scieee Open visual document viewer

A mismatch-insensitive high-accuracy high-speed continuous-time current comparator in low voltage CMOS

Río Fernández, Rocío del; Liñán Cembrano, Gustavo; Domínguez Castro, Rafael; Rodríguez Vázquez, Ángel Benito

Abstract

This paper presents a CMOS current comparator which employs nonlinear feedback to obtain high-accuracy (down to 1.5 pA) and high-speed for low input currents (8 ns@50 nA). This structure is much faster for low currents (below 10 /spl mu/A) than other previous nonlinear feedback comparators. Particularly, when compared to the fastest current comparator reported up to now, the new one operates at more that 100 times faster for a 1 nA current, with smaller area occupation and similar power consumption. In addition, the new comparator is virtually insensitive to mismatch and capable of operating with supply voltages as low as 1 V.

Full text

A Misma ch-Insensi i e High-Accu acy High-speed Con inuous-Time Cu en Compa a o in Low Vol age CMOS R. del Rio-Fe nandez, G. Liiian-Cemb ano, R. Dominguez-Cas o and A. Rod iguez- Vazquez Ins i u o de Mic oelec 6nica de Se illa, Cen o Nacional de Mic oelec 6nica, A da. Reina Me cedes s/n, Edi . CICA, E-4 10 12 Se illa Tel.: +34 5 4239923, FAX: +34 5 4231832, E-mail: [email p o ec ed] Abs ac This pape p esen s a CMOS cu en compa a o which emp ioys nonlinea eedback o ob ain high- accu acy (down o 1.5pA) and high-speed o low inpu cu en s (8ns050nA). This s uc u e is much as e o low cu en s (below 10pA) han o he p e- ious nonlinea eedback compa a o s [ 11, [2], [3], [4]. Pa icula ly, when compa ed o he as es cu en compa a o epo ed up o now [4], he new one goes mo e ha 100 imes as e o 1nA cu en , wi h smalle a ea occupa ion and simila powe consump- ion. In addi ion, he new compa a o is i ually insensi i e o misma ch and capable o ope a e wi h supply ol ages as low as 1V. 1. In oduc ion Analog cu en -mode echniques a e d awing s ong a en ion oday due o hei po en ial applica- ion in he clesign o high-speed mixed-signal p ocess- ing ci cui :; in low ol age s anda d VLSI CMOS echnologies [5]. Also cu en -mode ci cui s a e na u- al candida es o image senso y in o ma ion p ocess- ing whe e cu en sensing and compa ison is necessa y. Cu en compa a o s a e basic building blocks o nonlinea cu en -mode signal p ocessing and analog o digi al con e e s. The a ailabili y o la ge cu en anges is an appealing ea u e o bo h ields. Also, e icien small cu en le el de ec ion is undamen al o high accu acy and high speed appli- ca ions. Low le el, high speed cu en de ec ion is also equi ed in di e en ligh and adia ion sensing appli- ca ions: i.e. y-de ec o s using wide band gap semicon- duc o s [6], o con ollabili y and econ igu abili y issues in E-beam es ing o in eg a ed ci cui s [7]. Fo ins ance, in he la e he need a ises o de ec cu en le els as low as 1nA in a ew ps. Sub h eshold CMOS cu en mode massi e compu a ion a chi ec u es [8] also equi e e icien de ec ion o low cu en le els o as disc imina ing unc ion e alua ion. To high- ligh ano he applica ion, cu en de ec ion is also equi ed in IDDQ VLSI es ing app oaches [9]. In ecen yea s, many cu en compa a o ci cui s ha e been p oposed [ll, [21, [31, [41, [lo], [ill, [121. A possible classi ica ion o hem is o dis inguish be ween he ones whe e he inpu cu en is i s sensed a a low impedance node ( esis i e inpu ) and hem ampli ied using a ol age gain mechanism, and he ones whe e he inpu cu en is sensed a a high impedance node (capaci i e inpu ). The i s s p o ide high speed o high cu en le els bu a e somewha inaccu a e; he seconds ob ain enla ged esolu ion a he cos o inc easing ol age excu sions a he inpu node and consequen ly dec easing he ope a ion speed [ 121. Ano he possible classi ica ion is o dis inguish be ween he s uc u es ha use au oze oing ech- niques o a enua e o se o o educe p opaga ion delays by adding clock signal gene a ion ci cui y and he ones ha wo k on asynch onous mode. In his pape we p opose an asynch onous mode cu en compa a o ha uses nonlinea eedback o combine he ad an ages o capaci i e-inpu and esis- i e-inpu a chi ec u es. Because he p oposed ci cui is o se insensi i e by cons uc ion i does no equi e au oze o o o se a enua ion. We include also a compa a i e s udy wi h he ope a ion o he cu en compa a o which is claimed he as es among all 111 0-7803-42~#-2/97/$10.00 0 1997 IEEE hose epo ed o now [4]. 2. Ci cui beha iou Fig. 1 shows he schema ic o he so-called Cu en Swi ch Compa a o , whose ad an ages o low inpu cu en s ha e been demons a ed in [l2]. The ope a- ion o his ci cui in ol es wo di e en e,' =ions. I a posi i e cu en lows owa ds he inpu node, and assuming null ini ial condi ions (Vin = V,= E) ansis o s M, and Mp a e OFF, he e o e he equi a- len impedance a he inpu node is capaci i e and consequen ly he ci cui exhibi s a e y high esolu- ion (only limi ed by he leakage cu en s), he inpu cu en is in eg a ed in he pa asi ic inpu capaci o (Gin) p oducing a con inuous inc ease o he inpu ol age. The ampli ie causes Vo o dec ease as e han Vi, and so ha Mp u n ON d aining he inpu cu en and closing a nega i e eedback loop a ound he ampli ie . I ob ains a i ual g ound a he ampli- ie inpu , hus ixing he ol age a he inpu node. A simila si ua ion occu s o nega i e inpu cu en whe e M,l becomes ON. The ansien e olu ion is domina ed by he i s ope a ion mode ( he capaci i e inpu beha iou ), and i we conside a single pole model o he ampli ie s wi h a gain-bandwid h p oduc (GB) and linea capac- i o s o hei inpu and ou pu nodes one ob ains Howe e , due o he pa asi ic capaci o s be ween inpu and ou pu nodes (CM) he ansien esponse is slowed down e en o an ideal ampli ie ; his beha - iou is simila o he slew a e phenomena in Mille ope a ional ampli ie s. In he compa a o o Fig. 1 he coupling capaci o is o med by he Cgs capaci o s o Fig. I Cu en Swi ch Compa a o Schema ic. Mn and Mp, and e en ually he in insical coupling capaci o o he ampli ie . Due o C,, inpu and ou - pu nodes o he compa a o a e coupled, so ha he ansien esponse e olu ion ollows a linea cha ac- e is ic ins ead a quad a ic law 2 whe e c = CinCo + Cin C, + C,C, No e ha in (1) he ou pu ol age is di ec ly p o- po ional o he ampli ie GB, and hence he as e he ampli ie is he as e he cu en compa a o sol es. Howe e in (2) his dependence is anished. The esponse ime can be calcula ed om (1) and (2) by making AV, (T,) = E,, , whe e EOH deno es he logic noise ma gin. T =-- C, 'in ; (C, = 0) T =-- C, 'in Acco ding o his o mula, Fig.2 shows a compa ison o he esponse ime o he cu en swi ch compa a o o Fig.1 wi h and wi hou he coupling capaci o . We a e assuming EoH=l V, ypical capaci ances alues o Ci,=Co=50 F and g,=1001ul/ 2 o he ampli ie . Fo low cu en le els, i is seen ha he compa a- o beha iou becomes se e ely deg aded due o he coupling capaci o , e en o low alues o his capac- 1 Om , i In loop In 10n lOOn 1p l0cl 'U% 16, Fig. 2 Compa ison ime esponse , o he cu en compu u o wi h and wi hou coupling capaci o : 112 i o . On he o he hand, o la ge cu en s (in he o de o ens o PAS) smalle imes esponse a e ob ained in he coupled cc en compa a o . Fo hese high cu - en le els a simple in e e ea u es as ope a ion speed, wi h no eedback needed. To imp o e ansien esponse o he cu en swi ch compa a o o low inpu cu en s, ci cui s a egies a e equi ed o uncouple inpu and ou pu nodes o he ampli ie and hence, educing he cou- pling capaci ance. In o de o achie e as e esponse imes simple CMOS in e e s a e usually used as ampli ie s. To a oid he o e1,lap capaci o s in luence cascode an- sis o s could be used in he ampli ie ou pu node. Also, he eedback mechanism used o ob ain he non- linea d i ing-poin cha ac e is ic should be modi ied o a oid MI, and M, Cg.? in luence, so ha using a sou ce ollowe in he eedback loop mus be a oided. The new compa a o schema ic is shown in Fig.3. The ci cui hac; wo ope a ion zones. l a posi i e cu - en lows oua ds he inpu node, and assuming ha he ini ial condi ions a e: ansis o s Mn and Mp a e OFF, so ha he equi alen impedance a he inpu node is capaci i e (i.e., e y la ge in DC) and consequen ly he ci cui s exhibi s a e y high esc'lu ion (only limi ed by he leakage cu - en s), he inpu cu en is in eg a ed in he inpu capaci o (Gin) and p oduce a con inuous inc ease o he inpu ol age. The ampli ie causes V, o dec ease as e han Vi, inc easing Vs.y o he ansis o Mn and dec easing Vsg o M,.,, so ha M,, is d i en in o he ON s a e anc! a nega i e eedback loop is c ea ed a ound he ampli ie . I ob ains i ual g ound a he ampli ie inpu , ixing he ol age a he inpu node. A simila si ua ion occu s o nega i e cu en s, whe e MI, becomes ON. I Fix. 3 Cu en,, S ee ing Compa a o Schema ic. This s uc u e educes esponse imes o he Fig. 1 compa a o . An imp o ed a ound wo o de s o mag- ni ude is obse ed wi h he compa a o in [4] i we compa e he wo s cases (nega i e inpu cu en s). We can also ob ain a symbolic exp ession o he esponse ime pa ame e by ansien analysis. I we conside an inpu cu en s ep a =O om a nega i e o e d i e le el -J up o a posi i e o e d i e le el +J, he esponse ime can be w i en as ollows: Fig.4 shows he inal schema ic o he new cu - en s ee ing compa a o . As i can be seen he ampli- ie ha appea s in Fig3 has been implemen ed by a CMOS in e e wi h wo cascode ansis o s which a oid he capaci o s be ween inpu and ou pu nodes, so ha he ansien e olu ion p esen s a quad a ic ime dependence ins ead o a linea depcndencc, as occu s o he compa a o in [4]. The esponse imes simula ion measu es o he wo compa a o s o a ange o inpu cu en s a e shown in Fig.5. We ha e conside ed he wo s case, V , -L 2/o.8 1' 6/03 I '5 -1 I 7 T "n 1' 410.8 (b) Fig. 4 (a) New Cu en Conipu u o ii ES.? 0.8pn CMOS echnolo iy.. (hi Biasing Ci cui n. 113 13000 ID i.e. a s ep inpu cu en om a posi i e o e d i e le el +J o a nega i e o e d i e le el -J. The esul s show ha he g ea es speed imp o emen occu s o low inpu cu en s, and o less han lop4 he new compa- a o emains as e han he one in [4]. The quad a ic ime law p edic ed in (2) can be obse ed in Fig.6. I can be seen how his app oach is sui able in o de o desc ibe he main ansien e olu ion. I con as s wi h he beha iou obse ed in Fig.7 whe e he slew- a e phenomena go e ns he empo al esponse. The ci cui which p o ides he wa e- o m in Fig.7 is he same as in Fig.4 bu using he eedback scheme ha appea s in Fig. 1. A possible echnique o imp o e compa ison imes is o educe he ol age excu sions a he ampli- ie ou pu node. This app oach was applied by he au ho s in [4] o he ci cui in Fig.1. Howe e i p esen s a e y high sensi i i y o mis ma ch e ec s. In o de o a oid he mis ma ch in luence he excu - sions on he ou pu ol age mus be la ge han he expec ed maximum luc ua ions due o mis ma ch. .. -1 0: - -2.07 b) ; .......,.. .',... 1, ,I... ..... I ...................... I .........,.... I .... '...,, .............................. I Fig. 6 Response o he compa a o in Fig.4 . o U 50nA inpu pulse. ............. - - ....>......,...,..... I ..... > ....,.....,...... 1 .... , ...... I , .. I , .. : 1.0" 200.0" 400.0n 600.0n SO0 On 0. >me (1l"l Fig. 7 Response o he cu en swi ch compa a o in Fig. I , ii U 50d inpu pulse. Fig.8 shows a compa a i e s udy by Mon e Ca lo analysis o he new and [4] compa a o s designed in he same echnology (ES2 1.5pm). I is shown how he pe cen age o success ul compa isons o [4] dec eases as inpu cu en le el becomes lowe . How- e e , o he new compa a o his pe cen age becomes 100% as soon as he inpu cu en eaches he maxi- mum o se le el (<lSpA). In o de o ob ain he maximum imp o emen s he compa a o has inally been designed using ES2 0.8pm echnology. Fig.9 illus a es he layou o he compa a o and he biasing ci cui y. The main cha ac e is ics o he new compa a o a e : - 5V powe supply. - Inpu Cu en O se lowe han 1.2pA. - Maximum esolu ion in he o de o ew PA. - Maximum delay o a 50nA s ep inpu cu - - Maximum delay o a lpA s ep inpu cu en en 8ns. 2ns. A NA! Fig 8 Sensi i ies o he New and [4] Compa a o .7. 114 J/ Fig. I1 Low ol age ampl$e Fig. 9 Luyou o he compa a o and he biasing ci cui y. - A ea consump ion wi hou bounding pads - Powe consump ion 2.5mW. 0.005 nm2. Fig. 10 shows a compa ison be ween schema ic and ex ac ed esponse imes. I can be obse ed how he ex ac ed esponse imes a e wice slowe han schema ic imes due o he pa asi ic e ec s. A ca e ul layou mus be de eloped in o de o minimize he pa asi ic capaci o e ec s. Ano he impo an ea u e o his new s uc u e is i s capabili y o wo king a low ol age supplies. As i can be seen in Fig.4 he biasing ci cui y equi es whe eas he CMOS ampli ie need I allows xope ope a ion wi h a powe supply below 3V. I i. lowe powe supply is needed he bias- ing ci cui y should be modi ied, in his case (7) p o- ides he minimum V~D alue (lowe han 2V). This compa a o c,in ope a e e en below his limi i he I Fig. 10 Compu ison Be ween Schcm ic and Ex ac ed Delays. ampli ie in Fig1 1 is used. In his case he new limi becomes a ound 1V. Fig. 12 shows he esponse imes o he compa a- o in Fig.4 o wo di e en ail- o- ail ol ages. I can be obse ed how o low inpu cu en le els he maximum speed is ob ained using a 3.3V ail- o- ail ol age. I is due o lowe ol age excu sions a he ou pu node o he ampli ie . 3. Conclusions An analysis o he ansien e olu ion o nonlinea eedback cu en compa a o s demons a es ha i s esponse ime deg ades signi ican ly a low cu en s due o he in luence o a Mille coupling capaci o . This can be pa ially o e come by educing he ol - age excu sions, he p ice o be paid o his is a educ- ion o he ci cui ole ance o misma ches, specially signi ican a low- ol age supplies. The compa a o p esen ed in his pape uses ano he s a egy o ob ain la ge ope a ion speed wi hou deg ading he compa a- o esolu ion and keeping he compa a o ope a ion i ually misma ch-insensi i e, e en o ol age sup- plies as low as 1V. h 115 4. Re e ences [l] R. Dominguez Cas o, A. Rod iguez Viizquez and J.L. Hue as: “Sob e el Dise io de Compa a- do es de In ensidad”. DiseAo de Ci cui os In e- g ados 91, ISBN 84-87412-61-0, pp 61-66, Uni e sidad de Can ab ia 199 1. 121 R. Dominguez-Cas o, A. Rod iguez-Viizquez, F. Medei o and J.L. Hue as: “High-Resolu ion CMOS Cu en Compa a o s”. 1992 Eu opean Solid-s a e Ci cui s Con , ISBN 87-984232-0-7, pp. 242-245, Kand up 1992. [3] H. T a “No el App oach o High Speed CMOS Cu en Compa a o s”. Elec onics Le e s, Vol.28, No.3, pp. 310-312, 1992. [4] A.T.K.Tang, C.Toumazou: “High Pe o mance CMOS Cu en Compa a o ”. Elec onics Le - e s, Vo1.30, No.1, pp. 5-6, Jan. 1994. [5] C. Toumazou, E J. Lidgey and D. G. Haigh: “Analogue IC Design: The Cu en Mode App oach”. IEE Ci cui s and Sys ems, se ies 2, 1990. [6] P. Olmos e al: “Compu e Simula ions o he Beha io o he Pa ial Cha ge Collec ion Model in Thick HgI2 y-de ec o s”. Nuclea Ins . and Me hods in Physics Resea ch, A302, pp 91-104, 1991. [7] J. Mad enas and J. Cabes any: “E-Beam De ec o De ices o IC Con ollabili y”. Mic oelec onics Enginee ing 16, pp465-472, 1992. [SI A.G. And eou e al: “Cu en -Mode Sub h eshold MOS Ci cui s o Analog VLSI Neu al Sys ems”. IEEE T ans. on Neu al Ne wo ks, Vol. 2, pp 205- 213, Ma ch 1991. [9] J. Rius and J. Figue as: “P opo ional BIC Senso o Cu en Tes ing”. Jou nal o Elec onic Tes - ing, Vol. 3, pp 387-396, Dec. 1992. [ lO]G.Palmisano. 6. Palumbo, S. Pennisi: “A High- Accu acy High-speed CMOS Cu en Compa a- o ”. P oc. ISCAS’94, June 1994. [ 111 J.P.A. Ca ei a, J.E. F anca: “High Speed CMOS Cu en Compa a o ”.P oc. ZSCAS‘94, June 1994. [ 12lA. Rod iguez-Vazquez, R. Dominguez-Cas o, E Medei o, M. Delgado-Res i u o: “High Resolu- ion CMOS Cu en Compa a o s: Design and Applica ions o Cu en -Mode Func ion Gene a- ion”. Analog In eg a ed Ci cui s and Signal P o- cessing. 7, pp. 149-165, 1995. 116