BiCMOS Digital Circuit Design

Review of CMOS & NMOS Inverter Design - delay time - pair delay - driving large capacitive loads Features of BiCMOS Digital Circuit BiCMOS Inverters -basic type - delay time - improved type - full swing - design example a. size optimization b. driving large capacitive loads BiCMOS Gate Power-Supply Sensitivity - voltage scaling - low voltage gate I/O Interface - input - output - logic conversion

Tai-Haur Kuo, EE, NCKU, 1997 BiCMOS Circuit Design 3-1 NMOS DIGITAL CIRCUITS

Static inverters

Load

Vout ==> V in driver

A B C

IDS VGS=V(1) C

B A

VDS (driver) VDD Vth VDD D: in in

Vout VGS=V(1)

VGS=V(0)

V(0)

V(0) V(1) V(0) V(1) Vin

― Static power dissipation during Vout=V(0)

Tai-Haur Kuo, EE, NCKU, 1997 BiCMOS Circuit Design 3-2 Switching Characteristics of CMOS Inverter

CMOS inverter

VDD VDD T 2 V (t) o t

Vin(t) T1 CL

+VDD

Vin(t)

0 t

+VDD 0.9VDD 0.1VDD t td

tf tr

Trajectory of n- operating point during switching in CMOS inverter

Input transition : X1 X2

Output transition: X2 X3 Vds=Vgs-Vt UNSATURATED SATURATED STATE STATE X2 Vgs=VDD

Ids

OPERATING POINT AFTER COMPLETION OF SWITCHING INITIAL OPERATING POINT X3 X1 0 VDD Vo(t)

Tai-Haur Kuo, EE, NCKU, 1997 BiCMOS Circuit Design 3-3 Rise Time and Fall Time of CMOS Inverter > Equivalent Circuit

V DD VDD > Fall p-DEVICE p-DEVICE t=0 t=0

I  Vin ↑ c Ic Idsn R   V c C o Vo   n-DEVICE L n-DEVICE CL  Vo ↓ 

SATURATION: VVoD≥ D− Vtn 0 < VV≤ − V (a) SATURATION: 0 DD tn

> Rise VDD VDD p-DEVICE p-DEVICE Idsp Rc  Vin ↓   I Ic  Vo ↑  t=0 c Vo Vo n-DEVICE CL n-DEVICE CL

LINEAR (b) SATURATION

>tf = tf1 + tf2 ; fall time | tf1 | tf2 |

Vo=0.9VDD Vo=0.1VDD

Vo = VDD -Vtn

| tf1 | tf2 |

I = 0  DS()PMOS I DS()PMOS = 0   I DS()NMOS I  in saturation  DS()NMOS in

Tai-Haur Kuo, EE, NCKU, 1997 BiCMOS Circuit Design 3-4 Fall Time (Cont.)

t f

Vo : 0.9 VDD V-DD Vtn NMOS in saturation region 2 d o β n C V ( V V)= 0 L d t 2 DD tn

0.9V t= 2CL DD f1 2 d o n V β (VDD -)Vtn ∫V DD V tn

2CL (Vtn 0.1 V)DD = 2 β n (VDD -)Vtn

t f2

Vo : VDD Vtn 0.1 VDD NMOS in triode region

d o β n 2 C V = [ 2(V V)V V ] L d t 2 gs tn ds ds

VDD

V DD V tn d Vo t= CL f2 n 2 β (VDD -)Vtn ∫ 0.1V DD Vo Vo 2(VDD -)Vtn

Tai-Haur Kuo, EE, NCKU, 1997 BiCMOS Circuit Design 3-5 Fall Time (Cont.)

CL 19V-DD 20Vtn t f2 = ln( ) β n (VDD -)Vtn VDD

t=f t+f1 t f2 = 2CL β n (VDD -)Vtn V-0.1V 1 19V - 20V ×[ tn DD + ln( DD tn )] V-DD Vtn 2 VDD

e.g. V=DD 5V,V=tp -1V, V=tn 1V

4CL t~f ~ β nVDD Rise time 4CL Similarly, t~r ~ β p VDD

Tai-Haur Kuo, EE, NCKU, 1997 BiCMOS Circuit Design 3-6 Fall Time and Rise Time (Cont.)

4CL t f ≈ βn VDD

4CL t r ≈ βp[VDD

For equally sized n and p devices

βn = 2βp t t = r f 2

For tr =tf

⇒βp =βn ⇒Wp =2.5Wn For simplicity, 2 is used sometimes. t t = f df 2 t t = r dr 2

Tai-Haur Kuo, EE, NCKU, 1997 BiCMOS Circuit Design 3-7 Equivalent Circuit

Equivalent Resistance

― 4CL tf = =ReqCL βnVDD

― 4 for minimum sized NMOS Req = βnVDD

― 4 R (PMOS)= ≈2.5R eq β V eq forn minimumDD sized PMOS

Equivalent capacitance

Ceq =Cgs of minimum sized

Tai-Haur Kuo, EE, NCKU, 1997 BiCMOS Circuit Design 3-8 Inverter-pair Delay

>CMOS

(a) Wp=2.5Wn;Wn and L are minimum size

R eq Req Req

3.5Ceq 3.5Ceq Req Req Req

tinv-pair =tf +tr =3.5Req Ceq +3.5Req Ceq =7Req Ceq

(b) Wp=Wn; All minimum sized devices

2.5R eq 2.5Req 2.5Req

2Ceq 2Ceq Req Req Req

tinv-pair =5Req Ceq +2Req Ceq =7Req Ceq W =2.5W ==> V =2.5V same delay time p n inv } Wp=Wn ==>Vinv =2.2V Vinv ~10% variation

Tai-Haur Kuo, EE, NCKU, 1997 BiCMOS Circuit Design 3-9 Inverter-pair Delay (Cont.)

Driving same size inverter L,W L,W L,W

2Ceq Req L,W L,W L,W

L&W are minimum size Inverter-pair Delay=77Re qeC q= τ where τ = R e qeC q

nL,mW nL,mW nL,mW

nL,mW nL,mW nL,mW

n R mnCeq m eq

2 Inverter-pair Delay = 7nCRe qeq= 7n 2 τ where τ = R e qeC q (Depends on channel length independent of channel width)

gm µ u gs t Recall that W ==2 ()VV− CLg

Tai-Haur Kuo, EE, NCKU, 1997 BiCMOS Circuit Design 3-10 Driving Large Capacitive Loads

> CMOS L , f2W Lp, WP Lp, fWP p P ° ° ° CL Ln, Wn L , fW 2 n n Ln, f Wn

(i) If Lp & Ln are minimum channel length ,

tinv-pair = 7f τ

(ii) If Lp & Ln are not minimum channel length, additional calculation is required to obtain

tinv-pair - Delay per stage (minimum L)

(i) For Vin: (a) 3.5 τ (Wp= 2.5Wn), (b) 2 τ (Wp= Wn)

(i) For Vin: (a) 3.5 τ (Wp= 2.5Wn), (b) 5 τ (Wp= Wn)

> NMOS

tinv-pair= 5f τ

Delay per stage = f τ for Vin

Delay per stage = 4f τ for Vin

Tai-Haur Kuo, EE, NCKU, 1997 BiCMOS Circuit Design 3-11 Driving Large Capacitive Loads (Cont.)

CL Let y = = fN Cg

( Cg : gate capacitance of the first stage inverter) lny = Nlna lny N = lna

For N even , total delay N t = 7a = 3.5Na (CMOS) d 2 N (or) = 5a = 2.5Na (NMOS) 2

lny Delay α Nf = f lna (for both CMOS and NMOS)

==> For minimum td d lny ( f ) = 0 => obtain f = e to have the df lnf minimum value of Nf

Tai-Haur Kuo, EE, NCKU, 1997 BiCMOS Circuit Design 3-12 Driving Large capacitive Loads (Cont.)

Assuming f = e, then N = ln(y)

Overall delay td (i) N even: tNd = 25. eτ (NMOS) or tNd = 35. eτ (CMOS)

(ii) N is odd tNd =−[.25( 1)+1]eτ (NMOS) for tNd =−[.35( 1)+2]eτ (CMOS) } ∆Vin

tNd =−[.25( 1)+4]eτ (NMOS) } for tNd =−[.35( 1)+5]eτ (CMOS) ∇Vin

** For optimum speed f=e=2.71828 But values from 2 to 10 may be used to obtain more flexibility and reduce cost(# of stages)

Tai-Haur Kuo, EE, NCKU, 1997 BiCMOS Circuit Design 3-13 Features of BiCMOS Digital Circuit

Combines CMOS and bipolar -CMOS Low power dissipation High density

- Bipolar High drive capability Small swing logic

“System on a chip” is possible - Mixed analog/digital Analog circuit can be incorporated - High voltage, high power interface (e.g. sensor, drive, ...) can be incorporated if special process is developed

Tai-Haur Kuo, EE, NCKU, 1997 BiCMOS Circuit Design 3-14 BiCMOS Inverter Basic type

+

CMOS Bipolar(buffer)

Physical structure NMOS PMOS NPN S G D S G D B E C

n+ n+ p+ p+ p+ p+ n+ P-well N-well N-well p+-buried N+-buried N+-buried

P-sub

can be added to PMOS drain and NPN base increase latchup may be merged to reduce immunity (a) area (b) capacitance (c) # of contact

Tai-Haur Kuo, EE, NCKU, 1997 BiCMOS Circuit Design 3-15 Pull-up of BiCMOS Inverter

Basic type BiCMOS inverter

• • VDD M βIDP ° 1 Q IDP 1 Input IDN βI M DN 2 CL Q2 Pull-up Input assume M2 & Q2 are off (initial condition) V -V V = 0 DD BE BE1 V BE t

Input • VDD M βIDP Output ° 1 V -V I Q1 DD BE DP Output V V inv ° BE t CB CL+ CS1 tr td1 2 tdr CB is the capacitance at the Q1 base node CS1 is the capoacitance at the output node when CL= 0

VBECB td1 = ...time required to turn on Q IDN 1 (C +C ) tr Vinv L S1 = ...time required to change CL TO Vinv 2 βIDP ~ µpCOX W 2 where IDP ~ ( ) (Vdd-VBE-Vtp) 2 L M1 β vs. Ic β

Ic t delay time t = t + r dr d1 2 Tai-Haur Kuo, EE, NCKU, 1997 BiCMOS Circuit Design 3-16 Pull-down of BiCMOS Inverter Pull-down M Q are off assume 1 & 1 initial condition VBE2= 0

Input V -V Input ° Output DD BE βI M2 DN VBE t Q CL+CS1 IDN 2 Output CB VDD-VBE Vinv VBE t

td2 tf where CL is added capacitive load 2 t CS is the capacitance at the dr output node when CL= 0

V C V (CL+CS1) ~ BE B inv = + t tdf ~ + td2 f IDN βIDN 2

~ µnCOX W 2 where IDN ~ ( ) (VDD-2VBE-Vtn) 2 L M2 ~ VSB ~ VBE

β vs. Ic (body effect) β

Ic

Adjusting the size of M1 and M2, one can obtain tdr= tdf

Tai-Haur Kuo, EE, NCKU, 1997 BiCMOS Circuit Design 3-17 Delay Time of CMOS Inverter

CMOS inverter

Input Output Cs2 is stray capacitance

CL+CS2

Input

VDD

0 t Output

VDD

Vinv.

0 t

Vinv. (CL+CS2) td ≅ (rough estimation) ID

µCox W where I ≅ (V -V )2 D 2 ()L DD t

Tai-Haur Kuo, EE, NCKU, 1997 BiCMOS Circuit Design 3-18 Delay Time Comparison between BiCMOS and CMOS

td(tdr or tdf) CMOS

V Slope ≈ inv. ID BiCMOS

V Slope ≈ inv. βID

t1(td1 or td2)

CL -CS2

Rough estimation

VC()+C t + inv L s1 ;BICMOS 1 βI t = D d VC()+C inv L s2 ;CMOS I D

Tai-Haur Kuo, EE, NCKU, 1997 BiCMOS Circuit Design 3-19 Drawbacks of Basic-Type Inverter

No path to discharge the bases

Input

Output

Not full swing, i.e. not (0~VDD) swing: VBE〜 (VDD-VBE) Incompatible with CMOS logic As described in the introduction, most digital circuits of a large chip use CMOS logic instead of BiCMOS for two reasons: 1. high density 2. low power

Tai-Haur Kuo, EE, NCKU, 1997 BiCMOS Circuit Design 3-20 BiCMOS Inverter with MOS Discharge

Conventional BiCMOS inveretr First proposed by Hung-Chung Lin

Swing : VBE ~ (VDD -VBE)

V • DD• ° P • Q1

Input N • d1

Output • N Q • 2 Nd2

Q1 is reverse-biased when output = “low”

( i.e. VBE1< 0 when Output = VBE)

=> longer time to turn on Q1 when Output = “Low” “High”

Tai-Haur Kuo, EE, NCKU, 1997 BiCMOS Circuit Design 3-21 Improved BiCMOS Inverters

Discharges the bases Transistor discharge

=> Swing:VBE〜(VDD-VBE)

Input Output

Resistor discharge => Full swing

Input Output

Many other examples

Tai-Haur Kuo, EE, NCKU, 1997 BiCMOS Circuit Design 3-22 -Discharge BiCMOS Inverters

a. Path to discharge the bases b. Full swing

M 1 Q Input R 1 1 Output M 2 Q2 R 2 Input Pull-up Input M 1 t Q1 is on Q1 Output R1 Output Q turn off Q off 1 1 } Q 1 on t tv M2 turned off by input initially td1 2 V V -t/R1CL out(t) =-DD VBE e Q discharged initially -t/R1CL 2 Vout(t) = VDD (1- e ) Q 1turned off during (base is discharged through R 1 ) Pull-down Input Input Output M 2 C t Q2 L Output -t/R2 C L V DDe R2 Q 2 on -t/R2 C L } V BE e VBE t f t td2 2 Q 2 off Q 2 off Tai-Haur Kuo, EE, NCKU, 1997 BiCMOS Circuit Design 3-23 Alternative Full-Swing BiCMOS Inverters

[L. Wissel CICC 1992] CMOS-compatible BiCMOS logic Full-swing “finisher” is required Latch Increased power dissipation

CMOS Parallel High input capacitance Not partical for all functions

“Feedback” Added capacitance on base nodes

Tai-Haur Kuo, EE, NCKU, 1997 BiCMOS Circuit Design 3-24 Design Example [ L. Wissel CICC 1992 ]

BiCMOS circuit FET vs. NPN size optimization Model

WF

WE

WF

WE

Delay vs. FET width (Emitter width as parameter)

CL = 0.3pF CL = 3.0pF Delay(ns) Delay(ns)

.8 .8 WE =2µ

.6 .6

WE =2µ W =8µ .4 .4 E

WE =8µ .2 .2 10 20 40 10 20 40

WF = ( µ ) WF = ( µ )

Tai-Haur Kuo, EE, NCKU, 1997 BiCMOS Circuit Design 3-25 Design Example (Cont.) [L. Wissel CICC 1992] Driving large capacitive load - CMOS-BiCMOS comparison - layout and speed

A(CMOS) B(CMOS) C(CMOS) w = 20µm w = 40µm w = 12µm24 µm 48µm

w = 10µm w = 20µm w = 8µm 12µm 24µm

Performance Delay CMOS “A” CMOS “B” 0.8 (ns) 0.6 CMOS “C” 0.4 BiCMOS 0.2 C (pF) 1.0 2.0 Layout size

CMOS”A” CMOS”B”

CMOS”C” BiCMOS

Tai-Haur Kuo, EE, NCKU, 1997 BiCMOS Circuit Design 3-26 BiCMOS Gate

Basic type

VCC VCC

+ ≡

CMOS Bipolar BiCMOS (Buffer)

Similar methods as used in the inverter design 1. delay time calculation 2. path to discharge the bases 3. full-swing

Tai-Haur Kuo, EE, NCKU, 1997 BiCMOS Circuit Design 3-27 Power-Supply Sensitivity

VDD = 5.0V VDD = 2.5V

Delay(ns) Delay(ns) CMOS 0.8 CMOS 0.8 BiCMOS 0.4 0.4 BiCMOS C(pF) C(pF) 1.0 2.0 1.0 2.0

Input Waveform 5.0V 4.3 2.5V 1.8 0.0V 0.0V t t } } Critical switching window Critical switching window

4.3V 1.8V 1.2V 0.9V VGS=3.1V VGS=0.9V

Tai-Haur Kuo, EE, NCKU, 1997 BiCMOS Circuit Design 3-28 Voltage Scaling

BiCMOS (full swing) µC W ββIK=−()VV2 wherek=ox DP P DD tp 2 L 2 ββIKDN =−N ()VDD VBE −Vtn

2 V V =−βKV ()1 BE −tn 2 nDD (1) VDD VDD

CMOS 2 IKDP =−P ()VDD Vtp 2 IKDN =−n ()VDD Vtn

2 Vtn 2 =−KVnDD ()1 (2) VDD From (1) & (2) V V 1 −−BE tn V V I V ↓⇒()DD DD ↓⇒ DN ()BiCMOS ↓ DD V I 1 − tn DN ()CMOS V DD βI t ⇒↓DN ()BiCMOS ⇒d ()BiCMOS ↓ I DN ()CMOS t dC()MOS

CMOS BiCMOS IDN(CMOS) βIDN(BiCMOS)

Tai-Haur Kuo, EE, NCKU, 1997 BiCMOS Circuit Design 3-29 Low-Voltage BiCMOS Gates V IN H Conventional V BiCMOS BE V BE BiNMOS C BiCMOS V H NMOS PNP Q-C BiCMOS Q-PNP NPN

Tai-Haur Kuo, EE, NCKU, 1997 BiCMOS Circuit Design 3-30 Delay Time

[Y.Kobayshi] Delay Time (ps) 1000 100 200 400 600 800 L g = 0.3 QC-BiCMOS 2 µ m, f 3 T Supply Voltage (V) (npn) = 24 GH C-BiCMOS Darlington 4 CMOS 56 Z, BiCMOS f T (pnp) = 10 GH 3NAND(FO=5) Z

Tai-Haur Kuo, EE, NCKU, 1997 BiCMOS Circuit Design 3-31 Delay Time (Cont.)

[ Y.Kobayshi ]

300 CMOS 250

Darlington BiNMOS BiCMOS 200

C-BiCMOS 150 QC-BiCMOS

DELAY TIME (ps) 100

3 NAND 3.3V 50 01234567 FAN OUT

Tai-Haur Kuo, EE, NCKU, 1997 BiCMOS Circuit Design 3-32 TTL, ECL and CMOS Logic

Level conversion circuits for different logics

Input BiCMOS buffers of various logics

Output BiCMOS buffers of various logics

Example : ECL-CMOS interface

Tai-Haur Kuo, EE, NCKU, 1997 BiCMOS Circuit Design 3-33 High Speed BiCMOS IC

ECL I/O example

ECL Input Buffer

ECL to CMOS Translator BiCMOS IC

CMOS Logic

CMOS to ECL Translator

ECL Output Buffer

Tai-Haur Kuo, EE, NCKU, 1997 BiCMOS Circuit Design 3-34 ECL-CMOS Interface

ECL-to-CMOS converter

VDD= 3.6V 2.8V

2.2V 3.6V 3.6V 2.0V 2.0V 1.4V 3.6V 1.7V 0.0V 2.0V

CMOS-to-ECL converter

VDD= 3.6V

3.6V 3.6V 3.0V 2.8V

0.0V 2.2V

Tai-Haur Kuo, EE, NCKU, 1997 BiCMOS Circuit Design 3-35 Noise Coupling of Mixed ECL/CMOS Circuits

[ L.Wissel CICC 1992] Large CMOS waveform couples noise into ECL signal lines

Model 2 mm quiet ECL line, with partial adjacency to a switching CMOS line

NCMOS ∆VCMOS CC

CMOS

N1 N2 ECL

CL 1.6mm ∆VECL ECL 2.0mm

4.0 CL=0.4pF, CC=0.07pF V N1 C N ∆∆V = ()C V 2 ECL CMOS 2.0 CCLC+

N 0.0 CMOS

0.2 0.6 1.0 ns CMOS switching can induce noise as large as ECL signal

Tai-Haur Kuo, EE, NCKU, 1997 BiCMOS Circuit Design 3-36