Filters
381
One more point: The ideal low-pass filter has zero phase
shift for all frequencies in the passband. Zero phase shift is
important when the input signal is nonsinusoidal. When a fil-
ter has zero phase shift, the shape of the nonsinusoidal signal
is preserved as it passes through the ideal filter. For instance,
if the input signal is a square wave, it has a fundamental fre-
quency and harmonics. If the fundamental frequency and all Stopband Passband
significant harmonics (approximately the first 10) are inside
f
the passband, the square wave will have approximately the fc
same shape at the output.
Figure 23-2 Ideal high-pass response.
Av Av
Passband Stopband BW
f f
fc f1 f2
Figure 23-1 Ideal low-pass response. Figure 23-3 Ideal bandpass response.
382 Chapter 23
Filters 383
384 Chapter 23
Filters 385
Stopband
0 50 Hz Output 5 high
Voltage across load
CC
frequencies C5
Input voltage, Vin Output voltage, Vout
0.01 mF
Passband Input 5 RL
audio high and low
frequencies frequencies
(a)
Frequency
0.707 Vin
C C1 C2
Output voltage, Vout
fc
RL RL
Frequency
L L
(b)
(c) (d)
when f 5 0 Hz (dc) and f 5 ` Hz. At f 5 0 Hz, the capacitor
C has infinite capacitive reactance XC, calculated as
C 1
XC _____
2fC
RL 1
_____________________
L1 L2 2 0 Hz 0.01 F
`V
Figure 23-10a shows the equivalent circuit for this condi-
tion. Notice that C appears as an open. Since all of the input
(e) voltage appears across the open in a series circuit, Vout must
equal Vin when f 0 Hz.
Figure 23-8 High-pass filters. (a) The response curve for
an audio frequency filter cutting off at 50 Hz. (b) An RC coupling At the other extreme, consider the circuit when the fre-
circuit. (c) Inverted-L type. (d ) The T type. (e) The type. quency f is very high or infinitely high. Then XC 0 V,
calculated as
1
XC _____
23.5 Analyzing Filter Circuits 2fC
1
______________________
Any low-pass or high-pass filter can be thought of as a 2 ` Hz 0.01 F
frequency-dependent voltage divider, since the amount of
output voltage is a function of frequency. Special formulas 0V
can be used to calculate the output voltage for any frequency Figure 23-10b shows the equivalent circuit for this con-
of the applied voltage. What follows is a more mathematical dition. Notice that C appears as a short. Since the voltage
approach in analyzing the operation of the most basic low- across a short is zero, the output voltage for very high fre-
pass and high-pass filter circuits. quencies must be zero.
When the frequency of the input voltage is somewhere
RC Low-Pass Filter between zero and infinity, the output voltage can be deter-
Figure 23-9a shows a simple RC low-pass filter, and mined by using Formula (23-4):
Fig. 23-9b shows how its output voltage Vout varies with fre- X
Vout ___C Vin (23-4)
quency. Lets examine how the RC low-pass filter responds ZT
386 Chapter 23
In Fig. 23-9a, calculate (a) the cutoff frequency fc; (b) Vout at f c; (c)
at f c. (Assume Vin 10 Vp-p for all frequencies.)
Vin XC 5 ` V Vout 5 Vin when f 5 0 Hz
Answer:
a. To calculate f c, use Formula (23-6):
1
fc ______
(a) 2RC
1
_______________________
R 5 10 kV
2 10 kV 0.01 F
1.592 kHz
b. To calculate Vout at f c, use Formula (23-4). First, however, cal-
Vin XC 5 0 V Vout 5 0 Vp-p when f 5 ` Hz culate XC and ZT at f c:
1
XC _____
2fcC
1
__________________________
(b) 2 1.592 kHz 0.01 F
Figure 23-10 RC low-pass equivalent circuits. 10 kV
(a) Equivalent circuit for f 0 Hz. (b) Equivalent circuit for _______
very high frequencies, or f Hz. ZT R2 XC2
_______________
102 kV 102 kV
where 14.14 kV
_______
ZT R2 XC2 Next,
Filters 387
R5 R5
Input voltage, Vin Output voltage, Vout Input voltage, Vin 0utput voltage, Vout
1 kV 1.5 kV
(a) (a)
(b) (b)
Figure 23-11 RL low-pass filter. (a) Circuit. (b) Graph of Figure 23-12 RC high-pass filter. (a) Circuit. (b) Graph of
Vout versus frequency. Vout versus frequency.
since Z T 1.41R when XL R. The formula for the cutoff At very low frequencies where XC is very large, is approxi-
frequency of an RL low-pass filter is mately 908. At very high frequencies where XC approaches
zero, is approximately 08.
R
fc ____ (23-7)
2L To calculate the cutoff frequency fc for an RC high-pass
filter, use Formula (23-6). Although this formula is used to
The response curve in Fig. 23-11b shows that Vout 5 0.707Vin calculate fc for an RC low-pass filter, it can also be used to
at the cutoff frequency fc. calculate fc for an RC high-pass filter. The reason is that, in
both circuits, XC R at the cutoff frequency. In Fig. 23-12b,
RC High-Pass Filter
notice that Vout 0.707Vin at fc.
Figure 23-12a shows an RC high-pass filter. Notice that the
output is taken across the resistor R rather than across the RL High-Pass Filter
capacitor C. Figure 23-12b shows how the output voltage An RL high-pass filter is shown in Fig. 23-13a, and its re-
varies with frequency. To calculate the output voltage Vout at sponse curve is shown in Fig. 23-13b. In Fig. 23-13a, no-
any frequency, use Formula (23-8): tice that the output is taken across the inductor L rather than
across the resistance R. Like an RL low-pass filter, an RL
R V
Vout ___ (23-8)
ZT in high-pass filter is not widely used because of the inductor
disadvantages mentioned earlier.
where
_______
ZT R2 XC2 EXAMPLE 23-2
At very low frequencies, the output voltage approaches Calculate the cutoff frequency for the RC high-pass filter in
zero because the ratio RyZT becomes very small as XC Fig.23-12a.
and thus ZT approach infinity. At very high frequencies, Answer:
Vout is approximately equal to Vin, because the ratio RyZT Use Formula (23-6):
approaches one as ZT and R become approximately the 1
same value. fc _____
2RC
The phase angle of Vout with respect to Vin for an RC high- 1
_______________________
pass filter can be calculated using Formula (23-9): 2 1.5 kV 0.01 F
X 10.61 kHz
arctan ___C (23-9)
R
388 Chapter 23
RC Bandstop Filter 1
f N _______
4R1C1
A high-pass filter can also be combined with a low-pass fil-
ter when it is desired to block or severely attenuate a cer- 1
______________________
tain band of frequencies. Such a filter is called a bandstop 4 1 kV 0.01 F
or notch filter. Figure 23-14a shows an RC bandstop filter, 7.96 kHz
and Fig. 23-14b shows how its output voltage varies with
2R1 2 1 kV
frequency. In Fig. 23-14a, the components identified as 2R1
and 2C1 constitute the low-pass filter section, and the com- 2 kV
ponents identified as R1 and C1 constitute the high-pass filter 2C1 2 0.01 F
0.02 F
2R1 2R1
Filters 389
Answer:
EXAMPLE 23-5 a. At 0 Hz, Vout Vin 10 Vp-p, since the capacitor C appears as
an open. Therefore,
The input power to a filter is 100 mW, and the output power is
Vout
5mW. Calculate the attenuation, in decibels, offered by the filter. NdB 20 log ___
Vin
Answer:
Pout 10 Vp-p
NdB 10 log ___ 20 log ______
Pin 10 Vp-p
5 mW
10 log ________ 20 log 1
100 mW
20 0
10 (21.3)
213 dB 0 dB
b. Since 1.592 kHz is the cutoff frequency f c, Vout will be 0.707
V in or 7.07 Vp-p. Therefore,
The power gain or loss in decibels can also be computed
from a voltage ratio if the measurements are made across Vout
equal resistances. NdB 20 log ___
Vin
Vout
NdB 20 log ___ (23-12) 7.07 Vp-p
Vin
20 log ________
where 10 Vp-p
NdB gain or loss in decibels 20 log 0.707
Vin input voltage
Vout output voltage 20 (20.15)
The NdB values of the passive filters discussed in this chapter 23 dB
can never be positive because Vout can never be greater than Vin. c. To calculate N dB at 15.92 kHz, XC and Z T must first be determined.
Consider the RC low-pass filter in Fig. 23-15. The cutoff fre-
quency fc for this circuit is 1.592 kHz, as determined by For- 1
XC _____
mula (23-6). Recall that the formula for Vout at any frequency is 2fC
X 1
Vout ___C Vin __________________________
ZT 2 15.92 kHz 0.01 F
R 5 10 kV 1 kV
_______
Z T R2 XC2
Vin 5 10 VP-P C 5 0.01 m F Vout ______________
102 k 12 k
fC 5 1.592 kHz
10.05 k
Figure 23-15 RC low-pass filter.
390 Chapter 23
Filters 391
One octave
816 1 10
9
8
7
6
5
2
One octave One octave
2040 5,00010,000
1 1.0
9
8
7
6
5
One decade
3
0.11.0
One octave
0.10.2
1 0.1
9
8
7
6
5
2
One decade One decade
10100 1,00010,000
1 0.01
9
8
7
6
5
3
One decade
3003,000 2
1 0.001
10 100 1,000 10,000
Figure 23-16 Log-log graph paper. Notice that each octave corresponds to a 2-to-1
range of values and each decade corresponds to a 10-to-1 range of values.
392 Chapter 23
(a )
5 6 7 891
1,000,000
4
3
One decade
100,000
5 6 7 891
C
4
3
2
20 dB
octave
One
23 dB
10,000
B
5 6 7 891
6 dB
4
3
2
Frequency (Hz)
1,000
(b)
5 6 7 891
fc
4
3
2
100
A
5 6 7 891
4
3
2
10
5 6 7 891
4
3
2
1
1
210
220
230
240
250
260
0
NdB
Figure 23-17 RC low-pass filter frequency response curve. (a) Circuit. (b) Frequency response curve.
Filters 393
L C
C
Input RL
Series resonant
Input RL
(a )
(a ) RS
RS
L Parallel
Input L C RL
Series resonant
Input RL
resonant
(b) (b )
Figure 23-18 The filtering action of a series resonant Figure 23-19 The filtering action of a parallel resonant
circuit. (a) Bandpass filter when L and C are in series with R L . circuit. (a) Bandstop filter when LC tank is in series with R L .
(b) Bandstop filter when LC circuit is in shunt with R L . (b) Bandpass filter when LC tank is in shunt with R L .
394 Chapter 23
R2 R1
vin +
C1 vout
L2
C1
Input RL
C2 (a)
(a) C R2
R1
L3 C3
vin
vout
C
+
Input L4 C4 RL
(b)
though, the LC circuit has very high resistance for parallel Active Filters
resonance. Then most of the input voltage is across the LC Active filters are those that combine RC networks with op-
circuit and RL with little across RS. erational amplifiers (op amps). The amplifiers provide gain
to offset the normal loss of RC networks, and the feedback
L-Type Resonant Filter techniques provide improved selectivity. Any of the five
Series and parallel resonant circuits can be combined in L, basic types of filters can be implemented with active filter
T, or sections for sharper discrimination of the frequen- circuits.
cies to be filtered. Examples of an L-type filter are shown in Figure 23-21a shows a typical second-order, low-pass fil-
Fig.23-20. ter. Both R and C1 form one of the RC sections, while R
The circuit in Fig. 23-20a is a bandstop filter. The reason and C2 form the other. It provides a rolloff rate of 40 dB per
is that the parallel resonant L1C1 circuit is in series with the decade. The gain is usually set to one, but other op amps can
load, whereas the series resonant L2C2 circuit is in shunt with be added to boost that as required. Greater selectivity can
RL. There is a dual effect as a voltage divider across the input be achieved by cascading stages without the loss of gain. A
source voltage. The high resistance of L1C1 reduces voltage bandpass active filter is shown in Fig. 23-21b. High-pass and
output to the load. Also, the low resistance of L2C2 reduces notch filters can be formed in a similar way.
the output voltage. Most active filters are used at low frequencies to elimi-
For the opposite effect, the circuit in Fig. 23-20b is a nate the need for large expensive inductors. They are also
bandpass filter. Now the series resonant L3C3 circuit is in used at audio frequencies to set the frequency response of an
series with the load. Here the low resistance of L3C3 allows amplifier or other circuit. Active filters are not widely used at
more output for RL at resonance. Also, the high resistance of the higher frequencies but with wide-band op amps they can
L 4C4 allows maximum output voltage. function well into the 100-MHz RF range.
Filters 395
Interdigital Interdigital
transducers transducers
R
Surface acoustic
waves
Input
Piezoelectric ceramic
Figure 23-22 Equivalent circuit of a crystal filter and its
schematic symbol.
Figure 23-24 Basic construction of a surface acoustic
wave (SAW) filter.
occurs at this resonant frequency. The equivalent cir-
cuit of a crystal and its schematic symbol are shown in
Fig. 23-22. Depending on the frequency, the crystal may Surface Acoustic Wave Filters
operate as a parallel resonant circuit or as a series reso- A surface acoustic wave (SAW) filter is a special bandpass fil-
nant circuit. The Q of the crystal is very high, usually ter used primarily for radio-frequency selectivity. Figure23-24
over 10,000 or more. As a result, the crystal can be used shows the basic construction. The base is a piezoelectric ce-
to build very selective filters. ramic substrate such as lithium niobate. A pattern of interdigital
Most of these filters are bandpass filters. An example is fingers are made on the surface. The pattern on the left converts
shown in Fig. 23-23. The response curve has extremely steep the signals into acoustic waves that travel across the filter sur-
rolloffs, making the filter useful in separating closely spaced face. By controlling the shape, size, and spacing of the inter-
signals. Most crystal filters are used at radio frequencies digital fingers, the response can be tailored to any frequency or
from roughly 1 MHz to 100 MHz. desired shape. The interdigital pattern on the right in Fig. 23-24
Special ceramic materials, such as lead titanate, can also converts the acoustic waves back into an electronic signal.
be used like a crystal. Ceramic resonators are available to SAW filters are available with a frequency range of 10
make filters. Ceramic filters are smaller, with a lower Q, MHz 4 GHz. The attenuation is very high and in the 10- to
but still have better selectivity than larger LC or crystal 35-dB range. As a result they are used with appropriate am-
filters. Most are used at radio frequencies from 400 kHz plifiers. SAW filters are widely used in TV sets, cell phones,
to 50 MHz. and many types of wireless equipment.
DSP Filters
As you have seen in this chapter, filters are made up of com- DSP refers to a way to process analog signals with digital
ponents like resistors, capacitors, inductors, or op amps. circuits. The analog signal to be processed, in this case, is first
Special filters are made of ceramic and crystal resonators converted to digital form. Then a special processor or digital
and unique components, like SAW filters. All these filters computer processes the equivalent digital data to perform the
are still widely in use. However, a newer and more com- filtering function. Almost any type of filter, as described ear-
plex filter has found many applications in modern electronic lier, can be implemented with software to perform the filtering
circuits and systems. This is the digital signal processing digitally. Once processed, the data are then converted back
(DSP) filter. into analog form. The result is the desired filtering effect.
396 Chapter 23
expensive. That was once true, but today, thanks to modern Analog Data Filtered
signal to be analog
semiconductor technology, the circuitry and processors can filtered signal
ADC DAC
be made small enough and cheap enough to make DSP filters
practical. And they are easily integrated onto larger chips to
create a complete system on a chip (SoC). Better still, the digi-
tal filters are usually more effective than equivalent analog
Program
filters. They are usually more selective with steeper rolloff and
less attenuation, and with proper tailoring of the program, the Filter
Algorithm
phase and other characteristics can be controlled as desired.
Figure S23-1 shows a simplified block diagram of a DSP
filter. The analog signal to be filtered is first digitized in a Digital signal
circuit called an analog-to-digital converter (ADC). This processor
circuit produces a stream of binary numbers that represent
samples of the analog signal at closely spaced time intervals. Figure S23-1 The concept of digital signal processing (DSP).
The binary numbers are stored in a data memory.
The analog signal data are then processed by the digital signal. The output is similar to that obtained with an analog
signal processor (DSP). The processing may be done by a filter. Typically, the filtering action is superior in some way.
special microcontroller designed for DSP, or it can be any And the filter characteristics can be changed on the fly by
microprocessor or controller with DSP capability. using a new or modified processing algorithm.
Stored in the processors memory is the program that DSP filters are invisible, since they are implemented in-
does the processing. It is usually a special mathematical side a small processor or other chip. They are widely used in
algorithm that performs any one of the normal filter func- all forms of electronics, including TV sets, cell phones, MP3
tions like low pass, high pass, bandpass, or bandstop. The players, military radios, and many other types of equipment.
processed data are then stored back in data memory. Finally, The mathematics of DSP is well beyond the scope of this
the processed data are fed to a digital-to-analog converter text, but you should know of its existence because it is not
(DAC) that translates the digital data back into an analog experimental but very widely used in many applications.
Filters 397
398 Chapter 23
CHAPTER 23 PROBLEMS
Filters 399
2R1 5 36 kV 2R1 5 36 kV
2C1 5
0.002 F
R1 5 18 kV
Figure 23-29
400 Chapter 23