FOUNDATIONS

CHAPTER 1
1.1 Introduction
electronic circuits. Because you can’t touch, see, smell, or
The field of electronics is one of the great success stories hear electricity, there will be a certain amount of abstrac-
of the 20th century. From the crude spark-gap transmitters tion (particularly in the first chapter), as well as some de-
and “cat’s-whisker” detectors at its beginning, the first half- pendence on such visualizing instruments as oscilloscopes
century brought an era of vacuum-tube electronics that de- and voltmeters. In many ways the first chapter is also the
veloped considerable sophistication and found ready ap- most mathematical, in spite of our efforts to keep math-
plication in areas such as communications, navigation, in- ematics to a minimum in order to foster a good intuitive
strumentation, control, and computation. The latter half- understanding of circuit design and behavior.
century brought “solid-state” electronics – first as discrete In this new edition we’ve included some intuition-aiding
transistors, then as magnificent arrays of them within “in- approximations that our students have found helpful. And,
tegrated circuits” (ICs) – in a flood of stunning advances by introducing one or two “active” components ahead of
that shows no sign of abating. Compact and inexpensive their time, we’re able to jump directly into some applica-
consumer products now routinely contain many millions tions that are usually impossible in a traditional textbook
of transistors in VLSI (very large-scale integration) chips, “passive electronics” chapter; this will keep things inter-
combined with elegant optoelectronics (displays, lasers, esting, and even exciting.
and so on); they can process sounds, images, and data, and Once we have considered the foundations of electron-
(for example) permit wireless networking and shirt-pocket ics, we will quickly get into the active circuits (amplifiers,
access to the pooled capabilities of the Internet. Perhaps as oscillators, logic circuits, etc.) that make electronics the ex-
noteworthy is the pleasant trend toward increased perfor- citing field it is. The reader with some background in elec-
mance per dollar.1 The cost of an electronic microcircuit tronics may wish to skip over this chapter, since it assumes
routinely decreases to a fraction of its initial cost as the no prior knowledge of electronics. Further generalizations
manufacturing process is perfected (see Figure 10.87 for at this time would be pointless, so let’s just dive right in.
an example). In fact, it is often the case that the panel con-
trols and cabinet hardware of an instrument cost more than 1.2 Voltage, current, and resistance
the electronics inside. 1.2.1 Voltage and current
On reading of these exciting new developments in elec-
There are two quantities that we like to keep track of in
tronics, you may get the impression that you should be able
electronic circuits: voltage and current. These are usually
to construct powerful, elegant, yet inexpensive, little gad-
changing with time; otherwise nothing interesting is hap-
gets to do almost any conceivable task – all you need to
pening.
know is how all these miracle devices work. If you’ve had
that feeling, this book is for you. In it we have attempted Voltage (symbol V or sometimes E). Officially, the volt-
to convey the excitement and know-how of the subject of age between two points is the cost in energy (work done)
electronics. required to move a unit of positive charge from the more
In this chapter we begin the study of the laws, rules of negative point (lower potential) to the more positive
thumb, and tricks that constitute the art of electronics as we point (higher potential). Equivalently, it is the energy
see it. It is necessary to begin at the beginning – with talk of released when a unit charge moves “downhill” from
voltage, current, power, and the components that make up the higher potential to the lower.2 Voltage is also called
1 A mid-century computer (the IBM 650) cost $300,000, weighed 2.7 2 These are the definitions, but hardly the way circuit designers think of
tons, and contained 126 lamps on its control panel; in an amusing re- voltage. With time, you’ll develop a good intuitive sense of what volt-
versal, a contemporary energy-efficient lamp contains a computer of age really is, in an electronic circuit. Roughly (very roughly) speaking,
greater capability within its base, and costs about $10. voltages are what you apply to cause currents to flow.

1
2 1.2. Voltage, current, and resistance Art of Electronics Third Edition

potential difference or electromotive force (EMF). The In real circuits we connect things together with wires
unit of measure is the volt, with voltages usually ex- (metallic conductors), each of which has the same voltage
pressed in volts (V), kilovolts (1 kV = 103 V), millivolts on it everywhere (with respect to ground, say).4 We men-
(1 mV = 10−3 V), or microvolts (1 μ V = 10−6 V) (see tion this now so that you will realize that an actual circuit
the box on prefixes). A joule (J) of work is done in mov- doesn’t have to look like its schematic diagram, because
ing a coulomb (C) of charge through a potential differ- wires can be rearranged.
ence of 1 V. (The coulomb is the unit of electric charge, Here are some simple rules about voltage and current:
and it equals the charge of approximately 6×1018 elec-
1. The sum of the currents into a point in a circuit equals
trons.) For reasons that will become clear later, the op-
the sum of the currents out (conservation of charge).
portunities to talk about nanovolts (1 nV = 10−9 V) and
This is sometimes called Kirchhoff’s current law (KCL).
megavolts (1 MV = 106 V) are rare.
Engineers like to refer to such a point as a node. It fol-
Current (symbol I). Current is the rate of flow of elec-
lows that, for a series circuit (a bunch of two-terminal
tric charge past a point. The unit of measure is the
things all connected end-to-end), the current is the same
ampere, or amp, with currents usually expressed in
everywhere.
amperes (A), milliamperes (1 mA = 10−3 A), microam-
peres (1 μ A = 10−6 A), nanoamperes (1 nA = 10−9 A),
or occasionally picoamperes (1 pA = 10−12 A). A cur-
rent of 1 amp equals a flow of 1 coulomb of charge per A B
second. By convention, current in a circuit is considered
to flow from a more positive point to a more negative
point, even though the actual electron flow is in the op- Figure 1.1. Parallel connection.
posite direction.
Important: from these definitions you can see that cur- 2. Things hooked in parallel (Figure 1.1) have the same
rents flow through things, and voltages are applied (or ap- voltage across them. Restated, the sum of the “voltage
pear) across things. So you’ve got to say it right: always re- drops” from A to B via one path through a circuit equals
fer to the voltage between two points or across two points the sum by any other route, and is simply the voltage
in a circuit. Always refer to current through a device or between A and B. Another way to say it is that the sum
connection in a circuit. of the voltage drops around any closed circuit is zero.
To say something like “the voltage through a resistor This is Kirchhoff’s voltage law (KVL).
. . . ” is nonsense. However, we do frequently speak of the 3. The power (energy per unit time) consumed by a circuit
voltage at a point in a circuit. This is always understood to device is
mean the voltage between that point and “ground,” a com- P = VI (1.1)
mon point in the circuit that everyone seems to know about.
Soon you will, too. This is simply (energy/charge) × (charge/time). For V in
We generate voltages by doing work on charges in volts and I in amps, P comes out in watts. A watt is a
devices such as batteries (conversion of electrochemical joule per second (1W = 1 J/s). So, for example, the cur-
energy), generators (conversion of mechanical energy by rent flowing through a 60W lightbulb running on 120 V
magnetic forces), solar cells (photovoltaic conversion of is 0.5 A.
the energy of photons), etc. We get currents by placing volt- Power goes into heat (usually), or sometimes mechan-
ages across things. ical work (motors), radiated energy (lamps, transmitters),
At this point you may well wonder how to “see” volt- or stored energy (batteries, capacitors, inductors). Manag-
ages and currents. The single most useful electronic instru- ing the heat load in a complicated system (e.g., a large
ment is the oscilloscope, which allows you to look at volt- computer, in which many kilowatts of electrical energy are
ages (or occasionally currents) in a circuit as a function converted to heat, with the energetically insignificant by-
of time.3 We will deal with oscilloscopes, and also volt- product of a few pages of computational results) can be a
meters, when we discuss signals shortly; for a preview see crucial part of the system design.
Appendix O, and the multimeter box later in this chapter. 4 In the domain of high frequencies or low impedances, that isn’t strictly
3 It has been said that engineers in other disciplines are envious of elec- true, and we will have more to say about this later, and in Chapter 1x.
trical engineers, because we have such a splendid visualization tool. For now, it’s a good approximation.
Art of Electronics Third Edition 1.2.2. Relationship between voltage and current: resistors 3

Figure 1.2. A selection of common resistor types. Top row, left to right (wirewound ceramic power resistors): 20W vitreous enamel with
leads, 20W with mounting studs, 30W vitreous enamel, 5W and 20W with mounting studs. Middle row (wirewound power resistors): 1W,
3W, and 5W axial ceramic; 5W, 10W, 25W, and 50W conduction-cooled (“Dale-type”). Bottom row: 2W, 1W, 12 W, 14 W, and 18 W carbon
composition; surface-mount thick-film (2010, 1206, 0805, 0603, and 0402 sizes); surface-mount resistor array; 6-, 8-, and 10-pin single
in-line package arrays; dual in-line package array. The resistor at bottom is the ubiquitous RN55D 14 W, 1% metal-film type; and the pair
of resistors above are Victoreen high-resistance types (glass, 2 GΩ; ceramic, 5 GΩ).

Soon, when we deal with periodically varying voltages capacitors (I proportional to rate of change of V ), diodes
and currents, we will have to generalize the simple equa- (I flows in only one direction), thermistors (temperature-
tion P = V I to deal with average power, but it’s correct as dependent resistor), photoresistors (light-dependent resis-
a statement of instantaneous power just as it stands. tor), strain gauges (strain-dependent resistor), etc., are ex-
Incidentally, don’t call current “amperage”; that’s amples. Perhaps more interesting still are three-terminal
strictly bush league.5 The same caution will apply to the devices, such as transistors, in which the current that can
term “ohmage”6 when we get to resistance in the next flow between a pair of terminals is controlled by the volt-
section. age applied to a third terminal. We will gradually get into
some of these exotic devices; for now, we will start with
1.2.2 Relationship between voltage and current: the most mundane (and most widely used) circuit element,
resistors the resistor (Figure 1.3).
This is a long and interesting story. It is the heart of elec-
tronics. Crudely speaking, the name of the game is to make
and use gadgets that have interesting and useful I-versus- Figure 1.3. Resistor.
V characteristics. Resistors (I simply proportional to V ),
5 Unless you’re a power engineer working with giant 13 kV transformers A. Resistance and resistors
and the like – those guys are allowed to say amperage. It is an interesting fact that the current through a metal-
6 . . . also, Dude, “ohmage” is not the preferred nomenclature: resistance, lic conductor (or other partially conducting material) is
please. proportional to the voltage across it. (In the case of wire
4 1.2. Voltage, current, and resistance Art of Electronics Third Edition

PREFIXES

Multiple Prefix Symbol Derivation

1024 yotta Y end-1 of Latin alphabet, hint of Greek iota
1021 zetta Z end of Latin alphabet, hint of Greek zeta
1018 exa E Greek hexa (six: power of 1000)
1015 peta P Greek penta (five: power of 1000)
1012 tera T Greek teras (monster)
109 giga G Greek gigas (giant)
106 mega M Greek megas (great)
103 kilo k Greek khilioi (thousand)
10−3 milli m Latin milli (thousand)
10−6 micro μ Greek mikros (small)
10−9 nano n Greek nanos (dwarf)
10−12 pico p from Italian/Spanish piccolo/pico (small)
10−15 femto f Danish/Norwegian femten (fifteen)
10−18 atto a Danish/Norwegian atten (eighteen)
10−21 zepto z end of Latin alphabet, mirrors zetta
10−24 yocto y end-1 of Latin alphabet, mirrors yotta

These prefixes are universally used to scale units in is a milliwatt, or one-thousandth of a watt; 1 MHz is a
science and engineering. Their etymological derivations megahertz or 1 million hertz. In general, units are spelled
are a matter of some controversy and should not be con- with lowercase letters, even when they are derived from
sidered historically reliable. When abbreviating a unit proper names. The unit name is not capitalized when it
with a prefix, the symbol for the unit follows the prefix is spelled out and used with a prefix, only when abbre-
without space. Be careful about uppercase and lowercase viated. Thus: hertz and kilohertz, but Hz and kHz; watt,
letters (especially m and M) in both prefix and unit: 1 mW milliwatt, and megawatt, but W, mW, and MW.

conductors used in circuits, we usually choose a thick- values from 1 ohm (1 Ω) to about 10 megohms (10 MΩ).
enough gauge of wire so that these “voltage drops” will Resistors are also characterized by how much power they
be negligible.) This is by no means a universal law for all can safely dissipate (the most commonly used ones are
objects. For instance, the current through a neon bulb is a rated at 1/4 or 1/8 W), their physical size,7 and by other
highly nonlinear function of the applied voltage (it is zero parameters such as tolerance (accuracy), temperature co-
up to a critical voltage, at which point it rises dramatically). efficient, noise, voltage coefficient (the extent to which R
The same goes for a variety of interesting special devices depends on applied V ), stability with time, inductance, etc.
– diodes, transistors, lightbulbs, etc. (If you are interested See the box on resistors, Chapter 1x, and Appendix C for
in understanding why metallic conductors behave this way, further details. Figure 1.2 shows a collection of resistors,
read §§4.4–4.5 in Purcell and Morin’s splendid text Elec- with most of the available morphologies represented.
tricity and Magnetism). Roughly speaking, resistors are used to convert a
A resistor is made out of some conducting stuff (carbon,
or a thin metal or carbon film, or wire of poor conductivity),
with a wire or contacts at each end. It is characterized by 7 The sizes of chip resistors and other components intended for surface
its resistance: mounting are specified by a four-digit size code, in which each pair of
digits specifies a dimension in units of 0.010 (0.25 mm). For exam-
R = V /I; (1.2) ple, an 0805 size resistor is 2 mm×1.25 mm, or 80 mils×50 mils (1 mil
is 0.001 ); the height must be separately specified. To add confusion
R is in ohms for V in volts and I in amps. This is known to this simple scheme, the four-digit size code may instead be metric
as Ohm’s law. Typical resistors of the most frequently used (sometimes without saying so!), in units of 0.1 mm: thus an “0805”
type (metal-oxide film, metal film, or carbon film) come in (English) is also a “2012” (metric).
 Art of Electronics Third Edition 1.2.2. Relationship between voltage and current: resistors 5

RESISTORS ceramic-composition moldings; from metal foil or metal

Resistors are truly ubiquitous. There are almost as many wire wound on a form; or from semiconductor elements
types as there are applications. Resistors are used in am- similar to field-effect transistors (FETs). The most com-
plifiers as loads for active devices, in bias networks, and monly used resistor type is formed from a carbon, metal,
as feedback elements. In combination with capacitors or oxide film, and comes in two widely used “packages”:
they establish time constants and act as filters. They are the cylindrical axial-lead type (typified by the generic
used to set operating currents and signal levels. Resistors RN55D 1% 1/4 W metal-film resistor),8 and the much
are used in power circuits to reduce voltages by dissi- smaller surface-mount “chip resistor.” These common
pating power, to measure currents, and to discharge ca- types come in 5%, 2%, and 1% tolerances, in a standard
pacitors after power is removed. They are used in pre- set of values ranging from 1 Ω to 10 MΩ. The 1% types
cision circuits to establish currents, to provide accurate have 96 values per decade, whereas the 2% and 5% types
voltage ratios, and to set precise gain values. In logic cir- have 48 and 24 values per decade (see Appendix C). Fig-
cuits they act as bus and line terminators and as “pullup” ure 1.2 illustrates most of the common resistor packages.
and “pull-down” resistors. In high-voltage circuits they Resistors are so easy to use and well behaved that
are used to measure voltages and to equalize leakage cur- they’re often taken for granted. They’re not perfect,
rents among diodes or capacitors connected in series. In though, and you should be aware of some of their limita-
radiofrequency (RF) circuits they set the bandwidth of tions so that you won’t be surprised someday. The princi-
resonant circuits, and they are even used as coil forms pal defects are variations in resistance with temperature,
for inductors. voltage, time, and humidity. Other defects relate to in-
Resistors are available with resistances from 0.0002 Ω ductance (which may be serious at high frequencies), the
through 1012 Ω, standard power ratings from 1/8 watt development of thermal hot spots in power applications,
through 250 watts, and accuracies from 0.005% through or electrical noise generation in low-noise amplifiers. We
20%. Resistors can be made from metal films, metal- treat these in the advanced Chapter 1x.
oxide films, or carbon films; from carbon-composition or

voltage to a current, and vice versa. This may sound aw- R1 R2

fully trite, but you will soon see what we mean.
Figure 1.4. Resistors in series.
B. Resistors in series and parallel
From the definition of R, some simple results follow: R1

1. The resistance of two resistors in series (Figure 1.4) is R2

R = R1 + R2 . (1.3)
Figure 1.5. Resistors in parallel.
By putting resistors in series, you always get a larger re-
sistor.
resistor as a 1M resistor (or 1 meg).9 If these preliminaries
2. The resistance of two resistors in parallel (Figure 1.5) is
bore you, please have patience – we’ll soon get to numer-
R1 R2 1
R= or R = . (1.4) ous amusing applications.
R1 + R2 1 1
+
R1 R2 Exercise 1.1. You have a 5k resistor and a 10k resistor. What is
their combined resistance (a) in series and (b) in parallel?
By putting resistors in parallel, you always get a smaller
resistor. Resistance is measured in ohms (Ω), but in prac- Exercise 1.2. If you place a 1 ohm resistor across a 12 volt car
tice we frequently omit the Ω symbol when referring to battery, how much power will it dissipate?
resistors that are more than 1000 Ω (1 kΩ). Thus, a 4.7 kΩ Exercise 1.3. Prove the formulas for series and parallel resistors.
resistor is often referred to as a 4.7k resistor, and a 1 MΩ
9 A popular “international” alternative notation replaces the decimal
8 Conservatively rated at 1/8 watt in its RN55 military grade (“MIL- point with the unit multiplier, thus 4k7 or 1M0. A 2.2 Ω resistor be-
spec”), but rated at 1/4 watt in its CMF-55 industrial grade. comes 2R2. There is an analogous scheme for capacitors and inductors.
6 1.2. Voltage, current, and resistance Art of Electronics Third Edition

Exercise 1.4. Show that several resistors in parallel have resis- education is a fine way to prevent you from understanding
tance what’s really going on.
1
R= (1.5) In trying to develop intuition about resistance, some
1 1 1
+ + +··· people find it helpful to think about conductance, G = 1/R.
R1 R2 R3
The current through a device of conductance G bridging
Beginners tend to get carried away with complicated al- a voltage V is then given by I = GV (Ohm’s law). A
gebra in designing or trying to understand electronics. Now small resistance is a large conductance, with correspond-
is the time to begin learning intuition and shortcuts. Here ingly large current under the influence of an applied volt-
are a couple of good tricks: age. Viewed in this light, the formula for parallel resistors
Shortcut #1 A large resistor in series (parallel) with a is obvious: when several resistors or conducting paths are
small resistor has the resistance of the larger (smaller) connected across the same voltage, the total current is the
one, roughly. So you can “trim” the value of a resistor sum of the individual currents. Therefore the net conduc-
up or down by connecting a second resistor in series or tance is simply the sum of the individual conductances,
parallel: to trim up, choose an available resistor value G = G1 + G2 + G3 + · · ·, which is the same as the formula
below the target value, then add a (much smaller) series for parallel resistors derived earlier.
resistor to make up the difference; to trim down, choose Engineers are fond of defining reciprocal units, and they
an available resistor value above the target value, then have designated as the unit of conductance the siemens
connect a (much larger) resistor in parallel. For the lat- (S = 1/Ω), also known as the mho (that’s ohm spelled
ter you can approximate with proportions – to lower the backward, given the symbol ). Although the concept of
value of a resistor by 1%, say, put a resistor 100 times conductance is helpful in developing intuition, it is not
as large in parallel.10 used widely;11 most people prefer to talk about resistance
Shortcut #2 Suppose you want the resistance of 5k in par- instead.
allel with 10k. If you think of the 5k as two 10k’s in C. Power in resistors
parallel, then the whole circuit is like three 10k’s in par- The power dissipated by a resistor (or any other device) is
allel. Because the resistance of n equal resistors in par- P = IV . Using Ohm’s law, you can get the equivalent forms
allel is 1/nth the resistance of the individual resistors, P = I 2 R and P = V 2/R.
the answer in this case is 10k/3, or 3.33k. This trick is
Exercise 1.5. Show that it is not possible to exceed the power
handy because it allows you to analyze circuits quickly
rating of a 1/4 watt resistor of resistance greater than 1k, no matter
in your head, without distractions. We want to encour-
how you connect it, in a circuit operating from a 15 volt battery.
age mental designing, or at least “back-of-the-envelope”
designing, for idea brainstorming. Exercise 1.6. Optional exercise: New York City requires about
1010 watts of electrical power, at 115 volts12 (this is plausible:
Some more home-grown philosophy: there is a ten- 10 million people averaging 1 kilowatt each). A heavy power ca-
dency among beginners to want to compute resistor val- ble might be an inch in diameter. Let’s calculate what will happen
ues and other circuit component values to many signif- if we try to supply the power through a cable 1 foot in diameter
icant places, particularly with calculators and computers made of pure copper. Its resistance is 0.05 μ Ω (5×10−8 ohms)
that readily oblige. There are two reasons you should try to per foot. Calculate (a) the power lost per foot from “I 2 R losses,”
avoid falling into this habit: (a) the components themselves (b) the length of cable over which you will lose all 1010 watts, and
are of finite precision (resistors typically have tolerances of (c) how hot the cable will get, if you know the physics involved
(σ = 6 × 10−12 W/K4 cm2 ). If you have done your computations
±5% or ±1%; for capacitors it’s typically ±10% or ±5%;
correctly, the result should seem preposterous. What is the solu-
and the parameters that characterize transistors, say, fre-
tion to this puzzle?
quently are known only to a factor of 2); (b) one mark of a
11 Although the elegant Millman’s theorem has its admirers: it says that
good circuit design is insensitivity of the finished circuit to
precise values of the components (there are exceptions, of the output voltage from a set of resistors (call them Ri ) that are driven
from a set of corresponding input voltages (Vi ) and connected together
course). You’ll also learn circuit intuition more quickly if
at the output is Vout =(∑ Vi Gi )/ ∑ Gi , where the Gi are the conductances
you get into the habit of doing approximate calculations in
(Gi =1/Ri ).
your head, rather than watching meaningless numbers pop 12 Although the “official” line voltage is 120 V ±5%, you’ll sometimes
up on a calculator display. We believe strongly that reliance see 110 V, 115 V, or 117 V. This loose language is OK (and we use
on formulas and equations early in your electronic circuit it in this book), because (a) the median voltage at the wall plug is 3
to 5 volts lower, when powering stuff; and (b) the minimum wall-plug
10 With an error, in this case, of just 0.01%. voltage is 110 V. See ANSI standard C84.1.
Art of Electronics Third Edition 1.2.3. Voltage dividers 7

D. Input and output

Nearly all electronic circuits accept some sort of applied signal in
R1
input (usually a voltage) and produce some sort of corre-
sponding output (which again is often a voltage). For ex- signal out
ample, an audio amplifier might produce a (varying) output
volume R2
voltage that is 100 times as large as a (similarly varying) in-
(adjustable)
put voltage. When describing such an amplifier, we imag- A.
ine measuring the output voltage for a given applied input
voltage. Engineers speak of the transfer function H, the ra-
tio of (measured) output divided by (applied) input; for the
R1
audio amplifier above, H is simply a constant (H = 100). signal in

We’ll get to amplifiers soon enough, in the next chapter. volume signal out
However, with only resistors we can already look at a very
R2
important circuit fragment, the voltage divider (which you B.
might call a “de-amplifier”).
Figure 1.7. An adjustable voltage divider can be made from a fixed
1.2.3 Voltage dividers and variable resistor, or from a potentiometer. In some contempo-
rary circuits you’ll find instead a long series chain of equal-value
We now come to the subject of the voltage divider, one of resistors, with an arrangement of electronic switches that lets you
the most widespread electronic circuit fragments. Show us choose any one of the junctions as the output; this sounds much
any real-life circuit and we’ll show you half a dozen volt- more complicated – but it has the advantage that you can adjust
age dividers. To put it very simply, a voltage divider is a the voltage ratio electrically (rather than mechanically).
circuit that, given a certain voltage input, produces a pre-
dictable fraction of the input voltage as the output voltage.
(We’ve used the definition of resistance and the series law.)
The simplest voltage divider is shown in Figure 1.6.
Then, for R2 ,
R2
R1 Vout = IR2 = V . (1.6)
R1 + R2 in
Vout R2
Vin =
Vin R1 + R 2 Note that the output voltage is always less than (or equal
R2 Vout to) the input voltage; that’s why it’s called a divider. You
could get amplification (more output than input) if one
of the resistances were negative. This isn’t as crazy as it
Figure 1.6. Voltage divider. An applied voltage Vin results in a
sounds; it is possible to make devices with negative “incre-
(smaller) output voltage Vout .
mental” resistances (e.g., the component known as a tunnel
diode) or even true negative resistances (e.g., the negative-
An important word of explanation: when engineers impedance converter that we will talk about later in the
draw a circuit like this, it’s generally assumed that the Vin book, §6.2.4B). However, these applications are rather spe-
on the left is a voltage that you are applying to the circuit, cialized and need not concern you now.
and that the Vout on the right is the resulting output voltage Voltage dividers are often used in circuits to gener-
(produced by the circuit) that you are measuring (or at ate a particular voltage from a larger fixed (or varying)
least are interested in). You are supposed to know all this voltage. For instance, if Vin is a varying voltage and R2
(a) because of the convention that signals generally flow is an adjustable resistor (Figure 1.7A), you have a “vol-
from left to right, (b) from the suggestive names (“in,” ume control”; more simply, the combination R1 R2 can be
“out”) of the signals, and (c) from familiarity with circuits made from a single variable resistor, or potentiometer (Fig-
like this. This may be confusing at first, but with time it ure 1.7B). This and similar applications are common, and
becomes easy. potentiometers come in a variety of styles, some of which
What is Vout ? Well, the current (same everywhere, as- are shown in Figure 1.8.
suming no “load” on the output; i.e., nothing connected The humble voltage divider is even more useful, though,
across the output) is as a way of thinking about a circuit: the input voltage and
Vin upper resistance might represent the output of an amplifier,
I= .
R1 + R2 say, and the lower resistance might represent the input of
8 1.2. Voltage, current, and resistance Art of Electronics Third Edition

Figure 1.8. Most of the common potentiometer styles are shown here. Top row, left to right (panel mount): power wirewound, “type
AB” 2W carbon composition, 10-turn wirewound/plastic hybrid, ganged dual pot. Middle row (panel mount): optical encoder (continuous
rotation, 128 cycles per turn), single-turn cermet, single-turn carbon, screw-adjust single-turn locking. Front row (board-mount trimmers):
multiturn side-adjust (two styles), quad single-turn, 3/8 (9.5 mm) square single-turn, 1/4 (6.4 mm) square single-turn, 1/4 (6.4 mm)
round single-turn, 4 mm square single-turn surface mount, 4 mm square multiturn surface mount, 3/8 (9.5 mm) square multiturn, quad
nonvolatile 256-step integrated pot (E2 POT) in 24-pin small-outline IC.

the following stage. In this case the voltage-divider equa- current (when shorted) of 3 amps (which, however, will kill
tion tells you how much signal gets to the input of that last the battery in a few minutes). A voltage source “likes” an
stage. This will all become clearer after you know about open-circuit load and “hates” a short-circuit load, for obvi-
a remarkable fact (Thévenin’s theorem) that will be dis- ous reasons. (The meaning of “open circuit” and “short cir-
cussed later. First, though, a short aside on voltage sources cuit” sometimes confuse the beginner: an open circuit has
and current sources. nothing connected to it, whereas a short circuit is a piece
of wire bridging the output.) The symbols used to indicate
a voltage source are shown in Figure 1.9.
1.2.4 Voltage sources and current sources
A perfect current source is a two-terminal black box
A perfect voltage source is a two-terminal “black box” that that maintains a constant current through the external cir-
maintains a fixed voltage drop across its terminals, regard- cuit, regardless of load resistance or applied voltage. To do
less of load resistance. This means, for instance, that it this it must be capable of supplying any necessary volt-
must supply a current I = V /R when a resistance R is at- age across its terminals. Real current sources (a much-
tached to its terminals. A real voltage source can supply neglected subject in most textbooks) have a limit to the
only a finite maximum current, and in addition it generally voltage they can provide (called the output-voltage compli-
behaves like a perfect voltage source with a small resis- ance, or just compliance), and in addition they do not pro-
tance in series. Obviously, the smaller this series resistance, vide absolutely constant output current. A current source
the better. For example, a standard 9 volt alkaline battery “likes” a short-circuit load and “hates” an open-circuit
behaves approximately like a perfect 9 volt voltage source load. The symbols used to indicate a current source are
in series with a 3 Ω resistor, and it can provide a maximum shown in Figure 1.10.
Art of Electronics Third Edition 1.2.5. Thévenin equivalent circuit 9

resistor R in series with a single voltage source V . This

+5 is remarkable. Any mess of batteries and resistors can be
+15 110V
+ power mimicked with one battery and one resistor (Figure 1.11).
volts supply Vs
ac com (Incidentally, there’s another theorem, Norton’s theorem,
–
that says you can do the same thing with a current source
(battery) in parallel with a resistor.)
How do you figure out the Thévenin equivalent RTh and
Figure 1.9. Voltage sources can be either steady (dc) or varying VTh for a given circuit? Easy! VTh is the open-circuit voltage
(ac). of the Thévenin equivalent circuit; so if the two circuits
behave identically, it must also be the open-circuit voltage
of the given circuit (which you get by calculation, if you
know what the circuit is, or by measurement, if you don’t).
Then you find RTh by noting that the short-circuit current
1mA 1mA 1mA of the equivalent circuit is VTh /RTh . In other words,
Figure 1.10. Current-source symbols. VTh = V (open circuit),
V (open circuit) (1.7)
RTh = .
I (short circuit)
A battery is a real-life approximation to a voltage source
(there is no analog for a current source). A standard D-size Let’s apply this method to the voltage divider, which must
flashlight cell, for instance, has a terminal voltage of 1.5 V, have a Thévenin equivalent:
an equivalent series resistance of about 0.25 Ω, and a to-
1. The open-circuit voltage is
tal energy capacity of about 10,000 watt–seconds (its char-
acteristics gradually deteriorate with use; at the end of its R2
V = Vin .
life, the voltage may be about 1.0 V, with an internal se- R1 + R2
ries resistance of several ohms). It is easy to construct volt- 2. The short-circuit current is
age sources with far better characteristics, as you will learn
Vin /R1 .
when we come to the subject of feedback; this is a major
topic of Chapter 9. Except in the important class of devices So the Thévenin equivalent circuit is a voltage source,
intended for portability, the use of batteries in electronic R2
VTh = Vin , (1.8)
devices is rare. R1 + R2
V3 R9 in series with a resistor,
R1 R2
R8 RTh = . (1.9)
R1 + R2
R4
R10 (It is not a coincidence that this happens to be the paral-
R3 R5
V2 lel resistance of R1 and R2 . The reason will become clear
V6
R6 R Th later.)
R2
V1 =
V Th
R7 R Th
R1

R1 V4 V5 Vin V Th

R2 Rload Rload
Figure 1.11. The Thévenin equivalent circuit.

Figure 1.12. Thévenin equivalent of a voltage divider.

1.2.5 Thévenin equivalent circuit

From this example it is easy to see that a voltage divider
Thévenin’s theorem states12 that any two-terminal network is not a very good battery, in the sense that its output volt-
of resistors and voltage sources is equivalent to a single age drops severely when a load is attached. As an exam-
ple, consider Exercise 1.10. You now know everything you
12 We provide a proof, for those who are interested, in Appendix D. need to know to calculate exactly how much the output will
10 1.2. Voltage, current, and resistance Art of Electronics Third Edition

MULTIMETERS set of switchable ranges. Ideally, a current-measuring meter should have

There are numerous instruments that let you measure voltages and cur- zero resistance13 in order not to disturb the circuit under test, since it
rents in a circuit. The oscilloscope is the most versatile; it lets you “see” must be put in series with the circuit. In practice, you tolerate a few
voltages versus time at one or more points in a circuit. Logic probes tenths of a volt drop (sometimes called “voltage burden”) with both
and logic analyzers are special-purpose instruments for troubleshooting VOMs and digital multimeters. For either kind of meter, selecting a
digital circuits. The simple multimeter provides a good way to mea- current range puts a small resistor across the meter’s input terminals,
sure voltage, current, and resistance, often with good precision; how- typically of resistance value to create a voltage drop of 0.1 V to 0.25 V
ever, it responds slowly, and thus it cannot replace the oscilloscope for the chosen full-scale current; the voltage drop is then converted to a
where changing voltages are of interest. Multimeters are of two vari- corresponding current indication.14 Typically, multimeters provide cur-
eties: those that indicate measurements on a conventional scale with a rent ranges from 50 μ A (or less) to an amp (or more), full scale.
moving pointer, and those that use a digital display. Multimeters also have one or more batteries in them to power the
The traditional (and now largely obsolete) VOM (volt-ohm- resistance measurement. By supplying a small current and measuring
milliammeter) multimeter uses a meter movement that measures cur- the voltage drop, they measure resistance, with several ranges to cover
rent (typically 50 μ A full scale). (See a less-design-oriented electronics values from 1 Ω (or less) to 10 MΩ (or more).
book for pretty pictures of the innards of meter movements; for our Important: don’t try to measure “the current of a voltage source,” by
purposes, it suffices to say that it uses coils and magnets.) To measure sticking the meter across the wall plug; the same applies for ohms. This
voltage, the VOM puts a resistor in series with the basic movement. is a leading cause of blown-out meters.
For instance, one kind of VOM will generate a 1 V (full-scale) range
by putting a 20k resistor in series with the standard 50 μ A movement;
higher voltage ranges use correspondingly larger resistors. Such a VOM
Exercise 1.7. What will a 20,000 Ω/V meter read, on its 1 V
is specified as 20,000 Ω/V, meaning that it looks like a resistor whose scale, when attached to a 1 V source with an internal resistance
value is 20k multiplied by the full-scale voltage of the particular range of 10k? What will it read when attached to a 10k–10k voltage
selected. Full scale on any voltage range is 1/20,000 amps, or 50 μ A. divider driven by a “stiff” (zero source resistance) 1 V source?
It should be clear that one of these voltmeters disturbs a circuit less
on a higher range, since it looks like a higher resistance (think of the Exercise 1.8. A 50 μ A meter movement has an internal resis-
voltmeter as the lower leg of a voltage divider, with the Thévenin resis- tance of 5k. What shunt resistance is needed to convert it to a
tance of the circuit you are measuring as the upper resistor). Ideally, a 0–1 A meter? What series resistance will convert it to a 0–10 V
voltmeter should have infinite input resistance. meter?
Most contemporary multimeters use electronic amplification and
have an input resistance of 10 MΩ to 1000 MΩ when measuring volt- Exercise 1.9. The very high internal resistance of digital multi-
age; they display their results digitally, and are known collectively as meters, in their voltage-measuring ranges, can be used to mea-
digital multimeters (DMMs). A word of caution: sometimes the input sure extremely low currents (even though the DMM may not
resistance of these meters is very high on the most sensitive ranges, offer a low current range explicitly). Suppose, for example, you
dropping to a lower resistance for the higher ranges. For instance, you want to measure the small current that flows through a 1000 MΩ
might typically have an input resistance of 109 Ω on the 0.2 V and 2 V
“leakage” resistance (that term is used to describe a small cur-
ranges, and 107 Ω on all higher ranges. Read the specifications care-
rent that ideally should be absent entirely, for example through
fully! However, for most circuit measurements these high input resis-
tances will produce negligible loading effects. In any case, it is easy to the insulation of an underground cable). You have available a
calculate how serious the effect is by using the voltage-divider equation. standard DMM, whose 2 V dc range has 10 MΩ internal resis-
Typically, multimeters provide voltage ranges from a volt (or less) to a tance, and you have available a dc source of +10 V. How can
kilovolt (or more), full scale. you use what you’ve got to measure accurately the leakage re-
A multimeter usually includes current-measuring capability, with a sistance?

drop for a given load resistance: use the Thévenin equiv- Vin =30 V and R1 = R2 = 10k, find (a) the output voltage with
alent circuit, attach a load, and calculate the new output, no load attached (the open-circuit voltage); (b) the output voltage
noting that the new circuit is nothing but a voltage divider with a 10k load (treat as a voltage divider, with R2 and Rload com-
(Figure 1.12). bined into a single resistor); (c) the Thévenin equivalent circuit;
(d) the same as in part (b), but using the Thévenin equivalent cir-
Exercise 1.10. For the circuit shown in Figure 1.12, with cuit [again, you wind up with a voltage divider; the answer should
agree with the result in part (b)]; (e) the power dissipated in each
13 of the resistors.
This is the opposite of an ideal voltage-measuring meter, which should
present infinite resistance across its input terminals.
14 A special class of current meters known as electrometers operate with A. Equivalent source resistance and circuit loading
very small voltage burdens (as little at 0.1 mV) by using the technique As we have just seen, a voltage divider powered from some
of feedback, something we’ll learn about in Chapters 2 and 4. fixed voltage is equivalent to some smaller voltage source
Art of Electronics Third Edition 1.2.5. Thévenin equivalent circuit 11

examples in the chapters ahead. This high-resistance con-

dition ideally characterizes measuring instruments such as
10k voltmeters and oscilloscopes.
+ 5k A word on language: you frequently hear things like
30V “the resistance looking into the voltage divider” or “the
+
10k 15V output sees a load of so-and-so many ohms,” as if circuits
had eyes. It’s OK (in fact, it’s a rather good way of keeping
straight which resistance you’re talking about) to say what
actual Thévenin
part of the circuit is doing the “looking.”16

Figure 1.13. Voltage divider example. 1.0 10

in series with a resistor. For example, the output terminals 0.8 8

% Reduction from VOPEN

of a 10k–10k voltage divider driven by a perfect 30 volt

VOUT / VOPEN
battery are precisely equivalent to a perfect 15 volt bat- 0.6 6
tery in series with a 5k resistor (Figure 1.13). Attaching
a load resistor causes the voltage divider’s output to drop,
0.4 4
owing to the finite source resistance (Thévenin equivalent
resistance of the voltage divider output, viewed as a source
of voltage). This is often undesirable. One solution to the 0.2 2
problem of making a stiff voltage source (“stiff” is used in
this context to describe something that doesn’t bend under 0.0 0
load) might be to use much smaller resistors in a voltage 0.1 1 10 100
RLOAD / ROUT
divider. Occasionally this brute-force approach is useful.
However, it is usually best to construct a voltage source, or Figure 1.14. To minimize the attenuation of a signal source below
power supply, as it’s commonly called, using active com- its open-circuit voltage, keep the load resistance large compared
ponents like transistors or operational amplifiers, which we with the output resistance.
will treat in Chapters 2–4. In this way you can easily make
a voltage source with internal (Thévenin equivalent) resis-
tance as small as milliohms (thousandths of an ohm), with- B. Power transfer
out the large currents and dissipation of power characteris- Here is an interesting problem: what load resistance will
tic of a low-resistance voltage divider delivering the same result in maximum power being transferred to the load for
performance. In addition, with an active power supply it is a given source resistance? (The terms source resistance,
easy to make the output voltage adjustable. These topics internal resistance, and Thévenin equivalent resistance all
are treated extensively in Chapter 9. mean the same thing.) It is easy to see that either Rload =0
The concept of equivalent internal resistance applies to or Rload =∞ results in zero power transferred, because
all sorts of sources, not just batteries and voltage dividers. Rload =0 means that Vload =0 and Iload =Vsource /Rsource , so that
Signal sources (e.g., oscillators, amplifiers, and sensing de- Pload =Vload Iload =0. But Rload =∞ means that Vload =Vsource
vices) all have an equivalent internal resistance. Attach- and Iload =0, so that again Pload =0. There has to be a maxi-
ing a load whose resistance is less than or even compara- mum in between.
ble to the internal resistance will reduce the output con- Exercise 1.11. Show that Rload = Rsource maximizes the power in
siderably. This undesirable reduction of the open-circuit the load for a given source resistance. Note: skip this exercise if
voltage (or signal) by the load is called “circuit loading.” you don’t know calculus, and take it on faith that the answer is
Therefore you should strive to make Rload Rinternal , be- true.
cause a high-resistance load has little attenuating effect on
dio frequencies and transmission lines, you must “match impedances”
the source (Figure 1.14).15 We will see numerous circuit (i.e., set Rload =Rinternal ) in order to prevent reflection and loss of power.
See Appendix H on transmission lines.
15 16 The urge to anthropomorphize runs deep in the engineering and scien-
There are two important exceptions to this general principle: (1) a cur-
rent source has a high (ideally infinite) internal resistance and should tific community, despite warnings like “don’t anthropomorphize com-
drive a load of relatively low load resistance; (2) when dealing with ra- puters . . . they don’t like it.”
12 1.2. Voltage, current, and resistance Art of Electronics Third Edition

Lest this example leave the wrong impression, we through a resistor from a higher voltage available some-
would like to emphasize again that circuits are ordinarily where in the circuit, as in Figure 1.16.
designed so that the load resistance is much greater than
the source resistance of the signal that drives the load. I
1k
1mA
1.2.6 Small-signal resistance 4k

We often deal with electronic devices for which I is not

proportional to V ; in such cases there’s not much point in 1V V
talking about resistance, since the ratio V /I will depend on
V , rather than being a nice constant, independent of V . For
these devices it is sometimes useful to know instead the
slope of the V –I curve, in other words, the ratio of a small
change in applied voltage to the resulting change in cur- A.
rent through the device, ΔV /ΔI (or dV/dI). This quantity
I
has the units of resistance (ohms) and substitutes for re-
diode
sistance in many calculations. It is called the small-signal conduction
resistance, incremental resistance, or dynamic resistance.
real V
zener
A. Zener diodes
As an example, consider the zener diode, which has the ideal
zener
I–V curve shown in Figure 1.15. Zeners are used to cre- B.
ate a constant voltage inside a circuit somewhere, simply
Figure 1.15. I–V curves: A. Resistor (linear). B. Zener diode (non-
done by providing them with a (roughly constant) current
linear).
derived from a higher voltage within the circuit.17 For ex-
ample, the zener diode in Figure 1.15 will convert an ap-
plied current in the range shown to a corresponding (but R
+ +
fractionally narrower) range of voltages. It is important to
know how the resulting zener voltage will change with ap- Vin Vout

plied current; this is a measure of its “regulation” against

changes in the driving current provided to it. Included in Figure 1.16. Zener regulator.
the specifications of a zener will be its dynamic resistance,
given at a certain current. For example, a zener might have Then,
a dynamic resistance of 10 Ω at 10 mA, at its specified
Vin −Vout
zener voltage of 5 V. Using the definition of dynamic re- I=
R
sistance, we find that a 10% change in applied current will
therefore result in a change in voltage of and
ΔVin − ΔVout
ΔV = Rdyn ΔI = 10 × 0.1 × 0.01 = 10 mV ΔI = ,
R
or so
Rdyn
ΔV /V = 0.002 = 0.2%, ΔVout = Rdyn ΔI = (ΔVin − ΔVout )
R
thus demonstrating good voltage-regulating ability. In this and finally
sort of application you frequently get the zener current Rdyn
ΔVout = ΔV .
R + Rdyn in
17 Zeners belong to the more general class of diodes and rectifiers, im-
portant devices that we’ll see later in the chapter (§1.6), and indeed Aha – the voltage-divider equation, again! Thus, for
throughout the book. The ideal diode (or rectifier) acts as a perfect changes in voltage, the circuit behaves like a voltage di-
conductor for current flow in one direction, and a perfect insulator for vider, with the zener replaced by a resistor equal to its
current flow in the reverse direction; it is a “one-way valve” for current. dynamic resistance at the operating current. This is the
Art of Electronics Third Edition 1.2.7. An example: “It’s too hot!” 13

utility of incremental resistance. For instance, suppose in 10

the preceding circuit we have an input voltage ranging
between 15 and 20 V, and we use a 1N4733 (5.1 V, 1W

er

5.6V zener
r
8

2.4V zene

3.3 V zen
zener diode) in order to generate a stable 5.1 V power sup-
ply. We choose R = 300 Ω, for a maximum zener cur-

Current (mA)
6
rent of 50 mA: (20 V−5.1 V)/300 Ω. We can now estimate

2.50V ref
the output-voltage regulation (variation in output voltage),

1.25V ref
knowing that this particular zener has a specified maxi- 4
mum dynamic resistance of 7.0 Ω at 50 mA. The zener cur-
rent varies from 50 mA to 33 mA over the input-voltage 2
range; this 17 mA change in current then produces a volt-
age change at the output of ΔV = Rdyn ΔI, or 0.12 V. 0
It’s a useful fact, when dealing with zener diodes, that 0 1 2 3 4 5 6
the dynamic resistance of a zener diode varies roughly Voltage (volts)
in inverse proportion to current. It’s worth knowing, also, Figure 1.17. Low-voltage zeners are pretty dismal, as seen in
that there are ICs designed to substitute for zener diodes; these measured I vs. V curves (for three members of the 1N5221–
these “two-terminal voltage references” have superior per- 67 series), particularly in contrast to the excellent measured per-
formance – much lower dynamic resistance (less than 1 Ω, formance of a pair of “IC voltage references” (LM385Z-1.2 and
even at currents as small as 0.1 mA; that’s a thousand times LM385Z-2.5, see §9.10 and Table 9.7). However, zener diodes
better than the zener we just used), and excellent temper- in the neighborhood of 6 V (such as the 5.6 V 1N5232B or 6.2 V
1N5234B) exhibit admirably steep curves, and are useful parts.
ature stability (better than 0.01%/C). We will see more of
zeners and voltage references in §§2.2.4 and 9.10.
In real life, a zener will provide better regulation if it the lower leg of a voltage divider (R3 R4 ), whose output
driven by a current source, which has, by definition, is compared with the temperature-insensitive divider R1 R2 .
Rincr =∞ (the same current, regardless of voltage). But cur- When it’s hotter than 30◦ C, point “X” is at a lower voltage
rent sources are more complex, and therefore in practice than point “Y,” so the comparator pulls its output to ground.
we often resort to the humble resistor. When thinking about At the output there’s an LED, which behaves electrically
zeners, it’s worth remembering that low-voltage units (e.g., like a 1.6 V zener diode; and when current is flowing, it
3.3 V) behave rather poorly, in terms of constancy of volt- lights up. Its lower terminal is then at 5 V−1.6 V, or +3.4 V.
age versus current (Figure 1.17); if you think you need So we’ve added a series resistor, sized to allow 5 mA when
a low voltage zener, use a two-terminal reference instead the comparator output is at ground: R5 =3.4 V/5 mA, or
(§9.10). 680 Ω.
If you wanted to, you could make the setpoint adjustable
by replacing R2 with a 5k pot in series with a 5k fixed
1.2.7 An example: “It’s too hot!” resistor. We’ll see later that it’s also a good idea to add
Some people like to turn the thermostat way up, annoy- some hysteresis, to encourage the comparator to be deci-
ing other people who like their houses cool. Here’s a little sive. Note that this circuit is insensitive to the exact power-
gadget (Figure 1.18) that lets folks of the latter persuasion supply voltage because it compares ratios. Ratiometric
know when to complain – it lights up a red light-emitting techniques are good; we’ll see them again later.
diode (LED) indicator when the room is warmer than 30◦ C
(86◦ F). It also shows how to use the humble voltage divider
(and even humbler Ohm’s law), and how to deal with an 1.3 Signals
LED, which behaves like a zener diode (and is sometimes A later section in this chapter will deal with capacitors, de-
used as such). vices whose properties depend on the way the voltages and
The triangular symbol is a comparator, a handy device currents in a circuit are changing. Our analysis of dc cir-
(discussed in §12.3) that switches its output according to cuits so far (Ohm’s law, Thévenin equivalent circuits, etc.)
the relative voltages at its two input terminals. The temper- still holds, even if the voltages and currents are changing in
ature sensing device is R4 , which decreases in resistance by time. But for a proper understanding of alternating-current
about 4%/◦ C, and which is 10kΩ at 25◦ C. So we’ve made (ac) circuits, it is useful to have in mind certain common
14 1.3. Signals Art of Electronics Third Edition

+5V second. Just remember the important relation ω = 2π f and

red LED
VF = 1.6V you won’t go wrong.
R3
R1
10k
@ 5mA The great merit of sinewaves (and the cause of their
10k R5
Y – 680Ω perennial popularity) is the fact that they are the solutions
X
goes low to certain linear differential equations that happen to de-
+ TLC 372 when V X < V Y
comparator
scribe many phenomena in nature as well as the prop-
R2
8.06k R4 erties of linear circuits. A linear circuit has the property
thermistor: 10.0k @ 25°C that its output, when driven by the sum of two input sig-
#103JG1F 8.06k @ 30°C
6.53k @ 35°C nals, equals the sum of its individual outputs when driven
by each input signal in turn; i.e., if O(A) represents the
Figure 1.18. The LED lights up when it’s hotter than 30◦ C. The output when driven by signal A, then a circuit is linear
comparator (which we’ll see later, in Chapters 4 and 12) pulls its
if O(A + B) = O(A) + O(B). A linear circuit driven by
output to ground when the voltage at “X” is less than the voltage
a sinewave always responds with a sinewave, although in
at “Y.” R4 is a thermistor, which is a resistor with a deliberate neg-
ative temperature coefficient; that is, its resistance decreases with
general the phase and amplitude are changed. No other pe-
increasing temperature – by about 4%/◦ C. riodic signal can make this statement. It is standard prac-
tice, in fact, to describe the behavior of a circuit by its fre-
types of signals, voltages that change in time in a particu- quency response, by which we mean the way the circuit
lar way. alters the amplitude of an applied sinewave as a function
of frequency. A stereo amplifier, for instance, should be
characterized by a “flat” frequency response over the range
1.3.1 Sinusoidal signals 20 Hz to 20 kHz, at least.
The sinewave frequencies we usually deal with range
Sinusoidal signals are the most popular signals around;
from a few hertz to a few tens of megahertz. Lower fre-
they’re what you get out of the wall plug. If someone says
quencies, down to 0.0001 Hz or lower, can be generated
something like “take a 10 μ V signal at 1 MHz,” they mean
with carefully built circuits, if needed. Higher frequencies,
a sinewave. Mathematically, what you have is a voltage de-
up to say 2000 MHz (2 GHz) and above, can be gener-
scribed by
ated, but they require special transmission-line techniques.
V = A sin 2π f t (1.10) Above that, you’re dealing with microwaves, for which
conventional wired circuits with lumped-circuit elements
where A is called the amplitude and f is the frequency in
become impractical, and exotic waveguides or “striplines”
hertz (cycles per second). A sinewave looks like the wave
are used instead.
shown in Figure 1.19. Sometimes it is important to know
the value of the signal at some arbitrary time t = 0, in which
case you may see a phase φ in the expression: 1.3.2 Signal amplitudes and decibels
V = A sin(2π f t + φ ). In addition to its amplitude, there are several other ways to
characterize the magnitude of a sinewave or any other sig-
nal. You sometimes see it specified by peak-to-peak ampli-
A tude (pp amplitude), which is just what you would guess,
namely, twice the amplitude. The other method is to give
t the root-mean-square amplitude (rms amplitude), which is
√
1 1 3 Vrms = (1/ 2)A = 0.707A (this is for sinewaves only; the
2f f 2f ratio of pp to rms will be different for other waveforms).
Odd as it may seem, this is the usual method, because rms
Figure 1.19. Sinewave of amplitude A and frequency f . voltage is what’s used to compute power. The nominal volt-
age across the terminals of a wall socket (in the United
The other variation on this simple theme is the use of States) is 120 volts rms, 60 Hz. The amplitude is 170 volts
angular frequency, which looks like this: (339 volts pp).18
V = A sin ω t.
18 Occasionally you’ll encounter devices (e.g., mechanical moving-
Here ω is the angular frequency, measured in radians per pointer meters) that respond to the average magnitude of an ac signal.
Art of Electronics Third Edition 1.3.3. Other signals 15

A. Decibels say something like “an amplitude of 27 decibels relative to

How do you compare the relative amplitudes of two sig- 1 V rms,” or abbreviate “27 dB re 1 V rms,” or define a term
nals? You could say, for instance, that signal X is twice like “dBV.”19
as large as signal Y . That’s fine, and useful for many pur-
Exercise 1.12. Determine the voltage and power ratios for a pair
poses. But because we often deal with ratios as large as a of signals with the following decibel ratios: (a) 3 dB, (b) 6 dB, (c)
million, it is better to use a logarithmic measure, and for 10 dB, (d) 20 dB.
this we present the decibel (it’s one-tenth as large as some-
thing called a bel, which no one ever uses). By definition, Exercise 1.13. We might call this amusing exercise “Desert Island
dBs”: in the table below we’ve started entering some values for
the ratio of two signals, in decibels (dB), is
power ratios corresponding to the first dozen integral dBs, using
P2
dB = 10 log10 , (1.11) the results for parts (a) and (c) of the last exercise. Your job is to
P1 complete the table, without recourse to a calculator. A possibly
where P1 and P2 represent the power in the two signals. helpful hint: starting at 10 dB, go down the table in steps of 3 dB,
We are often dealing with signal amplitudes, however, in then up in a step of 10 dB, then down again. Finally, get rid of
which case we can express the ratio of two signals having yucky numbers like 3.125 (and its near relatives) by noticing that
the same waveform as it’s charmingly close to π .
A2 dB ratio (P/P0 )
dB = 20 log10 , (1.12)
A1 0 1
where A1 and A2 are the two signal amplitudes. So, for 1
instance, one signal of twice the amplitude of another is 2
3 2
+6 dB relative to it, since log10 2 = 0.3010. A signal 10
4
times as large is +20 dB; a signal one-tenth as large is 5
−20 dB. 6 4
Although decibels are ordinarily used to specify the ra- 7
tio of two signals, they are sometimes used as an abso- 8
lute measure of amplitude. What is happening is that you 9 8
are assuming some reference signal level and expressing 10 10
any other level in decibels relative to it. There are sev- 11
eral standard levels (which are unstated, but understood)
that are used in this way; the most common references are
1.3.3 Other signals
(a) 0 dBV (1 V rms); (b) 0 dBm (the voltage correspond-
ing to 1 mW into some assumed load impedance, which A. Ramp
for radiofrequencies is usually 50 Ω, but for audio is often The ramp is a signal that looks like the one shown in Fig-
600 Ω; the corresponding 0 dBm amplitudes, when loaded ure 1.20A. It is simply a voltage rising (or falling) at a
by those impedances, are then 0.22 V rms and 0.78 V rms); constant rate. That can’t go on forever, of course, even in
and (c) the small noise voltage generated by a resistor science fiction movies. It is sometimes approximated by a
at room temperature (this surprising fact is discussed in finite ramp (Figure 1.20B) or by a periodic ramp (known
§8.1.1). In addition to these, there are reference amplitudes as a sawtooth, Figure 1.20C).
used for measurements in other fields of engineering and
science. For instance, in acoustics, 0 dB SPL (sound pres- B. Triangle
sure level) is a wave whose rms pressure is 20 μ Pa (that’s The triangle wave is a close cousin of the ramp; it is simply
2×10−10 atm); in audio communications, levels can be a symmetrical ramp (Figure 1.21).
stated in dBrnC (relative noise reference weighted in fre-
quency by “curve C”). When stating amplitudes this way, C. Noise
it is best to be specific about the 0 dB reference amplitude; Signals of interest are often mixed with noise; this is a
catch-all phrase that usually applies to random noise of
For a sinewave the relationship is Vavg =Vrms /1.11. However, such me- thermal origin. Noise voltages can be specified by their
ters are usually calibrated so that they indicate the rms sinewave am-
plitude. For signals other than sinewaves their indication is in error; be 19 One of the authors, when asked by his nontechnical spouse how much
sure to use a “true rms” meter if you want the right answer. we spent on that big plasma screen, replied “36 dB$.”
16 1.3. Signals Art of Electronics Third Edition

V D. Square wave
A square wave is a signal that varies in time as shown in
Figure 1.23. Like the sinewave, it is characterized by am-
plitude and frequency (and perhaps phase). A linear cir-
cuit driven by a square wave rarely responds with a square
A. wave. For a square wave, the peak amplitude and the rms
t amplitude are the same.
V

1 1 2 3
B. t=0
2f f f f
t
V Figure 1.23. Square wave.

100%
90%
C.
t

Figure 1.20. A: Voltage-ramp waveform. B: Ramp with limit. C: 10%

0
Sawtooth wave.
tr

Figure 1.24. Rise time of a step waveform.

The edges of a square wave are not perfectly square; in

typical electronic circuits the rise time tr ranges from a few
nanoseconds to a few microseconds. Figure 1.24 shows the
Figure 1.21. Triangle wave. sort of thing usually seen. The rise time is conventionally
defined as the time required for the signal to go from 10%
to 90% of its total transition.
V
V

+
t

t
–
Figure 1.22. Noise.
Figure 1.25. Positive- and negative-going pulses of both polarities.
frequency spectrum (power per hertz) or by their ampli-
tude distribution. One of the most common kind of noise
is band-limited white Gaussian noise, which means a sig- E. Pulses
nal with equal power per hertz in some band of frequencies A pulse is a signal that looks like the objects shown in Fig-
and that exhibits a Gaussian (bell-shaped) distribution of ure 1.25. It is defined by amplitude and pulse width. You
amplitudes when many instantaneous measurements of its can generate a train of periodic (equally spaced) pulses, in
amplitude are made. This kind of noise is generated by a which case you can talk about the frequency, or pulse repe-
resistor (Johnson noise or Nyquist noise), and it plagues tition rate, and the “duty cycle,” the ratio of pulse width to
sensitive measurements of all kinds. On an oscilloscope it repetition period (duty cycle ranges from zero to 100%).
appears as shown in Figure 1.22. We will discuss noise and Pulses can have positive or negative polarity; in addi-
low-noise techniques in considerable detail in Chapter 8. tion, they can be “positive-going” or “negative-going.” For
Art of Electronics Third Edition 1.3.5. Signal sources 17

instance, the second pulse in Figure 1.25 is a negative- with provision for precise control of amplitude (using a
going pulse of positive polarity. resistive divider network called an attenuator). Some units
let you modulate (i.e., vary in time) the output amplitude
F. Steps and spikes (“AM” for “amplitude modulated”) or frequency (“FM”
Steps and spikes are signals that are talked about a lot but for “frequency modulated”). A variation on this theme is
are not so often used. They provide a nice way of describ- the sweep generator, a signal generator that can sweep
ing what happens in a circuit. If you could draw them, they its output frequency repeatedly over some range. These
would look something like the example in Figure 1.26. The are handy for testing circuits whose properties vary with
step function is part of a square wave; the spike is simply a frequency in a particular way, e.g., “tuned circuits” or
jump of vanishingly short duration. filters. Nowadays these devices, as well as most test
instruments, are available in configurations that allow you
to program the frequency, amplitude, etc., from a computer
or other digital instrument.
For many signal generators the signal source is a fre-
quency synthesizer, a device that generates sinewaves
step spike whose frequencies can be set precisely. The frequency is
set digitally, often to eight significant figures or more, and
Figure 1.26. Steps and spikes. is internally synthesized from a precise standard (a stand-
alone quartz-crystal oscillator or rubidium frequency stan-
dard, or a GPS-derived oscillator) by digital methods we
will discuss later (§13.13.6). Typical of synthesizers is the
1.3.4 Logic levels
programmable SG384 from Stanford Research Systems,
Pulses and square waves are used extensively in digital with a frequency range of 1 μ Hz to 4 GHz, an amplitude
electronics, in which predefined voltage levels represent range of −110 dBm to +16.5 dBm (0.7 μ V to 1.5 V, rms),
one of two possible states present at any point in the cir- and various modulation modes such as AM, FM, and ΦM;
cuit. These states are called simply HIGH and LOW, and it costs about $4,600. You can get synthesized sweep gen-
correspond to the 1 (true) and 0 (false) states of Boolean erators, and you can get synthesizers that produce other
logic (the algebra that describes such two-state systems). waveforms (see Function Generators, below). If your re-
Precise voltages are not necessary in digital electron- quirement is for no-nonsense accurate frequency genera-
ics. You need only to distinguish which of the two pos- tion, you can’t beat a synthesizer.
sible states is present. Each digital logic family therefore
specifies legal HIGH and LOW states. For example, the B. Pulse generators
“74LVC” digital logic family runs from a single +3.3 V Pulse generators make only pulses, but what pulses! Pulse
supply, with output levels that are typically 0 V (LOW) and width, repetition rate, amplitude, polarity, rise time, etc.,
3.3 V (HIGH), and an input decision threshold of 1.5 V. may all be adjustable. The fastest ones go up to gigahertz
However, actual outputs can be as much as 0.4 V away pulse rates. In addition, many units allow you to gener-
from ground or from +3.3 V without malfunction. We’ll ate pulse pairs, with settable spacing and repetition rate,
have much more to say about logic levels in Chapters 10 or even programmable patterns (they are sometimes called
through 12. pattern generators). Most contemporary pulse generators
are provided with logic-level outputs for easy connection
to digital circuitry. As with signal generators, these come
1.3.5 Signal sources
in the programmable variety.
Often the source of a signal is some part of the circuit
you are working on. But for test purposes a flexible sig- C. Function generators
nal source is invaluable. They come in three flavors: signal In many ways function generators are the most flexible
generators, pulse generators, and function generators. signal sources of all. You can make sine, triangle, and
square waves over an enormous frequency range (0.01 Hz
A. Signal generators to 30 MHz is typical), with control of amplitude and dc
Signal generators are sinewave oscillators, usually offset (a constant-dc voltage added to the signal). Many
equipped to give a wide range of frequency coverage, of them have provision for frequency sweeping, often in
18 1.4. Capacitors and ac circuits Art of Electronics Third Edition

several modes (linear or logarithmic frequency variation

versus time). They are available with pulse outputs (al- +
though not with the flexibility you get with a pulse gen-
erator), and some of them have provision for modulation. Figure 1.27. Capacitors. The curved electrode indicates the neg-
Traditional function generators used analog circuitry, ative terminal of a polarized capacitor, or the “outer foil” of a
but contemporary models generally are synthesized dig- wrapped-film capacitor.
ital function generators, exhibiting all the flexibility of
a function generator along with the stability and accu- tions of the treatment that follows are necessarily mathe-
racy of a frequency synthesizer. In addition, they let you matical in nature; the reader with little mathematical prepa-
program an “arbitrary” waveform, specifying the ampli- ration may find the math review in Appendix A helpful. In
tude at a set of equally spaced points. An example is any case, an understanding of the details is less important
the Tektronix AFG3102, with a lower frequency limit of in the long run than an understanding of the results.
1 microhertz, which can make sine and square waves to
100 MHz, pulses and “noise” to 50 MHz, and arbitrary
waveforms (up to 128k points) to 50 MHz. It has modula- 1.4.1 Capacitors
tion (five kinds), sweep (linear and log), and burst modes (1 A capacitor (Figure 1.27) (the old-fashioned name was
to 106 cycles), and everything is programmable, including condenser) is a device that has two wires sticking out of
frequency, pulse width and rise times, modulation, and am- it and has the property
plitude (20 mV to 10 Vpp); it even includes some bizarre
Q = CV. (1.13)
built-in waveforms such as sin(x)/x, exponential rise and
fall, Gaussian, and Lorentzian. It has two independent out- Its basic form is a pair of closely-spaced metal plates,
puts and costs about $5k. For general use, if you can have separated by some insulating material, as in the rolled-
only one signal source, the function generator is for you. up “axial-film capacitor” of Figure 1.28. A capacitor of C
farads with V volts across its terminals has Q coulombs of
stored charge on one plate and −Q on the other. The capac-
1.4 Capacitors and ac circuits itance is proportional to the area and inversely proportional
Once we enter the world of changing voltages and currents, to the spacing. For the simple parallel-plate capacitor, with
or “signals,” we encounter two very interesting circuit separation d and plate area A (and with the spacing d much
elements that are useless in purely dc circuits: capacitors less than the dimensions of the plates), the capacitance C is
and inductors. As you will see, these humble devices, given by
combined with resistors, complete the triad of passive
C = 8.85×10−14 ε A/d F, (1.14)
linear circuit elements that form the basis of nearly all
circuitry.20 Capacitors, in particular, are essential in nearly where ε is the dielectric constant of the insulator, and the
every circuit application. They are used for waveform dimensions are measured in centimeters. It takes a lot of
generation, filtering, and blocking and bypass applications. area, and tiny spacing, to make the sort of capacitances
They are used in integrators and differentiators. In com- commonly used in circuits.21 For example, a pair of 1 cm2
bination with inductors, they make possible sharp filters plates separated by 1 mm is a capacitor of slightly less than
for separating desired signals from background. You will 10−12 F (a picofarad); you’d need 100,000 of them just to
see some of these applications as we continue with this create the 0.1 μ F capacitor of Figure 1.28 (which is nothing
chapter, and there will be numerous interesting examples special; we routinely use capacitors with many microfarads
in later chapters. of capacitance). Ordinarily you don’t need to calculate ca-
Let’s proceed, then, to look at capacitors in detail. Por- pacitances, because you buy a capacitor as an electronic
component.
20 Readers of the scientific journal Nature (London) were greeted, To a first approximation, capacitors are devices that
in 2008, with an article titled “The missing memristor found” might be considered simply frequency-dependent resistors.
(D. B. Strukov et al., 453, 80, 2008), purporting to have found a hereto-
fore missing “fourth fundamental [passive circuit] element.” We are 21 And it doesn’t hurt to have a high dielectric constant, as well: air has
skeptical. However the controversy is ultimately resolved, it should be ε =1, but plastic films have ε =2.1 (polypropylene) or 3.1 (polyester).
noted that the memristor is a nonlinear element; there are only three And certain ceramics are popular among capacitor makers: ε =45
linear passive 2-terminal circuit elements. (“C0G” type) or 3000 (“X7R” type).
Art of Electronics Third Edition 1.4.1. Capacitors 19

dV
I =C . (1.15)
dielectric dt
(insulator)
So a capacitor is more complicated than a resistor: the cur-
rent is not simply proportional to the voltage, but rather
to the rate of change of voltage. If you change the voltage
across a farad by 1 volt per second, you are supplying an
outside foil
amp. Conversely, if you supply an amp, its voltage changes
by 1 volt per second. A farad is an enormous capacitance,
inside foil and you usually deal in microfarads (μ F), nanofarads (nF),
Figure 1.28. You get a lot of area by rolling up a pair of metal- or picofarads (pF).23 For instance, if you supply a current
lized plastic films. And it’s great fun unrolling one of these axial-lead of 1 mA to 1μ F, the voltage will rise at 1000 volts per sec-
Mylar capacitors (ditto for the old-style golf balls with their lengthy ond. A 10 ms pulse of this current will increase the voltage
wound-up rubber band). across the capacitor by 10 volts (Figure 1.29).

They allow you to make frequency-dependent voltage di- I

viders, for instance. For some applications (bypass, cou- 1mA
pling) this is almost all you need to know, but for other
applications (filtering, energy storage, resonant circuits) a t
deeper understanding is needed. For example, ideal capac-
10ms
itors cannot dissipate power, even though current can flow
through them, because the voltage and current are 90◦ out
of phase.
Before launching into the details of capacitors in the fol- V C=1μF
lowing dozen pages (including some necessary mathemat- Vafter
ics that describes their behavior in time and in frequency), 10V
Vbefore
we wish to emphasize those first two applications – by-
pass and coupling – because they are the most common
uses of capacitors, and they are easy to understand at the t
simplest level. We’ll see these in detail later (§§1.7.1C and
1.7.16A), but no need to wait – it’s easy, and intuitive. Be- Figure 1.29. The voltage across a capacitor changes when a cur-
cause a capacitor looks like an open circuit at dc, it lets rent flows through it.
you couple a varying signal while blocking its average dc
level. This is a blocking capacitor (also called a coupling When you charge up a capacitor, you’re supplying en-
capacitor), as in Figure 1.93. Likewise, because a capac- ergy. The capacitor doesn’t get hot; instead, it stores the
itor looks like a short circuit at high frequencies, it sup- energy in its internal electric fields. It’s an easy exercise to
presses (“bypasses”) signals where you don’t want them, discover for yourself that the amount of stored energy in a
for example on the dc voltages that power your circuits, as charged capacitor is just
in Figure 8.80A (where capacitors are suppressing signals 1
on the +5 V and −5 V dc supply voltages, and also on the UC = CV 2 , (1.16)
2
base terminal of transistor Q2 ).22 Demographically, these
where UC is in joules for C in farads and V in volts. This is
two applications account for the vast majority of capaci-
an important result; we’ll see it often.
tors that are wired into the world’s circuits.
Taking the derivative of the defining equation 1.13, you Exercise 1.14. Take the energy challenge: imagine charging up
get a capacitor of capacitance C, from 0 V to some final voltage Vf .
If you do it right, the result won’t depend on how you get there,
22 Ironically, these essential bypass capacitors are so taken for granted
that they are usually omitted from schematic diagrams (a practice we 23 To make matters confusing to the uninitiated, the units are often omit-
follow in this book). Don’t make the mistake of omitting them also ted on capacitor values specified in schematic diagrams. You have to
from your actual circuits! figure it out from the context.
20 1.4. Capacitors and ac circuits Art of Electronics Third Edition

Figure 1.30. Capacitors masquerade as anything they like! Here is a representative collection. In the lower left are small-value variable
capacitors (one air, three ceramic), with large-value polarized aluminum electrolytics above them (the three on the left have radial leads,
the three on the right have axial leads, and the specimen with screw terminals at top is often called a computer electrolytic). Next in
line across the top is a low-inductance film capacitor (note the wide strap terminals), then an oil-filled paper capacitor, and last, a set
of disc ceramic capacitors running down the right. The four rectangular objects below are film capacitors (polyester, polycarbonate, or
polypropylene). The D-subminiature connector seems misplaced – but it is a filtered connector, with a 1000 pF capacitor from each pin
to the shell. To its left is a group of seven polarized tantalum electrolytics (five with axial leads, one radial, and one surface-mount). The
three capacitors above them are axial-film capacitors. The ten capacitors at bottom center are all ceramic types (four with radial leads,
two axial, and four surface-mount chip capacitors); above them are high-voltage capacitors – an axial-glass capacitor, and a ceramic
transmitting capacitor with screw terminals. Finally, below them and to the left are four mica capacitors and a pair of diode-like objects
known as varactors, which are voltage-variable capacitors made from a diode junction.

so you don’t need to assume constant current charging (though closer spacing; the usual approach is to plate some conduc-
you’re welcome to do so). At any instant the rate of flow of energy tor onto a thin insulating material (the dielectric), for in-
into the capacitor is V I (joules/s); so you need to integrate dU = stance, aluminized plastic film rolled up into a small cylin-
V Idt from start to finish. Take it from there. drical configuration. Other popular types are thin ceramic
wafers (ceramic chip capacitors), metal foils with oxide
Capacitors come in an amazing variety of shapes and insulators (electrolytic capacitors), and metallized mica.
sizes (Figure 1.30 shows examples of most of them); with Each of these types has unique properties; for a brief run-
time, you will come to recognize their more common incar- down, see the section on capacitors in Chapter 1x. In gen-
nations. For the smallest capacitances you may see exam- eral, ceramic and polyester types are used for most non-
ples of the basic parallel-plate (or cylindrical piston) con- critical circuit applications; capacitors with polycarbonate,
struction. For greater capacitance, you need more area and polystyrene, polypropylene, Teflon, or glass dielectric are
Art of Electronics Third Edition 1.4.2. RC circuits: V and I versus time 21

used in demanding applications; tantalum capacitors are study of ac circuits in the time domain. Starting with §1.7,
used where greater capacitance is needed; and aluminum we will tackle the frequency domain.
electrolytics are used for power-supply filtering. What are some of the features of circuits with capaci-
tors? To answer this question, let’s begin with the simple
A. Capacitors in parallel and series RC circuit (Figure 1.31). Application of the capacitor rules
The capacitance of several capacitors in parallel is the sum gives
of their individual capacitances. This is easy to see: put dV V
C =I=− . (1.19)
voltage V across the parallel combination; then dt R
This is a differential equation, and its solution is
CtotalV = Qtotal = Q1 + Q2 + Q3 + · · ·
= C1V +C2V +C3V + · · · V = Ae−t/RC . (1.20)
= (C1 +C2 +C3 + · · ·)V So a charged capacitor placed across a resistor will dis-
charge as in Figure 1.32. Intuition serves well here: the cur-
or rent that flows is (from the resistor equation) proportional
Ctotal = C1 +C2 +C3 + · · · . (1.17) to the remaining voltage; but the slope of the discharge is
(from the capacitor equation) proportional to that current.
For capacitors in series, the formula is like that for resistors So the discharge curve has to be a function whose deriva-
in parallel: tive is proportional to its value, i.e., an exponential.
1
Ctotal = (1.18)
1 1 1
+ + +···
C1 C2 C3
C R
or (two capacitors only)
C1C2
Ctotal = . Figure 1.31. The simplest RC circuit.
C1 +C2
Exercise 1.15. Derive the formula for the capacitance of two ca-
pacitors in series. Hint: because there is no external connection V
to the point where the two capacitors are connected together, they V0
must have equal stored charges. V = V0e –t/RC

The current that flows in a capacitor during charging

37%
(I = C dV/dt) has some unusual features. Unlike resistive
current, it’s not proportional to voltage, but rather to the
t
rate of change (the “time derivative”) of voltage. Further-
more, unlike the situation in a resistor, the power (V ×I) as- A. t = RC
sociated with capacitive current is not turned into heat, but
is stored as energy in the capacitor’s internal electric field. V
1.0
You get all that energy back when you discharge the capac-
itor. We’ll see another way to look at these curious proper-
ties when we talk about reactance, beginning in §1.7.
0.1
1.4.2 RC circuits: V and I versus time
When dealing with ac circuits (or, in general, any circuits
that have changing voltages and currents), there are two 0.01
possible approaches. You can talk about V and I versus
time, or you can talk about amplitude versus signal fre- 0 RC 2RC 3RC 4RC 5RC
B. time (seconds)
quency. Both approaches have their merits, and you find
yourself switching back and forth according to which de- Figure 1.32. RC discharge waveform, plotted with (A) linear and
scription is most convenient in each situation. We begin our (B) logarithmic voltage axes.
22 1.4. Capacitors and ac circuits Art of Electronics Third Edition

A. Time constant V
The product RC is called the time constant of the circuit. Vf
For R in ohms and C in farads, the product RC is in seconds.
A microfarad across 1.0k has a time constant of 1 ms; if the 63%
Vout = Vf (1 – e –t / RC )
capacitor is initially charged to 1.0 V, the initial current is
1.0 mA.

R
0
t = RC t

+ Vout
battery, Figure 1.34. RC charging waveform.
voltage = Vf C

Vin
Figure 1.33. RC charging circuit.
t
Figure 1.33 shows a slightly different circuit. At time (lower frequency)
t = 0, someone connects the battery. The equation for the
circuit is then Vout
dV V −Vout
I =C = f ,
dt R t
with the solution
Vout = Vf + A e−t/RC . Figure 1.35. Output (lower waveforms) across a capacitor, when
driven by square waves through a resistor.
(Please don’t worry if you can’t follow the mathematics.
What we are doing is getting some important results, which within 1% of its final value in five time constants.) If we
you should remember. Later we will use the results of- then change the battery voltage to some other value (say,
ten, with no further need for the mathematics used to de- 0 V), V will decay toward that new value with an exponen-
rive them. For readers whose knowledge of math is some- tial e−t/RC . For example, replacing the battery’s step input
what, uh, rusty, the brief review in Appendix A may prove from 0 to +Vf with a square-wave input Vin (t) would pro-
helpful.) The constant A is determined by initial conditions duce the output shown in Figure 1.35.
(Figure 1.34): V = 0 at t = 0; therefore, A = −Vf , and
Exercise 1.16. Show that the rise time (the time required for going
Vout = Vf (1 − e−t/RC ). (1.21) from 10% to 90% of its final value) of this signal is 2.2RC.
Once again there’s good intuition: as the capacitor charges You might ask the obvious next question: what about
up, the slope (which is proportional to current, because it’s V (t) for arbitrary Vin (t)? The solution involves an inhomo-
a capacitor) is proportional to the remaining voltage (be- geneous differential equation and can be solved by stan-
cause that’s what appears across the resistor, producing the dard methods (which are, however, beyond the scope of
current); so we have a waveform whose slope decreases this book). You would find
proportionally to the vertical distance it has still to go – an  t
1
exponential. V (t) = Vin (τ )e−(t−τ )/RC d τ .
You can turn around the last equation to figure out the RC −∞

time required to reach a voltage V on the way to the final That is, the RC circuit averages past history at the input
voltage Vf . Try it! (Refer to Appendix A if you need help with a weighting factor of
with logarithms.) You should get
  e−Δt/RC .
Vf
t = RC loge (1.22)
Vf −V In practice, you seldom ask this question. Instead, you
deal in the frequency domain, in which you ask how much
B. Decay to equilibrium of each frequency component present in the input gets
Eventually (when t  RC), V reaches Vf . (Presenting the through. We will get to this important topic soon (§1.7).
“5RC rule of thumb”: a capacitor charges or decays to Before we do, though, there are a few other interesting
Art of Electronics Third Edition 1.4.2. RC circuits: V and I versus time 23

circuits we can analyze simply with this time-domain ap- D. A circuit example: time-delay circuit
proach. Let’s take a short detour to try out these theoretical ideas
on a couple of real circuits. Textbooks usually avoid such
R1
pragmatism, especially in early chapters, but we think it’s
fun to apply electronics to practical applications. We’ll
Vin V(t) need to introduce a few “black-box” components to get the
R2 C
job done, but you’ll learn about them in detail later, so don’t
worry.
Figure 1.36. Looks complicated, but it’s not! (Thévenin to the res-
We have already mentioned logic levels, the voltages
cue.) that digital circuits live on. Figure 1.37 shows an applica-
tion of capacitors to produce a delayed pulse. The triangu-
lar symbols are “CMOS24 buffers.” They give a HIGH out-
C. Simplification by Thévenin equivalents put if the input is HIGH (more than one-half the dc power-
We could go ahead and analyze more complicated circuits supply voltage used to power them), and vice versa. The
by similar methods, writing down the differential equations first buffer provides a replica of the input signal, but with
and trying to find solutions. For most purposes it simply low source resistance, to prevent input loading by the RC
isn’t worth it. This is as complicated an RC circuit as we (recall our earlier discussion of circuit loading in §1.2.5A).
will need. Many other circuits can be reduced to it; take, The RC output has the characteristic decays and causes the
for example, the circuit in Figure 1.36. By just using the output buffer to switch 10 μ s after the input transitions (an
Thévenin equivalent of the voltage divider formed by R1 RC reaches 50% output after a time t = 0.7RC). In an actual
and R2 , you can find the output V (t) produced by a step application you would have to consider the effect of the
input for Vin . buffer input threshold deviating from one-half the supply
voltage, which would alter the delay and change the output
Exercise 1.17. In the circuit shown in Figure 1.36, R1 = R2 =10k,
pulse width. Such a circuit is sometimes used to delay a
and C =0.1 μ F. Find V (t) and sketch it.
pulse so that something else can happen first. In designing
circuits you try not too often to rely on tricks like this, but
CMOS buffers
they’re occasionally handy.
A 15k B C
E. Another circuit example: “One Minute of Power”
Figure 1.38 shows another example of what can be done
‘1G17 1000pF ‘1G17 with simple RC timing circuits. The triangular symbol is a
comparator, something we’ll treat in detail later, in Chap-
ters 4 and 10; all you need to know, for now, is that (a) it is
an IC (containing a bunch of resistors and transistors), (b)
it is powered from some positive dc voltage that you con-
nect to the pin labeled “V+ ,” and (c) it drives its output (the
A: input wire sticking out to the right) either to V+ or to ground, de-
pending on whether the input labeled “+” is more or less
positive than the input labeled “−,” respectively. (These are
called the non-inverting and the inverting inputs, respec-
B: RC
tively.) It doesn’t draw any current from its inputs, but it
happily drives loads that require up to 20 mA or so. And a
comparator is decisive: its output is either “HIGH” (at V+ )
C: output or “LOW” (ground).
Here’s how the circuit works: the voltage divider R3 R4
holds the (−) input at 37% of the supply voltage, in this
10μs 10μs
case about +1.8 V; let’s call that the “reference voltage.”
Figure 1.37. Producing a delayed digital waveform with the help of
an RC and a pair of LVC-family logic buffers (tiny parts with a huge 24 Complementary metal-oxide semiconductor, the dominant form of dig-
part number: SN74LVC1G17DCKR!). ital logic, as we’ll see from Chapter 10 onward.
24 1.4. Capacitors and ac circuits Art of Electronics Third Edition

+5V that can activate pretty much any load you care to switch
on and off. The use of a relay has the important property
R3
620k
that the load – the circuit being switched by the relay – is
R1 comparator electrically isolated from the +5 V and ground of the timing
1k TLC 3702
–
V+ circuit itself.
+ 5V
0 300 LED
R2 R4
C1 + 6.2M 1 minute Panasonic
360k
10μF LN21RUQ
A.

Figure 1.38. RC timing circuit: one push → one minute!.

piezo buzzer
(w. drive ckt)
CUI CEP-2242 etc.
switch
closure
B.
+5

VC1 +1.8V (Vref)

relay
0 COTO 8L01-05
5V, 10mA coil
C. (use Vcc=6V)
comp
output
1 min solid-state relay
270 Crydom D2450 air
Figure 1.39. Producing a delayed digital waveform for the circuit compressor
of Figure 1.38. The voltage VC1 has a rise time of R1C1 ≈10 ms.

So if the circuit has been sitting there for a while, C1 is

fully discharged, and the comparator’s output is at ground. 110/220 vac
D. from wall plug
When you push the START button momentarily, C1 charges
quickly (10 ms time constant) to +5 V, which makes the Figure 1.40. Driving interesting stuff from the output of the timer
comparator’s output switch to +5 V; see Figure 1.39. Af- circuit in Figure 1.38.
ter the button is released, the capacitor discharges expo-
nentially toward ground, with a time constant of τ = R2C1 , Finally, for turning serious industrial machinery on and
which we’ve set to be 1 minute. At that time its voltage off, you would probably use a hefty solid-state relay (SSR,
crosses the reference voltage, so the comparator’s output §12.7), which has within it an infrared LED coupled to
switches rapidly back to ground. (Note that we’ve conve- an ac switching device known as a triac. When activated,
niently chosen the reference voltage to be a fraction 1/e of the triac acts as an excellent mechanical switch, capable
V+ , so it takes exactly one time constant τ for that to hap- of switching many amperes, and (like the electromechani-
pen. For R2 we used the closest standard value to 6 MΩ; cal relay) is fully isolated electrically from its input circuit.
see Appendix C.) The bottom line is that the output spends The example shows this thing hooked to an air compressor,
1 minute at +5 V, after the button is pushed. so your friends will get a minute’s worth of air to inflate
We’ll add a few details shortly, but first let’s use the their tires at your home “gas station” (literally!) after they
output to do some interesting things, which are shown in drop a quarter into your coin-initiated timer. You could do
Figures 1.40A–D. You can make a self-stopping flashlight an analogous thing with a coin-operated hot shower (but,
key fob by connecting its output to an LED; you need to hey, we get only one minute?!).
put a resistor in series, to set the current (we’ll say much Some details: (a) in the circuit of Figure 1.38 you could
more about this later). If you prefer to make some noise, omit R1 and the circuit would still work, but there would
you could connect a piezo beeper to beep continuously (or be a large transient current when the discharged capaci-
intermittently) for a minute (this might be an end-of-cycle tor was initially connected across the +5 V supply (recall
signal for a clothes dryer). Another possibility is to attach a I = C dV/dt: here you would be trying to produce 5 V of
small electromechanical relay, which is just an electrically “dV ” in roughly 0 s of “dt”). By adding a series resistor R1
operated mechanical switch, to provide a pair of contacts you limit the peak current to a modest 5 mA while charging
Art of Electronics Third Edition 1.4.3. Differentiators 25

the capacitor fast enough (> 99% in 5 RC time constants, could somehow use the current through C as our “output”!
or 0.05 s). (b) The comparator output would likely bounce But we can’t.25
around a bit (see Figure 4.31), just as the (+) input crosses 2. So we add a small resistor from the low side of the
the reference voltage in its leisurely exponential prome- capacitor to ground, to act as a “current-sensing” resistor
nade toward ground, owing to unavoidable bits of electrical (Figure 1.41B). The good news is that we now have an out-
noise. To fix this problem you usually see the circuit ar- put proportional to the current through the capacitor. The
ranged so that some of the output is coupled back to the in- bad news is that the circuit is no longer a perfect mathe-
put in a way that reinforces the switching (this is officially matical differentiator. That’s because the voltage across C
called hysteresis, or positive feedback; we’ll see it in Chap- (whose derivative produces the current we are sensing with
ters 4 and 10). (c) In electronic circuits it’s always a good R) is no longer equal to Vin ; it now equals the difference be-
idea to bypass the dc supply by connecting one or more ca- tween Vin and Vout . Here’s how it goes: the voltage across
pacitors between the dc “rail” and ground. The capacitance C is Vin −Vout , so
is noncritical – values of 0.1 μ F to 10 μ F are commonly d Vout
used; see §1.7.16A. I =C (V −Vout ) = .
dt in R
Our simple examples above all involved turning some If we choose R and C small enough so that dVout /dt 
load on and off. But there are other uses for an electronic dVin /dt, then
logic signal, like the output of the comparator, that is in
one of two possible binary states, called HIGH and LOW (in dVin Vout
C ≈
this case +5 V and ground), 1 and 0, or TRUE and FALSE. dt R
For example, such a signal can enable or disable the op- or
eration of some other circuit. Imagine that the opening of d
Vout (t) ≈ RC V (t).
a car door triggers our 1 min HIGH output, which then en- dt in
ables a keypad to accept a security code so you can start That is, we get an output proportional to the rate of change
the car. After a minute, if you haven’t managed to type the of the input waveform.
magic code, it shuts off, thus ensuring a certain minimum To keep dVout /dt  dVin /dt, we make the product RC
of operator sobriety. small, taking care not to “load” the input by making R too
small (at the transition the change in voltage across the ca-
Vin(t) I (t)
pacitor is zero, so R is the load seen by the input). We will
have a better criterion for this when we look at things in the
C frequency domain (§1.7.10). If you drive this circuit with a
square wave, the output will be as shown in Figure 1.42.
A.

C Vin

Vin (t) Vout (t)

R
Vout

B.
Figure 1.41. Differentiators. A. Perfect (except it has no output ter- Figure 1.42. Output waveform (bottom) from differentiator driven
minal). B. Approximate (but at least it has an output!). by a square wave.

Differentiators are handy for detecting leading edges

and trailing edges in pulse signals, and in digital cir-
1.4.3 Differentiators cuitry you sometimes see things like those depicted in Fig-
You can make a simple circuit that differentiates an input ure 1.43. The RC differentiator generates spikes at the tran-
signal; that is, Vout ∝ dVin /dt. Let’s take it in two steps. sitions of the input signal, and the output buffer converts
1. First look at the (impractical) circuit in Figure 1.41A: 25 Devotees of the cinema will be reminded of Dr. Strangelove’s outburst:
The input voltage Vin (t) produces a current through the ca- “The whole point of a doomsday machine is lost . . . if you keep it
pacitor of Icap = C dVin /dt. That’s just what we want – if we secret!”
26 1.4. Capacitors and ac circuits Art of Electronics Third Edition

A 100pF B C 1pF

50Ω
10k

differentiator
output
10mV/div

A: input
dV/dt = 1V/ns
0.5V/ns
input step 0.25V/ns
time constant
2V/div
= 1μs 4 ns/div
B: RC

Figure 1.44. Three fast step waveforms, differentiated by the RC

network shown. For the fastest waveform (109 volts per second!),
imperfections in the components and measuring instruments cause
deviation from the ideal.
C: output

broken connection somewhere, usually at the scope probe.

Figure 1.43. Leading-edge detector. The very small capacitance of the broken connection com-
bines with the scope input resistance to form a differentia-
the spikes to short square-topped pulses. In practice, the tor. Knowing that you’ve got a differentiated “something”
negative spike will be small because of a diode (a handy can help you find the trouble and eliminate it.
device discussed in §1.6) built into the buffer.
To inject some real-world realism here, we hooked up
and made some measurements on a differentiator that we
configured for high-speed signals. For this we used C=1 pF
and R=50 Ω (the latter is the world-wide standard for high-
speed circuits, see Appendix H), we drove it with a 5 V step
of settable slew rate (i.e., dV/dt). Figure 1.44 shows both
input and output waveforms, for three choices of dVin /dt.
At these speeds (note the horizontal scale: 4 nanoseconds
per division!) circuits often depart from ideal performance,
as can be seen in the fastest risetime. The two slower steps
show reasonable behavior; that is, a flat-top output wave-
Figure 1.45. Two examples of unintentional capacitive coupling.
form during the input’s upward ramp; check for yourself
that the output amplitude is correctly predicted by the for-
mula.
1.4.4 Integrators
A. Unintentional capacitive coupling If RC circuits can take derivatives, why not integrals? As
Differentiators sometimes crop up unexpectedly, in situa- before, let’s take it in two steps.
tions where they’re not welcome. You may see signals like 1. Imagine that we have an input signal that is a time-
those shown in Figure 1.45. The first case is caused by a varying current versus time, Iin (t) (Figure 1.46A).26 That
square wave somewhere in the circuit coupling capacitively input current is precisely the current through the
to the signal line you’re looking at; that might indicate a  capaci-
tor, so Iin (t) = C dV (t)/dt, and therefore V (t) = Iin (t) dt.
missing resistor termination on your signal line. If not, you
must either reduce the source resistance of the signal line 26 We’re used to thinking of signals as time-varying voltages; but we’ll
or find a way to reduce capacitive coupling from the of- see how we can convert such signals to proportional time-varying cur-
fending square wave. The second case is typical of what rents, by using “voltage-to-current converters” (with the fancier name
you might see when you look at a square wave, but have a “transconductance amplifiers”).
Art of Electronics Third Edition 1.4.4. Integrators 27

Iin (t) Vout (t) V Vin

C
Vout
A.
t

10% error
R
Vin (t) Vout (t) V at about
Vout 10% of Vin
C (straight)
B. t

Figure 1.46. Integrator. A. Perfect (but requires a current input sig-

Figure 1.47. Integrator approximation is good when Vout  Vin .
nal). B. Approximate (see text).

That’s just what we wanted! Thus a simple capacitor, with circuit. A large voltage across a large resistance approxi-
one side grounded, is an integrator, if we have an input mates a current source and, in fact, is frequently used as
signal in the form of a current Iin (t). Most of the time we one.
don’t, though. Later, when we get to operational amplifiers and feed-
2. So we connect a resistor in series with the more usual back, we will be able to build integrators without the re-
input voltage signal Vin (t), to convert it to a current (Fig- striction Vout  Vin . They will work over large frequency
ure 1.46B). The good news is that it works, sort of. The and voltage ranges with negligible error.
bad news is that the circuit is no longer a perfect integrator. The integrator is used extensively in analog computa-
That’s because the current through C (whose integral pro- tion. It is a useful subcircuit that finds application in control
duces the output voltage) is no longer proportional to Vin ; systems, feedback, analog–digital conversion, and wave-
it is now proportional to the difference between Vin and V . form generation.
Here’s how it goes: the voltage across R is Vin −V , so
dV V −V A. Ramp generators
I =C = in . At this point it is easy to understand how a ramp generator
dt R
If we manage to keep V  Vin , by keeping the product RC works. This nice circuit is extremely useful, for example in
large,27 then timing circuits, waveform and function generators, analog
oscilloscope sweep circuits, and analog-digital conversion
dV V circuitry. The circuit uses a constant current to charge a
C ≈ in
dt R capacitor (Figure 1.48). From the capacitor equation I =
or C(dV /dt), you get V (t) = (I/C)t. The output waveform is
 t
1 as shown in Figure 1.49. The ramp stops when the current
V (t) = Vin (t) dt + constant.
RC source “runs out of voltage,” i.e., reaches the limit of its
That is, we get an output proportional to the integral over compliance. On the same figure is shown the curve for a
time of the input waveform. You can see how the approxi- simple RC, with the resistor tied to a voltage source equal
mation works for a square-wave input: V (t) is then the ex- to the compliance of the current source, and with R chosen
ponential charging curve we saw earlier (Figure 1.47). The so that the current at zero output voltage is the same as that
first part of the exponential is a ramp, the integral of a con- of the current source. (Real current sources generally have
stant; as we increase the time constant RC, we pick off a output compliances limited by the power-supply voltages
smaller part of the exponential, i.e., a better approximation used in making them, so the comparison is realistic.) In the
to a perfect ramp. next chapter, which deals with transistors, we will design
Note that the condition V  Vin is the same as saying some current sources, with some refinements to follow in
that I is proportional to Vin , which was our first integrator the chapters on operational amplifiers (op-amps) and FETs.
Exciting things to look forward to!
27 Just as with the differentiator, we’ll have another way of framing this Exercise 1.18. A current of 1 mA charges a 1 μ F capacitor. How
criterion in §1.7.10. long does it take the ramp to reach 10 volts?
28 1.5. Inductors and transformers Art of Electronics Third Edition

output where L is called the inductance and is measured in henrys

I
V(t) (or mH, μ H, nH, etc.). Putting a constant voltage across an
C inductor causes the current to rise as a ramp (compare with
a capacitor, in which a constant current causes the voltage
to rise as a ramp); 1 V across 1 H produces a current that
increases at 1 amp per second.
Figure 1.48. A constant-current source charging a capacitor gen-
erates a ramp voltage waveform.

Figure 1.50. Inductors. The parallel-bar symbol represents a core

V of magnetic material.
Vsupply
“IC”
RC Just as with capacitors, the energy invested in ramping
up the current in an inductor is stored internally, here in the
form of magnetic fields. And the analogous formula is
t 1
UL = LI 2 , (1.24)
Figure 1.49. Constant-current charging (with finite compliance) 2
versus RC charging. where UL is in joules (watt seconds) for L in henrys and I
in amperes. As with capacitors, this is an important result,
1.4.5 Not quite perfect. . . one which lies at the core of switching power conversion
(exemplified by those little black “wall-warts” that provide
Real capacitors (the kind you can see, and touch, and pay
power to all manner of consumer electronic gadgets). We’ll
money for) generally behave according to theory; but they
see lots more of this in Chapter 9.
have some additional “features” that can cause problems in
The symbol for an inductor looks like a coil of wire;
some demanding applications. For example, all capacitors
that’s because, in its simplest form, that’s all it is. Its some-
exhibit some series resistance (which may be a function of
what peculiar behavior comes about because inductors are
frequency), and some series inductance (see the next sec-
magnetic devices, in which two things are going on: current
tion), along with some frequency-dependent parallel resis-
flowing through the coil creates a magnetic field aligned
tance. Then there’s a “memory” effect (known as dielectric
along the coil’s axis; and then changes in that field pro-
absorption), which is rarely discussed in polite society: if
duce a voltage (sometimes called “back EMF”) in a way
you charge a capacitor up to some voltage V0 and hold it
that tries to cancel out those changes (an effect known as
there for a while, and then discharge it to 0 V, then when
Lenz’s law). The inductance L of a coil is simply the ratio
you remove the short across its terminals it will tend to
of magnetic flux passing through the coil divided by the
drift back a bit toward V0 .
current through the coil that produces that flux (multiplied
Don’t worry about this stuff, for now. We’ll treat in de-
by an overall constant). Inductance depends on the coil ge-
tail these effects (and other oddities of real-world compo-
ometry (e.g., diameter and length) and the properties of any
nents) in the advanced topics Chapter 1x.
magnetic material (the “core”) that may be used to confine
the magnetic field. That’s all you need to understand why
1.5 Inductors and transformers the inductance of a coil of given geometry is proportional
to the square of the number of turns.
1.5.1 Inductors
Exercise 1.19. Explain why L ∝ n2 for an inductor consisting of
If you understand capacitors, you shouldn’t have great a coil of n turns of wire, maintaining fixed diameter and length as
trouble with inductors (Figure 1.50). They’re closely re- n is varied.
lated to capacitors: the rate of current change in an inductor
is proportional to the voltage applied across it (for a capac- We’ll get into some more detail in the Chapter 1x. But
itor it’s the other way around – the rate of voltage change is it’s worth displaying a semi-empirical formula for the ap-
proportional to the current through it). The defining equa- proximate inductance L of a coil of diameter d and length
tion for an inductor is l, in which the n2 dependence is on display:
dI d 2 n2
V =L , (1.23) L≈K μ H,
dt 18d + 40 l
Art of Electronics Third Edition 1.5.1. Inductors 29

Figure 1.51. Inductors. Top row, left to right: encapsulated toroid, hermetically-sealed toroid, board-mount pot core, bare toroid (two
sizes). Middle row: slug-tuned ferrite-core inductors (three sizes). Bottom row: high-current ferrite-core choke, ferrite-bead choke, dipped
radial-lead ferrite-core inductor, surface-mount ferrite chokes, molded axial-lead ferrite-core chokes (two styles), lacquered ferrite-core
inductors (two styles).

where K = 1.0 or 0.395 for dimensions in inches or cen- made in molds). See Figure 1.51 for some typical geome-
timeters, respectively. This is known as Wheeler’s formula tries.
and is accurate to 1% as long as l > 0.4d. Inductors find heavy use in radiofrequency (RF) cir-
As with capacitive current, inductive current is not sim- cuits, serving as RF “chokes” and as parts of tuned cir-
ply proportional to voltage (as in a resistor). Furthermore, cuits (§1.7.14). A pair of closely coupled inductors forms
unlike the situation in a resistor, the power associated with the interesting object known as a transformer. We will talk
inductive current (V ×I) is not turned into heat, but is stored briefly about them shortly.
as energy in the inductor’s magnetic field (recall that for a An inductor is, in a real sense, the opposite of a capac-
capacitor the power associated with capacitive current is itor.28 You will see how that works out later in the chapter
likewise not dissipated as heat, but is stored as energy in when we deal with the important subject of impedance.
the capacitor’s electric field). You get all that energy back
when you interrupt the inductor’s current (with a capacitor A. A look ahead: some magic with inductors
you get all the energy back when you discharge the voltage Just to give a taste of some of the tricks that you can do
to zero). with inductors, take a look at Figure 1.52. Although we’ll
The basic inductor is a coil, which may be just a loop understand these circuits a lot better when we go at them
with one or more turns of wire; or it may be a coil with in Chapter 9, it’s possible to see what’s going on with what
some length, known as a solenoid. Variations include coils we know already. In Figure 1.52A the left-hand side of in-
wound on various core materials, the most popular being ductor L is alternately switched between a dc input voltage
iron (or iron alloys, laminations, or powder) and ferrite (a Vin and ground, at some rapid rate, spending equal times
gray, nonconductive, brittle magnetic material). These are 28 In practice, however, capacitors are much more widely used in elec-
all ploys to multiply the inductance of a given coil by the
tronic circuits. That is because practical inductors depart significantly
“permeability” of the core material. The core may be in the from ideal performance – by having winding resistance, core losses,
shape of a rod, a toroid (doughnut), or even more bizarre and self-capacitance – whereas practical capacitors are nearly perfect
shapes, such as a “pot core” (which has to be seen to be (more on this in Chapter 1x). Inductors are indispensable, however, in
understood; the best description we can come up with is a switching power converters, as well as in tuned LC circuits for RF ap-
doughnut mold split horizontally in half, if doughnuts were plications.
30 1.5. Inductors and transformers Art of Electronics Third Edition

connected to each (a “50% duty cycle”). But the defin-

ing equation V =L dI/dt requires that the average voltage
across an inductor must be zero, otherwise the magnitude
of its average current is rising without limit. (This is some-
times called the volt-second balance rule.) From this it fol- Figure 1.53. Transformer.
lows that the average output voltage is half the input volt-
age (make sure you understand why). In this circuit C2
Transformers are quite efficient (output power is very
acts as a storage capacitor for steadying the output voltage
nearly equal to input power); thus, a step-up transformer
(more on this later, and in Chapter 9).
gives higher voltage at lower current. Jumping ahead for
Producing an output that is half the voltage of an input
a moment, a transformer of turns ratio n increases the
is not very exciting; after all, a simple voltage divider can
impedance by n2 . There is very little primary current if the
do that. But, unlike a voltage divider, this circuit does not
secondary is unloaded.
waste any energy; apart from non-idealities of the compo-
Power transformers (meant for use from the 115 V pow-
nents, it is 100% efficient. And in fact this circuit is widely
erline) serve two important functions in electronic instru-
used in power conversion; it’s called a “synchronous buck
ments: they change the ac line voltage to a useful (usually
converter.”
lower) value that can be used by the circuit, and they “iso-
But look now at Figure 1.52B, which is just a turned-
late” the electronic device from actual connection to the
around version of Figure 1.52A. This time, volt-second
powerline, because the windings of a transformer are elec-
balance requires that the output voltage be twice the in-
trically insulated from each other. They come in an enor-
put voltage. You can’t do that with a voltage divider! Once
mous variety of secondary voltages and currents: outputs
again, the output capacitor (C1 this time) serves to hold the
as low as 1 volt or so up to several thousand volts, current
output voltage steady by storing charge. This configuration
ratings from a few milliamps to hundreds of amps. Typical
is called a “synchronous boost converter.”
transformers for use in electronic instruments might have
These and other switching converters are discussed ex-
secondary voltages from 10 to 50 volts, with current rat-
tensively in Chapter 9, where Table 9.5 lists some fifty rep-
ings of 0.1 to 5 amps or so. A related class of transformers
resentative types.
is used in electronic power conversion, in which plenty of
power is flowing, but typically as pulse or square wave-
+ L + forms, and at much higher frequencies (50 kHz to 1 MHz
Vin C1
– C2 Vout = 1 Vin is typical).
– 2
A. Transformers for signals at audio frequencies and ra-
50% dio frequencies are also available. At radio frequencies you
+
sometimes use tuned transformers if only a narrow range
L
+
of frequencies is present. There is also an interesting class
C1 Vout = 2Vin
Vin C2 of transmission-line transformers. In general, transformers
– – for use at high frequencies must use special core materi-
B.
50% als or construction to minimize core losses, whereas low-
frequency transformers (e.g., ac powerline transformers)
Figure 1.52. Inductors let you do neat tricks, such as increasing a
dc input voltage.
are burdened instead by large and heavy cores. The two
kinds of transformers are in general not interchangeable.

1.5.2 Transformers
A. Problems, problems. . .
A transformer is a device consisting of two closely coupled This simple “first-look” description ignores interesting –
coils (called primary and secondary). An ac voltage applied and important – issues. For example, there are inductances
to the primary appears across the secondary, with a voltage associated with the transformer, as suggested by its cir-
multiplication proportional to the turns ratio of the trans- cuit symbol: an effective parallel inductance (called the
former, and with a current multiplication inversely propor- magnetizing inductance) and an effective series inductance
tional to the turns ratio. Power is conserved. Figure 1.53 (called the leakage inductance). Magnetizing inductance
shows the circuit symbol for a laminated-core transformer causes a primary current even with no secondary load;
(the kind used for 60 Hz ac power conversion). more significantly, it means that you cannot make a “dc
Art of Electronics Third Edition 1.6.2. Rectification 31

transformer.” And leakage inductance causes a voltage which is measured in the nanoamp range for a general-
drop that depends on load current, as well as bedeviling cir- purpose diode (note the hugely different scales in the graph
cuits that have fast pulses or edges. Other departures from for forward and reverse current), is almost never of any
ideal performance include winding resistance, core losses, consequence until you reach the reverse breakdown volt-
capacitance, and magnetic coupling to the outside world. age (also called the peak inverse voltage, PIV), typically
Unlike capacitors (which behave nearly ideally in most cir- 75 volts for a general-purpose diode like the 1N4148. (Nor-
cuit applications), the deficiencies of inductors have signif- mally you never subject a diode to voltages large enough to
icant effects in real-world circuit applications. We’ll deal cause reverse breakdown; the exception is the zener diode
with these in Chapter 1x and Chapter 9. we mentioned earlier.) Frequently, also, the forward volt-
age drop of about 0.5 to 0.8 V is of little concern, and the
diode can be treated as a good approximation to an ideal
1.6 Diodes and diode circuits
one-way conductor. There are other important characteris-
We are not done with capacitors and inductors! We have tics that distinguish the thousands of diode types available,
dealt with them in the time domain (RC circuits, exponen- e.g., maximum forward current, capacitance, leakage cur-
tial charge and discharge, differentiators and integrators, rent, and reverse recovery time; Table 1.1 includes a few
and so on), but we have not yet tackled their behavior in popular diodes, to give a sense of the capabilities of these
the frequency domain. little devices.
We will get to that soon enough. But this is a good time
to take a break from “RLC” and put our knowledge to use 20mA
with some clever and useful circuits. We begin by intro-
ducing a new device, the diode. It’s our first example of a
10mA “FORWARD”
nonlinear device, and you can do nifty things with it.
–100V –50V
1.6.1 Diodes
1V 2V
“REVERSE”
The circuit elements we’ve discussed so far (resistors, ca-
pacitors, and inductors) are all linear, meaning that a dou- 0.1μA
bling of the applied signal (a voltage, say) produces a dou- Note scale change!
bling of the response (a current, say). This is true even for
the reactive devices (capacitors and inductors). These com- 0.2μA
ponents are also passive, as opposed to active devices, the
Figure 1.55. Diode V –I curve.
latter exemplified by transistors, which are semiconductor
devices that control the flow of power. And they are all
two-terminal devices, which is self-explanatory. Before jumping into some circuits with diodes, we
should point out two things: (a) a diode doesn’t have a re-
anode cathode sistance (it doesn’t obey Ohm’s law). (b) If you put some
A K diodes in a circuit, it won’t have a Thévenin equivalent.
Figure 1.54. Diode.

1.6.2 Rectification
The diode (Figure 1.54) is an important and useful two-
terminal passive nonlinear device. It has the V –I curve A rectifier changes ac to dc; this is one of the simplest and
shown in Figure 1.55. (In keeping with the general phi- most important applications of diodes (which are some-
losophy of this book, we will not attempt to describe the times called rectifiers). The simplest circuit is shown in
solid-state physics that makes such devices possible.) Figure 1.56. The “ac” symbol represents a source of ac
The diode’s arrow (the anode terminal) points in the di- voltage; in electronic circuits it is usually provided by a
rection of forward current flow. For example, if the diode transformer, powered from the ac powerline. For a sine-
is in a circuit in which a current of 10 mA is flowing from wave input that is much larger than the forward drop
anode to cathode, then (from the graph) the anode is ap- (about 0.6 V for silicon diodes, the usual type), the out-
proximately 0.6 V more positive than the cathode; this is put will look like that in Figure 1.57. If you think of the
called the “forward voltage drop.” The reverse current, diode as a one-way conductor, you won’t have any trouble
32 1.6. Diodes and diode circuits Art of Electronics Third Edition

Table 1.1 Representative Diodes

V R (max) IR (typ, 25°C) V F @ IF Capacitance
Part # (V) (A @ V) (mV) (mA) (pF @ V R) SMT a p/n Comments
Silicon
PAD5 45 0.25pA 20V 800 1 0.5pF 5V SSTPAD5 metal + glass can
1N4148 75 10nA 20V 750 10 0.9pF 0V 1N4148W jellybean sig diode
1N4007 1000 50nA 800V 800 250 12pF 10V DL4007 1N4004 lower V
1N5406 600 <10μA 600V 1.0V 10A 18pF 10V none heat through leads
Schottky b
1N6263 60 7nA 20V 400 1 0.6pF 10V 1N6263W see also 1N5711
1N5819 40 10μA 32V 400 1000 150pF 1V 1N5819HW jellybean
1N5822 40 40μA 32V 480 3000 450pF 1V none power Schottky
MBRP40045 45 500μA 40V 540 400A 3500pF 10V you jest! Moby dual Schottky

Notes: (a) SMT, surface-mount technology. (b) Schottky diodes have lower forward voltage and zero
reverse-recovery time, but more capacitance.

understanding how the circuit works. This circuit is called which the diode drop becomes significant, you have to re-
a half-wave rectifier, because only half of the input wave- member that.29
form is used.

V load

ac Rload

Figure 1.59. Full-wave output voltage (unfiltered).

Figure 1.56. Half-wave rectifier.

1.6.3 Power-supply filtering

The preceding rectified waveforms aren’t good for much as
V load
they stand. They’re “dc” only in the sense that they don’t
change polarity. But they still have a lot of “ripple” (peri-
odic variations in voltage about the steady value) that has
Figure 1.57. Half-wave output voltage (unfiltered). to be smoothed out in order to generate genuine dc. This
we do by attaching a relatively large value capacitor (Fig-
ure 1.60); it charges up to the peak output voltage during
the diode conduction, and its stored charge (Q = CV ) pro-
vides the output current in between charging cycles. Note
that the diodes prevent the capacitor from discharging back
through the ac source. In this application you should think
Rload
of the capacitor as an energy storage device, with stored
energy U = 12 CV 2 (recall §1.4.1; for C in farads and V in
Figure 1.58. Full-wave bridge rectifier. volts, U comes out in joules, or equivalently, watt seconds).
The capacitor value is chosen so that
Figure 1.58 shows another rectifier circuit, a “full-wave RloadC  1/ f ,
bridge.” Figure 1.59 shows the voltage across the load;
note that the entire input waveform is used. The gaps at 29 The diode drop can be eliminated with active switching (or syn-
zero voltage occur because of the diodes’ forward voltage chronous switching, a technique in which the diodes are replaced by
drop. In this circuit, two diodes are always in series with transistor switches, actuated in synchronism with the input ac wave-
the input; when you design low-voltage power supplies, for form (see §9.5.3B).
Art of Electronics Third Edition 1.6.4. Rectifier configurations for power supplies 33

of accuracy, though, for two reasons. (a) The discharge is

+
an exponential only if the load is a resistance; many loads
full-wave + are not. In fact, the most common load, a voltage regula-
ac
bridge C load
tor, looks like a constant-current load. (b) Power supplies
– are built with capacitors with typical tolerances of 20%
or more. Realizing the manufacturing spread, you design
Figure 1.60. Full-wave bridge with output storage (“filter”) capaci- conservatively, allowing for the worst-case combination of
tor. component values.
In this case, viewing the initial part of the discharge as
(where f is the ripple frequency, here 120 Hz) in order to a ramp is in fact quite accurate, especially if the ripple is
ensure small ripple by making the time constant for dis- small, and in any case it errs in the direction of conservative
charge much longer than the time between recharging. We design – it overestimates the ripple.
make this vague statement clearer now.
Exercise 1.20. Design a full-wave bridge rectifier circuit to de-
liver 10 V dc with less than 0.1 V (pp) ripple into a load drawing
A. Calculation of ripple voltage
up to 10 mA. Choose the appropriate ac input voltage, assuming
It is easy to calculate the approximate ripple voltage, par- 0.6 V diode drops. Be sure to use the correct ripple frequency in
ticularly if it is small compared with the dc (see Fig- your calculation.
ure 1.61). The load causes the capacitor to discharge some-
what between cycles (or half-cycles, for full-wave rectifi- Vdc ≈ √2 Vsec (rms)
cation). If you assume that the load current stays constant
(it will, for small ripple), you have Vdc
 
I dV +
ΔV = Δt from I = C . (1.25)
C dt
Just use 1/ f (or 1/2 f for full-wave rectification) for Δt (this
estimate is a bit on the safe side, because the capacitor be- Figure 1.62. Bridge rectifier circuit. The polarity marking and
gins charging again in less than a half-cycle). You get30 curved electrode indicate a polarized capacitor, which must not be
allowed to charge with the opposite polarity.
Iload
ΔV = (half-wave),
fC
Iload
ΔV = (full-wave).
2 fC 1.6.4 Rectifier configurations for power supplies
A. Full-wave bridge
A dc power supply with the bridge circuit we just discussed
peak-to-peak ripple looks as shown in Figure 1.62. In practice, you generally
V load
output from filter, buy the bridge as a prepackaged module. The smallest ones
under load come with maximum current ratings of 1 A average, with
a selection of rated minimum breakdown voltages going
output with
no capacitor from 100 V to 600 V, or even 1000 V. Giant bridge rectifiers
are available with current ratings of 25 A or more.
t

Figure 1.61. Power-supply ripple calculation. B. Center-tapped full-wave rectifier

The circuit in Figure 1.63 is called a center-tapped full-
If you wanted to do the calculation without any approxi- wave rectifier. The output voltage is half what you get if
mation, you would use the exact exponential discharge for- you use a bridge rectifier. It is not the most efficient cir-
mula. You would be misguided in insisting on that kind cuit in terms of transformer design, because each half of
the secondary is used only half the time. To develop some
30 While teaching electronics we’ve noticed that students love to memo- intuition on this subtle point, consider two different con-
rize these equations! An informal poll of the authors showed that two figurations that produce the same rectified dc output volt-
out of two engineers don’t memorize them. Please don’t waste brain age: (a) the circuit of Figure 1.63, and (b) the same trans-
cells that way – instead, learn how to derive them. former, this time with its secondary cut at the center tap and
34 1.6. Diodes and diode circuits Art of Electronics Third Edition

rewired with the two halves in parallel, the resultant com- I

bined secondary winding connected to a full-wave bridge.
2.0 A
Now, to deliver the same output power, each half winding
in (a), during its conduction cycle, must supply the same
0
current as the parallel pair in (b). But the power dissipated ~1ms
in the winding resistances goes like I 2 R, so the power lost
to heating in the transformer secondary windings reduced Figure 1.64. Illustrating greater I 2 R heating with discontinuous
by a factor of 2 for the bridge configuration (b). current flow.
Here’s another way to see the problem: imagine we use
the same transformer as in (a), but for our comparison cir- +V

cuit we replace the pair of diodes with a bridge, as in Fig- +

ure 1.62, and we leave the center tap unconnected. Now, ground
ac
to deliver the same output power, the current through the in
winding during that time is twice what it would be for a true +
full-wave circuit. To expand on this subtle point: heating in –V
the windings, calculated from Ohm’s law, is I 2 R, so you
have four times the heating for half the time, or twice the Figure 1.65. Dual-polarity (split) supply.
average heating of an equivalent full-wave bridge circuit.
You would have to choose a transformer with a current rat- +V
ing 1.4 (square root of 2) times as large compared with the +
(better) bridge circuit; besides costing more, the resulting
supply would be bulkier and heavier. load
+
Exercise 1.21. This illustration of I 2 R heating may help you un-
derstand the disadvantage of the center-tapped rectifier circuit. com
What fuse rating (minimum) is required for passing the current
waveform shown in Figure 1.64, which has 1 amp average cur- Figure 1.66. Voltage doubler.
rent? Hint: a fuse “blows out” by melting a metallic link (I 2 R
heating), for steady currents larger than its rating. Assume for D. Voltage multipliers
this problem that the thermal time constant of the fusible link The circuit shown in Figure 1.66 is called a voltage dou-
is much longer than the time scale of the square wave, i.e., bler. Think of it as two half-wave rectifier circuits in se-
that the fuse responds to the value of I 2 averaged over many ries. It is officially a full-wave rectifier circuit because both
cycles.
halves of the input waveform are used – the ripple fre-
quency is twice the ac frequency (120 Hz for the 60 Hz line
C. Split supply voltage in the United States).
A popular variation of the center-tapped full-wave circuit Variations of this circuit exist for voltage triplers,
is shown in Figure 1.65. It gives you split supplies (equal quadruplers, etc. Figure 1.67 shows doubler, tripler, and
plus and minus voltages), which many circuits need. It is an quadrupler circuits that let you ground one side of the trans-
efficient circuit, because both halves of the input waveform former. You can extend this scheme as far as you want, pro-
are used in each winding section. ducing what’s called a Cockcroft–Walton generator; these
are used in arcane applications (such as particle accelera-
tors) and in everyday applications (such as image intensi-
fiers, air ionizers, laser copiers, and even bug zappers) that
require a high dc voltage but hardly any current.
+
ac
in 1.6.5 Regulators
By choosing capacitors that are sufficiently large, you can
reduce the ripple voltage to any desired level. This brute-
Figure 1.63. Full-wave rectifier using center-tapped transformer. force approach has three disadvantages.
Art of Electronics Third Edition 1.6.6. Circuit applications of diodes 35

+ regulator

ac
in + in out
+ ground

load
Figure 1.68. Regulated dc power supply.

A. doubler
10% of the dc voltage), then use an active feedback circuit
to eliminate the remaining ripple. Such a feedback circuit
+
“looks at” the output, making changes in a controllable se-
ac ries resistor (a transistor) as necessary to keep the output
in +
voltage constant (Figure 1.68). This is known as a “linear

load
regulated dc power supply.”32
+ These voltage regulators are used almost universally as
power supplies for electronic circuits. Nowadays complete
voltage regulators are available as inexpensive ICs (priced
B. tripler under $1). A power supply built with a voltage regulator
can be made easily adjustable and self-protecting (against
short circuits, overheating, etc.), with excellent properties
+
as a voltage source (e.g., internal resistance measured in
ac
+
milliohms). We will deal with regulated dc power supplies
in +
in Chapter 9.
load

+
1.6.6 Circuit applications of diodes
A. Signal rectifier
C. quadrupler
There are other occasions when you use a diode to make
Figure 1.67. Voltage multipliers; these configurations don’t require
a waveform of one polarity only. If the input waveform
a floating voltage source. isn’t a sinewave, you usually don’t think of it as a recti-
fication in the sense of a power supply. For instance, you
• The required capacitors may be prohibitively bulky and might want a train of pulses corresponding to the rising
expensive. edge of a square wave. The easiest way is to rectify the dif-
• The very short interval of current flow during each cy- ferentiated wave (Figure 1.69). Always keep in mind the
cle31 (only very near the top of the sinusoidal waveform) 0.6 V(approximately) forward drop of the diode. This cir-
produces more I 2 R heating. cuit, for instance, gives no output for square waves smaller
• Even with the ripple reduced to negligible levels, you still than 0.6 V pp. If this is a problem, there are various tricks
have variations of output voltage that are due to other to circumvent this limitation. One possibility is to use hot
causes, e.g., the dc output voltage will be roughly pro- carrier diodes (Schottky diodes), with a forward drop of
portional to the ac input voltage, giving rise to fluctua- about 0.25 V.
tions caused by input line voltage variations. In addition, A possible circuit solution to this problem of finite diode
changes in load current will still cause the output volt- drop is shown in Figure 1.70. Here D1 compensates D2 ’s
age to change because of the finite internal resistances of forward drop by providing 0.6 V of bias to hold D2 at the
the transformer, diode, etc. In other words, the Thévenin threshold of conduction. Using a diode (D1 ) to provide
equivalent circuit of the dc power supply has R > 0. the bias (rather than, say, a voltage divider) has several

A better approach to power-supply design is to use 32 A popular variant is the regulated switching power converter. Although
enough capacitance to reduce ripple to low levels (perhaps its operation is quite different in detail, it uses the same feedback prin-
ciple to maintain a constant output voltage. See Chapter 9 for much
31 Called the conduction angle. more on both techniques.
36 1.6. Diodes and diode circuits Art of Electronics Third Edition

+5V (when turned on)

C
D2
D1
R1 R2
+3V real-time clock
NXP PCF8563,
lithium Seiko S-35390A, etc
X2 V+
“coin”
Figure 1.69. Signal rectifier applied to differentiator output.
cell digital
SDA
link to
X1 G SCL microprocessor
C
100 pF D2
in out

+5V Figure 1.71. Diode OR gate: battery backup. The real-time clock
R1 R2 chips are specified to operate properly with supply voltages from
1.0k R3 1.0k +1.8 V to +5.5 V. They draw a paltry 0.25 μ A, which calculates to a
1.0k 1-million-hour life (a hundred years) from a standard CR2032 coin
+0.6V cell!
D1
+5V

Figure 1.70. Compensating the forward voltage drop of a diode 1N4148

1.0k
signal rectifier. in out

Figure 1.72. Diode voltage clamp.

advantages: (a) there is nothing to adjust, (b) the compen-
sation will be nearly perfect, and (c) changes of the for-
ward drop (e.g., with changing temperature) will be com- +15V

pensated properly. Later we will see other instances of

matched-pair compensation of forward drops in diodes, 2.0k
transistors, and FETs. It is a simple and powerful trick.
R
B. Diode gates signal out
in
Another application of diodes, which we will recognize
later under the general heading of logic, is to pass the 1.0k
higher of two voltages without affecting the lower. A good
example is battery backup, a method of keeping some-
thing running (e.g, the “real-time clock” chip in a com- Figure 1.73. Voltage divider providing clamping voltage.
puter, which keeps a running count of date and time) that
must continue running even when the device is switched clamping action; however, a side effect is that it adds 1 kΩ
off. Figure 1.71 shows a circuit that does the job. The bat- of series resistance (in the Thévenin sense) to the signal,
tery does nothing until the +5 V power is switched off; then so its value is a compromise between maintaining a desir-
it takes over without interruption. able low source (Thévenin) resistance and a desirable low
clamping current. Diode clamps are standard equipment
C. Diode clamps on all inputs in contemporary CMOS digital logic. With-
Sometimes it is desirable to limit the range of a signal (i.e., out them, the delicate input circuits are easily destroyed by
prevent it from exceeding certain voltage limits) some- static electricity discharges during handling.
where in a circuit. The circuit shown in Figure 1.72 will ac-
Exercise 1.22. Design a symmetrical clamp, i.e., one that confines
complish this. The diode prevents the output from exceed-
a signal to the range −5.6 to +5.6 V.
ing about +5.6 V, with no effect on voltages less than that
(including negative voltages); the only limitation is that A voltage divider can provide the reference voltage for
the input must not go so negative that the reverse break- a clamp (Figure 1.73). In this case you must ensure that the
down voltage of the diode is exceeded (e.g., −75 V for a resistance looking into the voltage divider (Rvd ) is small
1N4148). The series resistor limits the diode current during compared with R because what you have looks as shown
Art of Electronics Third Edition 1.6.6. Circuit applications of diodes 37

R VD 0 C
667Ω
+5V

signal R
out
in
Figure 1.77. dc restoration.
Figure 1.74. Clamping to voltage divider: equivalent circuit.

R
1.0k in out
in out

667Ω

+5.6V
Figure 1.78. Diode limiter.
Figure 1.75. Poor clamping: voltage divider not stiff enough.

the frequency domain, a subject we’ll take up shortly. For

now we’ll simply say that you can put a capacitor across
Vclamp the 1k resistor, and its stored charge acts to maintain that
point at constant voltage. For example, a 15 μ F capacitor to
ground would make the divider look as if it had a Thévenin
resistance of less than 10 Ω for frequencies above 1 kHz.
(You could similarly add a bypass capacitor across D1 in
Figure 1.70.) As we’ll learn, the effectiveness of this trick
Figure 1.76. Clamping waveform for circuit of Figure 1.73. decreases at low frequencies, and it does nothing at dc.
One interesting clamp application is “dc restoration” of
a signal that has been ac coupled (capacitively coupled).
in Figure 1.74 when the voltage divider is replaced with its
Figure 1.77 shows the idea. This is particularly important
Thévenin equivalent circuit. When the diode conducts (in-
for circuits whose inputs look like diodes (e.g., a transistor
put voltage exceeds clamp voltage), the output is really just
with grounded emitter, as we’ll see in the next chapter);
the output of a voltage divider, with the Thévenin equiva-
otherwise an ac-coupled signal will just fade away, as the
lent resistance of the voltage reference as the lower resistor
coupling capacitor charges up to the signal’s peak voltage.
(Figure 1.75). So, for the values shown, the output of the
clamp for a triangle-wave input would look as shown in
Figure 1.76. The problem is that the voltage divider doesn’t D. Limiter
provide a stiff reference, in the language of electronics. A One last clamp circuit is shown in Figure 1.78. This cir-
stiff voltage source is one that doesn’t bend easily, i.e., it cuit limits the output “swing” (again, a common electronics
has low internal (Thévenin) resistance. term) to one diode drop in either polarity, roughly ±0.6 V.
In practice, the problem of finite impedance of the That might seem awfully small, but if the next stage is an
voltage-divider reference can be easily solved by use of a amplifier with large voltage amplification, its input will al-
transistor or an op-amp. This is usually a better solution ways be near 0 V; otherwise the output is in “saturation”
than using very small resistor values, because it doesn’t (e.g., if the next stage has a gain of 1000 and operates from
consume large currents, yet it provides a voltage reference ±15 V supplies, its input must stay in the range ±15 mV
with a Thévenin resistance of a few ohms or less. Further- in order for its output not to saturate). Figure 1.79 shows
more, there are other ways to construct a clamp, using an what a limiter does to oversize sinewaves and spikes. This
op-amp as part of the clamp circuit. You will see these clamp circuit is often used as input protection for a high-
methods in Chapter 4. gain amplifier.
Alternatively, a simple way to stiffen the clamp circuit
of Figure 1.73, for time-varying signals only, is to add a so- E. Diodes as nonlinear elements
called bypass capacitor across the lower (1 kΩ) resistor. To To a good approximation the forward current through a
understand this fully we need to know about capacitors in diode is proportional to an exponential function of the
38 1.6. Diodes and diode circuits Art of Electronics Third Edition

V = A sin ωt R
Vin Vout
( 0.6V)

Figure 1.81. Approximate log converter.

A.

R B
Vin Vout ∝ log Vin
+0.6V
dV
= ωA D1
dt D2
A
– 0.6 V
–0.6V
B. R1

–V

Figure 1.82. Diode drop compensation in the logarithmic con-

+0.6V verter.

point A at about −0.6 V. Point B is then near ground (mak-

ing Iin accurately proportional to Vin , incidentally). As long
C. –0.6V
as the two (identical) diodes are at the same temperature,
Figure 1.79. A. Limiting large-amplitude sinewaves; B. details; and there is good cancellation of the forward drops, except, of
C. spikes. course, for the difference owing to input current through
D1 , which produces the desired output. In this circuit, R1
should be chosen so that the current through D2 is signif-
icantly larger than the maximum input current in order to
keep D2 in conduction.
Iin
In Chapter 2x we will examine better ways of construct-
Vout ∝ log Iin
ing logarithmic converter circuits, along with careful meth-
ods of temperature compensation. With such methods it is
possible to construct logarithmic converters accurate to a
few percent over six decades or more of input current. A
Figure 1.80. Exploiting the diode’s nonlinear V –I curve: logarith-
better understanding of diode and transistor characteristics,
mic converter.
along with an understanding of op-amps, is necessary first.
This section is meant to serve only as an introduction for
voltage across it at a given temperature (for a discussion
things to come.
of the exact law, see §2.3.1). So you can use a diode to
generate an output voltage proportional to the logarithm
of a current (Figure 1.80). Because V hovers in the region 1.6.7 Inductive loads and diode protection
of 0.6 V, with only small voltage changes that reflect input
current variations, you can generate the input current with What happens if you open a switch that is providing current
a resistor if the input voltage is much larger than a diode to an inductor? Because inductors have the property
drop (Figure 1.81). V = L dI/dt,
In practice, you may want an output voltage that isn’t
offset by the 0.6 V diode drop. In addition, it would be it is not possible to turn off the current suddenly, because
nice to have a circuit that is insensitive to changes in tem- that would imply an infinite voltage across the inductor’s
perature (a silicon diode’s voltage drop decreases approxi- terminals. What happens instead is that the voltage across
mately 2 mV/◦ C). The method of diode drop compensation the inductor rises abruptly and keeps rising until it forces
is helpful here (Figure 1.82). R1 makes D2 conduct, holding current to flow. Electronic devices controlling inductive
Art of Electronics Third Edition 1.6.8. Interlude: inductors as friends 39

+20V

A
100Ω
I
0.05μF
B

BOOM! Figure 1.85. RC “snubber” for suppressing inductive kick.

is proportional to the voltage across it. For applications

Figure 1.83. Inductive “kick.” in which the current must decay quickly (high-speed
actuators or relays, camera shutters, magnet coils, etc.), it
+ may be better to put a resistor across the inductor, choosing
its value so that Vsupply + IR is less than the maximum
allowed voltage across the switch. For the fastest decay
with a given maximum voltage, a zener with series diode
(or other voltage-clamping device) can be used instead,
Ion (at turnoff) giving a linear-like ramp-down of current rather than an
exponential decay (see discussion in Chapter 1x).
For inductors driven from ac (transformers, ac relays),
the diode protection just described will not work, because
Figure 1.84. Blocking inductive kick. the diode will conduct on alternate half-cycles when the
switch is closed. In that case a good solution is an RC
“snubber” network (Figure 1.85). The values shown are
loads can be easily damaged, especially the component
typical for small inductive loads driven from the ac power-
that “breaks down” in order to satisfy the inductor’s crav-
line. Such a snubber should be included in all instruments
ing for continuity of current. Consider the circuit in Fig-
that run from the ac powerline, because the power trans-
ure 1.83. The switch is initially closed, and current is flow-
former is inductive.33
ing through the inductor (which might be a relay, as de-
An alternative to the RC snubber is the use of a bidirec-
scribed later). When the switch is opened, the inductor tries
tional zener-like voltage-clamping element. Among these
to keep current flowing from A to B, as it had been. In other
the most common are the bidirectional “TVS” (transient
words, it tries to make current flow out of B, which it does
voltage suppressor) zener and the metal-oxide varistor
by forcing B to a high positive voltage (relative to A). In a
(“MOV”); the latter is an inexpensive device that looks
case like this, in which there’s no connection to terminal B,
something like a disc ceramic capacitor and behaves elec-
it may go 1000 V positive before the switch contact “blows
trically like a bidirectional zener diode. Both classes are de-
over.” This shortens the life of the switch and also gener-
signed for transient voltage protection, are variously avail-
ates impulsive interference that may affect other circuits
able at voltage ratings from 10 to 1000 volts, and can
nearby. If the switch happens to be a transistor, it would be
handle transient currents up to thousands of amperes (see
an understatement to say that its life is shortened; its life is
Chapter 9x). Including a transient suppressor (with ap-
ended.
propriate fusing) across the ac powerline terminals makes
The best solution usually is to put a diode across the
good sense in a piece of electronic equipment, not only to
inductor, as in Figure 1.84. When the switch is on, the
prevent inductive spike interference to other nearby instru-
diode is back-biased (from the dc drop across the induc-
ments but also to prevent occasional large powerline spikes
tor’s winding resistance). At turn-off the diode goes into
from damaging the instrument itself.
conduction, putting the switch terminal a diode drop above
the positive supply voltage. The diode must be able to han-
dle the initial diode current, which equals the steady current 1.6.8 Interlude: inductors as friends
that had been flowing through the inductor; something like Lest we leave the impression that inductance and induc-
a 1N4004 is fine for nearly all cases. tors are things only to be feared, let’s look at the circuit in
The only disadvantage of this simple protection circuit
is that it lengthens the decay of current through the 33 As explained in §9.5.1, you should choose a capacitor rated for
inductor, because the rate of change of inductor current “across-the-line” service.
40 1.7. Impedance and reactance Art of Electronics Third Edition

Figure 1.86. The goal is to charge up the capacitor from a

relative response (linear scale)

source of dc voltage Vin . In the top circuit (Figure 1.86A) compensating filter
we’ve done it the conventional way, with a series resis- 2
tor to limit the peak current demanded from the voltage speaker
source. OK, it does work – but it has a drawback that can
be serious, namely that half the energy is lost as heat in the
1 lowest
resistor. By contrast, in the circuit with the inductor (Fig- note on
piano highest
ure 1.86B) no energy is lost (assuming ideal components); note on
and, as a bonus, the capacitor gets charged to twice the middle C piano
input voltage. The output-voltage waveform is√a sinusoidal
20Hz 200Hz 2kHz 20kHz
half-cycle at the resonant frequency f = 1/2π LC, a topic
we’ll see soon (§1.7.14).34,35 audible frequencies
infrasonic ultrasonic
frequency (log scale)

R V(t) Vin /R Figure 1.87. Example of frequency analysis: “boom box” loud-
I
speaker equalization. The lowest and highest piano notes, called
+
C
A0 and C8, are at 27.5 Hz and 4.2 kHz; they are four octaves below
Vin
– A440 and four octaves above middle C, respectively.
V Vin

A. t
1.7 Impedance and reactance
L D V(t) tf
I Warning: this section is somewhat mathematical; you may
+ wish to skip over the mathematics, but be sure to pay atten-
Vin C
2Vin tion to the results and graphs.
–
V Circuits with capacitors and inductors are more com-
B. plicated than the resistive circuits we talked about ear-
t lier, in that their behavior depends on frequency: a “volt-
Figure 1.86. Resonant charging is lossless (with ideal compo- age divider” containing a capacitor or inductor will have
nents) compared with the 50% efficiency of resistive charging. a frequency-dependent division ratio. In addition, circuits
Charging is complete after tf , equal to a half-cycle of the resonant containing these components (known collectively as re-
frequency. The series diode terminates the cycle, which would oth- active components) “corrupt” input waveforms such as
erwise continue to oscillate between 0 and 2Vin . square waves, as we saw earlier.
However, both capacitors and inductors are linear de-
vices, meaning that the amplitude of the output waveform,
whatever its shape, increases exactly in proportion to the
34 A mechanical analogy may be helpful here. Imagine dropping pack- input waveform’s amplitude. This linearity has many con-
ages onto a conveyor belt that is moving at speed v ; the packages are sequences, the most important of which is probably the
accelerated to that speed by friction, with 50% efficiency, finally reach- following: the output of a linear circuit, driven with a
ing the belt speed v , at which speed they ride into the sunset. That’s re- sinewave at some frequency f , is itself a sinewave at the
sistive charging. Now we try something completely different, namely, same frequency (with, at most, changed amplitude and
we rig up a conveyor belt with little catchers attached by springs to the phase).
belt; and alongside it we have a second belt, running at twice the speed Because of this remarkable property of circuits contain-
(2v ). Now when we drop a package onto the first conveyor it com- ing resistors, capacitors, and inductors (and, later, linear
presses a spring, then rebounds at 2v ; and it makes a soft landing onto
amplifiers), it is particularly convenient to analyze any such
the second conveyor. No energy is lost (ideal springs), and the package
circuit by asking how the output voltage (amplitude and
rides off into the sunset at 2v . That’s reactive charging.
35 Resonant charging is used for the high-voltage supply in flashlamps
phase) depends on the input voltage for sinewave input at a
and stroboscopes, with the advantages of (a) full charge between single frequency, even though this may not be the intended
flashes (spaced no closer than tf ), and (b) no current immediately after use. A graph of the resulting frequency response, in which
discharge (see waveforms), thus permitting the flashlamp to “quench” the ratio of output to input is plotted for each sinewave
after each flash. frequency, is useful for thinking about many kinds of
Art of Electronics Third Edition 1.7.1. Frequency analysis of reactive circuits 41

waveforms. As an example, a certain “boom-box” loud- is

speaker might have the frequency response shown in dV
Figure 1.87, in which the “output” in this case is of course I(t) = C = Cω V0 cos ω t,
dt
sound pressure, not voltage. It is desirable for a speaker
to have a “flat” response, meaning that the graph of sound i.e., a current of amplitude ω CV0 , with its phase leading the
pressure versus frequency is constant over the band of input voltage by 90◦ . If we consider amplitudes only, and
audible frequencies. In this case the speaker’s deficiencies disregard phases, the current is
can be corrected by the introduction of a passive filter with V
I= .
the inverse response (as shown) within the amplifiers of 1/ω C
the radio.
As we will see, it is possible to generalize Ohm’s (Recall that ω = 2π f .) It behaves like a frequency-
law, replacing the word “resistance” with “impedance,” in dependent resistance R = 1/ω C, but in addition the current
order to describe any circuit containing these linear passive is 90◦ out of phase with the voltage (Figure 1.89).
devices (resistors, capacitors, and inductors). You could
I (t)
think of the subject of impedance (generalized resistance)
as Ohm’s law for circuits that include capacitors and V(t) = V0 sin ωt
inductors. C
Some terminology: impedance (Z) is the “generalized
resistance”; inductors and capacitors, for which the volt-
age and current are always 90◦ out of phase, are reactive; Figure 1.88. A sinusoidal ac voltage drives a capacitor.
they have reactance (X). Resistors, with voltage and cur-
rent always in phase, are resistive; they have resistance (R).
In general, in a circuit that combines resistive and reactive I
components, the voltage and current at some place will
have some in-between phase relationship, described by a
complex impedance: impedance = resistance + reactance, t
or Z = R + jX (more about this later).36 However, you’ll
see statements like “the impedance of the capacitor at this
frequency is . . . ” The reason you don’t have to use the word V
“reactance” in such a case is that impedance covers every-
thing. In fact, you frequently use the word “impedance” t

even when you know it’s a resistance you’re talking about;

you say “the source impedance” or “the output impedance” Figure 1.89. Current in a capacitor leads the sinusoidal voltage by
when you mean the Thévenin equivalent resistance of some 90◦ .
source. The same holds for “input impedance.”
In all that follows, we will be talking about circuits For example, a 1 μ F capacitor put across the 115 volts
driven by sinewaves at a single frequency. Analysis of cir- (rms) 60 Hz powerline draws a current of rms amplitude:
cuits driven by complicated waveforms is more elaborate,
115
involving the methods we used earlier (differential equa- I= = 43.4 mA (rms).
1/(2π × 60 × 10−6 )
tions) or decomposition of the waveform into sinewaves
(Fourier analysis). Fortunately, these methods are seldom Soon enough we will complicate matters by explicitly
necessary. worrying about phase shifts and the like – and that will get
us into some complex algebra that terrifies beginners (of-
1.7.1 Frequency analysis of reactive circuits ten) and mathophobes (always). Before we do that, though,
this is a good time to develop intuition about the frequency-
Let’s start by looking at a capacitor driven by a sinewave dependent behavior of some basic and important circuits
voltage source V (t) = V0 sin ω t (Figure 1.88). The current that use capacitors, ignoring for the time being the trouble-
36 But, in a nutshell, the magnitude of Z gives the ratio of amplitudes some fact that, when driven with a sinusoidal signal, cur-
of voltage to current, and the polar angle of Z gives the phase angle rents and voltages in a capacitor are not in phase.
between current and voltage. As we just saw, the ratio of magnitudes of voltage to
42 1.7. Impedance and reactance Art of Electronics Third Edition

current, in a capacitor driven at a frequency ω , is just -40

|V | 1
= , -30
|I| ωC

% error
which we can think of as a sort of “resistance” – the mag- -20
nitude of the current is proportional to the magnitude of
-10
the applied voltage. The official name for this quantity is
reactance, with the symbol X, thus XC for the reactance of 0
a capacitor.37 So, for a capacitor,
1
1 exact
XC = . (1.26)
ωC approx approx
0.8
This means that a larger capacitance has a smaller reac-
tance. And this makes sense, because, for example, if you

Vout / V in
0.6
double the value of a capacitor, it takes twice as much
current to charge and discharge it through the same volt- 0.4
age swing in the same time (recall I = C dV/dt). For the
same reason the reactance decreases as you increase the 0.2 highpass lowpass
frequency – doubling the frequency (while holding V con-
stant) doubles the rate of change of voltage and therefore 0
0.001 0.01 0.1 1 10 100 1000
requires that you double the current, thus half the reac- f/fo
tance.
So, roughly speaking, we can think of a capacitor as Figure 1.91. Frequency response of single-section RC filters,
a “frequency-dependent resistor.” Sometimes that’s good showing the results both of a simple approximation that ignores
phase (dashed curve), and the exact result (solid curve). The frac-
enough, sometimes it isn’t. We’ll look at a few circuits in
tional error (i.e., dashed/solid) is plotted above.
which this simplified view gets us reasonably good results,
and provides nice intuition; later we’ll fix it up, using the
correct complex algebra, to get a precise result. (Keep in
mind that the results we are about to get are approximate;
we’re lying to you – but it’s a small lie, and anyway we’ll Vout XC 1/ω C 1
tell the truth later. In the meantime we’ll use the weird sym- = = (approximate!)
Vin R + XC R + 1/ω C 1 + ω RC
bol instead of = in all such “approximate equations,” and (1.27)
we’ll flag the equation as approximate.)
We’ve plotted that ratio in Figure 1.91 (and also that of its
R
Vin Vout cousin, the highpass filter), along with their exact results
that we’ll understand soon enough in §1.7.8.
C You can see that the circuit passes low frequencies fully
(because at low frequencies the capacitor’s reactance is
very high, so it’s like a divider with a smaller resistor atop
Figure 1.90. Lowpass filter.
a larger one) and that it blocks high frequencies. In partic-
ular, the crossover from “pass” to “block” (often called the
breakpoint) occurs at a frequency ω0 at which the capaci-
A. RC lowpass filter (approximate)
tor’s reactance (1/ω0C) is equal to the resistance R: ω0 =
The circuit in Figure 1.90 is called a lowpass filter, because
1/RC. At frequencies well beyond the crossover (where the
it passes low frequencies and blocks high frequencies. If
product ω RC  1), the output decreases inversely with in-
you think of it as a frequency-dependent voltage divider,
creasing frequency; that makes sense because the reactance
this makes sense: the lower leg of the divider (the capaci-
of the capacitor, already much smaller than R, continues
tor) has a decreasing reactance with increasing frequency,
dropping as 1/ω . It’s worth noting that, even with our “ig-
so the ratio of Vout /Vin decreases accordingly:
noration of phase shifts,” the equation (and graph) for the
37 Later we’ll see inductors, which also have a 90◦ phase shift (though of ratio of voltages is quite accurate at both low and high fre-
the opposite sign), and so are likewise characterized by a reactance XL . quencies and is only modestly in error around the crossover
Art of Electronics Third Edition 1.7.1. Frequency analysis of reactive circuits 43
√
frequency, where the correct ratio is Vout /Vin = 1/ 2 ≈ without loss (attenuation). That determines the product
0.7, rather than the 0.5 that we got.38 RC: RC > 1/ωmin , where for this case you might choose
fmin ≈ 5 Hz, and so RC = 1/ωmin = 1/2π fmin ≈ 30 ms.
C
Vin Vout
Rout ≤ 1k Rin ≥ 100k
R C
3.3μF

R
Figure 1.92. Highpass filter. pre-amp 10k amplifier

f 3dB = 1 ≈ 5Hz
2πRC
B. RC highpass filter (approximate)
Figure 1.93. “Blocking capacitor”: a highpass filter for which all sig-
You get the reverse behavior (pass high frequencies, block
nal frequencies of interest are in the passband.
low) by interchanging R and C, as in Figure 1.92. Treating
it as a frequency-dependent voltage divider, and ignoring
phase shifts once again, we get (see Figure 1.91) You’ve got the product, but you still have to choose in-
Vout R R ω RC dividual values for R and C. You do this by noticing that
= = (approximate!) the input signal sees a load equal to R at signal frequen-
Vin R + XC R + 1/ω C 1 + ω RC
(1.28) cies (where C’s reactance is small – it’s just a piece of wire
High frequencies (above the same crossover frequency as there), so you choose R to be a reasonable load, i.e., not so
before, ω  ω0 = 1/RC) pass through (because the capac- small that it’s hard to drive, and not so large that the circuit
itor’s reactance is much smaller than R), whereas frequen- is prone to signal pickup from other circuits in the box. In
cies well below the crossover are blocked (the capacitor’s the audio business it’s common to see a value of 10 kΩ, so
reactance is much larger than R). As before, the equation we might choose that value, for which the corresponding C
and graph are accurate at both ends, and only modestly is 3.3 μ F (Figure 1.93). The circuit connected to the output
in error at the√crossover, where the correct ratio is, again, should have an input resistance much greater than 10 kΩ, to
Vout /Vin = 1/ 2. avoid loading effects on the filter’s output; and the driving
circuit should be able to drive a 10 kΩ load without signifi-
C. Blocking capacitor cant attenuation (loss of signal amplitude) to prevent circuit
Sometimes you want to let some band of signal frequencies loading effects by the filter on the signal source. It’s worth
pass through a circuit, but you want to block any steady dc noting that our approximate model, ignoring phase shifts,
voltage that may be present (we’ll see how this can hap- is perfectly adequate for blocking capacitor design; that is
pen when we learn about amplifiers in the next chapter). because the signal band is fully in the passband, where the
You can do the job with an RC highpass filter if you choose effects of phase shifts are negligible.
the crossover frequency correctly: a highpass filter always In this section we’ve been thinking in the frequency do-
blocks dc, so what you do is choose component values so main (sinewaves of frequency f ). But it’s useful to think
that the crossover frequency is below all frequencies of in- in the time domain, where, for example, you might use a
terest. This is one of the more frequent uses of a capacitor blocking capacitor to couple pulses, or square waves. In
and is known as a dc blocking capacitor. such situations you encounter waveform distortion, in the
For instance, every stereo audio amplifier has all its in- form of “droop” and overshoot (rather than the simple am-
puts capacitively coupled, because it doesn’t know what dc plitude attenuation and phase shift you get with sinusoidal
level its input signals might be riding on. In such a coupling waves). Thinking in the time domain, the criterion you use
application you always pick R and C so that all frequen- to avoid waveform distortion in a pulse of duration T is
cies of interest (in this case, 20 Hz to 20 kHz) are passed that the time constant τ =RCT . The resulting droop is
approximately T /τ (followed by a comparable overshoot
38 Of course, it fails to predict anything about phase shifts in this circuit. at the next transition).
As we’ll see later, the output signal’s phase lags the input by 90◦ at You often need to know the reactance of a capacitor at a
high frequencies, going smoothly from 0◦ at low frequencies, with a given frequency (e.g., for design of filters). Figure 1.100 in
45◦ lag at ω0 (see Figure 1.104 in §1.7.9). §1.7.8 provides a very useful graph covering large ranges
44 1.7. Impedance and reactance Art of Electronics Third Edition

of capacitance and frequency, giving the value of XC = the first stage is highpass with a breakpoint of 100 Hz, and the
1/2π fC. second stage is lowpass with a breakpoint of 10 kHz. Assume the
input signal source has an impedance of 100Ω. What is the worst-
D. Driving and loading RC filters case output impedance of your filter, and therefore what is the
This example of an audio-blocking capacitor raised the is- minimum recommended load impedance?
sue of driving and loading RC filter circuits. As we dis-
cussed in §1.2.5A, in the context of voltage dividers, you 1.7.2 Reactance of inductors
generally like to arrange things so that the circuit be-
Before we embark on a fully correct treatment of
ing driven does not significantly load the driving resis-
impedance, replete with complex exponentials and the like,
tance (Thévenin equivalent source resistance) of the signal
let’s use our approximation tricks to figure out the reac-
source.
tance of an inductor.
The same logic applies here, but with a generalized
It goes as before: we imagine an inductor L driven by
kind of resistance that includes the reactance of capacitors
a sinusoidal voltage source of angular frequency ω such
(and inductors), known as impedance. So a signal source’s
that a current I(t) = I0 sin ω t is flowing.40 Then the voltage
impedance should generally be small compared with the
across the inductor is
impedance of the thing being driven.39 We’ll have a pre-
dI(t)
cise way of talking about impedance shortly, but it’s cor- V (t) = L = Lω I0 cos ω t.
rect to say that, apart from phase shifts, the impedance of a dt
capacitor is equal to its reactance. And so the ratio of magnitudes of voltage to current – the
What we want to know, then, are the input and out- resistance-like quantity called reactance – is just
put impedances of the two simple RC filters (lowpass and |V | Lω I0
= = ω L.
highpass). This sounds complicated, because there are four |I| I0
impedances, and they all vary with frequency. However, if
So. for an inductor,
you ask the question the right way, the answer is simple,
and the same in all cases! XL = ω L. (1.29)
First, assume that in each case the right thing is being
Inductors, like capacitors, have a frequency-dependent
done to the other end of the filter: when we ask the input
reactance; however, here the reactance increases with in-
impedance, we assume the output drives a high impedance
creasing frequency (the opposite of capacitors, where it
(compared with its own); and when we ask the output
decreases with increasing frequency). So, in the simplest
impedance, we assume the input is driven by a signal
view, a series inductor can be used to pass dc and low fre-
source of low internal (Thévenin) impedance. Second, we
quencies (where its reactance is small) while blocking high
dispose of the variation of impedances with frequency by
frequencies (where its reactance is high). You often see in-
asking only for the worst-case value; that is, we care what
ductors used this way, particularly in circuits that operate
only the maximum output impedance of a filter circuit may
at radio frequencies; in that application they’re sometimes
be (because that is the worst for driving a load), and we care
called chokes.
about only the minimum input impedance (because that is
the hardest to drive). 1.7.3 Voltages and currents as complex numbers
Now the answer is astonishingly simple: in all cases the
worst-case impedance is just R. At this point it is necessary to get into some complex al-
gebra; you may wish to skip over the math in some of the
Exercise 1.23. Show that the preceding statement is correct. following sections, taking note of the results as we derive
So, for example, if you want to hang an RC lowpass fil- them. A knowledge of the detailed mathematics is not nec-
ter onto the output of an amplifier whose output resistance essary for understanding the remainder of the book. Very
is 100 Ω, start with R = 1k, then choose C for the break- little mathematics will be used in later chapters. The sec-
point you want. Be sure that whatever loads the output has tion ahead is easily the most difficult for the reader with
an input impedance of at least 10k. You can’t go wrong. little mathematical preparation. Don’t be discouraged!
As we’ve just seen, there can be phase shifts between
Exercise 1.24. Design a two-stage “bandpass” RC filter, in which
40 We take the easy path here by specifying the current, rather than the
39 With two important exceptions, namely, transmission lines and current voltage; we are rewarded with a simple derivative (rather than a simple
sources. integral!).
Art of Electronics Third Edition 1.7.4. Reactance of capacitors and inductors 45

the voltage and current in an ac circuit being driven by a I(t) = Re(Ie jω t )

sinewave at some frequency. Nevertheless, as long as the = Re(I) cos ω t − I m(I) sin ω t.
circuit contains only linear elements (resistors, capacitors,
inductors), the magnitudes of the currents everywhere in For example, a voltage whose complex representation is
the circuit are still proportional to the magnitude of the
V = 5j
driving voltage, so we might hope to find some general-
ization of voltage, current, and resistance in order to res- corresponds to a (real) voltage versus time of
cue Ohm’s law. Evidently a single number won’t suffice to
specify the current, say, at some point in the circuit, be- V (t) = Re[5 j cos ω t + 5 j( j) sin ω t]
cause we must somehow have information about both the = −5 sin ω t volts.
magnitude and phase shift.
Although we can imagine specifying the magnitudes
and phase shifts of voltages and currents at any point in 1.7.4 Reactance of capacitors and inductors
the circuit by writing them out explicitly, e.g., V (t) = With this convention we can apply complex Ohm’s law cor-
23.7 sin(377t + 0.38), it turns out that we can meet our rectly to circuits containing capacitors and inductors, just
requirements more simply by using the algebra of com- as for resistors, once we know the reactance of a capacitor
plex numbers to represent voltages and currents. Then we or inductor. Let’s find out what these are. We begin with
can simply add or subtract the complex number represen- a simple (co)sinusoidal voltage V0 cos ω t applied across a
tations, rather than laboriously having to add or subtract capacitor:
the actual sinusoidal functions of time themselves. Be-
cause the actual voltages and currents are real quantities V (t) = Re(V0 e jω t ).
that vary with time, we must develop a rule for convert- Then, using I = C(dV (t)/dt), we obtain
ing from actual quantities to their representations, and vice  
versa. Recalling once again that we are talking about a sin- V0 e jω t
I(t) = −V0Cω sin ω t = Re ,
gle sinewave frequency, ω , we agree to use the following − j/ω C
 j ω t 
rules. V0 e
= Re
ZC
1. Voltages and currents are represented by the complex
quantities V and I. The voltage V0 cos(ω t + φ ) is to be i.e., for a capacitor
represented by the complex number jφ
φ
√ V0 e . Recall that ZC = − j/ω C (= − jXC );
e = cos φ + j sin φ , where j = −1.
j

2. We obtain actual voltages and currents by multiply- ZC is the impedance of a capacitor at frequency ω ; it is
ing their complex number representations by e jω t and equal in magnitude to the reactance XC = 1/ω C that we
then taking the real part: V (t) = Re(Ve jω t ), I(t) = found earlier, but with a factor of − j that accounts for the
Re(Ie jω t ). 90◦ leading phase shift of current versus voltage. As an ex-
ample, a 1 μ F capacitor has an impedance of −2653 j Ω at
In other words, 60 Hz, and −0.16 j Ω at 1 MHz. The corresponding reac-
circuit voltage complex number tances are 2653 Ω and 0.16 Ω.41 Its reactance (and also its
versus time representation impedance) at dc is infinite.
−→ If we did a similar analysis for an inductor, we would
V0 cos(ω t + φ ) V0 e jφ = a + jb
←− find
multiply by
ZL = j ω L (= jXL ).
e jω t and
take real part A circuit containing only capacitors and inductors always
has a purely imaginary impedance, meaning that the volt-
(In electronics, the symbol j is used instead of i in the ex-
age and current are always 90◦ out of phase – it is purely
ponential in order to avoid confusion with the symbol i
reactive. When the circuit contains resistors, there is also
meaning small-signal current.) Thus, in the general case,
the actual voltages and currents are given by
41 Note the convention that the reactance XC is a real number (the 90◦
V (t) = Re(Ve jω t ) phase shift is implicit in the term “reactance”), but the corresponding
= Re(V) cos ω t − I m(V) sin ω t impedance is purely imaginary: Z = R − jX.
46 1.7. Impedance and reactance Art of Electronics Third Edition

a real part to the impedance. The term “reactance” in that we’ll analyze (correctly, this time) the simple but extremely
case means the imaginary part only. important and useful RC lowpass and highpass filter cir-
cuits.
Imagine putting a 1 μ F capacitor across a 115 volts
1.7.5 Ohm’s law generalized (rms) 60 Hz powerline. What current flows? Using com-
With these conventions for representing voltages and cur- plex Ohm’s law, we have
rents, Ohm’s law takes a simple form. It reads simply
Z = − j/ω C.
I = V/Z,
V = IZ,
Therefore, the current is given by

where the voltage represented by V is applied across a cir- I = V/Z.

cuit of impedance Z, giving a current represented by I. The
The phase of the voltage is arbitrary, so let us choose√V =
complex impedance of devices in series or parallel obeys
A, i.e., V (t) = A cos ω t, where the amplitude A = 115 2 ≈
the same rules as resistance:
163 volts. Then
Z = Z1 + Z2 + Z3 + · · · (series), (1.30)
I = jω CA ≈ −0.061 sin ω t.
1
Z= (parallel). (1.31) The resulting current has an amplitude of 61 mA (43 mA
1 1 1
+ + +···
Z1 Z2 Z3 rms) and leads the voltage by 90◦ . This agrees with our
previous calculation. More simply, we could have noticed
Finally, for completeness we summarize here the formulas
that the impedance of the capacitor is negative imaginary,
for the impedance of resistors, capacitors, and inductors:
so whatever the absolute phase of V , the phase of Icap must
ZR = R (resistor), lead by 90◦ . And in general the phase angle between cur-
ZC = − j/ω C = 1/ jω C (capacitor), (1.32) rent and voltage, for any two-terminal RLC circuit, is equal
to the angle of the (complex) impedance of that circuit.
ZL = j ω L (inductor). Note that if we wanted to know just the magnitude of
With these rules we can analyze many ac circuits by the the current, and didn’t care what the relative phase was, we
same general methods we used in handling dc circuits, i.e., could have avoided doing any complex algebra: if
application of the series and parallel formulas and Ohm’s
A = B/C,
law. Our results for circuits such as voltage dividers will
look nearly the same as before. For multiply-connected then
networks we may have to use Kirchhoff’s laws, just as with
dc circuits, in this case using the complex representations A = B/C,
for V and I: the sum of the (complex) voltage drops around
a closed loop is zero, and the sum of the (complex) cur- where A, B, and C are the magnitudes of the respective
rents into a point is zero. The latter rule implies, as with dc complex numbers; this holds for multiplication, also (see
circuits, that the (complex) current in a series circuit is the Exercise 1.18). Thus, in this case,
same everywhere. I = V /Z = ω CV.
Exercise 1.25. Use the preceding rules for the impedance of de-
vices in parallel and in series to derive the formulas (1.17) and This trick, which we used earlier (because we didn’t know
(1.18) for the capacitance of two capacitors (a) in parallel and (b) any better), is often useful.
in series. Hint: in each case, let the individual capacitors have ca- Surprisingly, there is no power dissipated by the capaci-
pacitances C1 and C2 . Write down the impedance of the parallel or tor in this example. Such activity won’t increase your elec-
series combination; then equate it to the impedance of a capacitor tric bill; you’ll see why in the next section. Then we will
with capacitance C. Then find C. go on to look at circuits containing resistors and capacitors
with our complex Ohm’s law.
Let’s try out these techniques on the simplest circuit
imaginable, an ac voltage applied across a capacitor, which Exercise 1.26. Show that, if A = BC, then A = BC, where A, B,
we looked at earlier, in §1.7.1. Then, after a brief look at and C are magnitudes. Hint: represent each complex number in
power in reactive circuits (to finish laying the groundwork), polar form, i.e., A = Aeiθ .
Art of Electronics Third Edition 1.7.6. Power in reactive circuits 47

where V and I are complex rms amplitudes (and an aster-

I (t) isk means complex conjugate – see the math review, Ap-
pendix A, if this is unfamiliar).
V(t)
Let’s take an example. Consider the preceding circuit,
V (t) I (t)
with a 1 volt (rms) sinewave driving a capacitor. We’ll do
everything with rms amplitudes, for simplicity. We have
t
sine
wave C V = 1,

V
I= = jω C,
− j/ω C
A B C D

Figure 1.94. The power delivered to a capacitor is zero over a full P = Re(VI∗ ) = Re(− jω C) = 0.
sinusoidal cycle, owing to the 90◦ phase shift between voltage and
That is, the average power is zero, as stated earlier.
current.

I C
1.7.6 Power in reactive circuits
The instantaneous power delivered to any circuit element
V0 cos ωt R
is always given by the product P = V I. However, in reac-
tive circuits where V and I are not simply proportional, you
can’t just multiply their amplitudes together. Funny things
can happen; for instance, the sign of the product can re- Figure 1.95. Power and power factor in a series RC circuit.
verse over one cycle of the ac signal. Figure 1.94 shows
an example. During time intervals A and C, power is being As another example, consider the circuit shown in Fig-
delivered to the capacitor (albeit at a variable rate), causing ure 1.95. Our calculations go like this:
it to charge up; its stored energy is increasing (power is the j
Z = R− ,
rate of change of energy). During intervals B and D, the ωC
power delivered to the capacitor is negative; it is discharg- V = V0 ,
ing. The average power over a whole cycle for this example V V0 V0 [R + ( j/ω C)]
I= = = 2 ,
is in fact exactly zero, a statement that is always true for Z R − ( j/ω C) R + (1/ω 2C2 )
any purely reactive circuit element (inductors, capacitors,
V02 R
or any combination thereof). If you know your trigonomet- P = Re(VI∗ ) = .
R + (1/ω 2C2 )
2
ric integrals, the next exercise will show you how to prove
this. (In the third line we multiplied numerator and denominator
by the complex conjugate of the denominator in order to
Exercise 1.27. Optional exercise: prove that a circuit whose cur- make the denominator real.) The calculated power42 is less
rent is 90◦ out of phase with the driving voltage consumes no than the product of the magnitudes of V and I. In fact, their
power, averaged over an entire cycle. ratio is called the power factor:
How do we find the average power consumed by an ar- V02
|V| |I| = ,
bitrary circuit? In general, we can imagine adding up little [R + (1/ω 2C2 )]1/2
2

pieces of V I product, then dividing by the elapsed time. In power

other words, power factor =
|V| |I|
 T
1
P= V (t)I(t) dt, (1.33) R
T 0 =
[R2 + (1/ω 2C2 )]1/2
where T is the time for one complete cycle. Luckily, that’s
42
almost never necessary. Instead, it is easy to show that the It’s always a good idea to check limiting values: here we see that
average power is given by P → V 2 /R for large C; and for small C the magnitude of the current
|I| → V0 /XC , or V0 ω C, thus P → I 2 R = V02 ω 2C2 R, in agreement at
P = Re(VI∗ ) = Re(V∗ I), (1.34) both limits.
48 1.7. Impedance and reactance Art of Electronics Third Edition

in this case. The power factor is the cosine of the phase Vin
angle between the voltage and the current, and it ranges
from 0 (purely reactive circuit) to 1 (purely resistive). A Z1

power factor of less than 1 indicates some component of

reactive current.43 It’s worth noting that the power factor Vout
goes to unity, and the dissipated power goes to V 2 /R, in
Z2
the limit of large capacitance (or of high frequency), where
the reactance of the capacitor becomes much less than R.
Exercise 1.28. Show that all the average power delivered to the
Figure 1.96. Generalized voltage divider: a pair of arbitrary
preceding circuit winds up in the resistor. Do this by computing
impedances.
the value of VR2 /R. What is that power, in watts, for a series circuit
of a 1 μ F capacitor and a 1.0k resistor placed across the 115 volt
(rms), 60 Hz powerline? Z2
Vout = IZ2 = Vin .
Z1 + Z2
Power factor is a serious matter in large-scale electrical
Rather than worrying about this result in general, let’s look
power distribution, because reactive currents don’t result in
at some simple (but very important) examples, beginning
useful power being delivered to the load, but cost the power
with the RC highpass and lowpass filters we approximated
company plenty in terms of I 2 R heating in the resistance of
earlier.
generators, transformers, and wiring. Although residential
users are billed only for “real” power [Re(VI∗ )], the power
company charges industrial users according to the power 1.7.8 RC highpass filters
factor. This explains the capacitor yards that you see behind
We’ve seen that by combining resistors with capacitors
large factories, built to cancel the inductive reactance of
it is possible to make frequency-dependent voltage di-
industrial machinery (i.e., motors).
viders, owing to the frequency dependence of a capacitor’s
Exercise 1.29. Show that adding a series capacitor of value C = impedance ZC = − j/ω C. Such circuits can have the de-
1/ω 2 L makes the power factor equal 1.0 in a series RL circuit. sirable property of passing signal frequencies of interest
Now do the same thing, but with the word “series” changed to while rejecting undesired signal frequencies. In this sub-
“parallel.” section and the next we revisit the simple lowpass and
highpass RC filters, correcting the approximate analysis
of §1.7.1; though simple, these circuits are important and
1.7.7 Voltage dividers generalized
widely used. Chapter 6 and Appendix E describe filters of
Our original voltage divider (Figure 1.6) consisted of a pair greater sophistication.
of resistors in series to ground, input at the top and output Referring back to the classic RC highpass filter (Fig-
at the junction. The generalization of that simple resistive ure 1.92), we see that the complex Ohm’s law (or the com-
divider is a similar circuit in which either or both resis- plex voltage-divider equation) gives
tors are replaced with a capacitor or inductor (or a more
R R(R + j/ω C)
complicated network made from R, L, and C), as in Fig- Vout = Vin = Vin 2 .
R − j/ω C R + (1/ω 2C2 )
ure 1.96. In general, the division ratio Vout /Vin of such a di-
vider is not constant, but depends on frequency (as we have (For the last step, multiply top and bottom by the complex
already seen, in our approximate treatment of the lowpass conjugate of the denominator.) Most often we don’t care
and highpass filters in §1.7.1). The analysis is straightfor- about the phase of Vout , just its amplitude:
ward: Vout = (Vout V∗out )1/2
Vin
I= , R
Ztotal = Vin .
[R2 + (1/ω 2C2 )]1/2
Ztotal = Z1 + Z2 Note the analogy to a resistive divider, where
R2
Vout = V .
43 Or, for nonlinear circuits, it indicates that the current waveform is not
R1 + R2 in
proportional to the voltage waveform. More on this in §9.7.1. Here the impedance of the series RC combination
Art of Electronics Third Edition 1.7.8. RC highpass filters 49

H
m
H

m
0
H
0

H
10
1H

10
10

10

1m
1MEG

0.

μH
1p
1p

0
R

10
100k

10
p

H
F

μ
10

H
reactance

10
0

1μ
10k

pF
1n

nH
F
10

0
10
n
Figure 1.97. Input impedance of unloaded highpass filter.

F
1k

10
0 nF

nH
1μ
F

10
100Ω

10
jZ

μ F
10
0
10Ω

μF
R |Ztotal | = √R 2+ 1
ω 2 C2
10Hz 100 1kHz 10 100 1MHz 10 100 1GHz
φ A. frequency
– j /ωC
φ = tan –1 ( –1/RωC )
10

0.
Ztotal = R – j /ωC
9

22
0.
10
8

0.
33
Figure 1.98. Impedance of series RC. 7

0.
8

47
6.
6

0.
7

68
4.
Vout 5
1.0
Vin
∝ω

C
3

=
3.
Vout + –3dB

1.
reactance 4

0
1.
2
Vin

2.

5
2.
3

2
0

5
ω 3dB = 1/RC ω

1.

0
1.
3.
3

=
L
4.

68
2
7
Figure 1.99. Frequency response of highpass filter. The corre-

0.
sponding phase shift goes smoothly from +90◦ (at ω = 0), through
6.

47
8

0.
+45◦ (at ω3 dB ), to 0◦ (at ω = ∞), analogous to the lowpass filter’s
10

33
phase shift (Figure 1.104).
.

0.
0

22
0.
(Figure 1.97) is as shown in Figure 1.98. So the “response” 1
of this circuit, ignoring phase shifts by taking magnitudes 1 2 3 4 5 6 7 8 9 10
B. frequency
of the complex amplitudes, is given by
R Figure 1.100. A: Reactance of inductors and capacitors versus
Vout = Vin
[R2 + (1/ω 2C2 )]1/2 frequency; all decades are identical, except for scale. B: A single
decade from part A expanded, with standard 20% component val-
2π f RC
= Vin (1.35) ues (EIA “E6”) shown.
[1 + (2π f RC)2 ]1/2
and looks as shown in Figure 1.99 (and earlier in Fig- very common; for instance, the input to the oscilloscope
ure 1.91). can be switched to “ac coupling.” That’s just an RC high-
Note that we could have gotten this result immediately pass filter with the bend at about 10 Hz (you would use ac
by taking the ratio of the magnitudes of impedances, as in coupling if you wanted to look at a small signal riding on
Exercise 1.26 and the example immediately preceding it; a large dc voltage). Engineers like to refer to the −3 dB
the numerator is the magnitude of the impedance of the “breakpoint” of a filter (or of any circuit that behaves like
lower leg of the divider (R), and the denominator is the a filter). In the case of the simple RC high-pass filter, the
magnitude of the impedance of the series combination of R −3 dB breakpoint is given by
and C.
f3 dB = 1/2π RC.
As we noted earlier, the output is approximately equal
to the input at high frequencies (how high? ω  1/RC) You often need to know the impedance of a capaci-
and goes to zero at low frequencies. The highpass filter is tor at a given frequency (e.g., for the design of filters).
50 1.7. Impedance and reactance Art of Electronics Third Edition

0.01μF
1.0
Vout + – 3dB Vout
Vin
∝ 1/ω
1.0k Vin

0
ω 3dB = 1/RC ω
Figure 1.101. Highpass filter example.
Figure 1.103. Frequency response of lowpass filter.
Figure 1.100 provides a very useful graph covering large
ranges of capacitance and frequency, giving the value of
|Z| = 1/2π fC. The lowpass filter’s output can be viewed as a signal
As an example, consider the filter shown in Fig- source in its own right. When driven by a perfect ac voltage
ure 1.101. It is a highpass filter with the 3 dB point44 at (zero source impedance), the filter’s output looks like R at
15.9 kHz. The impedance of a load driven by it should be low frequencies (the perfect signal source can be replaced
much larger than 1.0k in order to prevent circuit loading with a short, i.e., by its small-signal source impedance, for
effects on the filter’s output, and the driving source should the purpose of impedance calculations). It drops to zero
be able to drive a 1.0k load without significant attenuation impedance at high frequencies, where the capacitor dom-
(loss of signal amplitude) in order to prevent circuit load- inates the output impedance. The signal driving the filter
ing effects by the filter on the signal source (recall §1.7.1D sees a load of R plus the load resistance at low frequencies,
for worst-case source and load impedances of RC filters). dropping to just R at high frequencies. As we remarked in
§1.7.1D, the worst-case source impedance and the worst-
case load impedance of an RC filter (lowpass or highpass)
1.7.9 RC lowpass filters are both equal to R.
Revisiting the lowpass filter, in which you get the opposite
frequency behavior by interchanging R and C (Figure 1.90, 0.1fc
repeated here as Figure 1.102), we find the accurate result 0°
+ ≈ 6° = –tan-1 (f/fc)

phase shift
≈ 0.6°
1
Vout = Vin (1.36) f/fc
(1 + ω 2 R2C2 )1/2
0 0
as seen in Figure 1.103 (and earlier in Figure 1.91). The 0.1 – 5.7º
0.2 –11.3º
3 dB point is again at a frequency45 f = 1/2π RC. Lowpass – 45° 0.25 –14º
filters are quite handy in real life. For instance, a lowpass 0.5 –26.5º
1.0 – 45º
filter can be used to eliminate interference from nearby
radio and television stations (0.5–800 MHz), a problem
that plagues audio amplifiers and other sensitive electronic ≈ 6° + ≈ 0.6°
– 90°
equipment. fc 10fc

R 1.0
+ 0.707
Vin Vout

C
0.1
Vout
Vin
Figure 1.102. Lowpass filter. – 6dB/octave ~1/f
0.01 –20dB/decade

Exercise 1.30. Show that the preceding expression for the re-
sponse of an RC lowpass filter is correct. 0.001
0.01fc 0.1fc fc 10fc 100fc
44 One often omits the minus sign when referring to the −3 dB point.
45 As mentioned in §1.7.1A, we often like to define the breakpoint fre- Figure 1.104. Frequency response (phase and amplitude) of low-
quency ω0 =1/RC, and work with frequency ratios ω /ω0 . Then a use- pass filter plotted on logarithmic axes. Note that the phase shift is
ful form for the denominator in eq’n 1.36 is 1 + (ω /ω0 )2 . The same −45◦ at the −3 dB point and is within 6◦ of its asymptotic value for
applies to eq’n 1.35, where the numerator becomes ω /ω0 . a decade of frequency change.
Art of Electronics Third Edition 1.7.11. Inductors versus capacitors 51

In Figure 1.104, we’ve plotted the same lowpass filter The RC integrator (§1.4.4) is the same circuit as the low-
response with logarithmic axes, which is a more common pass filter; by similar reasoning, the criterion for a good in-
way that it’s done. You can think of the vertical axis as tegrator is that the lowest signal frequencies must be well
decibels, and the horizontal axis as octaves (or decades). above the 3 dB point.
On such a plot, equal distances correspond to equal ra-
tios. We’ve also plotted the phase shift, using a linear ver-
1.7.11 Inductors versus capacitors
tical axis (degrees) and the same logarithmic frequency
axis. This sort of plot is good for seeing the detailed re- Instead of capacitors, inductors can be combined with re-
sponse even when it is greatly attenuated (as at right); we’ll sistors to make lowpass (or highpass) filters. In practice,
see a number of such plots in Chapter 6, when we treat however, you rarely see RL lowpass or highpass filters. The
active filters. Note that the filter curve plotted here be- reason is that inductors tend to be more bulky and expen-
comes a straight line at large attenuations, with a slope of sive and perform less well (i.e., they depart further from
−20 dB/decade (engineers prefer to say “−6 dB/octave”). the ideal) than capacitors (see Chapter 1x). If you have a
Note also that the phase shift goes smoothly from 0◦ (at fre- choice, use a capacitor. One important exception to this
quencies well below the breakpoint) to −90◦ (well above general statement is the use of ferrite beads and chokes in
it), with a value of −45◦ at the −3 dB point. A rule of high-frequency circuits. You just string a few beads here
thumb for single-section RC filters is that the phase shift and there in the circuit; they make the wire interconnec-
is ≈ 6◦ from its asymptotic value at 0.1 f3 dB and at 10 f3 dB . tions slightly inductive, raising the impedance at very high
frequencies and preventing oscillations, without the added
Exercise 1.31. Prove the last assertion.
series resistance you would get with an RC filter. An RF
An interesting question is the following: is it possible choke is an inductor, usually a few turns of wire wound
to make a filter with some arbitrary specified amplitude re- on a ferrite core, used for the same purpose in RF circuits.
sponse and some other arbitrary specified phase response? Note, however, that inductors are essential components in
Surprisingly, the answer is no: the demands of causality (a) LC tuned circuits (§1.7.14), and (b) switch-mode power
(i.e., that response must follow cause, not precede it) force converters (§9.6.4).
a relationship between phase and amplitude response of re-
alizable analog filters (known officially as the Kramers–
1.7.12 Phasor diagrams
Kronig relation).
There’s a nice graphical method that can be helpful when
we are trying to understand reactive circuits. Let’s take an
1.7.10 RC differentiators and integrators in the example, namely, the fact that an RC filter attenuates 3 dB
frequency domain at a frequency f = 1/2π RC, which we derived in §1.7.8.
The RC differentiator that we saw in §1.4.3 is exactly the This is true for both highpass and lowpass filters. It is easy
same circuit as the highpass filter in this section. In fact, it to get a bit confused here, because at that frequency the re-
can be considered as either, depending on whether you’re actance of the capacitor equals the resistance of the resis-
thinking of waveforms in the time domain or response tor; so you might at first expect 6 dB attenuation (a factor
in the frequency domain. We can restate the earlier time- of 1/2 in voltage). That is what you would get, for example,
domain condition for its proper operation (dVout dVin ) in if you were to replace the capacitor with a resistor of the
terms of the frequency response: for the output to be small same impedance magnitude. The confusion arises because
compared with the input, the signal frequency (or frequen- the capacitor is reactive, but the matter is clarified by a pha-
cies) must be well below the 3 dB point. This is easy to sor diagram (Figure 1.105). The axes are the real (resistive)
check: suppose we have the input signal Vin = sin ω t. Then, and imaginary (reactive) components of the impedance. In
using the equation we obtained earlier for the differentiator a series circuit like this, the axes also represent the (com-
output, we have plex) voltage, because the current is the same everywhere.
So for this circuit (think of it as an RC voltage divider)
d the input voltage (applied across the series RC pair) is pro-
Vout = RC sin ω t = ω RC cos ω t,
dt portional to the length of the hypotenuse, and the output
and so dVout dVin if ω RC  1, i.e., RC  1/ω . If the voltage (across R only) is proportional to the length of the
input signal contains a range of frequencies, this must hold R leg of the triangle. The diagram represents the situation
for the highest frequencies present in the input. at the frequency where the capacitor’s reactance equals R,
52 1.7. Impedance and reactance Art of Electronics Third Edition

i.e., f = 1/2π RC, and√shows that the ratio of output voltage transfer functions in the complex frequency plane, known
to input voltage is 1/ 2, i.e., −3 dB. by engineers as the “s-plane.” This is discussed in the
advanced volume, in Chapter 1x.)
Z
Z A caution on multistage filters: you can’t simply cas-
R R R cade several identical filter sections in order to get a fre-
45°
2R
quency response that is the concatenation of the individual
– j/ωC
R√2
responses. The reason is that each stage will load the pre-
vious one significantly (since they’re identical), changing
RC filter at –3dB point resistive divider: the overall response. Remember that the response func-
R1 = R2 = R (– 6dB) tion we derived for the simple RC filters was based on a
A. B. zero-impedance driving source and an infinite-impedance
Figure 1.105. Phasor diagram for lowpass filter at 3 dB point. load. One solution is to make each successive filter sec-
tion have much higher impedance than the preceding one.
The angle between the vectors gives the phase shift from A better solution involves active circuits like transistor or
input to output. At the 3 dB point, for instance, the out- operational amplifier (op-amp) interstage “buffers,” or ac-
put amplitude equals the input amplitude divided by the tive filters. These subjects will be treated in Chapters 2–4,
square root of 2, and it leads by 45◦ in phase. This graph- 6, and 13.
ical method makes it easy to read off amplitude and phase
relationships in RLC circuits. You can use it, for example, 1.7.14 Resonant circuits
to get the response of the highpass filter that we previously
derived algebraically. When capacitors are combined with inductors or are used
in special circuits called active filters, it is possible to make
Exercise 1.32. Use a phasor diagram
 to derive the response of an circuits that have very sharp frequency characteristics (e.g.,
RC high-pass filter: Vout = Vin R/ R2 + (1/ω 2C2 ).
a large peak in the response at a particular frequency),
Exercise 1.33. At what frequency does an RC lowpass filter at- compared with the gradual characteristics of the RC filters
tenuate by 6 dB (output voltage equal to half the input voltage)? we’ve seen so far. These circuits find applications in vari-
What is the phase shift at that frequency? ous audio and RF devices. Let’s now take a quick look at
LC circuits (there will be more on them, and active filters,
Exercise 1.34. Use a phasor diagram to obtain the lowpass filter
response previously derived algebraically. in Chapter 6 and in Appendix E).

In the next chapter (§2.2.8) we’ll see a nice example of A. Parallel and series LC circuits
phasor diagrams in connection with a constant-amplitude
R
phase-shifting circuit. Vin Vout

1.7.13 “Poles” and decibels per octave L C

Look again at the response of the RC lowpass filter

(Figures 1.103 and 1.104). Far to the right of the “knee”
the output amplitude is dropping proportional to 1/ f . In Figure 1.106. LC resonant circuit: bandpass filter.
one octave (as in music, one octave is twice the frequency)
the output amplitude will drop to half, or −6 dB; so a First, consider the circuit shown in Figure 1.106. The
simple RC filter has a 6 dB/octave falloff. You can make impedance of the LC combination at frequency f is just
filters with several RC sections; then you get 12 dB/octave
1 1 1 1 ωC
(two RC sections), 18 dB/octave (three sections), and = + = −
ZLC ZL ZC jω L j
so on. This is the usual way of describing how a filter  
1
behaves beyond the cutoff. Another popular way is to say = j ωC − ,
ωL
a “three-pole filter,” for instance, meaning a filter with
three RC sections (or one that behaves like one). (The i.e.,
word “pole” derives from a method of analysis that is j
beyond the scope of this book and that involves complex ZLC = .
(1/ω L) − ω C
Art of Electronics Third Edition 1.7.14. Resonant circuits 53

In combination with R it forms a voltage divider. Because 10

of the opposite behaviors of inductors and capacitors, the R

Reactance (X/Xo)
Vin Vout
impedance of the parallel LC goes to infinity at the resonant
frequency 1.0 RLC
√ L
f0 = 1/2π LC (1.37) L C
√ C Q = 20
(i.e., ω0 = 1/ LC), giving a peak in the response there. 0.1
Q=∞
The overall response is as shown in Figure 1.107. 0.1fo fo 10fo
Frequency

61% (1/e in energy) Figure 1.108. LC notch filter (“trap”). The inductive and capacitive
37% (1/e in voltage)
reactances behave as shown, but the opposite sign of their com-
1.0
plex impedances causes the series impedance to plummet. For
ideal components the reactance of the series LC goes completely
t
to zero at resonance; for real-world components the minimum is
Δf 3dB non-zero, and usually dominated by the inductor.
Vout (time domain)
Vin fo
Q=
Δf 3dB
90º
60º
(frequency domain)

Phase shift
30º
fo =1/(2π√LC) f 0º
Figure 1.107. Frequency response of parallel LC “tank” circuit. The – 30º
inset shows the time-domain behavior: a damped oscillation (“ring-
– 60º
ing”) waveform following an input voltage step or pulse.
– 90º
In practice, losses in the inductor and capacitor limit the
sharpness of the peak, but with good design these losses 1.0
can be made very small. Conversely, a Q-spoiling resistor
is sometimes added intentionally to reduce the sharpness
Vout
of the resonant peak. This circuit is known simply as a par- Vin
allel LC resonant circuit (or “tuned circuit,” or “tank”) and
is used extensively in RF circuits to select a particular fre- Q=3
quency for amplification (the L or C can be variable, so 0 f
you can tune the resonant frequency). The higher the driv- 0 (linear scale)
fo =1/(2π√LC)
ing impedance, the sharper the peak; it is not uncommon to
drive them with something approaching a current source, Figure 1.109. Frequency and phase response of the series LC
trap. The phase changes abruptly at resonance, an effect seen in
as you will see later. The quality factor Q is a measure of
other resonator types (see for example Figure 7.36).
the sharpness of the peak. It equals the resonant frequency
divided by the width at the −3 dB points. For a parallel √
RLC circuit, Q = ω0 RC.46 at resonance ( f0 = 1/2π LC). Such a circuit is a “trap” for
Another variety of LC circuit is the series LC (Fig- signals at or near the resonant frequency, shorting them to
ure 1.108). By writing down the impedance formulas in- ground. Again, this circuit finds application mainly in RF
volved, and assuming that both the capacitor and inductor circuits. Figure 1.109 shows what the response looks like.
are ideal, i.e., that they have no resistive losses,47 you can The Q of a series RLC circuit is Q = ω0 L/R.48 To see the
convince yourself that the impedance of the LC goes to zero impact of increasing Q, look at the accurate plots of tank
and notch response in Figure 1.110.
46 Or, equivalently, Q = R/XC = R/XL , where XL = XC are the reactances
at ω0 .
47 We’ll see in Chapter 1x that real components depart from the ideal, 48 Or, equivalently, Q = XL /R = XC /R, where XL = XC are the reactances
often expressed in terms of an effective series resistance, ESR. at ω0 .
54 1.7. Impedance and reactance Art of Electronics Third Edition

Q=10
1.0
Q=3
Q=1
0.8
Vout / V in

0.6
Q=1
0.4
Q=3

0.2 Q=10

0
0 0.5 1 1.5 2
Normalized Frequency (f/fo)

Figure 1.110. Response of LC tank (dotted curves) and trap (solid Figure 1.111. There are six LC lowpass filters on this circuit board,
curves) for several values of quality factor, Q. part of the process of frequency conversion and digitizing for which
this “mixer-digitizer” was designed.

Exercise 1.35. Find the response (Vout /Vin versus frequency) for
the series LC trap circuit in Figure 1.108. RF bands; its design could occupy a book chapter. For now
just gaze at the lumpy filter in the oval (there are five more
These descriptions of LC resonant circuits are phrased on the board), comprising three inductors (the square metal
in terms of frequency response, i.e., in the frequency do- cans) and four capacitors (the pairs of shiny oblongs). It’s
main. In the time domain you’re generally interested in a a lowpass filter, designed to cut off at 1.0 MHz; it prevents
circuit’s response to pulses, or steps; there you see the sort ‘aliases” in the digitized output, a subject we’ll visit in
of behavior shown in the inset of Figure 1.107, an LC cir- Chapter 13.
cuit with Q=20. The signal voltage falls to 1/e (37%) in How well does it work? Figure 1.112 shows a “fre-
Q/π cycles; the stored energy (proportional to v 2 ) falls to quency sweep,” in which a sinewave input goes from
1/e (61% in amplitude) in Q/2π cycles. You may prefer to 0 Hz to 2 MHz as the trace goes from left to right across
think in radians: the energy falls to 1/e in Q radians, and the screen. The sausage shapes are the “envelope” of the
the voltage falls to 1/e in 2Q radians. LC resonant circuits sinewave output, here comparing the LC filter with an
are not unique in providing highly frequency-selective cir- RC lowpass filter with the same 1 MHz cutoff (1 kΩ and
cuit behavior; alternatives include quartz-crystal, ceramic, 160 pF). The LC wins, hands down. The RC is pathetic by
and surface acoustic-wave (SAW) resonators; transmission comparison. It’s not even good English to call 1 MHz its
lines; and resonant cavities. “cutoff”: it hardly cuts anything off.

1.7.15 LC filters 1.7.16 Other capacitor applications

By combining inductors with capacitors you can produce In addition to their uses in filters, resonant circuits, differ-
filters (lowpass, highpass, bandpass) with far sharper be- entiators, and integrators, capacitors are needed for several
havior in frequency response than you can with a filter other important applications. We treat these in detail later
made from a simple RC, or from any number of cascaded in the book, mentioning them here only as a preview.
RC sections. We’ll see more of this, and the related topic
of active filters, in Chapter 6. But it’s worth admiring now A. Bypassing
how well this works, to appreciate the virtue of the humble The impedance of a capacitor decreases with increasing
inductor (an often-maligned circuit component). frequency. This is the basis of another important appli-
As an example, look at Figure 1.111, a photograph cation: bypassing. There are places in circuits where you
of a “mixer-digitizer” circuit board that we built for a want to allow a dc voltage, but you don’t want signals
project some years back (specifically, a radio receiver with present. Placing a capacitor across that circuit element
250 million simultaneous channels). There’s lots of stuff on (usually a resistor) will help to kill any signals there.
the board, which has to frequency-shift and digitize three You choose the (noncritical) capacitor value so that its
Art of Electronics Third Edition 1.7.17. Thévenin’s theorem generalized 55

and the signal source (waveform, amplitude, and phase)

7-section from the open-circuit output voltage and the short-circuit
LC output current.
lowpass
filter
1.8 Putting it all together – an AM radio
In our circuit course we tie together the topics of this
chapter by hooking up a simple AM radio. The signal
RC
lowpass that’s transmitted is a sinewave at the station’s frequency
filter in the AM band (520–1720 kHz), with its amplitude var-
ied (“modulated”) according to the audio waveform (Fig-
ure 1.113). In other words, an audio waveform described
0 1MHz 2MHz by some function f (t) would be transmitted as a RF signal
input frequency
[A + f (t)] sin 2π fct; here fc is the station’s “carrier” fre-
Figure 1.112. Frequency sweep of the LC lowpass filter shown in quency, and the constant A is added to the audio waveform
Figure 1.111 compared with an RC lowpass filter with the same so that the coefficient [A + f (t)] is never negative.
1 MHz cutoff frequency. The dark outline is the amplitude envelope
of the fast swept sinewave, which achieves a sandpaper appear-
the audiofrequency
ance in this digital ’scope capture. waveform
(20Hz ~ 5kHz)

impedance at signal frequencies is small compared with

what it is bypassing. You will see much more of this in
later chapters. t

B. Power-supply filtering the modulated

radiofrequency carrier
We saw this application in §1.6.3, to filter the ripple from (~1MHz)
rectifier circuits. Although circuit designers often call them
Figure 1.113. An AM signal consists of an RF carrier (∼1 MHz)
filter capacitors, this is really a form of bypassing, or en-
whose amplitude is varied by the audio-frequency signal (speech
ergy storage, with large-value capacitors; we prefer the
or music; audible frequencies up to ∼5 kHz). The audio waveform
term storage capacitor. And these capacitors really are is dc offset so that the envelope does not cross zero.
large – they’re the big shiny round things you see in-
side most electronic instruments. We’ll get into dc power-
supply design in detail in Chapter 9. At the receiver end (that’s us!) the task is to select this
station (among many) and somehow extract the modulating
C. Timing and waveform generation envelope, which is the desired audio signal. Figure 1.114
As we’ve seen, a capacitor supplied with a constant current shows the simplest AM radio; it is the “crystal set” of
charges up with a ramp waveform. This is the basis of ramp yesteryear. It’s really quite straightforward: the parallel LC
and sawtooth generators, used in analog function gener- resonant circuit is tuned to the station’s frequency by the
ators, oscilloscope sweep circuits, analog–digital convert- variable capacitor C1 (§1.7.14); the diode D is a half-wave
ers, and timing circuits. RC circuits are also used for tim- rectifier (§1.6.2), which (if ideal) would pass only the pos-
ing, and they form the basis of delay circuits (monostable itive half-cycles of the modulated carrier; and R1 provides
multivibrators). These timing and waveform applications a light load, so that the rectified output follows the half-
are important in many areas of electronics and will be cov- cycles back down to zero. We’re almost done. We add small
ered in Chapters 3, 6, 10, and 11. capacitor C2 to prevent the output from following the fast
half-cycles of carrier (it’s a storage capacitor, §1.7.16B),
choosing the time constant R1C2 to be long compared with
1.7.17 Thévenin’s theorem generalized
a carrier period (∼1 μ s), but short compared with the pe-
When capacitors and inductors are included, Thévenin’s riod of the highest audio frequency (∼200 μ s).
theorem must be restated: any two-terminal network of re- Figures 1.115 and 1.116 show what you see when you
sistors, capacitors, inductors, and signal sources is equiva- probe around with a ’scope. The bare antenna shows plenty
lent to a single complex impedance in series with a single of low-frequency pickup (mostly 60 Hz ac powerline), and
signal source. As before, you find the (complex) impedance a tiny bit of signal from all the AM stations at once. But
56 1.9. Other passive components Art of Electronics Third Edition

antenna point “X”

audio
amplifier y
onl
R1
X D Y point “Y” R1+ C
2

point “X”
C1 R1 C2
L

speaker

Figure 1.116. Observed waveforms at point “Y” with R1 only (top)

Figure 1.114. The simplest AM receiver. Variable capacitor C1
and with smoothing capacitor C2 included (bottom). The upper pair
tunes the desired station, diode D picks off the positive envelope
is a single-shot capture (with the ∼1 MHz carrier appearing as
(smoothed by R1C2 ), and the resulting weak audio signal is ampli-
solid black), and the lower pair is a separate single-shot capture, in
fied to drive the loudspeaker, loudly.
which we have offset the rectified wave for clarity. Vertical: 1 V/div;
horizontal: 1 ms/div.
when you connect it to the LC resonant circuit, all the low-
frequency stuff disappears (because the LC looks like a
And one amusing final note: in our class, we like to
very low impedance, Figure 1.107) and it sees only the se-
show the effect of probing point “X” with a length of BNC
lected AM station. What’s interesting here is that the am-
(bayonet Neill–Concelman) cable going to a ’scope input
plitude of the selected station is much larger with the LC
(that’s how we start out, in the first week). When we do
attached than with nothing connected to the antenna: that’s
that, the cable’s capacitance (about 30 pF/ft) adds to C1 ,
because the resonant circuit’s high Q is storing energy from
lowering the resonant frequency and so tuning to a differ-
multiple cycles of the signal.49
ent station. If we choose right, it changes languages (from
English to Spanish)! The students howl with laughter – a
language-translating electronic component. Then we use
antenna an ordinary ’scope probe, with its ∼10 pF of capacitance:
(open-ckt) no change of station, nor of language.

1.9 Other passive components

antenna In the following subsections we would like to introduce
(parallel LC
to ground) briefly an assortment of miscellaneous but essential com-
ponents. If you are experienced in electronic construction,
you may wish to proceed to the next chapter.
Figure 1.115. Observed waveforms at point “X” from the bare an-
1.9.1 Electromechanical devices: switches
tenna (top) and with the LC connected. Note that the low-frequency
junk disappears and that the radio signal gets larger. These are These mundane but important devices seem to wind up in
single-shot traces, in which the ∼1 MHz radiofrequency carrier ap- most electronic equipment. It is worth spending a few para-
pears as a solid filled area. Vertical: 1 V/div; horizontal: 4 ms/div. graphs on the subject (and there’s more in Chapter 1x). Fig-
ures 1.117 and 1.118 show some common switch types.
The audio amplifier is fun, too, but we’re not ready for
it. We’ll see how to make one of those in Chapter 2 (with
A. Toggle switches
discrete transistors), and again in Chapter 4 (with opera-
The simple toggle switch is available in various configu-
tional amplifiers, the Lego™ block of analog design).
rations, depending on the number of poles; Figure 1.119
49 There are more complicated ways of framing this, but you don’t really shows the usual ones (SPST indicates a single-pole single-
want to know just yet. . . throw switch, SPDT indicates a single-pole double-throw
Figure 1.117. Switch Smorgasbord. The nine switches at right are momentary-contact (“pushbutton”) switches, including both panel-
mounting and PCB-mounting types (PCB, printed-circuit board). To their left are additional types, including lever-actuated and multipole
styles. Above them are a pair of panel-mounting binary-coded thumbwheel switches, to the left of which is a matrix-encoded hexadecimal
keypad. The switches at center foreground are toggle switches, in both panel-mounting and PCB-mounting varieties; several actuator
styles are shown, including a locking variety (fourth from front) that must be pulled before it will switch. The rotary switches in the left
column illustrate binary-coded types (the three in front and the larger square one), and the traditional multipole–multiposition configurable
wafer switches.

Figure 1.118. Board-mounted “DIP switches.” Left group, front to back and left to right (all are SPST): single station side-action toggle;
three-station side-action, two-station rocker, and single-station slide; eight-station slide (low-profile) and six-station rocker; eight-station
slide and rocker. Middle group (all are hexadecimal coded): six-pin low-profile, six-pin with top or side adjust; 16-pin with true and com-
plement coding. Right group: 2 mm×2 mm surface-mount header block with movable jumper (“shunt”), 0.1 ×0.1 (2.54 mm×2.54 mm)
through-hole header block with shunts; 18-pin SPDT (common actuator); eight-pin dual SPDT slide and rocker; 16-pin quad SPDT slide
(two examples).

57
58 1.9. Other passive components Art of Electronics Third Edition

switch, and DPDT indicates a double-pole double-throw codes its position as a 4-bit binary quantity, thereby saving
switch). Toggle switches are also available with “center lots of wires (only five are needed: the four bits, and a com-
OFF” positions and with up to four poles switched simulta- mon line). An alternative is the use of a rotary encoder, an
neously. Toggle switches are always “break before make,” electromechanical panel-mounting device that creates a se-
e.g., the moving contact never connects to both terminals quence of N pulse pairs for each full rotation of the knob.
in an SPDT switch. These come in two flavors (internally using either mechan-
ical contacts or electro-optical methods), and typically pro-
vide from 16 to 200 pulse pairs per revolution. The optical
varieties cost more, but they last forever.
SPST
SPDT
D. PC-mounting switches
It’s common to see little arrays of switches on printed-
DPDT circuit (PC) boards, like the ones shown in Figure 1.118.
Figure 1.119. Fundamental switch types.
They’re often called DIP switches, referring to the inte-
grated circuit dual in-line package that they borrow, though
contemporary practice increasingly uses the more compact
NC surface-mount technology (SMT) package. As the photo-
C
graph illustrates, you can get coded rotary switches; and
NO
because these are used for set-and-forget internal settings,
form A, NO form B, NC form C, SPDT you can substitute a multipin header block, with little slide-
on “shunts” to make the connections.
Figure 1.120. Momentary-contact (pushbutton) switches.

E. Other switch types

In addition to these basic switch types, there are available
B. Pushbutton switches various exotic switches such as Hall-effect switches, reed
Pushbutton switches are useful for momentary-contact ap- switches, proximity switches, etc. All switches carry max-
plications; they are drawn schematically as shown in Fig- imum current and voltage ratings; a small toggle switch
ure 1.120 (NO and NC mean normally open and normally might be rated at 150 volts and 5 amps. Operation with
closed). For SPDT momentary-contact switches, the ter- inductive loads drastically reduces switch life because of
minals must be labeled NO and NC, whereas for SPST arcing during turn-off. It’s always OK to operate a switch
types the symbol is self-explanatory. Momentary-contact below its maximum ratings, with one notable exception:
switches are always “break before make.” In the electrical since many switches rely on substantial current flow to
(as opposed to electronic) industry, the terms form A, form clean away contact oxides, it’s important to use a switch
B, and form C are used to mean SPST (NO), SPST (NC), that is designed for “dry switching” when switching low-
and SPDT, respectively. level signals;50 otherwise you’ll get noisy and intermittent
operation (see Chapter 1x).
C. Rotary switches
Rotary switches are available with many poles and many F. Switch examples
positions, often as kits with individual wafers and shaft As an example of what can be done with simple switches,
hardware. Both shorting (make-before-break) and non- let’s consider the following problem: suppose you want to
shorting (break-before-make) types are available, and they sound a warning buzzer if the driver of a car is seated and
can be mixed on the same switch. In many applications the one of the car doors is open. Both doors and the driver’s
shorting type is useful to prevent an open circuit between seat have switches, all normally open. Figure 1.121 shows
switch positions, because circuits can go amok with un- a circuit that does what you want. If one OR the other
connected inputs. Nonshorting types are necessary if the door is open (switch closed) AND the seat switch is closed,
separate lines being switched to one common line must not the buzzer sounds. The words OR and AND are used in a
ever be connected to each other. logic sense here, and we will see this example again in
Sometimes you don’t really want all those poles, you
just want to know how many clicks (detents) the shaft has
50
been turned. For that a common form of rotary switch en- These use gold contact plating.
Art of Electronics Third Edition 1.9.3. Connectors 59

+12V excitation, and coil voltages from 3 volts up to 115 volts

(ac or dc) are common. “Mercury-wetted” and “reed” re-
buzzer lays are intended for high-speed (∼1 ms) applications, and
giant relays intended to switch thousands of amps are used
by power companies.
seat
The solid-state relay (SSR) – consisting of a semicon-
ductor electronic switch that is turned on by a LED – pro-
vides better performance and reliability than mechanical
left right relays, though at greater cost. SSRs operate rapidly, with-
door door
out contact “bounce,” and usually provide for smart switch-
ing of ac power (they turn on at the moment of zero volt-
Figure 1.121. Switch circuit example: open door warning.
age, and they turn off at the moment of zero current). Much
more on these useful devices in Chapter 12.
Chapters 2, 3, and 10 when we talk about transistors and As we’ll learn, electrically controlled switching of sig-
digital logic. nals within a circuit can be accomplished with transistor
Figure 1.122 shows a classic switch circuit used to turn switches, without having to use relays of any sort (Chap-
a ceiling lamp on or off from a switch at either of two en- ters 2 and 3). The primary uses of relays are in remote
trances to a room. switching and high-voltage (or high-current) switching,
where it is important to have complete electrical isolation
between the control signal and the circuit being switched.

1.9.3 Connectors
Bringing signals in and out of an instrument, routing signal
and dc power around between the various parts of an in-
strument, providing flexibility by permitting circuit boards
and larger modules of the instrument to be unplugged (and
115Vac replaced) – these are the functions of the connector, an es-
sential ingredient (and usually the most unreliable part) of
Figure 1.122. Electrician’s “three-way” switch wiring.
any piece of electronic equipment. Connectors come in a
bewildering variety of sizes and shapes.52 Figures 1.123,
Exercise 1.36. Although few electronic circuit designers know 1.124, and 1.125 give some idea of the variety.
how, every electrician can wire up a light fixture so that any of
N switches can turn it on or off. See if you can figure out this A. Single-wire connectors
generalization of Figure 1.122. It requires two SPDT switches The simplest kind of connector is the simple pin jack or
and N−2 DPDT switches.
banana jack used on multimeters, power supplies, etc. It is
handy and inexpensive, but not as useful as the shielded-
1.9.2 Electromechanical devices: relays cable or multiwire connectors you often need. The humble
binding post is another form of single-wire connector, no-
Relays are electrically controlled switches. In the tradi-
table for the clumsiness it inspires in those who try to use
tional electromechanical relay, a coil pulls in an arma-
it.
ture (to close the contacts) when sufficient coil current
flows. Many varieties are available, including “latching”
B. Shielded-cable connectors
and “stepping” relays.51 Relays are available with dc or ac
To prevent capacitive pickup, and for other reasons we’ll go
51 In an amusing historical footnote, the stepping relay used for a century into in Appendix H, it is usually desirable to pipe signals
as the cornerstone of telephone exchanges (the “Strowger selector”) around from one instrument to another in shielded coaxial
was invented by a Topeka undertaker, Almon Strowger, evidently be- cable. The most popular connector is the BNC type that
cause he suspected that telephone calls intended for his business were
being routed (by the switchboard operators in his town) to a funeral 52 A search for “connector” on the DigiKey website returns 116 cate-
home competitor. gories, with approximately 43,000 individual varieties in stock.
60 1.9. Other passive components Art of Electronics Third Edition

Figure 1.123. Rectangular connectors. The variety of available multipin connectors is staggering. Here is a collection of common speci-
mens: the five connectors at lower left are multipin nylon power connectors (sometimes called Molex-type for historical reasons). Above
them are four dual-row box headers (0.1 spacing, shown with and without latch ejectors, and also with Wire-Wrap® and right-angle
tails), and to their right an open (“unshrouded”) 0.1 dual-row header, along with a pair of dual-row headers of finer pitch (2 mm and
1.27 mm). These dual-row male connectors mate with insulation displacement connectors (IDC) such as the one shown attached to a
short length of ribbon cable (just above the unshrouded header). Just below the ribbon are shown single-row 0.1 headers, with mating
shells (AMP MODU) that accept individual wire leads. At bottom right are several terminal blocks used for power wiring, and four “Faston”-
type crimpable spade lugs. Above them are USB connectors, and to their left are the common RJ-45 and RJ-11 modular telephone/data
connectors. The popular and reliable D-subminiature connectors are at center, including (right to left) a pair of 50-pin micro-D (cable plug,
PCB socket), the 9-pin D-sub, 26-pin high-density, and a pair of 25-pin D-subs (one IDC). Above them are (right to left) a 96-pin VME
backplane connector, a 62-pin card-edge connector with solder tails, a “Centronics-type” connector with latching bail, and a card-edge
connector with ribbon IDC. At top left is a miscellany – a mating pair of “GR-type” dual banana connectors, a mating pair of Cinch-type
connectors, a mating pair of shrouded Winchester-type connectors with locking jackscrews, and (to their right) a screw-terminal barrier
block. Not shown here are the really tiny connectors used in small portable electronics (smartphones, cameras, etc); you can see a fine
example in Figure 1.131.

adorns most instrument front panels. It connects with a inner (signal) conductor mates before the shield (ground)
quarter-turn twist and completes both the shield (ground) when you plug it in; furthermore, the design of the con-
circuit and inner conductor (signal) circuit simultaneously. nector is such that both shield and center conductor tend to
Like all connectors used to mate a cable to an instrument, make poor contact. You’ve undoubtedly heard the results!
it comes in both panel-mounting and cable-terminating va- Not to be outdone, the television industry has responded
rieties. with its own bad standard, the type-F coax “connector,”
Among the other connectors for use with coaxial ca- which uses the unsupported inner wire of the coax as the
ble are the TNC (“threaded Neill–Concelman,” a close pin of the male plug, and a shoddy arrangement to mate the
cousin of the BNC, but with threaded outer shell), the high- shield.53
performance but bulky type N, the miniature SMA and We hereby induct these losers into the Electronic
SMB, the subminiature LEMO and SMC, and the high-
voltage MHV and SHV. The so-called phono jack used in 53 Advocates of each would probably reply “This is our most modestly
audio equipment is a nice lesson in bad design, because the priced receptacle.”
Art of Electronics Third Edition 1.9.4. Indicators 61

Figure 1.124. Circular connectors. A selection of multipin and other “non-RF” connectors; the panel-mounting receptacle is shown to the
left of each cable-mounted plug. Top row, left to right: “MS”-type (MIL-C-5015) rugged connector (available in hundreds of configurations),
high-current (50 A) “Supericon,” multipin locking XLR. Middle row: weatherproof (Switchcraft EN3), 12 mm video (Hirose RM), circular
DIN, circular mini-DIN, 4-pin microphone connector. Bottom row: locking 6-pin (Lemo), microminiature 7-pin shielded (Microtech EP-7S),
miniature 2-pin shrouded (Litton SM), 2.5 mm power, banana, pin jack.

Components Hall of Infamy, some charter members of and they come with different lug styles according to the
which are shown in Figure 1.126. method of connection. You can solder them to a “moth-
erboard” or “backplane,” which is itself just another PCB
C. Multipin connectors containing the interconnecting wiring between the individ-
Very frequently electronic instruments demand multiwire ual circuit cards. Alternatively, you may want to use edge
cables and connectors. There are literally dozens of dif- connectors with standard solder-lug terminations, particu-
ferent kinds. The simplest example is a three-wire “IEC” larly in a system with only a few cards. A more reliable
powerline cord connector. Among the more popular are (though more costly) solution is the use of “two-part” PCB
the excellent type-D subminiature, the Winchester MRA connectors, in which one part (soldered onto the board)
series, the venerable MS type, and the flat ribbon-cable mates with the other part (on a backplane, etc); an example
mass-termination connectors. These and others are shown is the widely used VME (VersaModule Eurocard) connec-
in Figure 1.123. tor (upper right-hand corner of Figure 1.123).
Beware of connectors that can’t tolerate being dropped
on the floor (the miniature hexagon connectors are classic)
or that don’t provide a secure locking mechanism (e.g., the 1.9.4 Indicators
Jones 300 series). A. Meters
To read out the value of some voltage or current, you have
D. Card-edge connectors a choice between the time-honored moving-pointer type of
The most common method used to make connection to meter and digital-readout meters. The latter are more ex-
printed-circuit cards is the card-edge connector, which pensive and more accurate. Both types are available in a
mates to a row of gold-plated contacts at the edge of the variety of voltage and current ranges. There are, in addi-
card; common examples are the motherboard connectors tion, exotic panel meters that read out such things as VUs
that accept plug-in computer memory modules. Card-edge (volume units, an audio dB scale), expanded-scale ac volts
connectors may have from 15 to 100 or more connections, (e.g., 105 to 130 V), temperature (from a thermocouple),
62 1.9. Other passive components Art of Electronics Third Edition

Figure 1.125. RF and shielded connectors. The panel-mounting receptacle is shown to the left of each cable-mounted plug. Top row, left
to right: stereo phone jack, audio “XLR” type; N and UHF (RF connectors). Second row down: BNC, TNC, type F; MHV and SHV (high
voltage). Third row down: 2.5 mm (3/32 ) audio, 3.5 mm stereo, improved 3.5 mm stereo, phono (“RCA type”), LEMO coaxial. Bottom
row: SMA (panel jack, flexible coax plug), SMA (board-mount jack, rigid coax plug), SMB; SC and ST (optical fiber).

percentage motor load, frequency, etc. Digital panel meters placed with LEDs. The latter behave electrically like ordi-
often provide the option of logic-level outputs, in addition nary diodes, but with a forward voltage drop in the range
to the visible display, for internal use by the instrument. of 1.5 to 2 volts (for red, orange, and some green LEDs;
As a substitute for a dedicated meter (whether analog 3.6 V for blue54 and high-brightness green; see Figure 2.8).
or digital), you increasingly see an LCD (liquid-crystal When current flows in the forward direction, they light up.
display) or LED panel with a meter-like pattern. This is Typically, 2 mA to 10 mA produces adequate brightness.
flexible and efficient: with a graphic LCD display module LEDs are cheaper than incandescent lamps, they last pretty
(§12.5.3) you can offer the user a choice of “meters,” ac- much forever, and they come in four standard colors as well
cording to the quantity being displayed, all under the con- as “white” (which is usually a blue LED with a yellow fluo-
trol of an embedded controller (a built-in microprocessor; rescent coating). They come in convenient panel-mounting
see Chapter 15). packages; some even provide built-in current limiting.55
LEDs can also be used for digital displays, for example

B. Lamps, LEDs, and displays 54 The invention of the gallium nitride blue LED was the breakthrough
Flashing lights, screens full of numbers and letters, eerie product of a lone and unappreciated employee of Nichia Chemical In-
sounds – these are the stuff of science fiction movies, and dustries, Shuji Nakamura.
except for the last, they form the subject of lamps and dis- 55 And of course, for both residential and commercial area lighting, LEDs
plays (see §12.5.3). Small incandescent lamps used to be have now largely relegated to the dustbin of history the century-old
standard for front-panel indicators, but they have been re- hot-filament incandescent lamp.
Art of Electronics Third Edition 1.9.5. Variable components 63

For use inside an instrument, rather than on the front

panel, trimmer pots come in single-turn and multiturn
styles, most intended for printed-circuit mounting. These
are handy for calibration adjustments of the “set-and-
forget” type. Good advice: resist the temptation to use lots
of trimmers in your circuits. Use good design instead.

CCW CW

Figure 1.127. Potentiometer (three-terminal variable resistor).

Figure 1.126. Components to avoid. We advise against using

The symbol for a variable resistor, or pot, is shown in
components like these, if you have a choice (see text if you need Figure 1.127. Sometimes the symbols CW and CCW are
convincing!). Top row, left to right: low-value wirewound pot, type used to indicate the clockwise and counterclockwise ends.
UHF connector, electrical tape (“just say no!”). Middle row: “cinch- An all-electronic version of a potentiometer can be
type” connectors, microphone connector, hexagon connectors. made with an array of electronic (transistor) switches that
Bottom row: slide switch, cheap IC socket (not “screw-machined”), select a tap in a long chain of fixed resistors. As awk-
type-F connector, open-element trimmer pot, phono connector. ward as that may sound, it is a perfectly workable scheme
when implemented as an IC. For example, Analog Devices,
as 7-segment numeric displays or (for displaying letters as Maxim/Dallas Semiconductor, and Xicor make a series of
well as numbers – “alphanumeric”) 16-segment displays or “digital potentiometers” with up to 1024 steps; they come
dot-matrix displays. However, if more than a few digits or as single or dual units, and some of them are “nonvolatile,”
characters need to be displayed, LCDs are generally pre- meaning that they remember their last setting even if power
ferred. These come in line-oriented arrays (e.g., 16 charac- has been turned off. These find application in consumer
ters by 1 line, up to 40 characters by 4 lines), with a sim- electronics (televisions, stereos) where you want to adjust
ple interface that permits sequential or addressable entry the volume from your infrared remote control, rather than
of alphanumeric characters and additional symbols. They by turning a knob; see §3.4.3E.
are inexpensive, low power, and visible even in sunlight. One important point about variable resistors: don’t at-
Back-lighted versions work well even in subdued light, but tempt to use a potentiometer as a substitute for a precise
are not low power. Much more on these (and other) opto- resistor value somewhere within a circuit. This is tempting,
electronic devices in §12.5. because you can trim the resistance to the value you want.
The trouble is that potentiometers are not as stable as good
(1%) resistors, and in addition they may not have good res-
1.9.5 Variable components
olution (i.e., they can’t be set to a precise value). If you
A. Variable resistors must have a precise and settable resistor value somewhere,
Variable resistors (also called volume controls, poten- use a combination of a 1% (or better) precision resistor and
tiometers, pots, or trimmers) are useful as panel controls a potentiometer, with the fixed resistor contributing most
or internal adjustments in circuits. A classic panel type is of the resistance. For example, if you need a 23.4k resistor,
the 2-watt-type AB potentiometer; it uses the same basic use a 22.6k 1% fixed resistor (a standard value) in series
material as the fixed carbon-composition resistor, with a with a 2k trimmer pot. Another possibility is to use a series
rotatable “wiper” contact. Other panel types are available combination of several precision resistors, selecting the last
with ceramic or plastic resistance elements, with improved (and smallest) resistor to give the desired series resistance.
characteristics. Multiturn types (3, 5, or 10 turns) are avail- As we’ll see later (§3.2.7), it is possible to use FETs as
able, with counting dials, for improved resolution and lin- voltage-controlled variable resistors in some applications.
earity. “Ganged” pots (several independent sections on one Another possibility is an “optophotoresistor” (§12.7).
shaft) are also manufactured, although in limited variety, Transistors can be used as variable-gain amplifiers, again
for applications that demand them. Figure 1.8 shows a rep- controlled by a voltage. Keep an open mind when design
resentative selection of pots and trimmers. brainstorming.
64 1.10. A parting shot: confusing markings and itty-bitty components Art of Electronics Third Edition

Figure 1.128. Variable capacitor.

B. Variable capacitors
Variable capacitors are primarily confined to the smaller
capacitance values (up to about 1000 pF) and are com-
monly used in RF circuits. Trimmers are available for in-
circuit adjustments, in addition to the panel type for user
tuning. Figure 1.128 shows the symbol for a variable ca-
pacitor.
Diodes operated with applied reverse voltage can be
used as voltage-variable capacitors; in this application
they’re called varactors, or sometimes varicaps or epi-
caps. They’re very important in RF applications, especially
Figure 1.129. A powerline variable transformer (“Variac”) lets you
phase-locked loops, automatic frequency control (AFC), adjust the ac input voltage to something you are testing. Here a 5 A
modulators, and parametric amplifiers. unit is shown, both clothed and undressed.

C. Variable inductors 1.10 A parting shot: confusing markings and

Variable inductors are usually made by arranging to move itty-bitty components
a piece of core material in a fixed coil. In this form they’re
In our electronics course,57 and indeed in day-to-day elec-
available with inductances ranging from microhenrys to
tronics on the bench, we encounter a wonderful confusion
henrys, typically with a 2:1 tuning range for any given in-
of component markings. Capacitors in particular are just,
ductor. Also available are rotary inductors (coreless coils
well, perverse: they rarely bother specifying units (even
with a rolling contact).56
though they span 12 orders of magnitude, picofarads to
farads), and for ceramic SMT varieties they dispense with
D. Variable transformers any markings whatsoever! Even worse, they are still caught
Variable transformers are handy devices, especially the up in the transition from printing the value as an inte-
ones operated from the 115 volt ac line. They’re usually ger (e.g., “470” meaning 470 pF) versus using exponent
configured as “autotransformers,” which means that they notation (e.g., “470” meaning 47×100 , i.e., 47 pF). Fig-
have only one winding, with a sliding contact. They’re also ure 1.130 shows exactly that case! Another trap for the un-
commonly called Variacs (the name given to them by Gen- wary (and sometimes the wary, as well) is the date-code
eral Radio), and they are made by Technipower, Superior gotcha: the 4-digit code (yydd) can masquerade as a part
Electric, and others. Figure 1.129 shows a classic unit from number, as in the four examples in the photo. And, as com-
General Radio. Typically they provide 0 to 135 volts ac ponents become smaller and smaller, there’s precious lit-
output when operated from 115 volts, and they come in cur- tle room for all but the briefest of markings; so, following
rent ratings from 1 amp to 20 amps or more. They’re good the pharmaceutical industry, manufacturers invent a short
for testing instruments that seem to be affected by power-
line variations, and in any case to verify worst-case perfor- 57 Physics 123 (“Laboratory Electronics”) at Harvard University: “Half
mance. Important Warning: don’t forget that the output is course (fall term; repeated spring term). A lab-intensive introduction
not electrically isolated from the powerline, as it would be to electronic circuit design. Develops circuit intuition and debugging
with a transformer! skills through daily hands-on lab exercises, each preceded by class
discussion, with minimal use of mathematics and physics. Moves
quickly from passive circuits, to discrete transistors, then concen-
56 An interesting form of variable inductor of yesteryear was the variome- trates on operational amplifiers, used to make a variety of circuits
ter, a rotatable coil positioned within a fixed outer coil and connected including integrators, oscillators, regulators, and filters. The digi-
in series with it. As the inner coil was rotated, the total inductance tal half of the course treats analog–digital interfacing, emphasiz-
went from maximum (four times the inductance of either coil alone) ing the use of microcontrollers and programmable logic devices
all the way down to zero. These things were consumer items, listed for (PLDs).” See http://webdocs.registrar.fas.harvard.edu/
example in the 1925 Sears Roebuck catalog. courses/Physics.html.
Art of Electronics Third Edition 1.10.1. Surface-mount technology: the joy and the pain 65

Figure 1.131. We’re “all thumbs” when working with surface-

mount technology (SMT). This is a corner of a cellphone circuit
Figure 1.130. Confusion Central! The three ICs are each marked board, showing small ceramic resistors and capacitors, integrated
with both a part number (e.g., UA7812) and a “date code” (e.g., circuits with ball-grid connecting dots on their undersides, and the
UC7924, signifying the 24th week of 1979). Unfortunately, both are Lilliputian connectors for the antenna and display panel. See also
perfectly valid part numbers (a +12 V or a −24 V regulator). The Figure 4.84.
resistor pair (actually two views of identically marked resistors) suf-
fers from the same problem: it could be 7.32 KΩ±1%, or it could be
85.0 kΩ±5% (it’s the former, but who would know?). The pair of ce-
ramic capacitors are both marked 470K (470,000 of something?),
but, surprise, the “K” means 10% tolerance; and, bigger surprise,
the square cap is 47 pF, the round one is 470 pF. And what is one to
make of a black box labeled 80K000 (pronounced “eighty-koooh”),
or a diode with two cathodes (and no anode?), or a resistor with a
single black band in the center?

alphanumeric code for each component. And that’s all you

get. For example, National’s LMV981 op-amp comes in
several 6-pin packages: the SOT23 is marked “A78A,” the
smaller SC70 says “A77,” and the really tiny microSMD
blurts out a single letter “A” (or “H” if it’s free of lead).
Not much to go on.
Figure 1.132. How small can these things get?! The “01005”-size
SMT (0.016 ×0.008 , or 0.4mm×0.2mm) represents the industry’s
1.10.1 Surface-mount technology: the joy and the greatest insult to the experimenter.
pain
While we’re complaining, let’s whine just a bit about the a sense of the scale of these little components, and Fig-
difficulty of prototyping circuits with tiny surface-mount ure 1.132 displays the true horror of the tiniest of these –
components. From an electrical point of view they are ex- the “01005”-size chip components (0402 metric) that mea-
cellent: low inductance, and compact. But they are nearly sure 200μ m×400μ m: not much thicker than a human hair,
impossible to wire up in prototype breadboard fashion, in and indistinguishable from dust!
the way that was easy with “through-hole” (or “leaded” – Sometimes you can use little adapter carriers (from
pronounced lee -ded) components, such as resistors with companies like Bellin Dynamic Systems, Capital Ad-
axial leads (a wire sticking out each end), or integrated vanced Technologies, or Aries) to convert an SMT inte-
circuits in DIP (dual in-line) cases. Figure 1.131 gives grated circuit to a fake DIP. But the densest surface-mount
66 Additional Exercises for Chapter 1 Art of Electronics Third Edition

Figure 1.133. A taste of the world of passive components in surface-mount packages: connectors, switches, trimmer pots, inductors,
resistors, capacitors, crystals, fuses. . . . If you can name it, you can probably get it in SMT.

packages have no leads at all, just an array of bumps (up down at 10 kHz). Use the same source and load impedances as in
to several thousand!) on the underside; and these require Exercise 1.39.
serious “reflow” equipment before you can do anything
with them. Sadly, we cannot ignore this disturbing trend, 10k
because the majority of new components are offered only
in surface-mount packages. Woe to the lone basement +
10V battery 10k
experimenter–inventor! Figure 1.133 give a sense of the –
variety of passive component types that come in surface-
mount configurations.
Figure 1.134. Example for Norton equivalent circuit.

Additional Exercises for Chapter 1

+
Exercise 1.37. Find the Norton equivalent circuit (a current
source in parallel with a resistor) for the voltage divider in Fig- 0.5mA

ure 1.134. Show that the Norton equivalent gives the same output
voltage as the actual circuit when loaded by a 5k resistor. 10k

Exercise 1.38. Find the Thévenin equivalent for the circuit shown
in Figure 1.135. Is it the same as the Thévenin equivalent for Ex- 10k
ercise 1.37?

Exercise 1.39. Design a “rumble filter” for audio. It should pass

frequencies greater than 20 Hz (set the −3 dB point at 10 Hz). Figure 1.135. Example for Thévenin equivalent circuit.
Assume zero source impedance (perfect voltage source) and 10k
(minimum) load impedance (that’s important so that you can
choose R and C such that the load doesn’t affect the filter op- Exercise 1.41. How would you make a filter with R’s and C’s to
eration significantly). give the response shown in Figure 1.136?

Exercise 1.40. Design a “scratch filter” for audio signals (3 dB Exercise 1.42. Design a bandpass RC filter (as in Figure 1.137);
Art of Electronics Third Edition Additional Exercises for Chapter 1 67

1.0 6.3V ac (rms)

Vout 1k
Vin Vout
0.5

117V ac
0
0 ω0 ω

Figure 1.136. High-emphasis filter response.

Figure 1.138. Circuit for Exercise 1.43.
f1 and f2 are the 3 dB points. Choose impedances so that the first
stage isn’t much affected by the loading of the second stage. probe vertical
tip amplifier
(Zin = ∞)
Exercise 1.43. Sketch the output for the circuit shown in Fig-
ure 1.138.

1.0 100pF
20pF 1.0MΩ
Vout (cable)

Vin

0
0 ω1 ω2 ω scope input

Figure 1.137. Bandpass filter response. Figure 1.139. Oscilloscope ×10 probe.

Exercise 1.44. Design an oscilloscope “×10 probe” to use with a 20 dB (×10 voltage division ratio) attenuation at all frequencies,
scope whose input impedance is 1 MΩ in parallel with 20 pF by including dc. The reason for using a ×10 probe is to increase
figuring out what goes inside the probe handle in Figure 1.139. the load impedance seen by the circuit under test, which reduces
Assume that the probe cable adds an additional 100 pF and that loading effects. What input impedance (R in parallel with C) does
the probe components are placed at the tip end (rather than at your ×10 probe present to the circuit under test when used with
the scope end) of the cable. The resultant network should have the scope?
68 Review of Chapter 1 Art of Electronics Third Edition

Review of Chapter 1 power; 6 dB is a doubling of amplitude (or quadrupling

of power). Decibel measure is also used to specify ampli-
An A-to-H summary of what we have learned in Chapter 1.
tude (or power) directly, by giving a reference level: for
This summary reviews basic principles and facts in Chap-
example, −30 dBm (dB relative to 1 mW) is 1 microwatt;
ter 1, but it does not cover application circuit diagrams and
+3 dBVrms is a signal of 1.4 V rms amplitude (2 Vpeak,
practical engineering advice presented there.
4 Vpp).
Other important waveforms are square waves, triangle
¶ A. Voltage and Current. waves, ramps, noise, and a host of modulation schemes by
Electronic circuits consist of components connected to- which a simple “carrier” wave is varied in order to convey
gether with wires. Current (I) is the rate of flow of charge information; some examples are AM and FM for analog
through some point in these connections; it’s measured in communication, and PPM (pulse-position modulation) or
amperes (or milliamps, microamps, etc.). Voltage (V ) be- QAM (quadrature-amplitude modulation) for digital com-
tween two points in a circuit can be viewed as an applied munication.
driving “force” that causes currents to flow between them;
voltage is measured in volts (or kilovolts, millivolts, etc.); ¶ C. The Relationship Between Current and Voltage.
see §1.2.1. Voltages and currents can be steady (dc), or This chapter concentrated on the fundamental, essential,
varying. The latter may be as simple as the sinusoidal alter- and ubiquitous two-terminal linear devices: resistors, ca-
nating voltage (ac) from the wallplug, or as complex as a pacitors, and inductors. (Subsequent chapters deal with
high-frequency modulated communications waveform, in transistors – three-terminal devices in which a signal ap-
which case it’s usually called a signal (see ¶B below). The plied to one terminal controls the current flow through
algebraic sum of currents at a point in a circuit (a node) is the other pair – and their many interesting applications.
zero (Kirchhoff’s current law, KCL, a consequence of con- These include amplification, filtering, power conversion,
servation of charge), and the sum of voltage drops going switching, and the like.) The simplest linear device is the
around a closed loop in a circuit is zero (Kirchhoff’s volt- resistor, for which I=V /R (Ohm’s Law, see §1.2.2A).
age law, KVL, a consequence of the conservative nature of The term “linear” means that the response (e.g., current)
the electrostatic field). to a combined sum of inputs (i.e., voltages) is equal to
the sum of the responses that each input would produce:
¶ B. Signal Types and Amplitude. I(V1 +V2 )=I(V1 ) + I(V2 ).
See §1.3. In digital electronics we deal with pulses, which
are signals that bounce around between two voltages (e.g., ¶ D. Resistors, Capacitors, and Inductors.
+5 V and ground); in the analog world it’s sinewaves that The resistor is clearly linear. But it is not the only lin-
win the popularity contest. In either case, a periodic sig- ear two-terminal component, because linearity does not
nal is characterized by its frequency f (units of Hz, MHz, require I ∝ V . The other two linear components are ca-
etc.) or, equivalently, period T (units of ms, μ s, etc.). For pacitors (§1.4.1) and inductors (§1.5.1), for which there
sinewaves it’s often more convenient to use angular fre- is a time-dependent relationship between voltage and cur-
quency (radians/s), given by ω =2π f . rent: I=C dV/dt and V =L dI/dt, respectively. These are
Digital amplitudes are specified simply by the HIGH and the time domain descriptions. Thinking instead in the fre-
LOW voltage levels. With sinewaves the situation is more quency domain, these components are described by their
complicated: the amplitude of a signal V (t)=V0 sin ω t can impedances, the ratio of voltage to current (as a function
be given as (a) peak amplitude (or just “amplitude”)
√ V0 , of frequency) when driven with a sinewave (§1.7). A lin-
(b) root-mean-square (rms) amplitude Vrms =V0 / 2, or (c) ear device, when driven with a sinusoid, responds with a
peak-to-peak amplitude Vpp =2V0 . If unstated, a sinewave sinusoid of the same frequency, but with changed ampli-
amplitude is usually understood to be Vrms . A signal of rms tude and phase. Impedances are therefore complex, with
amplitude Vrms delivers power P=Vrms 2 /R
load to a resistive the real part representing the amplitude of the response
load (regardless of the signal’s waveform), which accounts that is in-phase, and the imaginary part representing the
for the popularity of rms amplitude measure. amplitude of the response that is in quadrature (90◦ out of
Ratios of signal amplitude (or power) are commonly ex- phase). Alternatively, in the polar representation of com-
pressed in decibels (dB), defined as dB = 10 log10 (P2 /P1 ) plex impedance (Z=|Z|eiθ ), the magnitude |Z| is the ra-
or 20 log10 (V2 /V1 ); see §1.3.2. An amplitude ratio of 10 tio of magnitudes (|Z|=|V |/|I|) and the quantity θ is the
(or power ratio of 100) is 20 dB; 3 dB is a doubling of phase shift between V and I. The impedances of the three
Art of Electronics Third Edition Review of Chapter 1 69

linear 2-terminal components are ZR =R, ZC = − j/ω C, and impedance at signal frequencies suppresses unwanted sig-
ZL = jω L, where (as always) ω =2π f ; see §1.7.5. Sinewave nals, e.g., on a dc supply rail; (b) blocking (§1.7.1C), in
current through a resistor is in phase with voltage, whereas which a highpass filter blocks dc, but passes all frequencies
for a capacitor it leads by 90◦ , and for an inductor it lags of interest (i.e., the breakpoint is chosen below all signal
by 90◦ . frequencies); (c) timing (§1.4.2D), in which an RC circuit
(or a constant current into a capacitor) generates a sloping
waveform used to create an oscillation or a timing interval;
¶ E. Series and Parallel.
and (d) energy storage (§1.7.16B), in which a capacitor’s
The impedance of components connected in series is
stored charge Q=CV smooths out the ripples in a dc power
the sum of their impedances; thus Rseries =R1 +R2 + · · ·,
supply.
Lseries =L1 +L2 + · · ·, and 1/Cseries =1/C1 +1/C2 + · · ·.
In later chapters we’ll see some additional applica-
When connected in parallel, on the other hand, it’s the
tions of capacitors: (e) peak detection and sample-and-hold
admittances (inverse of impedance) that add. Thus the
(§§4.5.1 and 4.5.2), which capture the voltage peak or tran-
formula for capacitors in parallel looks like the formula for
sient value of a waveform, and (f) the integrator (§4.2.6),
resistors in series, Cparallel =C1 +C2 + · · ·; and vice versa for
which performs a mathematical integration of an input sig-
resistors and inductors, thus 1/Rparallel =1/R1 +1/R2 + · · ·.
nal.
For a pair of resistors in parallel this reduces to
Rparallel =(R1 R2 )/(R1 +R2 ). For example, two resis-
¶ G. Loading; Thévenin Equivalent Circuit.
tors of value R have resistance R/2 when connected in
Connecting a load (e.g., a resistor) to the output of a circuit
parallel, or resistance 2R in series.
(a “signal source”) causes the unloaded output voltage to
The power dissipated in a resistor R is P=I 2 R=V 2/R.
drop; the amount of such loading depends on the load re-
There is no dissipation in an ideal capacitor or inductor,
sistance, and the signal source’s ability to drive it. The lat-
because the voltage and current are 90◦ out of phase. See
ter is usually expressed as the equivalent source impedance
§1.7.6.
(or Thévenin impedance) of the signal. That is, the signal
source is modeled as a perfect voltage source Vsig in series
¶ F. Basic Circuits with R, L, and C. with a resistor Rsig . The output of the resistive voltage di-
Resistors are everywhere. They can be used to set an op- vider driven from an input voltage Vin , for example, is mod-
erating current, as for example when powering an LED or eled as a voltage source Vsig =Vin R2 /(R1 +R2 ) in series with
biasing a zener diode (Figure 1.16); in such applications a resistance Rsig =R1 R2 /(R1 +R2 ) (which is just R1 R2 ). So
the current is simply I=(Vsupply −Vload )/R. In other appli- the output of a 1kΩ–1kΩ voltage divider driven by a 10 V
cations (e.g., as a transistor’s load resistor in an amplifier, battery looks like 5 V in series with 500 Ω.
Figure 3.29) it is the current that is known, and a resis- Any combination of voltage sources, current sources,
tor is used to convert it to a voltage. An important circuit and resistors can be modeled perfectly by a single volt-
fragment is the voltage divider (§1.2.3), whose unloaded age source in series with a single resistor (its “Thévenin
output voltage (across R2 ) is Vout =Vin R2 /(R1 +R2 ). equivalent circuit”), or by a single current source in paral-
If one of the resistors in a voltage divider is replaced lel with a single resistor (its “Norton equivalent circuit”);
with a capacitor, you get a simple filter: lowpass if the see Appendix D. The Thévenin equivalent source and re-
lower leg is a capacitor, highpass if the upper leg is a ca- sistance values are found from the open-circuit voltage and
pacitor (§§1.7.1 and 1.7.7). In either case the −3 dB tran- short-circuit current as VTh =Voc , RTh =Voc /Isc ; and for the
sition frequency is at f3dB =1/2π RC. The ultimate rolloff Norton equivalent they are IN =Isc , RN =Voc /Isc .
rate of such a “single-pole” lowpass filter is −6 dB/octave, Because a load impedance forms a voltage divider with
or −20 dB/decade; i.e., the signal amplitude falls as 1/ f the signal’s source impedance, it’s usually desirable for
well beyond f3dB . More complex filters can be created by the latter to be small compared with any anticipated load
combining inductors with capacitors, see Chapter 6. A ca- impedance (§1.2.5A). However, there are two exceptions:
pacitor in parallel with an inductor forms a resonant cir- (a) a current source has a high source impedance (ideally
cuit; its impedance (for ideal components)
√ goes to infinity infinite), and should drive a load of much lower impedance;
at the resonant frequency f =1/(2π LC). The impedance and (b) signals of high frequency (or fast risetime), trav-
of a series LC goes to zero at that same resonant frequency. eling through a length of cable, suffer reflections unless
See §1.7.14. the load impedance equals the so-called “characteristic
Other important capacitor applications in this chapter impedance” Z0 of the cable (commonly 50 Ω), see Ap-
(§1.7.16) include (a) bypassing, in which a capacitor’s low pendix H.
70 Review of Chapter 1 Art of Electronics Third Edition

¶ H. The Diode, a Nonlinear Component. reversal, as in Figure 1.84; and their exponential current
There are important two-terminal devices that are not lin- versus applied voltage can be used to fashion circuits with
ear, notably the diode (or rectifier), see §1.6. The ideal logarithmic response (§1.6.6E).
diode conducts in one direction only; it is a “one-way Diodes specify a maximum safe reverse voltage, be-
valve.” The onset of conduction in real diodes is roughly yond which avalanche breakdown (an abrupt rise of cur-
at 0.5 V in the “forward” direction, and there is some rent) occurs. You don’t go there! But you can (and should)
small leakage current in the “reverse” direction, see Fig- with a zener diode (§1.2.6A), for which a reverse break-
ure 1.55. Useful diode circuits include power-supply recti- down voltage (in steps, going from about 3.3 V to 100 V
fication (conversion of ac to dc, §1.6.2), signal rectification or more) is specified. Zeners are used to establish a volt-
(§1.6.6A), clamping (signal limiting, §1.6.6C), and gating age within a circuit (Figure 1.16), or to limit a signal’s
(§1.6.6B). Diodes are commonly used to prevent polarity swing.