IEEE TRANSACTIONS ON ELECTRON DEVICES, VOL 42, NO, 6, JUNE 1995 1149

Modeling the Positive-Feedback Regenerative

Process of CMOS Latchup by a Positive Transient

Pole Method-Part II: Quantitative Evaluation Ming-Dou Kef, Member, IEEE, and Chung-Yu Wu, Member, IEEE

Abstract-The positive-feedback regenerative process in a p-n­p-n structure during CMOS latchup transition has been modeled by a time-varying positive transient pole. The maximum peak value of the positive pole and the time required to first initiate the positive pole are adopted as two useful and meaningful parameters to quantitatively investigate the influence of de­vice parameters on the positive-feedback regeneration of CMOS latchup. Some design guidelines can be obtained to improve latchup immunity of CMOS IC's.

L INTRODUCTION

THE dynamic behaviors of CMOS latchup transition have

been analyzed by the lumped equivalent two-transistor

model and the time-varying large-signal base-emitter voltage

waveforms of the cross-coupled bipolar transistors have been

solved in [I]. In each time interval, the base-emitter voltage

waveforms can be further represented as two-pole functions

of time by the piecewise-linearized approximation. One of the

poles is found to increase from negative to positive during

the transition of latchup. After latchup occurs, this positive

transient pole will decrease from positive to negative again to

retain latchup in its stable low-impedance state.

Although a pole with positive value had ever appeared

in several early works to qualitatively explain the turn-on

transition of thyristors [2], [3], there was never an impor­

tant parameter used to quantitatively investigate the positive­

feedback regenerative process in a p-n-p-n structure. In this

paper, the positive-feedback regenerative process in a p-n-p­

n structure is quantitatively investigated by Pl(max) and tr parameters, which are the maximum peak value of the time­

varying positive pole during the latchup transition and the

minimum time period required to initiate this positive pole [I].

By using these two parameters, the influence of device param­

eters on the positive-feedback regenerative process in CMOS

latchup are quantitatively investigated. Some design guidelines

on the device parameters to prevent CMOS transient-induced

latchup can be highlighted.

Manuscript received September 7, 1994. The review of this paper was arranged by Associate Editor K. Tada, This work was supported by United Microelectronics Corporation (UMC), Taiwan, ROC, under Contract C82051.

The authors are with Integrated Circuits and Systems Laboratory, Institute of Electronics and Department of Electronics Engineering, National Chiao­Tung University, Hsin-Chu, Taiwan 300, Republic of China,

IEEE Log Number 9410586.

Pulse Width of the Trigger Current (nS)

(a)

8�r---------���7==='������ T�ggered by 11 = SmA 02=DmA).

6�

T�ggered by 12= O.SmA (11 =DmA).

RB=Q.8Krl, Rw=S.6Krl •

.=7.6nS 6.27nS

, Latchup t--" __ ��1I: 5 � -o-o.OOuu .. � .. _ ..... u ••••• -o ••••••••• �:.�.�� ••

5�4L---�---L

6--------�

8---------

1�O

------�12

Pulse Width of the Trigger Current (nS)

(b)

Fig. 1. The variations of Ca) Pl(max) and (b) tT when the p·n-p-n structure is triggered by a S-rnA It or a O.S-rnA h with different pulse widths.

II. QUA NTITATIVELY MODELING THE

POSmvE-FEEDBACK REGENERATIVE PROCESS

A. The Maximum Value of Positive Pole and The Time to Initiate Positive Pole

A p-n-p-n structure triggered by trigger currents with differ­

ent pulse widths or different pulse heights will lead to different

values of Pl(max) and lr. The curves in Fig. I(a) and (b)

show the variations on Pl(max) and t,. due to h = 5 rnA

or 12 = 0.8 rnA triggering with different pulse widths. The

device parameters in the p-n-p-n structure used for calculation

are listed in Table I of [I] with the substrate (well) resistance

of 0.8 Kf2 (5.6 Kf2). In Fig. l(a), as the pulse width of trigger

0018-9383/95$04.00 © 1995 IEEE

1150

u;­s. --

3.00+8r-----------::=;;;:;::;;""" ..... --�"'I IKF2; 4.8678-4

2.50+8

2.00+8

1.58+8

Rs�8000, Rw�5.6KO. 1.00+8

Irlggered by 11� SmA, Ian$.

_·-W-" b1ggered by 12= O.6mA. 10nS.

5·o.+;O�·7. -��l�O-; .• ,.....=:::;,O=.';:::::==l:O.::;:'==:,::;:O::;.'=:.......J,o.

IKF1 (A)

(a)

�.8r_-----------------__, IKFl � 6.9098-4

Trlggo.ed by II; SmA, IOn$. Irlggorad by 12= 0.6mA, IOnS.

'0" 1 0" 1 0"

IKF2 (A)

(b)

1 O' 10'

"r---r=7=�==========�==��===>--, IKF1 varying with. fixed IKF2=4.887H

10

10"

IKf2 varying wl1h. flxtclIKF1=6.90ge-4. IKFI varying wlUl • flxod IKF2=4.867.-1. IKF2 varying wllb • fixed IKF1=6.90ge-4.

Rs; 8000, R_ 5.6Kn.

Trlgger.d by 12=O.6mA, 10nS.

10"

IKF1 or IKF2 (A)

(c)

10' 10'

Fig. 2. The relations between (a) Pl(max) and IK Fl parameter of BJT Q" (b) Pl(max) and If{ F2 parameter of BIT Q2, (c) tr and the If{ Fl or IK F2 parameters, due to !he triggering of 1, = 5 rnA or 12 = 0.6 rnA with a lOonS pulse width.

current is greater than 7.6 nS for h =5 rnA (/2 = 0) or 7.2 nS for h = 0.8 rnA (h = 0) in the latchup range, Pl(max) is

held at a constant value about 2.2919 x J08 or 2.4286 x J08,

respectively. If the pulse width of trigger current decreases

from the critical time period, Pl(rnax) raises up sharply and then it drops to zero suddenly as the pulse width of trigger

current is much reduced. In Fig. I(b), tr is held at 6.27 nS

(5.40 nS) for the triggering of h = 5 rnA (/2 = 0.8 rnA) in

the latchup range. As the pulse width of trigger current is less

IEEE TRANSACTIONS ON ELECTRON DEVICES, VOL. 42, NO.6, JUNE 1995

'S' E

r::.,-;::

u;-S. ...t"

3�r----------------------------------, BF2:2T7.2 RI.aooo, RwQ.61ffi.

3.08+8 ____ --------..--0--<01

2..5e+8

T�gg.red by 11= 5mA, 1OnS. T�ggreed by 12= O.6mA, 10nS.

1.Oe.aL�-�.......'::=::::;===::::::;:==:::::;:=::.....-J

8.0

7.5

7.0

6.5

e.o

5.5 0

o 2. 4 6 8 ro The Forward Beta-Current Gain of BJT Q1

OF' (a)

1- Triggered by 11= SmA, ,ons. 1 n....... TrIggered by Ih G.8mA, IOnS ..

1--__ ........ -0 .. -0-0--••••• 0---.---0 ...•. --0---0 ...•. --.-.• ---0---

6 8

The Forward Beta-Current Gain of BJT Q1 BF1

(b)

10

Fig. 3. The influence of the forward beta gain PFI of BJT Q, on (a) Pl(max) and (b) tr due to Ihe triggering of '1 = 5 rnA or '2 = 0.6 rnA wi!h a lOonS pulse width.

than this constant value, tr is first held as a nearly constant value in the nonlatchup range but then it quickly increases

to infinite. The time tr raising up to infinite means that the

positive pole does not occur when the trigger current has a

too short pulse width.

The values of PI (max) and t,. in the latchup range of Fig. I(a)

and (b) are held as constants in spite of different pulse widths

of the trigger current. Therefore, for a trigger current with fixed

pulse height, there exist the corresponding constant Pl(max) and t,. in latchup case. These constant Pl(max) and tT can be

seen as one of inherent characteristics of a p-n-p-n structure.

Thus, they can be adopted as meaningful parameters to find

the influence of device parameters on the positive-feedback

regeneration of CMOS transient-induced latchup.

B. The Difference Between the Time to Initiate Positive Pole and the Time to Cause Latchup

By using iterative HSPICE simulations, the p-n-p-n structure

is found to be triggered into latchup only as the pulse width of 5-mA h is greater than 7.6 nS and that of 0.8-mA h greater

than 7.2 nS. It is very interesting to find the reasons why the

time tr of 6.27 nS (5.40 nS) to initiate the positive pole and

the minimum pulse width of 7.6 nS (7.2 nS) to cause latchup due to h = 5 rnA (/2 = 0.8 rnA) triggering are different.

KER AND WU: MODEUNG TIIE POSITIVE-FEEDBACK REGENERATIVE PROCESS OF CMOS LATCHUP-PART " 1151

3.O8+8,----------------------, BF1=I.104 •• 0······ Rs=8000, RW=5:�::�.4 •• 4

•• 4 •

• 4.-D-.�.-�--·D ...... -••

2.08+8

1- Trlgge"'" by 11= SmA, IDnS. 1 ------ Trine"'" by 12= 0 .6mA, IOnS. o.oe.o L�-��========��

o 100 200 300 400 !500

The Forward Beta-Current gain of BJT Q2 Bf2

(a)

grri:------��������� 1 - Triggeredbyl1=5mA,IOnS. 1 8

7

I�

i 1'---'- Trlgge"'" by 12 = o.8mA. IonS. � '\

" ..

.,."0'0'0"0 .... 0-...... 0

.....

.......... 0 ... 0 •• "'0 ........ "'0 •• -0 ........

0 ....... ...

BFI=1.104 Rs=BDOO, Rw=5.6KO.

6 �---� ____ -L ____ � ____ � __ �-J o 100 200 300 400

The Forward Beta-Current Gain of BJT Q2 BF2

(b)

500

Fig. 4. The influence of the forward beta gain (3P2 of BIT Q2 on (a) Pl(max) and (b) tr due to the uiggering of It = 5 rnA or I2 = 0.6 rnA with a lO-nS pulse width.

An analytical criterion based on the product of large-signal

transient current gains greater than unity has been developed

to judge the occurrence of latchup [4]. The criterion has been

expressed as

(1)

where

(2)

(3)

(J1tr ( t) and (J2tr (t) are heavily dependent upon the junction

capacitances in a p-n-p-n structure.

With this "(J1tr . (J2tr � I" criterion, the difference between

the minimum pulse width to cause latchup and the time tr to

initiate positive pole can be clearly explained. The time tn when the positive pole occurs, is the time that the positive­

feedback regeneration starts. Although the starting of the

positive-feedback regeneration will quickly push a p-n-p-n

structure into latchup with a double exponential regenerative

rate [I], it still needs a time period to cause the happening

.. .. .s

1:1,-

2.68+8

2Ae+8

2.21+11 Triggs"'" by l1=SmA, IOnS. Triggs"'" by 12= 0.6mA, lDnS.

2.08+8.1 10

Zero-Biased Well-Substrate Junction Capacitance Cjc1+Cjc2 (pF)

(al

1'.0,----------------:Jr-----,

9.0

ClehO.6pF, Cle2=1.3pF; R8=BODO, Rw=5.6KO. CIC1=Clc2

Triggered by l1=smA, IonS. Triggered by 12=O.6mA, Ions'

3.0 0 2 4 6 8 10

Zero-Biased Well-Substrate Junction Capacitance Cjc1+Cjc2 (pF)

(b)

Fig. 5. The influence of the well-substrate junction capacitance on (al Pl(m.x)and (b) tr, due to the uiggering of It = 5 rnA or h = 0.6 rnA triggering with a 10-nS pulse width.

of latchup. This time period is about 1.33 (1.8) nS for the

triggering of h = 5 rnA (12 = 0.8 rnA), A p-n-p-n structure which has a stronger positive-feedback regeneration needs a

shorter time period to cause latchup after the occurrence of positive pole. The time, when latchup happens, have been found as the time that the value of (J1tr (t) . (J2tT (t) reaches

to unity [4]. So, if the trigger current with a pulse width long enough to initiate the positive pole but not long enough to make the value of (Jltr (t) . (J2tr (t) reach to unity, the p-n-p-n structure is still not triggered into latchup state.

Moreover, there may be a question why the maximum positive pole in nonlatchup range has a greater value than

that in latchup range, as shown in Fig. I(a). The anomalous increase of Pl(max) in nonlatchup range is due to the happening

of physical self-regeneration in a p-n-p-n structure after the trigger current is off. This physical self-regeneration with the absence of trigger current in nonlatchup case causes the high­level injection effect, which occurs as the collector currents of BJT's Q1 and Q2 are greater than their IKF parameters, to occur much later. Since the self-regeneration in the base and collector currents without enough supporting from the trigger current can not push (J1tr . (J2tr to reach to unity, the collector currents of BJT's Q1 and Q2 raise up slower and then will gradually decrease because of the absence of latchup. Thus, the positive pole can grow up toward a larger value before the

high-level injection effect occurs in the nonlatchup case.

1152

2.68+8 r-----;:==:;===.:=.:====:::::;---, 1-- Triggered by 11= SmA,10nS. 1 .

.....

.. Triggered bV 12= O.6mA, 10nS. 1

2.58+8 ........

.

.............

......

....................

...............

.........

............. . 2.4e+8

2.38

+8 r�-o-----------_-_____ ....J Rs=800C, Rw-..5.6Kn; Cje2=1.3pF, Clc2=D.6pF; Cjcl =2.0pF. 2.2e+8L------�--�--�--------'

a 1 2 3 Zero·Blased Base-Emitter Junction Capacitance of BJT Q1

Cje1 (PF)

(a)

9.0 r;::::::::::::=;::;::�::::::;::;;�:7:;;::;;:=::;_------1 1 1-- TrllIII_byll=5mA,10nS. 1

I I··· ... ·· Triggered by 12: O.6mA, 10nS.

8.0

oO'"� .. 7.0

Ro=8OOn, Rw=5.6Kn; CJe2=1.3pF, CJc2=D.6pF; CJcl=2.OpF. • •••

� .

..

.

..

.

..

.

...

...

.......

..

... ..

� ...

...........

.........

.....

.. -

... .

--......-----I ......... 6.00L------�----- ........ --------'

Zero-Biased Base-Emitter JunctIon CapacItance of BJT Q1 Cje1 (pF)

(b)

Fig. 6. The influence of the base-ilmitter junction depletion capacitance of BJT Ql on (a) p, (max) and (b) tr due to the triggering of 1, = 5 rnA or 12 = 0.6 rnA with a lO·nS pulse width.

III. THE INFLUENCES OF DEVICE PARAME TERS

ON THE POSITIVE-FEEDBACK REGENERATION

In this section. the influence of device parameters on the positive-feedback regeneration during latchup transition is quantitatively investigated by the Pl(rnax) and tr parameters.

A. The Influence of Current·Gain Parameters

The value of Pl pole changing from positive to negative after latchup is due to the high-level injection effect which causes serious current-gain degradation in the parasitic BJT's Ql and Q2. The high-level injection effect, which is modeled by the knee-current I K F parameter in the Gummel-Poon model of BJT device [5]-[7], degrades the current gains of BJT's Ql and Q2 when these BJT's are heavily turned on during latchup transition in a p-n-p-n structure.

To quantitatively investigate the high-level injection effect on the positive-feedback regeneration of CMOS latchup, the trigger current of h = 5 rnA (or 12 = 0.6 rnA) with 10 nS pulse width is applied to the p-n-p-n structure under different 1KFl or 1KF2 parameters in BJT's Ql and Q2. The Pl(max) and tr are used to quantitatively observe the influence of this effect. With a fixed 1KF2 of 4.867 x 10-4 and variable 1KFl parameter, the influence of 1KFl on Pl(rnax) due to the high-level injection effect of BJT Ql in the p­n-p·n structure is shown in Fig. 2(a). In Fig. 2(b), 1K F2 is

IEEE TRANSACTIONS ON ELECTRON DEVICES. VOL. 42. NO.6. JUNE 1995

3Jle+8 Rs=8000, Rw=5.6KO; TF2=0.25nS.

Triggered by 11= SmA, 10nS. Triggered by 12: O.6mA, lOnS.

1Jle+8.Ll--------��1--��----�10�--���� Forward Transit Time of BJT Q1

TF1 (nS)

(a)

14.0 r;::::;::=;;::;:::=':;:::::;;::::;='=:;::;�;-----l 1-- Triggered by 11=5mA,10nS. 1 ........ Triggered by Ib MonA, lOnS .

. 12.0

10.0

8.0

6.0

4.0

Ra:80DO, Rw=5.6Kn; TF2=0.25nS.

·······O···'O··O_..,OOOO··· ... ··o .... "CI··a·'O"V�OOO-o·,o·-O-'O

�OL---------�u---------��--------�� .1 10 100

Forward Transit Time of BJT Q1 TF1 (nS)

(b)

Fig. 7. The influence of the forward transit time TFI of BIT Q, on (a) Pl(max) and (b) tr due to the triggering of 1, = 5 rnA or 12 = 0.6 rnA with a 10· nS pulse width.

a variable but 1KFl is fixed as 6.909 x 10-4 to observe the influence of high-level injection effect in the parasitic vertical BJT Q2. The influence of 1KFl or 1KF2 on tr is shown in Fig. 2(c). From these figures, it can be seen that larger 1KFl and 1KF2 parameters, which mean a weaker high-level injection effect, lead to a larger Pl(rnax) and a smaller tr. The larger Pl(max) implies a stronger positive­feedback regeneration during latchup transition. The smaller tr means that the positive-feedback regeneration occurs more early. Thus, a p-n-p-n structure with larger 1KFl and 1KF2

parameters is more sensitive to latchup. The variations of Pl(max) and tr in Fig. 2(a)-(c) are kept very small as 1KFl

and 1KF2 are greater than about Ix 10-2• This implies that the latchup immunity can be improved only when the 1KFl

and I K F2 parameters are less than some threshold levels. The device parameter related to the current gain of a BIT

is the ideal maximum forward beta gain denoted as {3 F in HSPICE [7]. The influence of the ideal maximum forward beta gain (3 F1 of BJT Q 1 on the positi ve-feedback regeneration is shown in Fig. 3(a) and (b) with a fixed {3F2 = 277.2, where fJF2 is the ideal maximum forward beta gain of BJT Q2 in the p-n-p-n structure. {3F2 in submicron CMOS technologies could be in the range about several hundreds, but {3Fl is only around 1-2. The influence due to (3F2 variation is shown in Fig. 4(a) and (b) with a fixed {3F1 = 1.104. As seen in Figs.

KER AND WU: MODELING THE POSITIVE-FEEDBACK REGENERATIVE PROCESS OF CMOS LATCH UP-PART II 1153

Ui S. .. �

1.08+8

Trlgganod by 11= 5mA, IOnS. Trlgganod by 12= 0. 8mA, IOnS. Rs=BODU, Rw=5.6KU; TF1=20nS.

0�.Ll------�----���

I-----L----------�

10

11.0

10.0

9.0

B.O

7.0

6.0 . 1

Forward Transit Time of BJT 02 TF2 (nS)

(a)

TrI_redbyll=5mA.10nS. I ! Trtggered by 12= O.8mA, IOnS. i

Rs=BODU, Rw=S.8KO; ,/ TF1=20nS. / /' "

,/ 4J�.d''''o

," •• � ..... _.CII .........

.o"",,"G

1 Forward Transit Time of BJT 02

TF2 (nS)

(b)

-

10

Fig. 8. The influence of the forward transit time TF2 of BIT Q2 on (a) Pl(max) and (b) tr due to the triggering of h = 5 rnA or [2 = 0.6 rnA with a lO-nS pulse width.

3(a) and 4(a), Pl(max) has a larger value if the p-n-p-n structure has a larger {3Fl or {3F2. The influence of {3F1 on tr is more sensitive as the p-n-p-n structure is triggered by the substrate current h, but t,. only varies a little as it is triggered by the well current 12. On the contrary, the influence of (3F2 on tT is more sensitive as the p-n-p-n structure is triggered by 12, but tr only varies a little by the triggering of h. A p-n-p-n structure with larger beta gains generally leads to a larger Pl(max) but a smaller t,., so it is easy triggered into its latching state.

B. The Injiuence of Device Capacitance Parameters

The device capacitances have strong relations to the time­varying transient poles [1]. Fig. Sea) and (b) show the depen­dence of the positive-feedback regeneration on the zero-biased well-to-substrate junction capacitance in the p-n-p-n structure of CMOS IC's. The well-substrate junction capacitance is the sum of the base-collector junction capacitances in BJT's Ql and Q2. The Cjel (Cjc2) presents the zero-biased base­collector junction capacitance of BJT QI(Q2). The trigger currents in Fig. Sea) and (b) are h = S rnA or 12 = 0.6 rnA with the pulse width of 10 nS. It is shown that a p-n-p-n structure with a larger well-substrate junction capacitance has a smaller PI (max) but a larger tT• This means that it has a higher latchup immunity against the substrate or well current triggering. This result is quite different to the latchup transition

'iC .. .s

r:., ....

2.7_

2.5e+8

2._

2.1_

1.geHI

1.7_

1.58+8 .1

Rw=5.6KO ....... ..................... __ .... ..... _ ............. -..... .

". .......... / -0000 .............. -••• 0 .. 0•0 •• 0•

J Qa ..... • .... • ..

...... .,g

.. ·o .. qoO

/ I

p' .' � ..

Triggered by 11= 10mA, IOnS. Trt� by 11= SmA. IOnS. Trl_red by 12= I.2mA, IOnS. Trtggered by 12= 0.8mA, IOnS.

10

Substrate Resistance Rs (KG)

(a)

9r---�------IC==�TrI� g� g.�m�bY�II�.� 10rnA��.I

���S.

�

8

4

3 .1

TrIggered by 11= 5mA, IOnS. TrIggered by 12= l.2mA. IOnS. Triggered by 12= 0.8mA, Ions.

........ " ..... _=----------�---...... o-o�

Rw:5.6KO "0"000 ••••• _-'0 .•.•.• 0-.-0.1:100 __ ..... _0 .

......... '0.'0.0._0_

10

Substrate Resistance Rs (KG)

(b)

100

Fig. 9. The influence of the substrate resistance on (a) PI (max) and (b) tr due (0 It = 10 rnA or 5 rnA and h = 1.2 rnA or 0.6 rnA triggering with a lOonS pulse width.

induced by the power-up ramp (dVDD/dt) [8]-[10]. There is an anomalous increase in the gradually decreasing curves in Fig. Sea). This anomalous increase is due to the nonlatchup range with the pulse width of trigger current not longer enough as that in Fig. lea). A little anomalous increase due to the same reason also appears in the gradually decreasing curves in Figs. 2(b) and 4(a) under 12 = 0.6 rnA triggering. In the latchup range of Fig. S(b), t,. is found to have a nearly linear relation with the well-substrate junction capacitance.

The dependence of base-emitter junction depletion capac­itances of BJT Ql on the positive-feedback regeneration is shown in Fig. 6(a) and (b), where the Cjel(Cje2) is the zero-biased base-emitter junction depletion capacitance of BJT Ql(Q2). The dependence of Cje2 on the PI (max) and tT is similar to that in Fig. 6(a) and (b). Generally, a larger base­emitter junction capacitance leads to a smaller PI(max) but a larger tr. A weaker dependence of Plimax) on the junction capacitance Cjel duo to h triggering and Cje2 due to lz triggering in the latchup range is observed. The anomalous raise due to nonlatchup condition also appears in the curves due to lz = 0.6 rnA triggering. It is also found that the dependence of tr on the zero-biased base-emitter junction capacitances Cjel and Cje2 is nearly linear.

Another important device parameter which also has an obvious contribution on the junction diffusion capacitances is

1 154

3.0l1+li r-;:;R;::"s=--;;:80;;;O:::n:---------------, �.OOOOOODClO-.---O--.. ""' ........ -............. O .... -O ••

I� 2.0e+8� t � i ; 1.0e+8 I 1 t .. _.. Triggered by 11z SmA, 10nS. 1

I" Triggered by 12= O.6mA, lOnS.

O·oe.ol!------�===========�J 10 100

Well Resistance Rw (Kn)

(a)

10 I-';-r.��;;::::::::;;:�:;::::::::;::;:::;-----l \ 1-- Triggeredbyll.5mA,10nS. I " ... _.. Triggered by 12= 0.1ImA, 10nS . .

" 0 ........... ."0 .... -0 ................. "0 .......... _ ••

Rs= 8000 5�1---------------�10�--�-------� 100 Well Resistance Rw (KQ)

(b)

Fig. 10 . . The influence of the well resistance on (a) Pl(max) and (b) tr due to the tnggenng of h = 5 rnA or I2 = 0.6 rnA with a lO-nS pulse width.

the forward transit time, denoted as TF [7]. The influence of the forward transit times, TFI and TF2, of BIT's Ql and Q2 on the positive-feedback regeneration are shown in Figs. 7 and 8,

respectively. The curves indicate that a p-n-p-n structure with

larger forward transit times has a smaller Pl(max) but a larger

t,.. However, tr is found to be nearly independent of TFl and TF2 in Figs. 7(b) and 8(b) under the triggering of 12 = 0.6 rnA and II = 5 rnA, respectively.

The curves in Figs. 6--8 show that a p-n-p-n structure has higher latchup immunity against the substrate and well current triggering if its base-emitter junction depletion and diffusion

capacitances are larger.

C. The Influences of Device Resistance Parameters

According to (23) of [1], the substrate and well resistances

have important effects on CMOS latchup transition. The

curves in Figs. 9 and 10 show their effects on the positive­feedback regeneration, In Fig. 9(a), PI (max) under h = 5 rnA or h = 10 rnA triggering is slowly decreased as the

substrate resistance increases from 0.2 to 3.0 K n, and then

it is held nearly constant as the substrate resistance greater

than 3.0 Kn. But, PI(max) under h = 0.6 rnA or 12 =

1.2 rnA triggering continuously increases to its final value as the substrate resistance increases. Although the variation

tendency of PI(max) in Fig. 9(a) due to h and h triggering

is different, Fig. 9(b) shows that tr decreases as the substrate resistance increases in spite of the trigger current. In Fig. 9(a),

IEEE TRANSACTIONS ON ELECTRON DEVICES. VOL. 42. NO. 6. JUNE 1995

" .. oS

r:.,. ..

'iii S. .:-

2.5e+8

2._

2.30+8

2.28+8

2. 10+8

2_ 0

R8=8000, RW�

.

�:::.

:: .....

.....

.......................

.....

...

.. ..

....

r l Triggered by 11= SmA, IOnS.

Triggered by 12. 0.8mA, 10n5.

4

VDD (Volt)

(a)

8

8.0 I'---��;;;;:;::::::;:::;::;:::;::;;::;! Trlgg __ by 11 = 5mA, IOnS.

7.0

6.0

5.0

"\ Trlgge_ by 12= 0.8mA, 10n5.

... \" ".

'O..�·o.

•

.

.. o···�·

···--···_· ..

·•·· ...

...... o ____

.

...

___ ..

Rs=0.8KO, Rw=5.6KO.

4.DO��--:------4:-�----':'-----8�---.--Jl0

VDD (Volt)

(b)

10

Fig.11. The influence of the power supply V'DD on (a) Pl(max) and (b) te due to the tnggenng of 1, = 5 rnA or I2 = 0.8 rnA with a lO·nS pulse width.

the nonlatchup condition causes two anomalous increases in the curves of PI(rno.x) under 12 = 0.6 rnA and II = 5 rnA triggering as the substrate resistance is smaller. Similarly, an

anomalous increase also appears in Fig. lOCal because the IO­nS pulse width of the 0.6-rnA trigger current 12 not long

enough to cause the occurrence of latchup when the well

resistance is less than 5 Kn. A p-n-p-n structure with larger well or substrate resistances

generally has a larger Pl(max) and a smaller tf' during latchup transition. This causes CMOS latchup more easy to be initiated by the transient-induced current triggering in substrate or well.

D. The Influence of VDD Power Supply In above various calculations, the VDD supply is all set to

5 volts as in the conventional CMOS IC's. The effect of VDD on the positive-feedback regeneration in a p·n·p-n structure

is also investigated and the results are shown in Fig. II to

observe the influence of the scaled down VDD on CMOS

latchup. The curves show that a lower VDD supply leads to a

smaller Pl(max) and a larger tr during latchup transition. Thus, the scaled down VDD has a benefit to prevent CMOS latchup.

IV. CONCLUSION

Through the detailed investigations on the variations of device parameters, it is found that a p-n-p-n structure with

stronger high-level injection effect, larger well-substrate ca-

KER AND WU: MODELING THE POSITIVE-I'EEDBACK REGENERATIVE PROCESS OF CMOS LATCHUP-PART II 1155

pacitance, larger forward transit time, larger base-emitter junc­

tion capacitance, but smaller well and substrate resistance, and smaller forward beta gain in the parasitic vertical and lateral BJT's generally has a smaller value of the maximum positive pole and a longer time to initiate the positive pole. Thus, it has higher latchup immunity against transient-induced current triggering in substrate or well of CMOS IC's.

The characterization of the positive-feedback regeneration using the time-varying positive transient pole is useful and convenient in CMOS latchup analysis under any type of triggering. This new developed method is also suitable for

transient analysis on the turn-on process of thyristors or

REFERENCES

[II M.-D. Ker and c.-Y. Wu, "Modeling the positive-feedback regenerative process of CMOS latchup by a po.itive tran.ient pole method: Part I-Theoretical derivation," IEEE Trans. Electron Devices, vol. 42, no. 6, pp. 1141-1148, 1995.

[21 T. Misawa, "Turn-on transient of p-n-p-n triode," J. Electron. Contr .. vol. 7, no. 6, pp. 523-533, 1959.

[3) G. D. Bergman, "The gate-triggered turn-on process in thyristors," Solid-State Electron., vol. 8, pp. 757-765, 1965.

(4) M.-D. Ker and c.-y. Wu, "Transient analysis of submicron CMOS latchup with a physical criterion," Solid-State Electron., vol. 37, no. 2, pp. 255 -273, Feb. 1994.

[5) l. E. Getreu, Modeling the Bipolar Transistor. Oregon: Tektronix, 1976.

(6) P. A ntognetti and G. Massobrio, Semiconductor Device Modeling with SPICE. New York: McGraw-Hili, 1988.

l7J HSPICE User's Manual. CA: Meta-Software, Inc., 1990. [8) R. R. Troutman and H. P. Zappe, "A transient analysis of latchup in

bulk CMOS," IEEE Trans. Electron Devices, vol. ED-30, no. 2, pp. 170-179, 1983.

[9] __ , "Layout and bias con.iderations for preventing transiently triggered 1atchup in CMOS," IEEE Trans. Electron Devices, vol. ED-3I, no. 3, pp. 315-321, 1984.

[10) K. Y. Fu, "Transient latchup in bulk CMOS with a voltage-dependent well-substrate junction capacitance," [EEE Trans. Electron Devices, vol. ED-32, no. 3, pp, 717-720, 1985.

Ming-Dou Ker (S'92-M'94), for photograph and biography, please see p. 1148 of this issue of this TRANSACTIONS.

Chung-Yu Wu (S'75-M'77) for photograph and biography, please see p. 1148 of this issue of this TRANSACTIONS.