0% found this document useful (0 votes)
104 viewsRead - Gregorian - Analog MOS IC For Signal Processing
Uploaded by
rahulpikaleCopyright
© © All Rights Reserved
Available Formats
Download as PDF or read online on Scribd
0% found this document useful (0 votes)
104 viewsRead - Gregorian - Analog MOS IC For Signal Processing
Uploaded by
rahulpikaleCopyright
© © All Rights Reserved
Available Formats
Download as PDF or read online on Scribd
/ 307
ANALOG MOS
INTEGRATED CIRCUITS
Desi Manag Aptos FOR SIGNAL PROCESSING
ators: Robert A Johnson aad Gorge Scemirma
Mechanical Filters in lctonis
Reber 4 Joheon
LC-Fiters: Design Testing, and Manufacturing
rich Chron FOUBIK GREGORIAN
sierra Semiconductor
Analog MOS Integrated Circus for Signal Processing San Jose, Caiforia
‘Roubik Gregorian and Gabor C. Tomes
GABOR ©. TEMES
Department of Electical Engineering
University of California, Los Angeles
[A Wiley-intersclence Publication
JOHN WILEY & SONS.
NewYork Chichester Brisbane Toronto Singapore Copy © 196 J Wie Sons. I
A rene, Pied sma a Cana
Repodon wat fay par his wk
‘ond tht permite y Secon oo Toke
{Ste Unie! Sts Copa Ac tat te permis
Sie cnet it Ref
{Se Pemos Depart en Wey & So,
Livery of Congr Cali iain Date
“Rleg MOS mete ey fo spl penn
SR Wey ners pbtiton~
pean
feo Men one one Te
To our wives,
i Temes
and Agnes Gregorian,
in gratitude for ther patience and help SERIES PREFACE
‘The primary chjctve ofthe Wiley Series Oa Filters ist bring together theory
and Industral prasioe in series of volumes written for iter users a well 3s
‘hose involved in filler design and manufacturing. Alvhough thi isa iicul
task, the authors in this eras are well qualified fr the job. Tey bring both
sHrong academic credentials and many Yeas of industrial experience to thei
books. They have all designed ter, have been involved in manufacturing,
ind have had experince in interacting with the titer user
ach ofthe hooks covers a wide range of subjects including filter specifi
tions, design, theory, parts and materials, maraactoring. tuning, texting,
specie applications, and help in using the fier in ccut, The books also
provide @ broad view of each sujet based on the auth’ own work and
frvolvement with filer experts from around the word
"The most oustanding feature ofthis serie isthe broad audience of fier
turers and users, which it addresses This includes fer research and
development engincss, fer designers, and. material specialist, a well as
industri, quality contol and sales engineers. On the ater use’ sd, the
books are of help t0 the crcl designer, the system engineer, as well as
applications reliability, and component Lest experts, and spastietons and
Standards engineers
Roose A. Jounsow
Groner Sersrisens PREFACE
The purpose of this book sto describe the operating principles of analog MOS
integrated circuits and to teach the reader how to dexgn and whe such crits,
[Examples ofthese devices include svitched-cpacito ters, analog-to-igital
and digitao-analog converters, amplifiers, modulators, osilators, and 9 9
‘The main eraphais son the phsial operation and on the design proces. 1s
hoped tht the hook wil be used as 8 senior: of gradutedeve txt i the
lecical engineering curioolam of universities and also as taining asd
feference material or india ict designers, To increase the useful of
the book asa text fr classroom teaching, numerous problems ae isluded at
the end ofeach chapter: these problems may be used for homework assign
‘ments To enhance its value ay + design reference, tables and numerical design
examples are include to clay the step-by-step process involved. The Hest
three chapters provide a cones, hasiclevel and (4s hope) clear description of
the general properties of analog MOS integrated ieuts, and the required
background in mathematics and Senjcondistor device physi. The remainder
fof the book is devoted to the design of the seta creuts, the practical
Problems encountered and thie solutions, and some examples of stem
Spplications
"This book evolved froma set of lecture notes written originally for shor.
courses presented seve tines annually since 179 inthe United States nd in
Western Europe both asa publi offering at UCLA, the Fedral lastiute of
‘Technology of Switerand, the Univers of Sig, and so on, and as 39
inchouse taining cours for high-technology semiconductor, communication,
tnd computer companies, fered through the Coatinuing Eatin Insite
‘Low Angeles, Late, this material formed the isi of » graduate course
tered om analog MOS integrated circuits at UCLA. The organization of the
‘material was therefore inftuenced by the need to make the presentation suitable
for audiences of widely varying backgrounds. Hence, we tried t0 mabe the
book reasonably selieontaned, and the presentation is atthe simplest lve
alflorded by the topics discussed Only limited amount of preparation wae
"assumed onthe part ofthe reader: mathematics oa the junior level, and ee
to introductory level courses in letonics and semicondclor phys ae the
‘minimum requirements. “The origi of the book also influenced the dtl choice ofits subject
snater. Since the original short couree Was intended to tn indus en
tincers i the design ef aalog MOS stcuits, the theoretical topics dncused
‘kere restricted to the minimum needed for the prc design proces. Also
in those situations where 4 number of design techniques wore avaible 10
accomplish 9 given task. we deseribed only the one that was most extensively
tenn practical applications. Hence many ingenious and elective desig
procedures were ignored
oth authors have had considerable ndusral experienc and also extensive
teaching background. We hope tht this experience is delectable in Our
Approach tothe tresnient of our subject.
"The hoch conti eight chapers. Chapter 1 gies a hase introduction to
switchdheapictor Gcuts, compares tbe analog MOS ciicuis with ter
Signal processce implementations, and describe (bt doesnot expain in any
‘et some typical applications. This material ean he covered im ome lecture
{vorhour lectures are assumed ere and thoughout this preface.
(Chapter 2 deserves the Laplae. Fourier, and =ransforms and introduces
the important sto. transformations needed vo desi a sampled-data sstem
from ao analog “model.” Depending onthe mathematical background othe
stunts. this mateial may require to to tyes tworbour lectures.
‘Chapies 3 ves a brief deserption of the physics of MOS devices, diseusses
the lmarized models of MOSFETs, and’ desehes MOS capacitors and
ssvitches. The technology usd to fabricate MOS devies i also briefly de
feribed. Once apn. depending on the askground ofthe audience S40 or
‘hee lctures should suffice to cover the contet of this chapter
‘Chapter 4 duces the cet design techniques fr teazing MOS oper-
ational ampli. The most common sir coatguations a well a5 their
‘design and limitations, are included, and a devgn example is worked tn
detail. Complete coverage of all topics inthis chapter requires about five
itor; this time ean be duced by resting the discussions, faving out
some specialized subjets sich as those scused in Sections 10, 4.11 and
‘113, and assigning the design example for reading.
Chapter 5 dals with sultehed-apactor iter design and hence represents
\he fecal point of the book. AS already mentioned. the desgh techniques
tisussed fre esrced tothe “minsream” ones: thse that have been most
thoroughly ested in practical applicaons. The design of the two commonly
used confguations cascade and laler crisis discussed detail and
ilustrated with a numerical design example Some special ecu, such 96
Switched-capacior Ncpath flrs and simblated-eistor active RC iter, ate
also describe. A fll eeverage ofa topics i ths chapter needs about five
Tectures: by omiting Sections 5.8 and 3.8, this rater can be reduced to three
Chapter 6 deals with nonfitering applications of switched capacitor its
Sueh important circuits as voltage amplifier, digitlLosanalog and analog -
Signal cooverters, comparators, mesulators, and osilators are discussed on
fan intrductory level complete coverage ofall opie my require fo
prerece a
ecu: an be presen in go Ist i he detailed desions of
Sections 62 and 6s ae condensed
cee tonaina deta dscson of the fondle ocung in
oe pace crete This material of uumost ingatance to the
SEEMS aigner af paces and Rene (in spite fs seemingly
aa et man) should be covered Test bi enn an unde
aaa eee snus soul be sant for + bit prea
Chapter dices Some of the stems aspects of anilog MOS signal
(rae shits he ws in Commerc incprated stems. The Ht
ao ee dat wh the peice and pone esiemeatsof
rea Af0S cca seule dessa nthe casroom. The remainder of
aa ies aecach ons deve ee sme pete applictons ad an
a se ening Hence the fate in this chapter mi) De
tra ed intone of to ares depending on how the application examples
te weated
Spending on the depth ofthe psetaton, the fall overage ofl
ral he book may rege many a 28 two hau eas a Fe as
aan scons (cha a sho couse oa taining cv) the
rose ee ea fay ben ore nf ay 9h i str ors pet da.
mpi Moat a to ou eles Drs DUXEYa, S.C. Fan, B. Fathi B
annie ses. Marin, T Catto and WE, Orchard, a wll sour
cone, ner sents JN, Baber, F. Dunlap, T. Hos.
Me, JB, Shyu and F.. Wang or dacusions review and rom, Nest
Lite dime vine tsk was done BY Ms Loci Loberman. We are
tu fr her ctl and pakeg help. The artwork was done (excel
Ey Me Kayan Abana of he Deparmet of lec Engine
wey GEL tase but ot las we would ke to expres our rade tot
‘Chester gas slonng neh daring the wnsing of his work
Rovank Garaosas
‘Gavor ©. Tots
on ange calor CONTENTS
Chapter One. Introduction 1
M1.
1a,
13,
‘The Use of Analog MOS Integrated Circuits for
Signal Procesing 1
Comparison of Analog MOS Signal Processors with
Other Implementations. 6
Examples of Analog MOS Signal Processing Circuits
and Systems. 13
Problems 15
ferences 18
Chapter To. Transformation Methods »
2
‘Analog, Digital, and Sampled.Aaalog Signals
and Systems “19.
22. The Laplace Transformation 24
23, The Fourie Transformation 29
218 Sampling andthe Transform 31
25, The'Specea of Sampled-Data Signals 39
36. The Design of Sainled-Data
Continous-Time Modis 3
Probleme 60.
References 62
Chapter Three. MOS Devices as Circuit Elements “4
31. Semiconductors 64
32. The MOS Transistor 68
33. MOS Transistor Types; Body EMect_ 75
34 SrullSionl Operation and Bauivalent Circuit
of MOSFETs. “78
3S, MOS Capacitors 86
Re The MOS swuen_ 98
37, Noise in MOSFETS 96
38, Analog MOS Technology 99
Problems 118
References 120, a contents.
Chapter Four. MOS Operational Amplifiers nm
4
Operational Amplirs 121,
42. Blas Circuits in MOS Technology 126,
43. MOS Gain Sues 133
44. MOS Source Followers and Output Bullers 14
43, MoS Dillerentsl Amplifiers 146
46, The Frequency Response of MOS Amplifier Stages 156
47. Uncompensiied CMOS Operational Ampiers 168
48. Stability Theory and Compensation of
CMOS Ampifers 172
49. Frequency Respons, Transient Response, and Slew Rate
ofthe Compensated CMOS Op-Amp 182
4.10, Noise Performance of CMOS Amplifiers 186
411, An EnhancerenteMode NMOS Op-Amp 189
412. NMOS Op-Amps with Depletion Lands” 196
413, The Noise Performance of NMOS Op-Amps 206
4414, Practical Considerations in MOS Op-Amp Design 208
415. Op-Amp Design Teehaigues and Examples 222
416. High-Performance CMOS Op-Amps 36
Problems 259.
References 264
Chapter Five. Switched-Capacitr Filters 265
SA. Integrate Fiters 265
52. Switced-Copactor Integrators 290
53. Sira-tnsnstveImegatoes 277
54. Second-Order Sesion: Cascade Filler Design 280
55, Switched-Capactor Laer Filters 296
56. The Scfing of Switched Capacitor Filters 339
57. Design samples 349
38. Switced-Capacitor A-Path Filters 363,
59, Integrated Actve-RC Files Using
MOSFET Resistors 387
Problems 304
References 409
Chapier Siz. Nonfitering Applications of
‘Switched-Capacitor Circus, an
Switched-Capacitor Guin Stages 412
Programmable Capacitor Atay: Digital-toAnaog and
Analog to Digital Converirs 417
MOS Comparators 25
Modulators, Recties, and Detectors 437
6s.
Contents =
Switehed-Cupactor Oscillators 445,
Problems 48
References 460
Chapter Seven. Nonidal Efets in Switched-Capacitor Circuits 462
a
Nonideal Ets inthe Switches 462
72. Capacitanee Inaceuacis 474
73, Nonideal Op-Amp Circuit Eifects 483,
14 Noise Generated in Swtched-Capactor Circuits $00
75. Layout Considerations in Swithed-Capacitor Circuits S13
Problems 524
References 528
Chapter Eight. System Considerations and.tppicaions 10
Bi. Prefkering Resuirement for Swiched-Capacivor
Filters 530
82. Postiterng for Switched Capacitor Filters $39
83. The Application of Switebed-Capactor Circuit ia
Digital Voice Transmission Systems. $45
84. The Applicaton of Switched-Cspacior Cicats in
Data Modems 561
85. The Appliction of Swichad-Capacioe Filters in
Tone Recsivers 56
86. Programmable Switched-Capacior Filters S71
2. The Applicaton of Suiched-Capacitor Circuits in
Speech Prosesing, S76
Problems 589
References 591
Index 593 Chapter One
INTRODUCTION
In this chapter, the basic coacept of a switched capacitor pesforming 2 a
simblted resistor is introduced, Some ofthe physi properties of ritched-
fapuctor cieuits ae abo briefly dscused. Then, 2 comparison is made
between signal processors using suitehes-capacito sie and some alterna
ve implementations, such as discrete analog cel, digital ers, and analog
bipolar integrated cteuits Finally, «few repreenttve example are given of
siruits and: systems utizing MOS analog signal processing techniques, 10
istrate the great potential ofthese cruisin tlecommuniaton systems and
related applications,
1.1. THE USE OF ANALOG MOS INTEGRATED CIRCUITS.
FOR SIGNAL PROCESSING."
letra signal pocesors are usally divided into two categories: analog and
igual sjstems. ‘An analog stem cartes signals in the foem of voltages,
‘currents, charges, and so on, which ae conics Fonction of the continua
ime variable. Some typical ctampls of analog signal processors are audio
amplifiers, passve-or active RC files, and s0 on. By contrast, in 2 dal
stem each signal i represented by a sequence of numbers. Since these
rumbers can only contain finite nomber of dips (pial, coded in the
{orm of binary digits or ts) they can only tke on rete ales. Also these
umber are the sampled vais ofthe signal, taken at derive me instances
‘Thus, both the dependent and independent vaviables of digital signal are
siscrete. Sine the proesing ofthe digital bits is usually performed synchro-
ously, a timing or lock cut iba important part ofthe digital system. The
‘lock provides one or more clock signals. cach containing accurately timed
pulses which operate or synchronize the operation of the eomponents ofthe
system. Typical cramps of digtal systems are a general-purpose digital 2 mutmoouCTION
“compute, of a Sptsiabpanpose compute: Getianed 0 (sy) celating he
Fourier wansfor of signal va the fast Fourie apsform (FFT), o gt
‘ler used in speech analysis, and $9 on
[Most of the eitcuis considered in this Book fl ito category which i in
between the twa main claslations desctibed above. hie othe eategory
Sampledsdate analog sstome, For such ystems, the sina is epreseated by the
‘uncoded amplitude of an electrical quantity (normaly, voltege) a8 in 2
analog sytem. However, the sistem coniins a cee, and the signal amplitude
{x sensed only at discrete tame instances. a in a git system Prior to he
‘development of the MOS cieuite discus in this Work the most importa
sampled-data analog systems were the charge-transfer devices, such as charge
‘coupled devices (CCDs) and bucket-beigade devices (BED). I thes, the
‘gnal ampltod i presented by the amount of charge shift fom cell 0
‘a. Since, wth vey few exceptions, thee devices didnot contain Feedback
loops; they were inherently nonrecunive in nature, Therefore, te) were more
suited for such applications as sampled-data delay ines, mulipeters, corel
{ors and soon Which did not require accurately coatolled pest well 35
2 far the comanool nested equensy aleve seg tasks. NO.
they require rpecil fabrcationtechnolwy, rather than the sandard MOS
proses used 1b manufactore digital MOS ccs, and usally need some
Specialized peripheral (input and output) cuit, For thse reasons, thei uses
were restricted to reatnely few large companies, whee the reausite seal
design background and technology could be developed and maitaind
BY contax, the euts eonsdered i this Book canbe abscated wing
standard digital MOS techology, and hence can also be placed om the same
‘hip with diptal circuitry. This later aspect is of great importance, fo
‘example, in moder tlecommunication systems, where bath anslog and digital
functions are-afen needed within the same functional block. Furthermore,
these cuits contain only fw standard building Docks: amplifiers, swith
‘capacitor, and, in rae instances, reson, Once these have been developed
Sod standardized is the Wealy gwafable vechnclogy, a large aumiber of
applications can be accommedated using only lightly diferent configurations
and/or dimensions.
‘To understand the basic concepts of the most commonly vsed eoafgur
tions of analog MOS circus, consider the simple sala ane Canton
Vogl) ’
felt) Par wa)
1s easy to verily thatthe RLC circlt shown im Fig. 1.10 can realize this
function (Problem 1), While this crcuit i ear 1 desi, bul nd test, the
presence of the inductor n the Grek makes the faristion in an iterated
fem practical in fae, fr low-frequency applications, thi c¥euit may well
seuire very largealued, and hence bulky, inductor and capaci, To
SEOF ANALOGMOS NTEGRATEDRCUITS. 3
FIGURE 1. Sco ions) pase set (0 athe RC Se
overome this problem, he sie may desde intend 0 eae the desired
SST ann eng an sane cea Har rey be shown hat te
SERIrGCre Lan which ues tee prina nlfer apabl of
peoiing the amr fenctonspcied nF (1) This eat meds ao
Pedt and aye ead with sale iret component 3 wide
Waray of specrcaos (Pele 12). Hrs out, howe, that Whe te
Tcgricon a ts cet en an MOS hp yin pone fen nce te
st rer and capacitors nde ca a be erat), ee ae
sameor paca sc otepaton Tet inade te ery age hp
SEE hy the RC components aswel ste vngentsraee and
Shay requtenets for ths sements Tes requirenens canna be edly
TERNS Simepaedcomponcns ne ner the abated ales mo the
Schpcnureimded wane of te sete an te capaci een
SCP cater The reuling oleae sistas ae 10 age Fr mes
‘foteatons (This sujet wal be Gucci deal Secon 5.1 of
Ser’)
Man efuine sty which can she bo he ae ad the matching
test to tpi exch restr ini cc by 8 combination =
pur an eps, Conse the banc show ini 12 Her the
iSuravtches $y Sand Scope nd close peel tate wich
seks than atthe vrs ofthe fermi slags a4
Beis SSS operate sychronou th cach oer. Bot I epposte
hus withstand STs whe Sand, ae clove. San Sy a open ” F J patel
and vice versa, Now when 5, and 5; lose, Cis charged. When ext and
Sy vpen, and $, and S close, Cis recharged 0 the vllage fe = 0 = Tis
fuses a charge q = Cley~ eto ow through the branch of Fig, 12° Nene
is agin dischavaed by, and Sand son. I this yee i repeated every T
seconds (where T isthe, swching period or clock pend) then the average
urreat though the branch is threfore
ac
oe Cha aay
‘Thus iy is proportional to the ranch voltage ~ op. Sima, fora branch
containing a resistor R, the branch curent ("= (1/RXey ~ ty), Th. the
average cutent flows in these two branches are the sume i the sation
R= T/C hols
Wis plausible therefore that the branch of Fig, 1.2 can be wd to teplice all
resistors inthe erat of Fig. 1.18. The rsuking stages shown in Fig 13. la
this iret, switches which belong to different “resistors” but perfrm ent.
‘al task, have Been combined. Furthrmeve, the second operational api
the
4
USL OF ANALOG MOS WTEGRATED CMCUE
(opamp) in Fig 1.18, which acted mercly as & phase inverter, has been
eliminated. This was posnble ance hy simply changing the phasing of two of
the switches associated with capacitor Cy the reguied phase inversion could
‘be accomplished without an opamp. The details ofthe transformation of the
rei of Fig, to that of Fig 1 ate dscucad in Chapter 5, Seton 54
‘As Fig. 12 illustrates, the transformed eiet contains onl) capacitors,
switches, and op-amps. A major advantage of this new arrangement i that
‘ow al ime consams,previeusly determined by the poorly oniralit RC
prodocts, will br given by expresions of the form (7/C,)C, ~ M1Cx/C,).
Here, the clock period vs usally determined by 4 quartcrytal-contolod
‘lock circuit ants hence yery accurate and sable. The other fator of the
time const is Cy/Cy, that ithe rato of two on-chip MOS eapastances.
Using seme simple usin the layout ofthese elements (described in Section
5.3}. is possible to obtain an acuraty and stability of the onder of 0.1% for
ihis ato. The esulting overall accuracy is at less hundred times etter than
‘wat cane achieved with an on-chip ressor and capacitor forthe RC time
"A dramatic improvement is also achievable for the area equied by the
passive elements. To achieve a time constant inthe audio requeney range (35
4 kead/s) even witha lage (10 pF) capacitor a resistance of 10 MD is
required. Such a resistor will occupy an area of about 10" pin? which is
proibiively large itis neatly 108 ofthe area of an average chip. By contrast,
Fora typical clock period of 10 ps, the capacitance ofthe switched capacitor
realising 2 10-MO rexstor is C= T/R™= 10 9/107 = 10." F= 1 pr The
tea requized to realize this capacitance is about 2500 ym? or only 025% of
that nesses by the resistor whic it replaces.
‘Using the thice types of components (op-amps, capacitors and switches)
shown in Fig 1.3. a lng quantity of signal processing circuitry canbe placed
‘on single chip. A highsualiy op-amp can be fabricated on an area of
Sc 10" to 10! pa while a switch needs typically only about 80 je, Since
the area of large chip may be around 8% 10” ym such a chip ean really
accommodate, sa 10 op-amps, 300 capacitors, and SO) switches, Extrapolat-
ing from the circuit of Fig 1.3 which realizes the second-order transfer
function given in Ea (11). it ean be seen thatthe signal procesing capability
fof such 2 chip is sufcient to implement transfer fonctions with combined
fonder of 100 lfc, since the op-amps can be time shared (multiplexed) for
low-frequency signal, even higher-order functions may be realized a speech
‘anaszer chip implementing switched-capaciorfltes with a total of 308
poles, in addition to a substantial quantity of on-chip distal circuitry was
‘evenly deseribed!
Ta addition to frequency selective Hltring which hasbeen the most common
pplication ofthe switched-eapcitor (SC eet introduced in Fig. 13, there
fre many other functions which such circus can perform. These include
fanalogo-digial (A/D) and digtal-to-analog (D/A) data conversion, pro
{rammable gain amplifation fr AGC and other applications, a wells such nonlinear operations as multiplication, modulation, detection, rectification,
Zero-rossing detection, and soon. They have also been sod extensively in
large mixed’ analog-dipital systems such as codecs, modems, and. specch
processors. I in expected tha this range wil expand further a the auaity
bandwidth. dynamic range. ete) of the components, especially oP-amPS,
Improve. nd as etter circuit wchnigues ae introduced
12. COMPARISON OF ANALOG MOS SIGNAL PROCESSORS WITH
(OTHER IMPLEMENTATIONS
‘is important to define the areas of applications in which analog MOS signal
procestor are conpetiive with, or even superior a ear implementations of
Signal processing systems. To do that, we list below some sient features of|
analog MOS (ypcally,switchedapactor)efeuits, and contast them with
‘hose of alternative realizations.
1. Switched. Capacitor Crate Are Integrated Coit This property has
1 profound effect on the economy of is applications. The development
(eoretical design, computer simulation, layout and fsbrication, esting and
troubleshooting) of such acct may require a combined inital dein lor
‘of ane or more man-yes, at cost of $80,000 oF more. Afer thi initial
expenditure the devices can be mas produced at alow peut cost, sy $5 of
less. Hence, compared to 2 discrete implementation costing (sy) $15 and
‘having a negligible design expenditure the integrated eairation x economical
ifthe inequality
50.000 + SW < 15N 09
holds. Here i the total numberof units requied, For the values wed, this
fives _N'> S000. Clearly the actual figures depend’ on the sxpenence
‘guipmen. aplication, and soon; however, the orders of magnitude ven are
fairly pica
‘Ore features associated with the intepate-creit (IC) character f SC
sireuits are small ize, light weg, high reliably, and small de bias power
Feduited. These may also have agieat msporance i, x4. aerospace appli
2 Switched-Capactor Cirats Are Sampled-Data Stems. As mentioned
before, the signal values ae evaluated only at pero time instants im an SC
‘cireuit, and the sampling perio is determined by a cystal-conelld lock
“This feature makes it possible to have ll pole and 2x0 valves dependent ony
‘on capacitance ratios (rather thin on absolute values. it ths permits the
Fealzation of highselciviy vesponses with good accuracy and stabil
Furthermore, sine only periodic samples of the signal are of interest it
possible to time share (aulple) the whole circuit of parts such as the
t %
FIGURE 14. Swish ipso handphone =
‘St ancy ih un BW ae nll wen
2 a te
‘opamps of the circuit, among several sgnal changes resulting in highly
‘cient multichannel ystems
Finally, all me constants of an SC crcit are proportional to the clock
pecod 7. As ares. the overall gain versus frequency response H() can
Feaily be sealed by changing the clock frequency J, = 1/T. A an illsiration,
Fig. 1 shows the responses of an SC bandpass fiter fortwo dierent cock
frequencies. Clearly, changing the vale of f rom fa 10 2, simply expands
the response curve horizontally by these factor 2. This Besa valuable tol
forthe fine wing of the response for applications suchas volagecontroled
‘osillatrs (VCO), apa ters, rackng ltrs, and soon.
“The above features (ieyrated realization, clock-coneolled sampled-data
‘operation are shared with chargeseansfes devices such as CCDs and BBDs, as
well as with digital filter, but not ih the other commonly used signal
processor implementations.
1. Switched Capacitor Circuit Are Analog Stems Insite of thir
sampled-data characters SC ccuts handle sighals in analog foros: thus, the
‘mmplteds ofthe sampled wolages are the signal values, without the us any
encoding. This makes the Basie operations (multiplication, addition, delay)
‘oeded in signal processing much simpler to pecform than in digital crus.
Hence, the density of operations on the chip can be much higher than for
Aig signal processors Without any molilesing liner filter sections reali
{ing 100 or even mote poles can be accommodated on single chip
‘De tothe bose simplicity ofthe eet wsed in SC systems the sped with
which the signal processing tasks can be accomplished is mach higher than for
{igi stems. The realtime ering of signals with frequencies up £0 about
(03 Mlig is possible curently, and this fire is continuously rising 38
‘aaproved technology and design technigues become available.
‘Doe i its much simpler stcture, along with the chip area needed, the de
power requitet fora given signal processing tasks also considerably less for 3 SC implementation than fora digital one. A pial vale 1 mW de power
teed pr fer ple rez. abd ths vale am Be drsialy ree i
in cots to these advange, when compared vith digi inal
resto, SC stm ako hve sve iporian advantage, Ges hat
Timid accuracy. Ar menue eae, the capacitance ration (which de
{crn ie coficient and has the aees and poles he afr anton)
‘ante made arurateand stable to soot O18 0 hei nominal vl: Whe
this a very impressive prsion fran analog sem only coresponds fo
{ost Reag-potaccrsy in dpa cra. For some appatios. tht
prston sno acento) a1 accuracy is mandatory. preci
{he wea analog methods
“Anrher potetal problem tea soar dynamic range. Duc oa reatiely
large level f aie orgnsting frm the operational anphfiers an utes
tnd alo coupled from he ek andthe sap ines and so othe Gyan
range (bas the rsuimumigal one rato) of SC cea sedom
{Soteds 100000 or 100 dB vale of 70" 90 dB ar como The one
Sponding rang of itl es uch ree
"Yt aother aang of digtl al powers ocr hi aa couter
parts thr superior Rexbiity and progsmmabliy, Whe SC orcas can
{oe made programmable, fora digtal sem he hunging of he aac
itis canbe performed simply by sect ire coset om 4 ROM
JT eae of operation Fok sso forthe mulileng of gia ina
proceso among seve sign chanel,
nally an imporant practi avanigs 0 gal sigs processor i de
to the fact that tes ne Walt sing stanford meroprcesor techno.
Hence. they can take advange of the sandra is oa er denn
infomation amase for mropocesr. Bese of he le volume at
Bence high pony of micoprcesor manufacturing, ere sully ample
‘lable information concerning te technology. te basic cts, accuse
vice mode for somputer sms nd 0.0m which mayor may nt
‘he cise forthe alg Gren using He sme tchnoog) Alo, the neitable
‘hangs inthe fabrication recess ich a he scaling Sow Of the nim
Tnewnth edu of trol aed sappy voltages and 50 on ae aad
primarily at improving the performance of the higherwokme dpa ICs
‘While change sme athe tr epetation of ig Gros (ch wang
down thir denon) tend wo pone nar improvement of the slog
Siete sharing the same techno, they mayan inreduce or agrawal
Donidel phenswena sich sshrehanal nd arow-channel eles ch
‘ten fl ths analog ccats more han te Sig ees
‘A the above arguments demonstate, thes no slearout and gentle
{or choeing analog SC tmpreniton ovr Sigil one, oe es.
Whenever simpy, spesd of operation, sal cp arse salle Bis
omer ae o rine importance, ad the eque op and utp sis ae
Tne spa io ature the analog ipementton maybe prefer oner
2 digit one Vie vena. ie ingatand/or oatpa sgal are aes pt
CCOUPARISON OF ANALOG MOS SIGNAL PROCESSORS ©
crt esi fe esr or rig oF mating vita
eee etn and or haan ngs eens re er sng oF
GOLESI Rtaposs RON BAM stmt unis, A/D and D/A
[QS Secarendy alae ssi reser may the lapel
coc
1 cenly develope signal proveing ships G2 €8. Ret 9 the
pag ats an uay ide ercon slg and dial eet.
Te uc in any ese mete bythe tl signal processors or
AILS! (4/D and D/A comenios, rerig and ponting
rete dig ease aed by SC Stes ls
sree Sen, programmn.aiatin, exis 30
SLE devon teaver analy and apt Seas nota Bay One
Seah are eed al poser The man ise i fe te
a ete foe as othe mon appropiate (ial oF analog)
‘inca! Dek
7. Sitch Capcr Ciro Are Rled MOS Techlg* AB
anda ents atte fms, the ipl ne. Bor abt
ere cbted © dar sprain of the bipolar and MOS tela
aera ih nc rte. MOS technol, wih spends
SERPS atelmonty fort lg sd meer appieatons we all
ser oe fangs a as ampleaen, ein. aa omen)
cee rd wang bib inept ec, sch as ole oP amp.
Serr ne howe ra progres re im MOS abrcton etnies
som St}. | lowpass ru 2, euren our
Ehaad wate) > a on
t *.
aw =
at le
rockin 5
veins a
1403030300 oF
HOURE 14. Te Hck diag fhe AMI S32 pune lop le, ‘i
‘COMPARISON OF ANA:G MOS SIGNAL PROCESSORS 11,
‘made it possible 19 sanufacture much more complex and flexible chips. In
‘ig, ew developments cocurred in communication technology (sich 38
Aigial telephony, data tansmssion via telephone lines, adaptive communica
tion channels, ete) which required analog and digital signal processing ci
‘try inthe same functional Blocks. The analog functions most often needed
‘sre lieing (or antnasng, smoothing, band Separation, cs), amplieaio,
Sample-and-hold opeations volage comparison, and the generation aswell a=
reise scaling of voltages and currents for daa conversion. The separation of
these analog fuetions fom the digital ones merely because ofthe diferent
fabscation technologies used is undesirable, ince it icreases both the pack
ing costs and the space fequzements and also, de to the additonal arco
rections required, deprades the performance. Hence, tere mas sig moti
tion to develop novel MOS circ which ean perform these analog functions
tnd which can also share the area othe same chip withthe digital iret.
“The MOS technology fas bth advantages and disadvantages as compared
withthe Bipotar one, An MOS device hasan extremely high impedance a is
‘pat (at) terminal, which enables i to sense the vllage across capacitor
‘without discharging it, Also, there sn ietent offset votage across the MOS
Gevie when it i used as a conducting switch. Furthermore, high-quality
‘apactors canbe fabricated reliably on an MOS chip. Thexe fentores make the
realization of soc eyeuits as precision sample-and:hold stages feasible on a8
MOS chip." Tiss usually not posible in bipolar technology,
‘On the negative side, the ansconductaee of MOS transistors is inherently
lower than that of bipolar ones. A typical wanscondoctance vale for 3
fe ©
6 ———
FreouDcr O2) Imoderatesize MOS device is around 0.5 mA/V; fora bipolar transistor, it
‘ay be about a hundred times larger. This leads wo a higher oft voltage foe
fan MOS amplifier than for bipolar one. (AC the same time, however the
‘input capacitance ofthe MOS transistors ypeally much sale than that of
{bipolar one.) Also, the aoise generated in'an MOS device is much higher,
spell at low frequencies, than in a bipolar transistor. The coclsion is
thatthe behavior ofan amplier realized on an MOS chip tends tobe inferior
to an equivalent bipolar realization in tems of let voltage, noise, and
“dynamic range. However, itcan havea much highes input impedance than its
bipolar counterpare.
"as a esult of these properties, switched capacitor circuits are especially
suitable for linear applications, where elementvalue ccuray is important, but
the signal frequency i ao too high, andthe dynamic range required is n0t
‘excesive. Voice and audiosrequency ering and data comversion ae inthis
‘ategry, and they eepresen the bulk of past applications.
Yn
EXAMPLES OF ALOGMOS CIRCUITS. 12
© 20 [ao 0] 90 700 120 [19 ie
Vea eater ener oentar
1.3. EXAMPLES OF ANALOG MOS SIGNAL PROCESSING.
CIRCUITS AND SYSTEMS.
In this section, @ few selected examples of practical analog MOS sign
proceso are sven, Of cours, the reads should ot exp’ o understand the
‘ets of these systems at thi tage, However the drums may sve an idea
ofthe potentials ofthese devies in modern telconniation stems. In the
later chapters of the book (Chapters S-8), the design of such cies willbe
dlscusod in deta
Figure 1.5 show the Block diagram ofa digitally programmable switched
capacitor low-pass ter marketed a $1528 by American Microwytem, Ine)
By choosing the stbit digital word (Dy, B= D,)al the dial input the
passband and stopband edge frequenses can be selected infil smal eps
‘The circuit diagram of the seventvarderswtched-apacitor tris showa in
Fig. 1.6 i measure responses shown i Fig. 4h, The measored passband
fin variation was fxs than 006 dB. This represets a superior performance,
‘hich could not have bes achieved without extensive trimming wing aay
lather analog iter technology:
Figure 1 llustaes the cieit diagram ofa simple narrow-band switches
capacitor fit” Figure L8a shows te veal loss response and Fig, 18b the
pussand details. The elaine bandwith was les than 02%, Among analog
Sirus only #caefUlly trimmed eystal or mechanial ter could paral his
performanet Figure 1.9 shows the circuit diagram of « switched-apacitor sine-wave
oscillator capable of generating ow-stortion waveforms with very acura
controled amplitude and frequency values" Is performance i eomparable
‘ha of high-quality tunable osilatr realized bya diet ire.
Figure 1.10 shows the Block diagram of the AN S306 codec” This circuit
contains four ters as well as a nolinearanalogtodigial converter, 3
sonlneardigital-to-analog converter, and an on-hip voltage reference circuit,
‘waddiion to a considerable amount of digital loge ccuity
Figure 1.11 illustrates the block diagram of the dua-ione muliequency
(DIME) receiver designed by Slcon Systems, Inc!" This chip eoalins
ozen ltrs, two multiplexed zero-crossing detectors, a power regulator and
‘voltage reference ict as well extensive digital ccaity. The total number
of operational amplies onthe chip i 57
Finally Fig. 1.12 shows the diagram ofa recent spech spectrum analyzer
“This ici splits the speech specu into 20 requency bands and measures
the power in each band. The resulting 20 values are then converted its
nine-it digital words and stored inthe output buffer memory from which
they can be asynchronously recalled. BY using time-divion muluplexing
(enade possible by the sampled-data crac of the sitehed-capacitor er)
4 otal number of 308 fers poles is realized using “only” 30 operational
“pliers. In addition, the chip contains two voltage comparator, 39-0
random-access memory (RAM) and S00 digital gates.
|As the above examples illustrate present-day analog MOS cits have
reached a certin amount of matuity.Altady, almost any analog. signal
processing task im the voice-oraudi-fequency range has «possible solution
"sng such circuits, With the expected improvements in oth fabrication
technology and circuit design techniques. wis likely that the speed and
[A signa is funtion —its independent variables. ouF applications, ime,
the dependent variable i» pica quantity such as vllage, charge, oF
current A continuum signal is signal which has a well-defined value a
‘very point inthe time interval of satfest (Fig. 21a) A dertesime signal
bas value only at discrete (osaly equally spaced) time instances (Fg. 2.15);
‘tis unspecified at anyother ime. Often the discrete-time signal obtained by
samplings continvoustime one, Thus, the signal of Fig: 21D sats es the
‘elton
WaT) = o(joare = O12.
withthe Signal of Fig, 2.1, Here, Tis the sampling intro, closely elated Wty
pr,
FQUME 21. Sil ps () cmt i (6 ieee sac) sample
ite sot
signal type isthe sampledand-held signal illstrated in Fi, 216 I sais
ene
owt), wPcr< (m4 A)T-
“Thos. a smpled-and-beld($/H) signals a coninsoustme signal whase value
changes only at discrete-time instanors, I completly defined by T and the
Salus o(nT) for = 0.1.2,
[A ital signal is 2 Sequence of numbers. Fach number ean be regarded 38
1 representation of value n(n7) of 2 dscretetime sal, However, the
‘umber inthe digital signal can only have aie aumber of digs (i, he
‘Signal s inary coded), Hence, thy can oly assume discrete vals, whi age
‘lil ofthe smallest valle digit By contrast, the functions (0). (a7),
land ty) my have any values; they wil be called analog sia. The ast,
two functions, e(47) and ty) are als ten elle sompledanalg signals,
‘or ample. data anaog sina
‘A Stem ia physieal device capable of processing signals. 1 wil be clled
alg or digital, comlinuous of dicete time, and 30 on, depeadiag on the
{ype of the Signal which it can handle. An example ofa digital system i a
digital computer; an example of a continaosttime analog sjstem is an
amir. Mos ofthe Geis discussed inthis hook proces sampled-and eld
nil signals such as that shown in Fig. 1. They ae thus continoou-ime
nal systems ofa special type: they are used otansform a $/H input signal
Sno another $/4 signal the output signal
‘An analog system which process clcrcal signals in the form of voltages,
currents ten and charges isan elecric cra network, Tis analy is
Petformod by consructing and solving the menor equation, These ate
{ypically obtained by using Kirhhofs colae law (KVL) which stats thatthe
sum of voltage in any closed lop af the erie zero; Kirchhoff # current lw
(KCL which tates hat the sum of caren enteg any nod zero, and the
‘ranch relations which ae satsied by the specie cet elements (Tso
‘ipactos, amplifiers ete) sed in the network. Aa excellent introduction 40
‘eit analysis is given in Ref. 2
‘The KVL and KCL invake only the addition and subtraction of voltages
and corens. The Branch elatins, by contas, may ivole dilerentation of
integration of these quantities wih espect to time. Thus, for an inductor L
the voltage (7) across the deve and the current“, (1) owing theough i
sausty the equation
ao)
tre
MO oe yen Efalerarsio. 2)
ate
Similarly, fora capacitor C, the elation
HAY oe ely Eficlsdar tec) 22)
tele)
hos. Ths, the cieuit equations aren grea, integro-diferentil relations
For the circuit shown in Fig. 22, fr example, the cutent 17) may be found
tysoving the uation
a(n, ee
AP + mie) + filed ar = 00) ~ re) 03)
For the solution, the inital values 0) and (0) must be known,
“The solution of such integroiferental equations canbe performed using
lost mathematical methods Those tend to be complicated, ower. 18
ws 12 TRANSFORMATION METHODS
‘much more convenient to use soaton techniques based on the Laplace
transformation. to be diseased nthe next section
‘Simiay, the analyss ofa disretesime s)stem involves the construction
and solution of network equations, To construct these equations fopolopcl
relations which invohe only additions and subtractions (and are hence analo-
{00 to the Kirciho aws) mast he combined with branch relations” The
Tatter may involve multiplication by 2 constant, or a delay by the sampling
ineral 7 Consider as an example, the digital system shown in Fig. 2 eis
easy 1 show that the input-output relation Is
MOT) ~ aya 7) ~
Ja = 27) belnr), (24)
This i eifference equation; it can be solved by direct methods, ifthe initial
conditions (0) and. y(~T) ae given. It much simpler, however to use
‘tansformation, tobe disused in Seton 24. the press
For a linear timesnvariant continvous time analog system, an important
characteristic s is impube reponse. The impute function or. Dir dela
{fancion 81) can be toparde a the limiting case of the pulse function 0)
shown in Fig. 24 a8 ¢> 0. The impulse response h(2) the output ofthe
system (with ze etal energy) f8() es input signal Fig 2.50)
mo
o
me ag yon sit
1 now the same zeroenergy system is driven by an input signal (+) (Fi
2.8h) then ican he shows thatthe ouput i given by
Ha) = [ile = ale) a 5)
where 3(1) = h(¢)=0 for #<0 is asumed. Thos the impulse response
etines the response of the stem to other signals as wel The operation
‘between x(r) and (0), indicated i (25), fen denoted by vr) (0). tis
called the (unilateral conoluion of (1) and (1). Cea, exept Tor very
Simple functions it isa complicated proces to carr out. Once aan. however
the use of Laplace tassormation can make the ask much easier, a8 wll be
iscussed in the next seston,
Similar result can be derived for linear timeivatian discrete sy
tems. The impulse function is now defined by
arya {te neo 3
wary =(e nre eo)
‘The impulse response Mn) isthe output ofthe sytem (vith ze ini
condos) st Bn) inp
Fora erent inp sal the output ofthe sytem with eo
condos given bythe comet (or comeltin on)
vat) = xtaryentar) = E mar )<(or— kr), (23)
luhote x(a) = h(nT) = 0 foe m <0 i assumed, This operation, performed
iret is tedious for all but the simplest «(a7 and A(AT). However. a wll,
‘pe shown later, the use of transformation reduces Ihe necessary efor
considerably.
"We conclide that the analysis of iear ime-nvariat systems is made much
casi by the us of transformation methods. Hence, in the next sections. these
Techniques wil be explored 2.2, THE LAPLACE TRANSFORMATION"
“The Laplace transform of anetion /() is etined by
Fi) = [ned (28)
Equation (28) defines the waveral Laplace transform: foe the bteral
Laplace transform, the lower lit ofthe iteration is ~ 20. We will assume
that + isa general complex varable, with values Which allow the integral to
‘converge Also it wil be assumed that the time origin 1 = x ehosen sch that
f= Ofer <0,
“The Laplace transformation is a tnear operation: if fe) = ky fil +
af. where ky ky ate constant, then is Wansform is US) = KyA(S) +
iA), where F, and F, are the teansforms of fy and f,respeaivey, In
“aditon, thas thee key properties
1. IF the Laplace transform of fe) is Fs), then that of dfn is
sF(2) ~ 00), i a selFexplanatory shorthand aoation,
iH) Fo. then OO se) 10.
2. tn he same notation,
f10) FL). then [Je dr
3. Finally
0) Fo) and 0) = Bsn
AVAL = [Hla hE) dr FUSVEG.
est ropenies 1) nd 2) promise 1 spi he oan of he nero
‘iferental equation inifcantly, For example, applying them to Eq, 3)
with (0) = 0,
belay] et) +E Dey) asp
resus Here, as elsewhere capital leuers denote tassormed functions. This
THE LAPLACE TRANSFORMATION. 25
TABLE 21. Key Properties ofthe Laplace Transformation
Tine Fection Laplace Transform
io, Fs)
wore HF) f0)
fuorer a
Heese, Rybor
fe~ aiatt~ oe 2 0 oR)
Ho, )
enn
ht 00)
«am easily be solve for I
a) +400)
1) RAC)
7 (210)
Hence, afer the operation of Laplace transformation is inverted so that i()
obtained from 1(3), the problem is solved
Both Laplace transformation and inverse transformation can usually be
cari oot using tables sting the mast commonly nesed traasforms and
‘ules. Table 2.1 gives some Key properties ofthe Laplace transformation; Table
22 the most commonly oecuring fonctions. The reader rlered to Refs
2 and Yor much more extensive dicussions ofthese important topic
‘TABLE 22. Some Commonly Used Laplace-Transtorm Pairs
10 Fo)
0
Kor ku) 2% ‘TaNsronmATION METHODS
The unilateral Laplace tansform ofa constant K is, as Table 22 indicates,
-K/s. This the same asthe Laplace transform of Ku(t), where w(1) is the ep
finuongiven'vy
Woe (ye [20 ean
‘The reson is thatthe unilateral Laplace transformation of Ea. (28) takes into
account only the behavior of (0) for 1 On that range Kut) = K,
Te should be noted thatthe time functions listed in Table 2.2 include all
‘commonly oocentg input signals for cotingou-sie cuits. Their Laplace
transforms ae al rational frcions (ie, ratios of polynomials) in s; thet
ener form
Fa) = @2)
Property 3. alo included as the third rule of Table 21, is of great
importance for solving problems involving coavolution, Appiyng 10 Ed
25)
Ys) = xt) en
results. Here, Y(s), X(0) and (9) are the Laplace transforms of ye), «(2
dnd. (). respectively. Thus, the integral relation of (25) becomes simple
‘uligcation in the + domain. Hs) often elle the rans fncion OF
syste ction
nthe Laplace transform domain, assuming zero init cemiins. ha is
{q(0) = 0 and tc(0)~ 0, the relations (2.1) and (22) become
Hs) 3b 4()
and
1
Vela) ~ leo ey
‘This indicates tht voltgee and curtets of the inductors and capacitors
satisfy relations akin Ohnts aw i the Laplace transform domain. The role
of resistance i played by the impedance Z: this is 2, ~ al (oe an indtor,
while Z. = 1/2C for a capacitor The analysis of sontinvoustime cre
‘with zero inital conditions and X(s) =1 can be performed, for example,
‘sing nodal analysis This voles the solution of system linear equations,
where each coefficient is constructed from Zand. Z,_ by rmltpheation,
division addition, and subraction Coasequenty, ll oeicients ate eaionl
unctions of «and ence (a ean be deduce by cosiering Cramer rl) 50
TWELAPLACE TRANSFORMATION 27
'& Y(), Since for X() = 1 we have Y(s) = H{s), we conc that MCs) i
tbo a rational function of
"As mentioned eli, ll commonly used iat functions (4) have atonal
Laplace transforms X(s}. Hence all commonly occurring ¥(s) are also r=
ional anetions
view of the ao, ts sficent to discuss the inversion ofthe Laplace
leansormation for rational fnctions 3). The spent tehigue in this ase
is based on parilracton expansion, The proces wil be fst usraed by an
‘example and subsequently Eeneraied Consider the iverson of,
feeeed
OU brrs rates
a1)
We want to break Fs) into sum of simpler terms, preferably such as are
contained in Table 22, that can hence Be inverted hy inspection. Since
SSS Tee Bm (ds AXE T= (52 JOKE KS Dh
G
Fs)
" a8
whore the poles ares; = =2.+ 2,5; = -2—jfvand s, = —1. Cleary
o.0|
an
Note that since F(s) contains s~ sy in its denominator, this factor ean bs
‘anole in (5 ~ 5,)F(3). Hence, setting ¢ = 5 18 2
Ga [=a FoL
[aa] om
Simi
Werle [Ea] ee em
wt
6.- toa )roYion= [EZ] 20)
wt, 4 WHANGFORUATION METHODS
“The fist wo terms canbe combined to eliminate imaginary quantities
i wh
FFT” (saa oy
and hence
rue 2 om)
Tanea aT
From Table 22, therefore
Ma) = Ye Msin 21 +e (22)
{Csrly, in the general ease ofan F(x) sive by Eg. (212) with m=, we
‘an use the partial faction expansion
Fu
att
e2)
where the 5, te the zeros ofthe devominator of As, and hence the poles of
FFU). The residues Care given by Cy= 44/5, for m=, xer0 if m= nd
Galen sl Moy Fe Men (225)
Terms corresponding to complex conjugate poles s, may be pairwise combined,
and then Table 22 used to oblai the inverse transforms of all erm
Equations 224) and (25) hold for simple poles only
1 fllows from the decribed procedure, and fom the ast thee entries of
‘Table 22 that corresponding 0 cach pole so, + jh of Hs) there will bea
term ofthe frm es re! cos. of e* sin Bin the impulse response!)
fof the system. I the weal pact 2, of the poles postive, then the mageited of
{he comesponding tem is exponentially increasing with time, Thus. for any
input x(0), ye) -» 20 as 1 ce. We conclude therefore thatthe transfer
function (1) of 4 stale sistem eannat tae pols with postive real parts
“Thus all poles are inthe lethal of the s plane (Fig. 2.6) Any imple pole
with vero eel part resus in term cos bo sin tin the transient which
Continues with constant amplitode ‘Tis is posible, for exatple, Tor a
Teslss ect? Malile imaginary poles, however, are not possible
“Thee ip aso a general formula avalable* for nding the inverse Laplace
transform of F(a) Let the fal part of the complex varihle #~ 9 + 0 De
‘hose larger than the ral pars ofall singularities (eg, pos) of Fa) then
(0) ean be found from
fo
liters ex)
for 1> 0. This formula dificult tows for actual caleltions; however a5
ssl be shown in the next ston, it has important theoretical implications
“The Laplace tansformation, as shown above, is a eletive tool forthe
‘mathematical analysis of continuous-time systems. It does Do, Bows¥e. pe
‘ide much physical insight ito the operation of the system. slight mafia
‘ion of the defining formula. which allows physical interpretation. results ia
another important tasfermatin, tobe discussed next
23. THE FOURIER TRANSFORMATION'«
“The ia Fourier transform o function fs eed by
Fuh = [" fem, em
il its inverse can be shown tobe
KO= EI” Fiche de (28)
‘A comparison of these equations with (28) and (226) shows thatthe
Fourier transform of /(1) ists (lateral) Laplace transform, evaluated for
=e (ie, for 9 0} in fact, these two transforms are essetialy identical
they dir onl im notation and application. However, the Fourier transform
also has @ move direct interpretation. Regarding the itepral 3s the liming
form of summation, (228) shows that (7) can be regarded a6 4 sum of
Snusoidal funetions e™"~ cosa + jsin the weight (or amplitide ofeach
Sinusoidal is F(ju)/2e. Hence, F(ju}'8 often called the spectrum of f(D TABLE 23. Key Properties othe Four Transtoraton
1 Fs)
dave teh)
fl pover Shri + arse)
faery Hwee
Snorar ef rere do
Went) Riaie
fine Fe fa
Mito bate, EAU) Bie)
Due to the bas similarities ofthe Laplace and Fourier teansforms, many of.
the properties ofthe Laplace Wansformation, wiscassed in Section 22, have
Similar counterparts forthe Fourier transformation. A numberof key proper.
tis are listed im Table 23. Alo, afew important wransforms ae incladd in
“Table 24. Note that the convolution (0) (7) iicated in Table 2.3 isthe
nlateradconvoton defined by
Howe = [Alo nle~»)de 2)
“The formula given in Table 24 fr the Fourier transform off fs shows tht
TABLE 24, Some Commonly Used Foute-Transiorm als
fo. Fue) -
au 1
we s+
pair @/o)sinw
oa) yopinaty
sings fh lee
Lo. wie a
BeBe)
deatay
Filan) a8 tat)
SAMPLING AND THE TRANSFORM 31
‘tthe spectrum of the inp signal is Xo), then that of the outst signals
Y(jo) = MLjo) Xe), (230)
Here, H{ jo) i the spectrum of the impulse response) ofthe et,
often cll the fregueney response of the network. and is usually ity single
‘most important charters. It determines, as Bq. (2.30) shows, ow the
‘Stent affects the various frequeney components of he signal transmitted from
{input to its output terminals
the next seton, the use of transformation methods for sampled-data
signals wil he dieansed. Firs, however, an iterpretation ofthe sampled ta
Signal a liming ease of a continuous-time one wll be given
24, SAMPLING AND THE TRANSFORM! -<*
Consider the circuit shown in Fig. 22, The switches Sand. 5, make
momentary comet atthe istanes shown a8 and fy, respectively in Fig
27h. Each time 5 closes, C charges to the instantaneous vale fn) f {0
‘When + seconds Inter S; close, C is discharged. The volage ses0ss C is
amplied by the butler anplier wth a voltage gain of k. As a rel the
‘ontinoous input signal (1 is raformed ito the pulse tain f°). Cleat,
the ath pals, located between nT and nT +1. can be written in terms of the
step function defined in (211) a8
Alt) = (aT u(t = aT) = u(r = eT 1)], (231)
Hence, suming 4) = for <0
10) = E404 E sorte ~ a) ~ ale aro) 30
Next, the Laplace transform F¥(s) of {2 wil be ealeuated. Using the
relations f(0~ ayuda) ees) trom Table 21, and ul) + 1/5 from
Table 22,
Ps
AE pearly = sderns]
ES Bron ex
rests
“The simple cicut of Fig. 2.7 i a sampling stage. It may oF may not be
sell fora given practic’ aplication; however itcan Be used usa model for 432 TRANSFORMATION METHODS
1,00”
o
[HGURE 27. Sani (a sampling si: () comin ga) 0 i amp
fone 4
‘costal discussions For example forthe generation of the sampled signal
(nT sown in Fig 212, we can assume tat the intra + BEEWEE fa
{smade very sm Then the factor prceding the summation in (233 sais
the approximation
(os)
oy)
“To simply the discussion, we choose k = 1/r. Then the area, eather than the
amplitude, ofeach pulse of /°2) equals /(n7) With these asumptions. (233)
becomes
r)= Eparyent= Ear (23
where the shorthand notation
er 039)
is use. 1s customary to denote the right-hand side of 2.35) by F(z) its
Called the (unilateral sranform of the set of samples f(nT) (For the
‘lateral z-transform, the lower iit of is ~ 2c.) Not that he z-transform is
relted onfy tothe samples /(nT) of ft independent of the function
‘aus between the samples. The set (f(T) often called seqwence or me
‘The process leading from f(0) to F(z) wll be illustrated by the simple
example /(7)= ¢-"u()- Thon the tne series contains the samples
ye leoh nz0 :
sory (en 228 en
and hens, by (2.35, transform Ben By
Aay= Efren Siete ye 238)
‘The infinite series in (238) converges only if Jes) > 1, thats
For such 2 value
Fa)= ae ex)
Clearly, the seansform of a sampled function dies fom iy Laplace
teanfarm only by the ue of the shortind notation (2.36). Hence, the
properties of 2-ranforms are readily deduced fom those of Laplace trans
Forms. For our purpose. the most important ones ae the following
1. te the stransorm ofa sequence fn) i F(z), then that of he delayed
Sequence /(n? ~ KT) is? #F(=), Here, mand & are positive integers,
Wiis assumed that f(a) = 0 for m <0, Thus if fT) +> FC)
then fin KT) #2-"FU3).
2. Afnty o F(2)and fa7) % A=) then heir convolution sats?
Aor fat) ® par yar ~ kT)
FEL) ea)
Property 1 canbe derived directly from the defnition (238) of the = trans
form. Aleratively, it can be seen fom the siting theorem f(t a)
fe F(s) of Laplace transformation, wth a = kTand e” = ¢ 7 =z
Ins obviously uel forthe analysis of sampled-lata systems, soc as that
Shown in Fig. 23. Applying it Bg. (24). we get
'y(2) = 6x(2) (at)
M3) 92"¥(2) ~ aye 1M TRANSFORMATION METHODS
‘TABLE 25. Key Properis ofthe =Transtormation
— a
Cia a)
Theat) shhh) Re) + KC
which cam be solved to give
ve) ea)
Property 2 is appicble to the input-output elations (2.7) of linear
time-invariant stmpladdata stems,
"These. and some other Key properties ofthe =-ransfrmation age collected
in Table 25. Some importnt sequences and their -ansforms até listed in
‘Table 76, Since unilateral =Uansforms are dieussed, all equenses ate as
Samed #0 be zero foe neptievales of, of eauvalenily tobe multiplied by
thedsetetetime unitsep sequence
worya {he 228 ea
‘We sote that all entries on the right-hand side of Table 26 are rational
fonctions of 2 Since the sates onthe left-hand sd include all commonly
ted input sequences, the sansfrmed input X(2) ofa diseretesime system
{s usually rational function. I the aperstiss performed in the system ate
Fesriced to additions, mullipications bya constant, and dey (ass the ease
for liner fiers) hen all signal nthe sytem will be atonal f X(z) i. Thus,
3nd H() are alo rational
TABLE28. Some Commonly Used 2Transtorm Pairs
iO) Fe
kaa
(7 sing)
Geteoait)e re”
con)
(Gate e+ a
Applying Ea, (240) of Property 2 27
¥
HE)X) ea
‘esas. I the inp signal isthe impulse function given in (2.6) then (rom
Table 26), X(2}= 1 and hence Y(z) = (=). We ths comm that (2)
the zransform of the impulse response nT) defined in Seton 2.
"Equation (2.48) gives an easy method fr finding the e-tansoemed output
signa ofa discrete-time system i simp signal and imple response ine
iowa. To obtait the output sequnce ia the time domain however, the
iver operation of the sransformation must be performed. This tsk i
"ually made simple by te rational character ofthe signal F(z) at hand. As
‘wth the Laplace transforms, the simplest available technique i then paral:
{tacion expansion. The purpose itn break F(z) ino stand second-orde
terms, which corespnd dr 4 the entries given in Table 26. Sine al of
these (except the HES! one whichis constant contin a Tacor is toe
expedient to expand F(2)/2 rather than F(2). The process wil be illustrated
by an example. Let T= Tang ‘4 TRANSFORMATION METHODS
“To find the Cy the tems on the sight-hand side ae recombined, and the
fe Me) +e) 4 GE =F
is equated to 8:2 ~ 302 +19, From the equity ofthe coeticiens of like
power, the linear equations
Gan
256, + G26
46-464 G21
rena, Solution gives C, = 5, y= I, and Cy = 3. Hence,
aes
vey
FC)
Using Table 26 for T= 1 the solution
fT) = (540434 )U(nr)
is obtained
‘Toilluseate the se ofthe etransformaton inthe analysis of a dsretetime
system the system of Fig. 2.3 and it input-output relations wil be revs
“The ber was piven inthe Gime domain by Eq, (24) and inthe 2 domain by
GQ) Let T= 2, b= 1, a, = Teand a; = ~05. The sep response gt) wll
be calculated as follows.
‘Since now the -ranformed input is (from Tabi 2)
tion in (242 ives
FS)
%d- Ea
Partataction expansion ves
|SAUPLINGAND THE #TRANGFOR. a7
For the second term, comparison of the coeskents with dose of the fourth
and Ath entries of Table 28 ves
-0s
20" cosu\T = — 2105 cos esl
Hence, @= 2° and cosuyP = sn? = (2/2, Ts, w= 2/4
Partioning the send term inthe form
sos os
pros * Poy 08
weet
ax(2~05) 22105)
. Porvos *F=2+03
Hence, from Table 26, the step response i
aT ~ ala s+ (V2./2)"(~ 20082 /4 + 2sin nn /)]
From the lst four enties of Table 26, its clear that for any pole 2, of
(2) there wll be a factor a™" = 2" in the corresponding wm of the
{eansent response I |e > I, then this tert wll increas exponentially with
‘Thus, ven fr a finite (bounded) input. the output wil be unbounded It
fan therefore Be concluded tata stale dsret-tmessstem cannot hive a8)
poe swith & magntode greater than I, Ths, al pols of sucha system must
Hein the shaded aea(ften ale the imi of the wat cle othe unit sk)
in the »plane (Fig. 28).
AN simple pole for whch [z= 1 les ov the unit circle. By Table 26, such
pole gies fae 0 a transient of the fOr 0s nayT oF sin mag. This. Its
possible for an undamped system. However, multple poles with [sve
‘ncreosing transients and ate thus not posse Cor stable system,
“The preceding discussion, starting With Eq (238, rele 1 the sampled
signal (17) shown in Fig. 21h. However. the cicut of Fig. 27 cam ls be
ned © generate the sampledand-eld signal ty) of Fig. 2c. merely by
sliminating the svitch 5. Then (0) will remain constant between 80
fdjacent sampling instances nT and (+ I)7. The resing time and
‘domain functions can be obtained by substituting + = Tinto Eqs (2.32) and
(235). Thos, using k= 1
Jusbth= EMT lala) — leat TI 248) Viti
and
Fal E poarye™ 04)
results.
‘Comparing Eqs (246) ad (2.38) cea that for given fr) —and thus
‘athe Laplace ranforms of the sampled and $/1 signals dite ool in
the factor
Hyyls) #2 oa)
Hence, while Fy) cannot now be wren simply a ational faneton (ora
jpomer serie) ins, canbe witen a6 H,,(3) mulled by sucha function
[so considera ine sbstem where the inpot signal X(s) isin sampled
finde form, and te internal operations consist only of additions, mul
ations by constant and delaying the signal by T: For such a ystems, all
Stenals incling the outpa Y. all be sampled and-beld ones. The tanser
function if T/X will not contain Hy) since that cancel in the
ts. by the same argoment for sampled-data systems, H wil Be 3a
function of 2 Then Y= HX will be Hf) times a rational function of 2 38
il be al internal signals.
'As was the cast for continuous-time systems the physica epation can be
beter undcrtoal wing spect analysis than general rasformation mets.
‘This wil be disused i the next Section
“THE SPECTRA OF SAMPLED.OATASONMLE 29
2.5. THE SPECTAA OF SAMPLED-DATA SIGNALS
Consider the spectrum ofthe sampled signal desried by (2.35.11 obtained
by replacing « by je or eguivalety, = Bye). Thea
Pus) = Efoen (248)
‘To emphasize the fact that w appears only through the exponentials his
spectrum is often written as F*(e), We shal se both notations
is evident that Fe") ia periodic function of with a peviod 2x,
since ig is replaced ts w + 29/7 ia (2.48, each exponential Factor stises
eomtente, om)
now since @ 7" = 1, FA(e"T) remains unchanged, It Renee as & period
2/7. In fact, it ean be shown (Problem 220) that the spectrum of the
Sampled signal °() i related to the spectrum of the cominuous-time signal
FU) trom ich iis obtained (Fg. 2.7) by
1B rw-soern es
Pe) =
Equstion (230) has some vey important implications. Consider the con-
tinaoatme signal spectra FJ) ia Fig. 29a, 1s flly Band aed, that
‘sitvanishes outside the bounds 4, and w,, where & < 2/7 Hence, whe
(jo) is tepcaad with 4 repetition period 24/7 a8 dicated by (2.50), de
repeat forming F7( jo) do net overlap (Fig. 29H). Thus, thereis a one-o-on€
reaton between the values of Fj) 8 #0)
By contrast the tals” of the broader specrum Py) extend beyond the
bounds 42/7 (Fig. 294) 0. uy Thus, when itis Feplicated Fg. 29D)
‘these tis overlap, andthe value of FR(ju) a any frequen w i nuenced
bythe values of Fj) at several diferent frequencies. This phenomenon s
called lasing of oling. 183 nonlinear distortion
For Fj) the spectrom and thus the tne faneton f(r) canbe obtained
from the Sampedstenl spectrum uniques, and in straightforward Wa.
Specifically, ian ideal low-pass filter hth qransfer Fnetion
Hijo) = {ts Wels 9/7
fo. jel> 2/7 es
' wed to proses the Sampled signal 721), then only the “main lobe” of 40 TRANSFORMATION METHODS
©
FUME 29S pa fr) miusine .on (6semlbn
Fx(s0), Wing in the —2/T < w = #/T range, will pass trough the fer
Since thi mai lobe is (apart fromthe unimportant factor 7) identical to
Fe), the continuous-time signal f() i ths recovered. By contrast no
sich operation can regain #y(u) fom Fy jo), ince no part of the latter
Fetaine the orginal shape of Fx 70.
‘We can thus cole that if the sampling frequency 22/7 is large than
twice the maxima Fequency (4, nthe speci ofthe comtinoue-time
signal, then no atiasng distortion will tke place. The eiginal specuum
Fis) can then be revered undistorted from the samplessigna spectrum
Dy using aToapassfter with» cutoff frequency 1/7. I onthe other hand,
25/7 is ks than twice the band limit of the onginal signa, then aliasing
‘oocurs, snd the continuous-time signal iretnevably lost
“This observation was Srst made by Nyquist who ao showed that if
15/T> oy then the feel for recovering J() from the samples (nT) is
sinl(2/T\e~ nT]
Ho= & por 32)
guaton 252 (oll sip rom Fg. 29 (Problem 221)
“hematin on the sping rate
eee 2.9)
Fone, es)
ifn aad N's etre oer ond 2, om the sping
uence the yg en psp set band
lead hemes euro; oy abn Seated ach
that he amped [F.6d sow ihm eqn waa oe
ioe
"So far, the discussions involved a sampled signal with zero-width samples.
(70, For simp (8/H) tonal te aps ban hy
cing += ne Laplace nsorm 8)
E parye (239)
Fay Jo) =
‘A comparison with (248) shows that Fy( 0) = Hy jo) P*(jo), where the
factor
teen sine /2)
or?
Hy Je) ®
Tem (233)
is due wo the holding operation, Hy jo) can be regard as the transfer
function of the holing device. Not that Hay(jo) its a near phase Is
ample response is show schematically in Fig. 210 1s often called the
(sin x)/ response,” and x characerisic of smpld-and- hed signal spec
‘Next, the spectrum of a continuous-time signal (1) wil be compared with
that ofthe S/H signal obtained from it. Assuming that the Nyquist erin
(2:53) holds, the siuation is a8 shown in Fig. 211. The orginal spectrum
FLJs)ischanged into
sneray 3
Fogo) = nt SST
a oe
15x combinaton of Eqs. (2.4) and (250) shows. Fj) is thus replicated and se ae
OURE 230, The pe nope Ml = 2unt6/
‘multiplied by the “(sin x)/+ response, Asa 1s, he main lobe ia 2/7 <
G'cia/T is no longer quite the same as F(Jo), and the sidelobes centered
round 420/T, 242/T,... age greatly reduce,
“The sistortin ofthe mui lobe is simply due othe factor Hay je). His
ths leary distri as oppose tothe morfincar distortion which abising
Ingoduces. In fac F{j) can be roovered foe Fyy( Jo). if Nyguss
triterion fs satised, in two steps The Hea lowpass ber wih the ansler
function H(ja) given in (251) can be used to select the main lobe ae
suppres all sde bes: then an “arplitde equlier™ stage with a wanser
function Hj) = 1/1) wll store Fe).
‘In practice, Nyquia' criterion often sais not simply by choosing she
sannpling rate 20/T higher than 2ey but rather by reducing w,. Ts before
Sampling the continuous-time signal /() is pased tough a lowpass fer
oe)
®
am f | \ a.
ot l as
Youre 25% syn pct (oo 0 (8 an
Sa ea
THE SPECTRAOF SAMMLED.OATASIGNALS 19
PURE 2, Ht dan ft em whined ap
(called an amtoing te) which reduces its bandwidth tothe ~#/T < w <
19/7 range This ean be accomplished by the eal low-pass filter of a (25).
Having thus “cutoff the as” of the specu Fj) which extended beyond
‘Ea/T._ fla) can be sampled (or sampled-and-eld) without aliasing The
sampled Signal can now be prcesed in a sampled-data stem, Finally. if
‘ecesary, continuous anal can again be eine by selecting the main
lobe ofthe sampled-data output specter with anther low-pass ltr given by
(2.51), This second filter is often called morhig ir, since it provides
smooth continuous-ste signal f(t) from a sampled (28/8) one. Wen $/11
System i wed, then (as discussed ear an equalize may also be needed 19
rode the (sn x)/ amplitade distortion.
‘A schematic representation ofa sampled-data system with continuous-time
input and output signals therfore of the form shown in Fig. 212. The
mialasing and smoothing filters ean be denial, and should ideally have the
Uansfer function given in Eg. (281) Ip practice of couse. the infinitely sharp
‘tof implied Py (251) can only be approximated by areal ter. Hence, &
margin of 84.20% or more is allowed in the selection of the cut Trequeacy
ofthe filters so thao, = G8=/T can be chose.
Inthe foregoing discussions, ana teciques have been deseibed for
continuous-ume and sampled-data systems. Next, the mathematical tools
Alveloped in the course of the discussions will be used forthe design of
sampled-data ters,
2.6. THE DESIGN OF SAMPLED-DATA FILTERS FROM.
CONTINUOUS-TIME MODELS
‘Consider the important case when the sampled-data sytem of Fig. 2.12 16 to
perform asa filer Sampled-dta fiers can be divided into two categories
Jniesipulre-respone (FURY so infnteopule reponse (J1R) ters, As the
‘ames imply, i the impulse 8(r7) of Eq (26) is applied to the input of an
FIR ter, then its output signal i its impulse taponse will
soniain only a Bait number of samples, So that h(n?) sof fine duration. By
onitast, the impulse response of IK iltersis infinitely loag Physclty, an UR.
iter as feedback loopn in which a signal, ace injected by the inp 827),
an eiealate forever An FIR fier typiily has no such Toop, although ia
principle itis posible to design an FIR iter which contains feedback and
Fis on the cancellation of signals omit he dration of nT). Mathemai
Cally. as Table 26 shows, any term inthe pats-racton expansion of (2) «4 TRansronmarion METHODS
ih ae of theo K/(2~ a te! +a +) oer to sequences
cee etc" Sac sah oma nnd yt pos of 3)
Sen een can, ne cnet a pols ofthe aster
st a cr mat cir ats = Oars Hence for sich
en) ures in wih i umber fens Ths
Haye Bae es)
ere nn <
‘Since Hs ‘the z-transform of the impulse response (rT), that is,
me) = Eater
238)
easly h, = h(n?) 8 the coon in (257) are the simples ofthe impulse
response nd a> ™ 20
Tra highly. sletve ter is requited, then (2) needs poles (natural
frequencies) neat the etl of the Hite, Henes TR fers are preferable in
ck applications. For an UR filter, (=) a vainal fonction, that is the
Tati of two polyoma N(z) and D(z):
Ne)
ME) Dey
Hence, some pols are ite and nonzero.
Tan senspecteaom th anstrfonton (2) ofan HR ker
is pov found ively. Instead. advantage taken of the fact that large
Sate of wansfer function has been compiled ove th years for contin
‘Shrine bites Ge, ¢, Refs S and 7-9). Thee functions Fs) etn alo be
the, ath minor Modcations Sore design of sampled-data ltrs using the
teansformaton methods which wil Be dese inthis ection,
"Al statis which wll be discussed below for obtaining, (=) from @
mel” tanner Tunetion ae based on 3 transfmaton of the frequency
‘arable. Thus, It the transfer function H,(s,) of a continuoustime fer be
[Phen (Heres and in what follows, the subscript i used to distinguish the
arises ofthe "sod! filters stands for “analog” but foe our purposes i
Jmplies“contiwus ime”) Thea, we wl obain A) simpy by replacing 5,
by some fonction
fe) oo)
in Ths the desired transfer function is
Nar Snoros= MLMAD 261)
MC
What are the requirements for /(2)? As mentioned eat, W(s,) a etional
funstion of, while H(2) i rational function of +. Hence f(z) mus eal
tea rational expression in. Futhertore for an analog filter, the fequens)
response H,( Jo, ofthe model has some desitable properties (usu selet*-
iy) which we would presimably ike to obtain alo forthe frequency response
12?" ofthe sampleddata iter to be designed. Hence, the frequency ax
5, =, ofthe splane mist be tansfred into the unit citcle 2 =e?" of
the plane. Finally the H,(2,)foncins are tabulated ely or stable (pas
sv) ler, snd (ones an osilator is required) the Gal smpledta Bier
‘designed from #7(2) must abo be stable, From the stability conditions dis-
‘used ear (cf, Fin. 26 and 2.8, we know therefore thatthe pols, of
H.Asq) ae inside the ltl of the 5, pine, while the ples of (=), en
ty the elation
su Slee ea)
‘must be inside the unit ctcle inthe z-plane. We conclu that the mapping
(260) must teansform the left hall of the splane andthe zplane unit disk
nto cach other The conditions of f(z) are theseore teflon
1.2) must be a rations fnetion of 2
2 For|2|= 1, f(z) must be pure imaginary: /(e7) = ja, Vice vers if
5, = f(2) is imaginary, then [|= 1 must bold
3. Fo [e< 1 the rel pat of sy = /(2) aust be negative. Vice vers, if
Res, 0, then te cvesponding ? must have at absolute vale less
than’
Rational funetions which satisfy these conditions ean be found heurisital's
However, a systematic derivation, based on numerical integration tchnigus,
is als possible. The latter wl be described here
‘Consider a model cotinuoustime filter Which has the ransfer funtion
1,{2,) ts response can be determined fom a system of firstonde diferent
sguations
ast)
de
= a0. ae, 0.63)
Here the x, ate the sae cori ofthe fer (Ref. 2, p. $01-522), and the
relations in (2.63 ae its state equations. The) are linear functions ofthe
State variables 1) andthe inpot smal Using Laplace ansformation, (263)
becomes
8K.) ~ Gls,
Bessa W Qe
Where) ~ Ofor 1 ¢ is assumed and the subscript used to inde that
the expression refers to the analog mode 446 TRANSFORNATION METHODS,
Next, we will derive the state equations fr some sampled-data systems
sshich ave approximately the same properties asthe continuous-time fie,
Since the required relations for the sampled-data sjsems are Gist-order
tiferece equations, (263) is ategrated between (0 ~ 1)P and nT:
pr at)
fo AIP a= sory — sor 7)
wf sl FAN (8)
“This transforms the LHS of the equation ito the proper diference form. To
caluate the
Forward Euler Integration
“Tis poses is based on the approximation. dv to Euler,
[stows rer—1) (206)
than 2m tuna it
ana ts katy ain) fe een og)
Sacer te rand te a
Sotho tpi
ecantonsppitg 28)
ftom (2.66) into (268), the diference
ada) — (eT 7) = Tao 7) ea)
ress. Using z-trasformation, (267) ves
x(a) KI
[Note that (2) = XL /()] 168 Affero function from % 4s.) een though,
for simplicity, the same notation is used.) From (268),
Te 62). (268)
xe) =
08)
T
‘A comparison of (2.64 and (2.69) suggest the transformation
(220)
[Next the transformation of (2.70) wil be tested with espe to Contons
1-3 Tis clearly 9 eatnal fonction of = as required Condition 1 Testing
FROURE 212. Nemec nye sigs (0) forse nega: (0) ake
for Condition 2 weet 5, = j. Then
pesrt
yer
and hence [2|~ 1 holds only for, = 0. Figures 214 and 2.148 show the
image ofthe jo, anisin the eplane. Since the image sa vertical ine tangential
tothe watch for ja,7]-« weave [2 = 1. Thus, Condition 2 met oly
pproximately, and only for [a 1/7. The igure also shows three poles in
‘he 1lane and their images in the s-plane. lea ll three poles elons toa
stable continvousime titer. However. inthe =plane, while i sable.
‘season the unit crle and 2 outside it. Thus, the (2) obtained Wing
the “forvaré-Ealer™ transformation of (2.70) from H,(4,) = 1/1 ~ ba
(3,212 ~ 3.) isin this ample unstable, Hene, Condition 3 soled
for this mapping. Even if all poles of a #8.) are mapped inside the unit
‘rte, usull the “dominant” poe Le, that closest to Ye janis) will nove
‘ery close to the unit etl inthe =-plne- Hence, 2 peatung of the passband
{ain occurs inthe sarpleat iter (F952 18a and 2.155). Alsi the 2205
OC H,(S,) (oe. the tansmision 208) ae on the Jans hase of (=) will
hot be Om the wait cite, Hence, H(e?™") wll aot he vero Tor any’ ab hus
the stopband response will dtesorate (Figs. 2154 and 2155) OF course, » 7
© cma
FHOURE 214. tron thr # maine (a pe and urea ce pe
{hn oe ene aor) em
(e!") is periodic in w. with a pesos 24/7: Only part of the (0.28/7)
period i shown nthe Figure
Backward Euler Integration
Using the rectangular approximation shown in ig, 2.13 forthe area under
the gu) cure, we ob
em
{pen ean ae th ria! nd
Substiution into (268) gives now
dnt) ~ (aT 7) en
Hence, by =transformaton,
e33)
om) 0 TRansFonMATION METHODS
“This is called the beckward-Euler transformation, 118 sin rational, 38
quired by Condition 1. However, fo 5 = iy
1 tier
- as
“ 7 =jet” 14 eT @as)
Cea, el 1 and Condon 2s te In fast x andy sasy
(ere Gy e
scan ely bse Prblem 25, Hes, the ima ofthe as inom
PTAA eas os emer ena? (ig. 2). The tee
sual og syand 2 nw the mapped, and eee
Mem AP elt table HC) mapped now able 2) and
ees Shots The damian ples ow move vay ro the un ci
see ee naman the pasta spose els (Fie 2130) Since the
seer Poe an off te utc, tepand response wil a Be
ey ced, Me!) cone, pain the ee shows ony
dar tone pod
“Traperoviel Integration
[A better approximation than those hasol on the Euler formulas can be
tained ifthe area under the (1) curve for (n= WT s 1
You might also like
