Table Of ContentANALOG CIRCUIT
DESIGN
Volt Electronics; Mixed-Mode Systems:
Low-Noisc and RF Power Amplifiers
for Telecommunication
Edited by
Johan Huijsing
Det ane of Tetley Bt The Neri
Rudy van de Plassche
vats Unies of Techs Bader Te Nehnds
end
‘Willy Sansen
Kelis thiveriet iwi Beige,
*
KLUWER ACADEMIC PUBLISHERS,
IRNSTON / DORDRECHT / LONDON
‘A.C. Cala eid oe this ki ral fra i Libary of Cones
15 0 7502 HO
Publ by Kies cadens Paso,
‘pow 17,3900 AR Doro, The Nema.
Sold nd ste a Nor, Cel an Suh Amon
Py Alone aod Pps,
101 Pip eis Morel MA CASES A
Prine ses ee oper
Aa Riba Reseed
2199 Kpnwer Adee Pairs sim
‘peat i mo psy he ay mabe
ead photopic bya oom ap nd
selva syeeo. Widon ween pcan ea aca et
Prime the Neel
‘Table of Contents
reface
Part 1:2-ValtElectrontcs
abeduetioa
fe Teautlinear Circuits
‘Wouter A. Senin, Jan Mulder, Paul Poort, Michiel Kouwenhoven, ie van
Staveron and Art ILM, van Roermad
eV LegeDomata Fters
(Girt Ee and Mantred Punzenherger
1V pricned-capactiorfters
A.Baaohirowo nd X. Carlo.
AV AE AD Converters
V.Bebaoo, ML Seyoer and W. Sansea
Jefe ciecut Deg or gee
BA CAL eee
DOC Conversa, the key low power sousemption,
BE. Oimans and FU. Shae
‘art: Design and implementation of Mixed Modes Sysreais
Tovadiction
‘Sobwtrate Hounce la Mixed-Mfode CMOS ICs
‘Sewn Nevia wd Gion Tlougzaad
‘Fesatory tmpacts on
Frapéais TR. Clement
Dene Techniques te Reduce Substrate Nolet
‘Taliellac.
12644 CML transcelver with IM CMOS AIMSDH Processor ia
BIEMOS Moncchip
ugowror
Matas None Cculng in ved Sant RFCs
‘Nithadh K. Verghese wa a
trate Naive
3
°
33
m
185
‘Fop-Dawn Design of Mined-Mode Systeme: Challenges and Soften
Gorges GF. Gielen, . .
Part IIt—Low-notse and RF power Amplifis forthe communication
Iatrwieton
‘The Dsiga of Narrowband CMOS RF Low-Notse Any
Thomas H. Lee
Design of Bruudband Low-Neise Amplifiers in Deep-Submicren CMOS
Technologies
hao Fansena and Michis Steyaor
Put your power tuto SOA LNAG!
eter Bates
Ratio ‘Trausceiver Circuits Silos Germann
‘aa Seveshaus and Bars Vestraten
Madelng for tiple Power Aap
Stephen Weber
Design Considerations for GaAs MESFRT RF Power Amplifiers
Stewan Taylor
ws
Bisfce
This book contains the revised contributions of 18 rural spenkers al
the seventh AACD 798 in Copenhagen, April 28-30, 1998. The
onference was organized by Ole Olesen, of the Technical Universi
of Denreark. The program committee consisted of Johan H. Iluijsing
from Detit University of Technology, The Netherlamils, Willy Samsen
from the Katholieke Universitet Leuven, Belgium and Rudy J. van ile
Piassche, Philips Research, The Netherlands.
‘The program was voncentrsted around throc important topics i
analog citcuit dexign, Each of these thrce topies hus been covered by
x papers. Each of the three chaplers ofthis book contains the siz
pers of one topic. The thrce topins are
[Volt Electronics
Decign and implemeniation of Mixed Modes Systems,
Low-Noise and RF power Amplifies far the communication,
Other topics, which have been covered in this series before are:
1992 Opamps
ADC's
Analog CAD.
1993 Mixed-Muile A/D design
Sensor Iuterfuoes
(Communication eit
1994 TLow.Power low-Valtage
Integrated Fitters
Smart Power.
1995 _Low-Noise, Low-Power, Lave
Mixed Mode with CAD ‘irale
‘Voltage, Current and Time References,
‘okape
1996 RFCMOS ciponit design
Bandpass Sigma Delia and other Converters
‘Translinear circuits,
1997 RFA-D Converters
Semwoe and Actuator Interfaces
Tow-noise Oscillators, PLL’s and and Synthesizers
‘We hope w serve the analoy design community with these series of
‘hooks and plan wo catinge Tis series in the future
Johay H. Hoijsing
Preface
1 Volt Elevtronies
‘The strive for more clectronics un a chip requires smaller transistor
dimensions, This in tur tesults in lower breakdown voltages, in the
order of 3 or 2 Volt far mainstream CMOS processes, Hence, tyether
‘with low-cast battery use in wireless application, ihe supply voliages
tne forced in the direction of 1.8 Volt and 89 Volt.
_Atsuch low supply voltages not only digital signal processing but alse
snalog signal processing bas to be performed. The Tatler inaposes a
sdallenge on the analog cizcuit designer who finds the dynamic range
of analog signals squeezed between the supply voltage coof and the
‘aise Door. These circuit design challenges are mit by the following 6
tiapers on J- Volt electronics
Inthe frst paper by Wouler Serdijn, TU Delft, the Netherlands,
(gereral approach for |- Vall truastinear circuit design is presented.
‘The seeond paper by Christian Huz and Manired Ponvenberger
‘Rockwell, USA, evaluates this approach for I- Vall Jog dorasin
‘ler design,
third paper by Andrea Haschiratio and Rinuldv Castello, Univ. af
sia, Italy, deseribes how a switched capacitor iter can he made
Thactioning ata supply voltage of 1 Volt using a switched opamp
nique, The fourth paper by Vincenzo Pclusa and Michiel Steyaer,
KU, Leuven, Belgium, shows how sigma della modulator can be
Aetigned for | Volt.
‘As lowcout battery usc for pagers ik desired, the design ofa 1 Volt
‘Somplete RF font end is presented by Ed Callaway, Motorola, LSA,
inthe ith paper
Finally to brcake aviay from the low-voltage baulery supply problem,
the sisth paper by Carel Dijkmans, Philips Rescarch, the Netherlands.
describes the design of DC - DC upeonverters fat may Cunctiom al a
primary voltage of | Volt,
Joban H, Huijsing
Dynamic Translinear Circuits
Wouter A. Serdijn, Jan Mulder,
Paul Poort, Michiel Kouwenhoven,
Arie van Staveren and Arthur ILM. van Roermund
Delft, University of Technology,
Faculty of Information ‘Technology and Systems/DIMES
Electronics Research Laboratory
Mekelweg 4, 2628 CD Delft, The Netherlands
phone: +31-15-2781715, fx: +31-15-2786922,
email: W.A Serd/jaG@its.tudelft.nl
Abstract
A promising neve approgch to shortan the design trajec-
tory of analog inegrated eizewits without glviny ap fane-
signality is formed by the clas of dynannie tranatinesr cit.
cults. This paper prawaty a structured design method for
shin young, yet rapidly developing, eirenit paradigm,
2 design example, a LV 1.6-pA claseAB translinewr sinh
Integrator for andio fiver applications, in presented
Introduction
Biectronies design can be considered te be the mapping of a set
‘& mathematical functions onto silicon, For discrete-time signal-
Processing systems, of which the digital signal procevsots (DSPs)
‘day are by far the most populsr this comes down to the implemen
‘etion of ¢ number of diference equations, whereas for continuous
‘Hone signal-ymcessing systems, often dencted by the term analog,
4
differenti equations aze the starting points. In mixed analog
digital systems, the analog pacts, however, often occupy Tess thi
tea percent of the complete, i, the mixed anslog-digitl circuitry,
whereas their design trajectory is offen substantially longer and
therefore more expensive than of their digital counterpart. Whete
does this disezepancy erise from? ‘This can be partially expleined
by the fact that, af circuit level, for analog circuits far more compo-
nents play an important role; various types of transistors, diodes,
resistors and capacilors, to mention a few; sometimes also induc-
tors, resonators, and others. Whereas far digital circuits, the com
plete fanctionality is covered by transistors ont
From the above, it automatically follows thal, if we restrict our
selves to the use of as few different types of eomponents as possible,
‘without giving up fonctionality, we can shorten the analog design
trajectory contiderably, in the same way as this is done for digital
circuits, One successful approach, a we wil aee in this paper, is
given by the class of circuts called dynamic transtinear circuits.
Dynamic translinenr (DTL) ciccuits, of which reoratly am all-
‘eucompasting curren(-mode analysis and synthesis theory has heen
developed in Delt [1-3] axe based on the DTL principle, which can
‘be regarded as a generalisation of the well-known ‘static’ tranalin-
fear principle, formulated by Gilbert in 1975 4]. ‘The first DTL
circuit was cxiginally introduced by Adams in 1979 [SI, being a
first-order lowpass filter, Although aot recognized then, this was
actually the first time a Grst-order linear diflereatial equation wes
implemented using translinear (TL) circuit techniques. In 1980,
Seevinck introduced a ‘companding current-mode integrator’ [6]
and since then the principle of TT. iting has brew extensively
studied by Frey (7-16), Punzenberger and Enz (17-81), Toumazoa
et al [92-51], Roberts et sl. [52-87], Tsividia [58-62], Mulder and
Serdiin [63-84] and others ‘85.
However, the DTL principle ie not limited to filters ie, linear
differential equations. By using the DTL principe, it is possible to
"Tv must be noted What for higher frequencies or bitrates, abo the iners
connects coe inte plag- However, ie iaOuease is comiderd to be equally
Important for analog aa wall wm dig syetems
5
{implement Uinear and nonlinear differential oquatious, using, tran-
sistors and capacitors only. Hence, « bigh functioual density eau be
obtained, and the absence of lage resistors makes them expecially
interesting for ultsalow-power applications [76
DIL ciecuts are inherently coenpanding (the voltage swings axe
logarithmically related to the currents), which is beneficial with re-
spect (o the dynamic ange ia low-voltage enviramuciés [87,88]. In
Addition, DTL circnits are easily implemented in class AB, which
‘tails a larger dynamic range and a reduced ayetage curreat eou-
sumption. Further, owing to the mall voltage swings, DTL citeuite
facilitate relatively wide bandwidth operation, At high frequencies
‘though, considerable care has to be taken regarding the influence of
parasitic cepacitances and resistances, which afect the exponential
behavior of the transistor.
DIL circuits are excellently tunable acsoss a wide rauge of sev.
eral parameters, ech as cut-off frequency, quality factor and gain,
‘which increases theitdesignability and makes them attractive to be
ted ac standard cols or progcormmuble building blocks.
The OTT principle can be applied to the stractured design
of both linear differential equations, ic. filters, and von-tnear
Aiferential equations, e.g, RMS.-DG converters (89-91), oscila-
ton [92-103], phaselock loops (PLLs) [90-82] and even chaos, Tn
fact, the DTT. principle laciitates a dizee mapping of any function,
described by diferential equations, oxto wlicon,
Application ereas where DTL circuits ean be successfully used
inslade andio fiers, high-fequency filters, high-frequency ovcle-
te, demodulstors,infea ted Lzout-ends and low-voltage ultra-low.
‘power applications
4 This paper aims to present a stractured design method for DTL
Geouits. “The static and dynamic TL principles are reviewed io
Section 2. ‘The general clase of DTL circuits contains several dif
forent types. In Section 3, the comespondences and diferences of
Jogedomiain, tanh and siub citcnits are teeaved. Finally, Section 4
Prevente the design method, applied to the design of @ DTL in-
fegrstor, starting from a dimensionless differential equation that
describes the integrator behavice in the time domain, Afler four
6
hierarchical desiga steps, being dimension transformation, the in
troduction of capaciteace custeats, TL decomposition and circait,
implementation, a complete eiecuit diagram results, Measurement
renults of the thus obtained DIL integrator, are presented.
2 Design principies
‘TL circuits can be divided into two auajor groups: static and dy-
nemie TL eitcnits. The first group can be applicd to realize a wide
variety of linear and non-linear static transfer functions. All kinds
of frequency-dependent functions can be implemented by circuits of
the second group. The underlying principles of static and dynamic
TTL circwite are zeviewed in this section,
2.1 Static translinear principle
Tr, civevits are based on the exponential relation between voltage
and current, characteristic for the hipolar transisior and the MOS
teansistor in the weak inversion region, In the following diseus-
sion, bipolar transistors axe aasuned. The eolleclor current Zo of a
bipolar transistor in the active region iv given by:
de = Iyeten™, i)
where all symbols have their nsnal meaning,
‘The TL principle applies 10 loops of semiconductor junctions.
ATL Joop is characterized by an even number of junctions [£. The
smumber of devies with a clackwise orientation equals the mumher of
counter-clockwise oriented devices. An example ofa fonr-transistor
‘TL loop in showa in Figure 1. is assumed that the transistors are
somehow biased at the collector currents f, through Jy. When all
levies are cxpuivalet and operute at the same temperature, this
yields the familine representation of TL luops in terms of products
of eurrents:
Dla hale @)
Figure 1: A four-(ransistor translinear loap.
‘This geuerie TT. equation is the bavis for a wide variety of static
lectronic functions, which are theosetieally temperature and pr
eras independent.
2.2 Dynamic translinear principle
‘The static TL principle is limited to frequency-indepeadent teans-
fer functions. By almitting capacitors in the TL loops, the TL
Dlinciple can be generalized to include frequency-depeudent tran:
fer functions. ‘The Lerm ‘Dynamic Teanslineat’ was vvined in [89]
to describe the remulting class of ciccuite. Tn conteast to other
‘numer proposed in literature, such as 'lag-dozaain’ [5], ‘companding
current-mode’ [6], “exponentia? state-space’ [7], this term empha
sines the TT, mature of these circuits, which is a distinct advuntage
‘with respect to structured analysis and sythesis
The DEL principle can be explained with reference (o the su
Sieuit shown in Figure 2. Using a curtent-mode aypmiach, this
incuit is descrihed in terms of the calicetar current Zg and the
rapacilunce Zo, flowing thzough the eapacitance C., Nott that the
ae voltage soutoe Viens 608s not affect fey Aw expression £0r Leap
ean be derived from the time derivative of (1) 16,89):
to
fag OVE, @
where the dot represents differentiation with reapeet to time.
___ Equation (3) shows that Zoay is a non-linear function of Zc and
ite time derivative Jc. More insight in (3) is obtained. by slightly
Figure 2: Principle of dymasie translivear cireuita
rewriting it
OV pic = hepde. @)
‘This equation directly states the DIL principle: A time derwative
af a cervent can be mapped onto a product of currents, Ax this point,
the conventional TL principle comes into play, since the product of
crurrenta on the right-hand side (RES) of (4) can be realized very
elegauly by means ofthis principle, Thus, the implementation of
(part of) ¢ differential equation (DE) tecomes equivalent to the
implementation of a product of currents
‘The DIL principle ean be usc to implement a wide warioty of
DBs, deserbing signal processing functions, For example, filters are
desctibed by Yinear DES. Exasnplea of non-livear DEa art harnwonie
amd chaotic vellatoe, PLLs and RMS-DG converters.
3 Classes of dynamic translinear circuits
Ia al DIL crcl, the wollages ate logically sand to the
current. Therefore, these citcuite are in some way instantaneous
companding, Figure 0 shows the geueral block schematic of an
inatentonvots eompanding inegratar (6. Fn DTL cleats, the in
ternal integrator i a near eapacttance, The expander E expand
the outpit vllage ofthis iniegeatoe into a curt, eplitng the
exponentel VT teansivion teamsor function, Several types of DTL
cireuit can be dininguished within the geaeral cass of DT. ci
cuits based on the particular implementation of B. Next ta the
rmoxt prevalent class of log-domain czcuits, the two cases of txts
9
and sinh cireuits have been proposed by Frey [12". In this section,
wwe describe their characterises, which can be derived from abe
generic output structures, depicted in Figure 4,
-G fT} feof os
Figure 3: General block schematic of ar instantaneous coupanding
integrator.
Figure 4: Generic output structares of (a) log-domein, (b) tanh,
sand (0) sinh elreuits
8.1 log-domain circuits
Most published DTT, circuits aze bused on the common-emitter
(CE) output stage showa in Figure 4{a),characteratc forthe cass
of og-domain citents. The traaser funetion from the cepactance
vollage Vag £0 the output current Ju is given by the wellknown
exponential law (1). Taother words, # equals exp, The compand.
lng characterinics of « DIL circuit cau be dexived fr: she second
onder derivative of F with teapeet to 2, denoted by F". Without
less of gencrality, z= 0's considered lo be the quietcent paint of the
Jategrator shown in Figure 3. Figure displays forthe output
w
stages shown in Figure 4 Applying a stsict definition af compand-
ing, B" should be suictly positive for z > 0 and strietly negative
for z <0, For Ing-damain eizcuits, a comparison of £ — expe
with the strict definition of compamuding reveals that these circuite
are indeed romping for 2 > 0; however, for x < 0 tha expo-
nential function constitutes a corapression instead of am expansion.
Kor a symmetrical outpat current, the overall behaviour of the CE
output stage implies a compression rather thar on expansion of the
peak-to-peak signal swings [88
Riguse 5: The second-order derivatives of the Vf transfer functions
of the output stages shown in Figure 4.
Brom a current-mode point of view, the most imporlant charac-
teristic of a DIEL output structure is the current-mode expression
for the capacitance current Tony. For log-domain filters, dog is given
bby Equation (3), where io = tar + Jn. As shown in Section 2,
Tineae desivative iu is obtained by wltiplying Leap by fac | Jone
‘A favorable property of log-domain circuits is that « linear
damping term can be implemented by the conuection of a de cur.
rent sotrwy Z, in parallel to a capacitance, This can be explained
from Equation (4), If instead of Jap, Jap { 4 1s multiplied by
ag + Tage, ta additional term Z,-(Jac-+ Tot) 8 generated. ‘The first
term ola: represents a de offset cutrent, The second term lala
renults in a nite negative pole,
‘Typically, log-downain cirenits operate in clags A. The actual ac
signal Ine is superposed on a dc bigs current Zac. AR 8 coneeqmence,
n
‘the output signal swing is limiled 40 Joe > —Iie, Note that this
limitation is single sided, whieh is advanlageous if seytumetrical
input wave-forms have to be processed. Thiy characteristic can
tbe exploited to enable class AB operation °6,9), Using a class AB
set-up, see Figure &, the dynamic range can be enlarged without in-
creasing the quiescent power consumption, Using a enrrent splitter,
‘the input current fais divided into two enrrents fin, aud Sis, Which,
are both strictly positive, aud related to Ln by: fa ~ Fat —Toa. The
current splitter impresses » constant geametric or harmonic mean
(on fig and dina. Next, Zinn aa Zin cam be processed by two clase A
Jog-domain circuits. 1 is imporiant to note that class AB operated
log-domain circuits do satisfy the strict definition of companding
foe to the fact Lhat only positive currents are procestett, ie, 238
sever aegative
alae
igure 6: Set-up for clase AB operation,
8.2 tanh circuits
Instead of a single transistor in CH configuration, the class of tanh
‘Suits is characterized by a differential pair vutput struct [17],
ee Figare 4(b). The uame of this clas of circuits is derived from the
‘well-known hyverbolic tangent V-1 transfer function, The second-
order derivative 5" is shown in Figure 5 and demonstrates thnt
tanh cirouits are uot companding af all 1, The differential pair
implements a compression function,
‘The tail carrent of the differential pair ie ade curremt Lie, and
Merefore, tanh circuits wu operate in class A, ‘The output current
Tous isthe difference of the two collector currents. The omtput swing
irs
fg Hmited to —Zae < Jous ~ Jan. Since this interval is symmetrical
the class AB set-up shown in Figure 6 caanot be spplied to tanh
sieuits
‘Fromm Figure 4(b), the capacitance cunt Jag ia found to be:
w= ove ( pte - jie) @
A linear derivative Jaa is obtained by multiplying this equation by
Uae + Fou} Tae ~ Zot}
WV pFackoor = Lapllae + Fase Mae — Zot ®
Comparing Equations (4) and (8), we can see that the RIS of
(8) is third-order, whereas the RHS of (4) is only yecoud-arder.
Consequently, in geuezal, TL loops of 8 higher order are required
to implement a tanh circuit, resulting in a more complex circuit
In addition, « lincar Loas exanot be implemented by a de current
sontce connected in parallel to, capacitance. This leads us to the
concluvion hat tank circuits do not seem to live any advantages
‘over log-domain cirenits
3.3 sinh cirenits
‘The third clase of DTL circuits proposed in literature it formed by
(he sinh ciraits (12). The outpet structure, shown in Figure 4(ch,
is & complete second-order TL loop. It implements the geouetsic
mean function 13, = ZounFous- The actual omtpiit current Zou is the
Ablfecence of Fagt Saul foyer. Since both Ioan and Joua are alovays
positive, the sinh output structure operates in class AB, which is
beneficial with respect to the dynamic range. The V-1 transfer
fanction of the outpat stracture is a hyperbolic sine friction. Fig
are 5 displays £" = sinh 2 and chows thet the sink output stage
implements a genuine expansion function.
‘The current-mode expression for the capacitance current fea is
a
given by:
Ine = ®
(3)
= (a)
Foy
= OMe om i Tom (a0)
A linear derivative sis obtained ty mlliplying fag by the sun
To ~ Jous. Ts inteesting to note that the voltage Vay and the
turrent Zogt—Zaa ate related through a hyperbole cosine fiction;
‘the first-order derivative of # with respect to z.
4 Structured design of a class-AB dy-
namic translinear integrator
Synthesis ofa dynamic crust, bet near or non-linear, start with
DE or witha eet of DEs describing its function. Ofte, i is moze
oavenieat io ase astae-space description, which # mathematically
equivalent. ‘The structared aynthest meltod for DTU cienits i
Ulustrated hereby the design ofa fstonder integrator, described
inthe time domain by:
4
dr ay
‘Thin equation desentbes the integrator output signal y a8» faction
of fhe input sigaal 2. +i the dimensionless time ofthe integrator.
4.1 Transformations
a the pure mathematical domain, equations are dimensiontess
However, ay soon at we enter the electronics demain to find an
Description:tors, resonators, and others g Equation (3) shows that Ieep is a non-linear function of Ic and IEE colloquium on Analog Signal Processing, pp.