scieee Open visual document viewer

Behavioral modeling of PWL analog circuits using symbolic analysis

Fernández Fernández, Francisco Vidal; Pérez Verdú, Belén; Rodríguez Vázquez, Ángel Benito

Abstract

Behavioral models are used both for top-down design and for bottom-up verification. During top-down design, models are created that reflect the nominal behavior of the different analog functions, as well as the constraints imposed by the parasitics. In this scenario, the availability of symbolic modeling expressions enable designers to get insight on the circuits, and reduces the computational cost of design space exploration. During bottom-up verification, models are created that capture the topological and constitutive equations of the underlying devices into behavioral descriptions. In this scenario symbolic analysis is useful because it enables to automatically obtain these descriptions in the form of equations. This paper includes an example to illustrate the use of symbolic analysis for top-down design.

Full text

Beha io al Modeling o PVVL Analog Ci cui s using Symbolic Analysis F ancisco V Fe ndndez, Bel& Pe' ez-Ve du' and Angel Rod iguez-Vdzquez Ins i u o de Mic oelec dnica de Se illa Cen o Nacional de Mic oelec 6nica-C.S.I.C. Edi icio CICA-CNM, A da. Reina Me cedes sln, 41 0 12-Se illa, SPAIN FAX:: 34 5 4231832 Phone: 34 5 4239923 email: [email p o ec ed] ABSTRACT^ Beha io al models a e used bo h o op-down design and o bo om-up e i ica ion. Du ing op-down design, models a e c ea ed ha e lec he nominal beha io o he di e en analog unc ions, as well as he cons ain s imposed by he pa asi ics. In his sce la io, he a ailabili y o symbolic mod- eling exp essions enable designe s o ge insigh on he ci - cui s, and educes he compu a ional cos o design space ex- plo a ion. Du ing bo om-up e i ica ion, models a e c ea ed ha cap u e he opological and cons i u i e equa ions o he unde lying de ices id o beha io al desc ip ions. In his sce- na io symbolic analykis is use ul because i enables o au o- ma ically ob ain hesi desc ip ions in he o m o equa ions. This pape includes an example o illus a e he use o sym- bolic analysis o op-down design. 1. INTRODUCTION Ci cui analysis is he co ne s one o elec onic ci cui en- ginee ing. On he on1 hand, i p o ides he keys o unde - s anding he in ica e mechanisms unde nea h he ci cui op- e a ion. On he o he ^ hand, designe s use analysis o ob ain models o he ci cui lbeha io - he basis on which his be- ha io can be p edicded. Howe e , manual analysis o e en he simples ci cui s kncoun e ed in p ac ical applica ions is a complica ed, ime-consuming, and e o -p one ask. Sym- bolic analyze s a e in ended o elie e designe s o sys ema - ic manual analysis hsks, hus le ing hem concen a e on c ea i e issues. I The las gene a io? symbolic analyze s a e able o handle up o a ound 400 di b en symbols [ 11. This pe mi s, o ex- ample, o analyze di cui s as complex as he ail- o- ail CMOS opamp o Fig!l(a) 121 wi h he small-signal ansis o model o Fig. l(b) [3]1-[5]. The analyze s capabili ies include he calcula ion o s-dbmain exp essions o all ypes o d i - ing-poin and ans e cha ac e is ics, he simpli ica ion o hese exp essions o e ain only he dominan e ms, he ex- ac ion o hei poles^ and ze oes, e c. Recen ly, di e en au ho s ha e ocused also on he sym- hose weak nonlinea - I bolic analysis o no ci cui s [6][7]. The mos impo - a ew e ms ( ypically 1. This wo k has been pa ially suppo ed by he Spanish C.I.C.Y.T. unde con ac TIC97-0580 ind he EEC in he p ojec ESPRIT AMADEUS. I I I I B Figu e 1. (a) Rail- o- ail opamp; (b) small-signal MOS ansis o model. 3) o a powe se ies expansion. In [6] sys ema ic echniques a e p oposed o he au oma ic calcula ion o he Vol e a ke nels [8] ha cha ac e ize he beha io o hese weakly- nonlinea ci cui s in he equency domain. Howe e , he symbolic analysis o ha dly-nonlinea ci cui s is ye in ex- plo a o y phase 171. To da e, no sys ema ic echnique is a ailable o au oma e he calcula ion o a se o symbolic equa ions go e ning he la ge signal nonlinea beha io , ei- he in he s a ic o in he dynamic case. This lack o nonlinea analysis limi s he modeling capa- bili ies o s a e-o - he-a symbolic analyze s. Howe e , in many p ac ical applica ions hese limi a ions can be o e - come by eso ing o piecewise-linea (PWL) ep esen a- ions. Fo ins ance, hese ep esen a ions su ice o s udy he quali a i e beha io o many p ac ical dynamical ci cui s such as oscilla o s and compa a o s, as well as o app oxi- ma e he ansien esponse o opamps and compa a o s. VI- 17 0~7803-4455-3/98/%10.~0 0 1998 lEEE 2. BEHAVIORAL MODELING A beha io al model is a se o equa ions ha cap u e he ope a ion o a ci cui om i s e minals. Beha io al models can also be gi en as ci cui al ep esen a ions o hese equa- ions. Depending on he na u e on he exci a ions and e- sponses, he se o equa ions and i s associa ed ci cui al ep- esen a ion can be s a ic o dynamic, linea o nonlinea . Fo example, Fig.2 shows di e en beha io al models o an opamp. The i s one is s a ic and he o he s dynamic; he i s wo a e linea and he o he s nonlinea . Ci cui al ep e- sen a ions a e included o each model. The cons uc ion o a beha io al model does no s ic ly equi e o know he de ailed ci cui schema ic. Nei he he model ci cui al ep esen a ion mus e lec he ci cui opolo- gy. Models can be buil by, i s , applying a sui able se o ex- ci a ion- esponse pai s and, hen, inding and uning a ma h s uc u e ha ep oduce hese pai s wi hin some p esc ibed e o ma gin. Conside o example he OTA shown in Fig.3(a) and assume ha we a e in e es ed only in i s linea AC beha io . This can be modeled by h -pa ame e s, which can be expe imen ally calcula ed in he lab by using a ne - wo k analyze in he con igu a ions o Fig.3(c). Al e na i ely, beha io al models can be cons uc ed h ough he elec ical analysis o he in e nal ci cui schema - ic. Fo ins ance, assuming ha he OTA schema ic o A dd . Figu e 3. (a) Simple OTA; (b) simpli ied small-signal MOS an- sis o model; (c) black box modeling p ocedu e o a wo-po . Fig.3(b) and he ansis o model o Fig.3(c), a las gene a- ion symbolic analyze such as SYMBA [9] would e u n he ollowing h -pa ame e app oxima ed exp essions, hi(s) = "b - + - gm3(gml+ gm2) 2 - - scg.~2gm1gm3 Cgs2[gml(Cgs3 'gs4) + gm3Cgsl] s[gm3(cgsl + 'gs2) + (gml + gm2)(Cgs3 + Cg. dI sCgs2gmlgm3 +S Cgs2[gml(cgs3 + Cg34) + gm3CgslI h, W = + 2 + n-; + FVO Vi-+ u VIP '"b - gdslgm3 +sgd.~l(Cg.sl + cgs3 + Cg.J gmlgml+ S[gml(Cg. 3 + Cgs4) + gm3Cgsll gds2 gm1 +sCgsl gds1gm3 + Sgdsl(Cgs1 + cgs3 + Cgs4) gmlgm3 +s[gin1(cgs3 + cg.~4) +~m3Cg. 1] - - - h,(s) = Vi-+ ,+ i; h,(s) = (1) In p ac ice, designe s use hei expe ience o build modu- la models, whe e unc ional subs uc u es a e used o ep e- sen he nominal ci cui beha io as well as he pa asi ics. Thus, o example, Fig.2(a) cap u es he nominal opamp be- - - Figu e 2. Opamp beha io al models. VI- 18 ha io ; ha o Fig.2 o he dynamics ass1 and Fig.2(c) models y. These h ee mod< o he hand, Fig.2(d) ing-poin beha io s The di e en un ha io al model can su emen s o h oug he case o PWL mc nonlinea subs uc u ed h ough hei syi p esen ed in he sec in e es o hese PW 3. A PWL MOD OPA Conside he SC opamp ep esen ed I ansien obse ed ii he in eg a ing capac pa . The linea pa gml( V,(S) = - C,(C VJS) = - whe e, yc = cic2. yielding, whe e he coe iciei pa ame e s [lo]. Th lowing condi ion is 1) includes a i s -o de app oxima ion :ia ed o any ol age gain mechanism; ie i s s age ansconduc o nonlinea i- ; ocus on he ans e beha io , On he ip u es also o a i s app oach he d i - he ci cui e minals. onal subs uc u es pe aining o a be- e uned ei he h ough blackbox mea- analysis o he unde lying ci cui . In els, measu emen s a e used o une he s while hose linea ones a e ep esen - bolic ans e unc ions. The example )n below illus a es abou he p ac ical models. L FOR TOP-DOWN TRANSIENT [P OOPTIMIZATION i eg a o o Fig.4(a) and assume he he PWL model o Fig.4(b) [lo]. The he ans e ing o he inpu ol age o o con ains a linea pa an a nonlinea in be calcula ed symbolically om, ic, + C2C, + c2c, + c,c, , = ci + c, + c, (3) cp(-a )cos@ + C,, exp(-a )sinp up(-a )cosp + Cllexp(-a )sinp (4) cp(-a )cosp + CUI exp(-a )sinp s a e gi en as unc ions o he model e equa ions emain alid while he ol- Hilled, 8mzI * (q I, (5) VI- 19 Figu e 4. (a) SC in eg a o and, (b) opamp model. A e wa ds, he second model s age en e s in o sa u a ion and he ansien is desc ibed by he ollowing equa ions, whe e. and he emaining coe icien s a e gi en as unc ions o he model pa ame e s [ 101. This second ansien pe sis s while he condi ion (5) holds. A e wa ds, he ansien is gi en by, yielding, ,( ) = A,, +B,,exp(-a )cosp + C,,exp(-a )sinp (10) whe e he coe icien s a e, as in he p e ious exp essions, gi en as unc ions o he model pa ame e s. The model has been alida ed by compa ing i s ou pu wa e o m o ha o ob ained h ough de ailed elec ical sim- ula ion. The ampli ie consis ed o a ully-di e en ial old- ed-cascode OTA whose co e schema ic and summa ized pe - o mance a e gi en in Fig.5. Said ampli ie was designed o ha e small phase ma gin (a ound 4Sdeg) o educe powe dissipa ion. Fig.6(a) shows he in eg a o ou pu ol age du - ing he in eg a ion phase ob ained h ough elec ical simula- ion (HSPICE) and ha ob ained using he model. Model pa- ame e s a e also enclosed in Fig5 A good conco dance be- ween bo h app oxima ions is obse ed. A mo e gene al esul is gi en in Fig.6(b) whe e he di e ence be ween he inal alue o he in eg a o ou pu ol age and i s ideal alue is shown as a unc ion o he inpu le el. V- --I GBW(3.2pF) = 106.7MHz PM (3.2pF) = 42.2O I, = 500pA Figu e 5. Fully di e en ial OTA. 0.4 . . 1 0.3 2 0.2 2 0.1 2 0.0 c c 0 0) +- -0.1 -0.2 - -0.40 5.0e-09 1 .Oe-08 13-08 2.0e-08 2.5e-08 Time (s) (4 - HSPICE 1 --Model 0.5 1 .o 1.5 2.0 -0.1 ‘0 Vol age s o ed in Ci (V) Figu e 6. Simula ed and calcula ed esponses. (b) 4. REFERENCES 11 F.V. Fe nsindez, A. Rod iguez-Vsizquez, J.L. Hue as and G. Gielen, eds., Symbolic Analysis Techniques. Ap- plica ions o Analog Design Au oma ion. IEEE P ess, 1998. 21 W.S. Wu, W. Helms, J.A. Kuhn and B.E. By ke , “Dig- i al-compa ible high-pe o mance ope a ional ampli ie wi h ail- o- ail inpu and ou pu anges,” IEEE J. Sol- id-s a e Ci cui s, Vol. 29, pp. 63-66, Jan. 1994. [3] P. Wambacq, EV. Fe nsindez, G. Gielen, W. Sansen and A. Rod iguez-VBzquez, “E icien symbolic compu a- ion o app oxima ed small-signal cha ac e is ics o ana- log in eg a ed ci cui s,” IEEE J. Solid-s a e Ci cui s, Vol. 30, pp. 327-330, Ma ch 1995. [4] Q. Yu and C. Sechen, “A uni ied app oach o he app ox- ima e symbolic analysis o la ge analog in eg a ed ci - cui s,” IEEE T ans. Ci cui s and Sys ems I, Vol. 43, No. [5] F.V. Fe nsindez, 0. Gue a, J.D. Ga cia and A. Rod iguez-VBzquez, “Symbolic analysis o analog in eg a ed ci cui s: he nume ical e e ence gene a ion p oblem,” IEEE T ans. Ci cui s and Sys ems, 1998, o appea . [6] P. Wambacq, Symbolic Analysis o La ge and Weakly Nonlinea Analog In eg a ed Ci cui s. Ph.D. Disse a- ion, Ka holieke Uni e si ei Leu en, Belgium, 1996. [7] C. Bo che s, L. Hed ich and E. Ba ke, “Equa ion-based model gene a ion o nonlinea analog ci cui s,” P oc. 33 d Design Au oma ion Con$, pp. 236-239, 1996. [8] S. A. Maas, Nonlinea Mic owa e Ci cui s, A ech House 1988. [9] IMSE-CNM, Me hods o SBG and SDG. ESPRIT P ojec 21812 AMADEUS, In e nal Repo , Ma ch 1997. [ 101 E Medei o, XA-modula o In e~aces o Mixed-Signal CMOS IC’s: Modeling, Simula ion and Au oma ed De- sign. Ph.D. disse a ion, Uni e si y o Se illa, Spain, 1997. 8, pp. 656-669, Aug. 1996. VI-20