +
102
Materials are insulators, but have a capacitance Equivalent circuit is a current source driving a high-pass filter with a time constant t=RLC... For low frequency sensitivity RL should be as large as possible
no d.c. sensitivity detect a.c. stress
C RL Ri
Q Rf 1 + C Ri
Vo
Q C
V=
Rf
Vo =
103
= 159k
Q V = e RC C
t
= -6 - 20dB = -26dB
104
ADC
ADC
Ri
Rf
Rf
RL
Ri Rf
c +
ADC
d
output type
ADC
Resistive
a,d c,b a a a b
current
Parameter-sensitive resistor
background knowledge/common sense its equivalent circuit determine its knee frequency (a.k.a. -3dB point, cut-off, pole) from its time constant draw a bode plot for high-pass and low-pass sensors and incorporate these into a Bode plot for a complete system
106
107
Qualitative description of the relationship between time-domain and frequency domain Concept of a Linear Time Invariant system Characterisation of a LTI by its impulse response Convolution
108
to view the presentation with animations and sounds click view then slide show or the side show icon bottom left click the loudspeaker icons or movies at any time to hear the sound or to make the movie move
109
amplitude
0.5
- 0.5 -1
1 0.5 0
power
dB
amplitude
2 1000Hz 8 4 6 milliseconds
10
- 0.5 -1
1 0.5 0
amplitude
2 2000Hz 4 6 8 milliseconds
10
power
dB
0.5 1 frequency
2 kHz
- 0.5 -1
2 4 6 8 milliseconds
10
power
dB
0.5 1 frequency
2 kHz
5
110
Low-pass filter
111
High-pass filter
+
112
500 Hz
10 0 - 10 - 20 - 30 - 40 0.2 0.5 1 frequency 2 kHz 5
volts
-1 -2
2 3 msecs
power power dB
volts
volts
1 V500 4
2 1 0 -1 -2
0.2
20 dB/decade 6 dB/octave
10 0 - 10 - 20 - 30 - 40 0.2 0.5 1 frequency
0.5 1 frequency
2 kHz
volts
power
2 3 msecs
2 kHz
5
114
Volts
Resultant:
10 5 0 -5 - 10
-4
-2
0 msec
Volts
4 msec
10
115
+
10
10 2p
Volts
- 10
10
-5
0 2 4 6 msec 8 10
power
- 10 - 20
0.1 0.2 0.5 1 frequency 10 2 5 kHz 10
-p
-2 p
2p
dB
Volts
- 10
-5
0 2 4 6 msec 8 10
power
- 10 - 20
0.1 0.2 0.5 1 frequency 2 5 kHz 10
-p
-2 p
116
phase
phase
dB
+
10
10 2p
dB
Volts
- 10
-5
0 2 4 6 msec 8 10
power
- 10 - 20
0.1 0.2 0.5 1 frequency 10 2 5 kHz 10
-p
-2 p
2p
10
Volts
power
- 10
-5
0 2 4 6 msec 8 10
- 10 - 20
0.1 0.2 0.5 1 frequency 2 5 kHz 10
-p
-2 p
117
phase
dB
phase
Volts
1 2 f max
-1 - 0.5
f max = 10kHz
0 msec
0.5
The width of the pulse is approximately equal to 1/f max If fmax = 1 kHz, the pulse is 1 msec long
118
0 msec
0.5
Volts
power
- 10 - 20
0.1 0.2 0.5 1 frequency 2 5 kHz 10
-p
-4
-2
0 msec
-2 p
119
phase
dB
Volts
-4
10
-2
0 msec
4
2p
10
power
- 10 - 20
0.1 0.2 0.5 1 frequency 2 5 kHz 10
-p
- 10 - 20
0.1 0.2 0.5 1 frequency 2 5 kHz 10
-2 p
-p
-2 p
We know what the frequency response of the output looks like How does the output pulse, that is, the impulse response, look?
120
phase
2p
dB
When the single-frequency components at the output are summed the amplitude and phase have been modified. Considering first the amplitude reduction of the high-frequency components:
1 0.5 0 - 0.5 -1
Volts
-1
- 0.5
0 msec
0.5
-1
- 0.5
0 msec
0.5
1
121
(of course, because it is an inverting low-pass amplifier, both low- and high- frequency terms suffer an additional p phase shift that is neglected below)
Volts
-1
- 0.5
0 msec
0.5
Volts
-1
- 0.5
0 msec
0.5
1
122
High frequency components suffer a both (relative) amplitude reduction and (relative) phase shift of up to p/2
1 0.5 0 - 0.5 -1
Volts
-1
- 0.5
0 msec
0.5
-1
- 0.5
0 msec
Thus, the (complex) frequency response of the low-pass filter has distorted the input pulse as shown above Can we obtain the frequency response back from the impulse response ?
123
Volts
Obtaining the frequency response from the impulse response There is an algorithm that tells us the amplitude and phase with which we need to add up the frequency components to give us the pulse we Amplitude (dB) observe.
- 10 - 20
0.1 0.2 0.5 1 frequency 2 5 kHz 10
The amplitude and phase of these components relative to the input is the frequency response (transfer function) of our filter
Volts
power
-p
-2 p
Fourier transform
-1
- 0.5
0 msec
0.5
1
124
phase
The algorithm is called a Fourier transform more about that next year
10
2 Phase (radians)p
dB
And for the inverting high-pass filter Low-frequencies are attenuated and advanced by p/2
1 0.5 0 - 0.5 -1
Volts
-1
- 0.5
0 msec
0.5
Volts
-1
10
dB
power
- 10 - 20
0.1 0.2 0.5 1 frequency 2 5 kHz 10
-p
-2 p
phase
- 0.5
p
2p
0 msec
0.5
Fourier transform
Provided it is short (much shorter than the impulse response), the result is identical
Volts
This is what is obtained from summing many sinusoids of equal amplitude, but with a finite maximum frequency As will be seen, it is more (and appropriate) to consider a rectangle impulse
time
126
time
What does the output of a low-pass or highpass filter look like when you apply a waveform that is neither a sine wave or an impulse ?
volts
The impulse response is the response to an impulse, what does it say about the response to a more typical signal ?
time
We can approximate any signal as a sum of many impulses: In the limit of an infinitely narrow pulse the approximation becomes perfect
volts
time
volts
time
127
To obtain the response of a circuit to an arbitrary signal we add up the impulse responses due to the individual impulses
making sure we get the right timing and the right amplitude
volts
time
Response to impulse
For simplicity, let us assume the impulse response looks like a smooth bell-shaped curve This process is called convolution:
volts
time
volts
time
128
Example applications
input
volts volts
Impulse response
volts
Output
time
time
time
Thus we have a circuit with a bell-shaped impulse response, we apply a pulse train at the output convolution of the pulse train with impulse response yields the output: a sequence of impulse responses We can represent this as
Convolution operator
129
input
* *
Impulse response
output
= =
* *
=
130
=
131
From page 123, the true impulse response of a low-pass filter has the form
Application convolution shows that a step input to a low-pass filter has the expected charging capacitor form - believe me!
There is an easy way to calculate this convolution - but this is for next year This convolution is presented only for completeness
132
We have tacitly throughout the course that if we apply two inputs to an amplifier, the output is simply the sum of the two multiplied by the frequency response
or equivalently v(t) is convolved with the impulse response:
v1(w)+v2(w)
G(w)
That is, that they are linear or (suppressing the w dependence for clarity): We have also assumed that the frequency response (and the equivalent impulse response) do not vary with time These systems are called Linear Time Invariant (LTI)
G v1 + v2 = Gv1 + Gv2
we used the same impulse response for at all points of our time-varying voltage
h t = h t +T
133
Examples of LTI systems Amplifiers Loudspeakers and microphones Springs Imaging devices
the above devices in saturation magnetic materials (think of B-H curve) Auditory perception (an impulse response does not sound the same as white noise, even though they have the same frequency content - only the phase is different)
134
Because LTI systems of many different types can be represented by the same mathematical formula It is convenient in analysis of LTI systems to represent a LTI component by a box representing the function and neglect the physical implementation (i.e. whether it is an electronic circuit, a chemical reaction, mechanical etc)
e.g. an electronic low-pass filter and a thermometer
135
Go to 78 Go to 106
apply a symmetrical impulse response to some simple inputs demonstrate an understanding of the concept of an LTI
129-131
136