PROFESSORS NOTES 12.

1 THE BIPOLARJUNCTION TRANSISTOR: BASIC PRINCIPLES The bipolarjunction transistor is one of the earliest types of semiconductor transistors. It likely would be the type that your mother used when she was a young college student. The BJT is made up of two pn junctions backto back, formed from three layers of semiconductor. It is therefore either of the form pnp or npn. The BJTs, like your forebears, were rugged and strong, and this endurance and strength are important for many categories of contemporary circuits. Most BJTs are formed by a planar technology in which an epitaxial layer is laid down on a substrate. This epilayer is of opposite gender to that of the substrate, e.g. ntype on a ptype substrate. The word epitaxial is one of those words coined by the semiconductor industry, meaning that an epi, or surface layer is grown on top of the crystalline substrate, acquiring the same crystalline taxial structure as the substrate. By means of implants, encirclement ptype diffusions are applied to the epilayer to create little islands of nmaterial in a ptype sea. If we had started with an ntype substrate and created a ptype epilayer, the islands would have been ptype islands in an ntype sea. Each island is a place where a BJT may be constructed in isolated splendor. This structure with an implanted npn BJT is represented by Figure 12.11.

Figure 12.11 Fabrication of the BJT. An npn transistor in crosssection, surrounded by ptype within its little island, is represented. The substrate is ptype and the epilayer is ntype. The figure shows that the three layers of the BJT are three successively shallower layers, each counterdoped from the previous layer. The epilayer is the outermost of these three layers. If we take a slice of npn (or pnp, if the epilayer is ptype), across this structure, we have a basic 3layer BJT, as represented by the inset to figure 12.11. Note that the uppermost layer becomes the emitter of the BJT. This is an advantage, inasmuch as it helps to create a BJT with large forward current gain F. Operation of the BJT in the forwardactive mode is such that the baseemitter (BE) junction is forwardbiased while the basecollector (BC) junction is reversebiased. Under this condition, carriers flow through the npn slice in the manner represented by Figure 12.12. Because the junctions are very close to one another, the carriers injected across the BE junction by the forward bias find themselves in the vicinity of the strong Efield of the reverse biased BC junction. This strong Efield is from the collector toward the emitter, which is just exactly the direction

120

1 6'@ 8 6! 3 90BA) 8 9754

6 (B27G S 79FB1 Hyw xQQHvt@ 72027raq 4G %' @ ) GG 61 @ % E S E # I w u ) )1 )s )1 ) % )GE! C 7HFD0    W  V       1 )' % #! 20(' &$" 

S  )' E1'G P I ! 2) U ' T R0HF(4Q94G

p di ` 3  f f d c 3 `X eFYYh9f S QgeDbaYC T

of field that excites the mobile () charge carriers, and sends them racing from the base region into the collector region. Some of the carriers, of course, dont make it across the base region, due to recombination. These casualties are carried off by the base terminal. Note that current flows in the opposite direction to the electron flow. This flow is indicated by Figure 12.12.

Figure 12.12: Flow of charge carriers and current in the BJT. An npn transistor is represented From this figure it is pretty clear that
8

IE and that

IC

IB

tor current I C with only a very small amount of base current I B . Equation (12.2.3) may also be expressed in the form:
E

As we will see in section 12.6, equation (12.14) is handy for determining the effects of our fabrications on the transistor current gain. The BJT is, in principle, bilaterally symmetric. An np pn junction structure exists, whether it is approached from lefttoright or from righttoleft. The transistor may be inverted, with the BC junction forwardbiased and the BE junction reversebiased, in which case we will have
F @ AGF 8

IC and

121

which is large provided that

is close to 1. The more ideal transistors are able to control a large amount of collec-

1
F

R E

C D8

(12.11) and (12.12), are sufficient to define the forward gain current gain, I C I B
@ C

where

the forward current factor, which for a good quality device, is almost equal to 1. These two equations,
F

@ A8

IC

F E

 

9 7   6

5 #2 ' $"     (& %#!    4 2 !   ) 1310

(12.11)

(12.12)

. (12.13)

(12.14)

(12.15)

where I C and I E are the equivalent emitter and collector currents for the transistor in the inverted mode. The principle of operation is the same. But if we compare reverse current gain R with forward current gain, we would find that it is weaker by about two orders of magnitude. Typical values are:

F

100 whereas

0.5

The inverted mode of the BJT also is weak in another respect. The BC junction, due to its relatively light doping, is able withstand a reverse bias of approximately 25 volts. However, the BE junction, since it is more heavilydoped, can withstand a reverse bias of only about 7 volts. Therefore a BJT biased in an inverted mode, in which the BE junction is reversebiased, is in serious risk of junction breakdown if voltage rails of more than about 7 volts are used.

12.2 EMITTER EFFICIENCY AND EMITTER DEFECT Since the emitterbase junction is usually placed in forward bias, we must pay close attention to the features of this junction. For any junction in forward bias, the ntype carriers injected into the ptype region will be of magnitude

np

and the current due to these carriers will be diffusion current

Jn

qD n

Figure 12.21 represents the BE junction in forward bias. The forward emitter current will be ntype carriers injected into the ptype base, of current density

J nE

qD nB

Figure 12.21: Excess minority carrier levels injected into the base region by BE junction forward bias. An npn transistor is represented.

122

dn p dx

qD nB n (e VBE W B pB

dn p dx

  

" #!

n p0(e V

Vt

1)

qD n n (e V L n p0

VT

1)

VT

1)

(12.21)

1
R

(12.16)

 

% 

 

Note that we need to use a regiondependent syntax. D nB represents the diffusion coefficient for ntype carriers injected into the ptype base, and n pB represents the equilibrium minoritycarrier level of ntype carriers in the ptype base. J nE represents emitter current density of ntype carriers injected into the ptype base from the emitter region. In this case, for convenience, a linear approximation has been made for n(x), as represented by the plot of n p (x) vs depth within the base region. Assuming the linear approximation for n(x), the slope dn p dx is approximately n 1 W B , where (12.22) n1 n pB(e VBE VT 1) n pBe VBE VT where V BE V T is assumed. WB is the thickness of the base region, also called the basewidth. If we wish to be more particular, the slope dnp /dx is actually a hyperbolic function that is matched to the boundary condition n = n1 at the emitter end of the base region to n = n2 at the collector end of the channel. The carrier density n2 at the collector side is: n2 n pBe VBC VT and therefore is practically zero since, in the forward active mode, VBC is negative. These hyperbolic functions do not provide much in the way of added insight, so we will stick to equation (12.21).

LB BD nB , also called the diffusion length. B is the recombination lifetime of the ntype carriers in the p type base region. The recombination distance represents the characteristic distance for the level of remaining carriers after penetrating deep into the ptype territory. Note that equation (12.21) may also be written in terms of the base doping NB , as

where we will usually assume ni = 1.5

1010, for silicon at room temperature (T = 300K).

Current is also injected from base to emitter by the forward bias, and will be of form similar to equation (12.23)

The use of emitter width WE in (12.24) is only valid if W E L E , where L E ED pE = the recombination distance for the ptype carriers in the ntype emitter region. Otherwise we must either resort to hyperbolic functions or, if the emitter is of infinite thickness, we can use W E L E , so that

As far as the operation of the transistor is concerned, only that portion of the emitter current that is injected into the base, JnE, will be able to link to the collector. Therefore the emitter will have an efficiency of
E

Using equations (12.23) and (12.25), the ratio JpE / JnE , which is also called the emitter defect, will be:

so that the emitter efficiency is of the form

J pE J nE

D pE W B N B D nB L E N E

123

J nE

J pE

 

J nE JE

J pE

J nE

qD pE n 2 V i e BE LE NE

VT

1 J pE J nE

J pE

J nE

qD pE n 2 V i (e BE WE NE

qD nB n 2 V i e BE WB NB

We can avoid the hyperbolic functions as long as W B

L B , where LB is the recombination distance

VT

VT

1)

(12.23)

(12.24)

(12.25)

(12.26)

Note that we have dropped the subscripts relating to type of carriers, i.e. DpE DE and DnB DB . This choice is made, both to reduce clutter and to recognize that exactly same generic form will exist for both npn and pnp transistor types. Unless we like to challenge ourselves with excess notation, which is very easy to do in dealing with BJT transistors, we will ALWAYS assume that terms like DE, DB, E, and B represent diffusion coefficients and mobilities for the minority carriers, for the emitter and base regions, respectively, in this case. If we had used the more exact hyperbolic functions, the emitter efficiency E would have been of the form[12.21] W D EN B L B tanh B LB D BN E L E
1

For WB /LB 1, tanh(x) x and equation (12.28) collapses to equation (12.27). E will be close to 100% efficiency if NE > > NB . This requirement is consistent with the construction shown by Figure 12.1.1. One of the reasons for bypassing the more exact hyperbolic analysis is that equations (12.27) and (12.28) assume that the base and emitter layers are uniformlydoped. This condition is usually NOT the case. Because of the fabrication process, both the emitter and the base dopings vary considerably NB = NB (x) and NE = NE (x). The emitter defect and efficiency should then be defined in terms of Gummel numbers GB and GE, where
WB

GB

N B(x) dx D B(x)

A similar equation may be identified for the emitter Gummel number, GE. The emitter defect will then be GB GE

and the emitter efficiency will be

1 GB GE

Note that, in like manner, we could define a defect factor for the collector,

For very high impurity doping of the emitter, which is fairly common for highcurrent devices, the bandgap is narrowed by a small but significant amount EG . This reduces the level of injected current, and therefore increases the emitter defect so that

Since EG may be on the order of .08eV at NE = 10 19#/ cm 3, the bandgap narrowing can increase the emitter defect by a factor on the order of 25 greater than that indicated by equation (12.26).

# " !

GB GC

Dc WB NB DB Lc N c

EG kT

124

 

 

 

 

DE WB NB DB LE NE

(12.27)

(12.28)

(12.29)

(12.210)

(12.211)

(12.212)

(12.213)

Due to the bandgap narrowing and other effects, higher power BJT devices usually do not have as high a value of F as do signalprocessing types of BJTs.

12.3 BASE TRANSPORT FACTOR AND BASE RECOMBINATION DEFECT The carriers which are injected across the emitter junction into the base region fall into the collector as result of the strong Efield in their favor at the collector junction. This strong field sweeps the excess chargecarriers into the collector region and reduces the excess carrier level to nearly zero at the collector boundary. The behavior is represented by Figure 12.31.

All of the carriers do not make it across the base region because they are passing through territory where they are subject to heavy attrition (due to recombination). For the linear approximation to n(x), represented by Figure 12.31, the carriers lost to recombination in traversing the base region are

Figure 12.31

Excess carrier levels in the base region of the npn transistor.

since the injected carrier density across the base region is approximately triangular. These carriers represent a recombination current density of 1q
B

WB

The fraction of carriers that succeed in making it across the base region are then

J nE

where, using equations (12.31), (12.2.1), and (12.2.2) the fraction of carriers lost to recombination is

125

 

J nR J nE

qn 1W B 2 B

WB qD Bn 1

J nC J nE

J nE

J nR

J nR

n(x)dx

qn 1W B 2 B

J nR J nE

W2 B 2 BD B

WB

n(x)dx

1n W 2 1 B

(12.31)

(12.32)

where, as before, we have dropped the subscript nB and replaced it by B, to make the formula generic and less cluttered. The small factor B is called the base recombination defect (or base defect). The fraction of carriers that survive the transit of the base region, also known as the basetransport factor, is given by

Since B DB = L 2 , where LB is the recombination length in the base, then the basetransport factor can also be B expressed as W2 B (12.34) 1 1 T B 2L 2 B If we had developed the basetransport factor in terms of the hyperbolic functions we would have obtained sech WB LB

In this case, the hyperbolic form might be slightly simpler than the linear approximation. However, the canonical simplicity of equation (12.34) will, in general, be of more value to our being able to interpret the physical effects of our BJT device. The forward current ratio F = JC / JE , is then a matter of taking ratios of currents, for which

If we have uniformlydoped base and emitter regions, which is at least our first approximation, T and E are represented by equations (12.34) and (12.2.6), respectively. More sophisticated devices and device analyses might have need of equations (12.2.9) and (12.35).

********************************************************************************** EXAMPLE 12.31: Determine F and F for an npn transistor with (uniform) doping levels of NE = 5 10 17, NB = 5 10 16, and NC = 5 10 15. Assume base width WB = 0.1LB and recombination time constants E = B = 50ns. The mobilities within these regions are given by table E12.31. Table E12.31: Representative mobilities for an npn transistor mobility
5

NE
1017

NB
5 1016

Note that the ratio DB /DE = B /E, since the mobilty is proportional to the diffusion coefficient, i.e n = Dn /VT .
B E

and

Then

Also note that LB /LE

950 190

5 in this case, since LB

DB

, LE

126

 '!&$ # !" 

# !% ! 

) 0 ( (0

$   

JC JE

J nC J nE

J nE JE

J nC J nE

W2 B 2 BD B

(12.33)

(12.35)

 

T E

(12.36)

NC
1015

units
#/cm3 cm2/Vs cm2/Vs

DE E ,

The base transport factor T is

Even though it is not a necessary part of these calculations it should be noted that

*********************************************************************************

12.4 LINKING CURRENT AND THE EARLY EFFECT For the npn transistor which we have used so far to tell us the story, the ntype carriers injected into the base from the emitter, JnE are coupled directly to the collector current JnC by the diffusions and fields within the base region. Since this is a semiconductor layer, current includes both drift and diffusion terms,

nB

and
pB

Jp = 0 since none of the current across the base is due to Jp . It is all a great rampage of ntype carriers flowing across the base. But the equation for Jp defines a relationship between E, the electric field, and the gradient of p(x). We can apply this relationship to equation (12.41),

Now according to Einsteins law, for which D = VT,

nB

pB

so that

This equation is called the linkingcurrent relationship, since it defines Jn (x) anywhere within the channel. Solving this differential equation and evaluating at x = WB , for which Jn = JnC ,
WB

127

"

qD n e

2 nB i

V BC V T

VBE VT

J nC

pdx

qD nB[np(x

W B)

Jn

D nB dp q p n dx

D pB

D nB

p dn dx

D q pnB d (np) dx

np(x

nB

 

Jn

Jp

pE

Jn

nE

qD nB dn dx

qD pB

dp dx

D pB dp pBp dx

qD nB dn dx

0)] (12.44)

WB = 0.1LB = 0.1

VT

which gives

F = 0.9955

0.995 = 0.9906, and F = 105.

950

.02585

W2 B 2L 2 B

0.1 2 2

0.995

10

1 5

1 10

0.1

1 1.0045

0.9955

1.1 m.

(12.41)

(12.42)

(12.43)

Therefore the collector current IC is of the form IC AJ nC q 2AD nBn 2 V i [e BE Qb

e VBC

where A is the crosssectional area of the junction. As indicated by the sign, the current flows from collector to emitter, since it is due to ntype charge carriers (electrons flowing form emitter to collector). Equation (12.45) neglects recombination. It primary virtue is that it recognizes that JnC is inversely proportional to the somewhat malleable basecharge per area,
WB

WB

QB

pdx
0

N B(x)dx
0

NB = NB (x) need not necessarily be uniform. QB is voltage dependent, since it represents the majority carrier charge in the zone that is between two depletion zones, each of which is voltage dependent. This situation is represented by Figure 12.41. For the usual case, in which the transistor is in a forward active mode, variation of the collector junction VBC has a small but noticeable effect and the emitter junction does not change. We can evaluate increments by examining the derivative behavior of IC . Due to the charge dependence of the (reversebiased) BC junction, IC will change slightly with respect to VCB as:

We have let dVCB = dVCE since, in forward bias, VBE is approximately constant. This change is handy, since dIC / dVCE represents the small slope go , as shown by Figure 12.41, and defines the Early voltage VA :

dQB /dVCB is negative, since an increase in reverse bias VCB on the BC junction increases the reach of the junction depletion charge QBC into the base, thereby decreasing QB by the same extent. This effect is illustrated by Figure 12.41

Figure 12.41:

Basewidth modulation and the Early effect.

128

"

 ! 

9 2 75 2 !4380(64310)' & ' (& 9 ' G@FD) B2' A@7 7E C

dI C dV CE

dI C dV CB

I C dQ B Q B dV CE

g0

IC VA

VT

(12.45)

(12.46)

(12.47)

(12.48)

   $#"

In fact, since both QB and QBC are proportional to the density NB of the base doping, then dQB = dQBC . If we make this substitution, equation (12.2.7) can be rewritten as

For convenience, we have put the charge effects in terms of CJC , the depletion capacitance of the BC junction (capacitance due to the uncovered charge QBC ) since CJC = dQBC /dVCE . This equation tells us that the base charge QB forms a relationship between the Early voltage VA and CJC : VA

Q B C JC

If we had biased the transistor in the reverse mode, with VBC forwardbiased and VBE reverse biased, the analysis would have been completely the same, except in terms of the BE junction. Therefore an analogous relationship for the emitter junction exists of the form: VB

Q B C JE

where VB is the reverse Early voltage. VA and VB also may have aliases such as as VAF and VAR. The model used by SPICE, the GummelPoon model, makes use of the Early effect relationships with the depletion charges of the BE junction and BC junctions, for voltage modulation of the basecharge QB .

12.5 BASE TRANSIT TIME

The current linking the collector and emitter is due to the charges that were injected into the base, moving with velocity v, as given by IC AJ nC qAn pv (12.51) qAn p dx dt If we evaluate this differential equation, we find that it defines a basetransit time, F, and a charge QF of forward transport carriers, according to 1 IC

dt

qAn pdx
0

For the forward active mode, np = np (x) is approximately by the linear distribution n(x) = n1 ( 1 x/WB ). Applying this approximation to equation (12.52), 1 q J nC

n(x)dx
0

where we have used JnC = IC /A. If we neglect the recombination in the base region, assuming WB is small, then equation we can use equation (12.2.1) to define JnC , J nC J nE qD B n1 WB

129

WB

1 1 qn W J nC 2 1 B

dI C dV CE

I C dQ BC Q B dV CE

IC C Q B JC

(12.49)

(12.410)

(12.411)

Qf . Ic

(12.52)

for which

Does this look familiar? It should! In section 12.2.3 we showed that the ratio of recombination current JnR to injected current JnE was
B

Therefore, tracing this definition through to its application to the base transport factor T,, and using equation (12.53), we can write the base transport factor, in alternative form, as

12.6 APPROXIMATIONS Because the equations for T and E are relatively simple, we can make some informative approximations for evaluating one of the principal parameters that we need for circuit evaluation, F. In order to make these approximations it is handy to use equation (12.1.4), which was

Keep in mind that

depends on the relationships for T and E ,

where E is the emitter defect, defined by equation (12.2.6). This approximation tells us that F is proportional, among other factors, to NE /NB . More explicitly, if we take E to be the more general form given by equation (12.2.10), F = GE /GB , where GB and GE are the Gummel numbers. As an alternate approximation, we might assume
E

1, which makes

130

2L 2 B W2 B

Therefore, the forward current gain would be

1
B

1 W 2 2L 2 B B

W2 B 2L 2 B

  

D EN BW B D BN EL E

D EN BW B D BN EL E

W2 B 2L 2 B

As a first approximation, suppose T

1, then

D EN BW B D BN EL E , and

J nR J nE

W2 B 2 BD B

1 1 qn W J nC 2 1 B

W2 B 2D nB

(12.53)

F B

(12.54)

(12.61)

W2 B 2L 2 B

(12.62)

, so that (12.63)

where B is the base defect. In equation (12.63) we made the approximation 1/(1 x) (12.63) tells us that F is proportional to L 2 W 2 . B B We can combine the emitter and base defects to give the approximation

Note that we can use equation (12.5.4), for T, which tells us that, in approximation of high emitter efficiency,
F

B F

This approximation is particularly useful for analysis of chargecontrol effects and timedomain response. As an approximation, both emitter and base defects can be combined under equation (12.65) by incorporating the emitter defect into either B or F. ********************************************************************************** EXAMPLE 12.61: Using the values of example 12.3.1, determine the base defect B and the forward and reverse emitter defects, E and C = Erev, (for the emitter and collector junctions respectively). From these values find F and R using equation (12.64). Assume all recombination times, E , B , and C = 50 ns, and that the emitter and collector regions to be large relative to their recombination lengths. In example 12.3.1 the doping levels were: NE = 5

and the base width WB = 0.1 LB . The mobilities were given as: mobility
5

NE

NB
5

1017

1016

Since DE / DB = E / B and DC / DB = C / B , and since L B L E result of all of the recombination times being equal, then

and L B L C

From the table, the mobility ratios E /B = 190/950 = 1/5 and C /B = 450/950 = 9/19. The ratios of the doping levels are NB / NE = (5 NB / NC = (5 1016 ) / (5 1015 ) = 10 This gives emitter defect

and collector defect, which is the same as the reverse emitter defect.

Erev

131

DC DB

NB NC

LB LC

WB LB

9 19

10

0.1

0.4472

19 9

DE DB

NB NE

LB LE

WB LB

1 5

1 10

1016 ) / (5

1017 ) = 1/ 10 and

10

0.1

0.6882

' ' (

' ' (

& #%  "!

 $"  

# 

10 17, NB = 5

10 16, and NC = 5

10 15.

NC
1015

1 + x for x << 1. Equation

(12.64)

(12.65)

units
#/cm3 cm2/Vs cm2/Vs

 

as

The base defect is

which tell us that

for which Similarly

= 105.5. This result is almost the same as the result found under example 12.3.1.

for which R = 1.44. Note that almost all of the current defect for the inverted mode of the transistor is due to the collector defect. (= reverse emitter defect) *********************************************************************************

12.7 CHARGECONTROL ANALYSIS In section 12.5, we identified the base transit time F as a relationship between collector current and a forward carrier charge in the base, QF. This charge, which, for the npn transistor, is associated with injected ntype carriers, n(x), forms an approximately triangular distribution within the base layer, as shown by figure 12.71. This charge current relationship, for steadystate, is: IC QF
F

Figure 12.71

Carrier charge QF in the base under forwardactive bias.

The figure also shows that IB plays a role in control of the forward carrier charge. The change of QF with respect to time, dQF /dt, should increase as result of base current and decrease as result of the recombination of carriers. This can be represented by

dQ F dt

IB

QF
B

132

2 0 (  "  31) $' & %$# !

5 64

(0.6882

.005)

 

7 84

0.6932

(0.4472

0.5)

10

0.9472

 

0.5

W2 B 2L 2 B

0.1 2 2

10

10

(12.71)

Note that for steadystate, dQ F

0 and Q F

Note that this result is just like equation (12.6.4), where we were analyzing charge in terms of recombination using the basetransport factor. If the collector junction were forwardbiased and the emitter junction reversebiased, a triangular distribution of carrier charge Q R would result, of a reverse orientation. Just like the analysis above, we could identify, in a similar way.

where R the reverse transit time. SPICE uses R TR. This equation is a little deceptive, in that R is low primarily because of a poor junction efficiency, NOT because of transit times. But, assuming that we can equivalently embed the junction effects E and C into the transittimes, then equations (12.72) and (12.73) are perfectly alright. If both junctions are forward biased, these triangular distributions could be superimposed, as shown in figure 12.72, to form a trapezoidal distribution. This situation corresponds to the saturation mode. This distribution creates something of a problem for the carriers, since only the upper triangular part of the trapezoid has a slope which can define a collector current. This upper triangular distribution, which we have labeled as QF , is identical in behavior to the triangular distribution that we considered in figure 12.71. It therefore can be represented by time constant F , so that saturation current I CS Q F F.

Figure 12.72: When the BJT is in saturation, carrier charge QF and carrier charge QR overlap to form a trapezoidal distribution. The charge distribution below the upper triangular distribution, QF represents a charge which is not swept across the base, since there is no slope dn/dx that can be associated with it. This part of the distribution therefore represents a stored charge of carriers, QS , in the base. The only means by which this charge may be diminished is by recombination, or by the feed currents IB and IBS , as represented by figure 12.72(b). Following the model represented by this figure, the behavior of charge QS is define by a chargecontrol equation, similar to (12.71), of the form

133

# !

% " &$

" 

! 

 " &$

 '$

 

IC IB

IC IB

# !

Then
B F

(12.72)

B R

(12.73)

 

 

 

 

where IBS represents the base current needed to supply the saturation I CS F. Other than the relatively slow recombination process, QS can only be diminished by reversal of IB , such that it drains the charge out of the base. This process is indicated by the logicswitch example shown by figure 12.73.

Figure 12.73: Saturated logic: The switch switches IB from IBF, which is large, and sufficient to put the transistor in saturation, to IBR , which acts as a basedrain to remove the stored charge QS . This case is fairly important. For t < 0, QS is assumed to be in steadystate, for which dQS / dt = 0. Now, using IB (t < 0 ) = IBF, equation (12.74) gives an initial boundary condition on QS ,
%

Q S(t

0)

I BF

I BS

At t = 0, the switch is toggled, and IB > IBR . The differential equation that results, using (12.74), is then

dQ S dt

I BR

I BS)

Q S(t)
b

This equation is subject to the boundary condition on QS given by equation (12.75). Solution is straightforward, since equation (12.76) is just another one of those standard decaytime equations that we always encounter with circuits. Its solution is Q S(t) (I BR I BF)e t B (I BR I BS) (12.77)
B

for which we can get the storage time ts , tS 1n

The storage time is an important aspect of BJT logic circuits where the transistor is allowed to go into saturation. It represents a finite time delay on the order of several ns. Even though it is a wellrecognized problem, for which a number of good solutions have been applied, it is a circuit design aspect that must always be addressed for any of the

134

&

(I BR (I BR

I BF) I BS)

&

) 0(

'

&

(I BR

I BF)e

tS

(I BR

I BS)

This equation tells us that after some time, t then

Note that for t

0, QS

IBF IBS , which is the boundary condition (12.75). ts , the stored charge QS 0. If we let QS 0 in equation (12.77)

&

  

) 0( '

 

# ! $"

 

&

dQ S dt

(I B

I BS)

QS
B

(12.74)

(12.75)

(12.76)

(12.78)

BJT logic circuit families, including such mixed technologies as BICMOS.

12.8 THE SMALLSIGNAL PARAMETERS

The smallsignal model is readily developed from the chargecontrol analysis. It is necessary that the transistor be in the forward active mode, for which collector current IC and the forward carrier charge QF, are related according to QF (12.81) IC where, according to the definition of QF given by equation (12.5.2) and the definition for n1 given by (12.2.2),

QF

Q F0e VBE

where QF0 is a constant, = 1/ 2 qAnpB WB if we happen to have a uniformly doped base. The smallsignal change of IC , from equation (12.81), is then dI C

IC dV V T BE dV BE, this equation represents the smallsignal voltagetocurrent

where

Equation (12.82) also defines a capacitance due to the excess charge carriers injected into the base, of the form

CD

Since QF is a diffusion charge, CD is usually called the diffusion capacitance of the emitter junction. We have already shown, from equation (12.4.8) that

g0 where VA is the forward Early voltage.

IC VA

A basecollector diffusion capacitance CA can also be defined, as follows:

CA

dQ F dV CB

dQ F dV CE

Since smallsignals i C transducer (VCT) form

dI C and

BE

iC

gm

gm

IC VT

dQ F dV BE

d ( I ) dV BE F C

dI C dV CE

dQ F

Q FOe

V BE V T F

BE

135

VT

(12.82)

dV BE VT

(12.83)

dI C dV BE

F m

(12.84)

(12.85)

dI C dV CE

F 0

(12.86)

Since go is approximately 4 orders of magnitude smaller than gm , then CA is likewise very much less than CD , being on the order of fF. Junction depletion capacitances CJE and CJC are larger than CA by several orders of magnitude, typically on the order of pF. Other smallsignal terms relate to the base current, IB, = IC / F. In this respect there are two base conductances, g and g given by: g and g

g represents such a small conductance that it is usually omitted from the smallsignal model. Even the relatively small conductance go is often neglected, for simplicity. The complete smallsignal BJT model, also called the hybridpi model, is shown by figure 12.81.

Figure 12.81:

The hybridpi smallsignal model of the BJT

Both depletion capacitances and diffusion capacitances contribute to C and C . C junction, C C D C JE whereas C is associated with the BC junction,

Since we have welldefined capacitances and conductances, the hybridpi smallsignal model of figure 12.81 has a welldefined highfrequency corner. If we look at the case under which we get maximum output current, in which case the output is shorted, the hybridpi circuit has a pole at g C

Since the current gain is F, the gainbandwidth product is

136

CA

C JC

C JC

gm C



'

 

 

% &"

53 4

dI B dV CE

1 dI C F dV CE

 

$#"

)("

dI B dV BE

1 dI C F dV BE

gm
F

(12.87)

g0
F

(12.88)

is associated with the BE (12.89)

(12.810)

(12.811)

is usually called the unity gainbandwidth product, and is often a specification of the transistor upper fT T 2 frequency limit at some test current IC. Note that if we use equations (12.89) and (12.810), and neglect CA , then

or, we can determine F from gm , CJE , and CJC , as follows:

Our favorite circuit simulation software, SPICE, requires input values for F (TF), CJE (CJE), and CJC (CJC) to be included in the .MODEL statement in order to make a reasonable assessment of circuit frequency response.

representing a cutoff floor for the frequency response of a transistor. ********************************************************************************** EXAMPLE 12.81: A BJT has test specifications fT = 318 MHz at IC = 1 mA. The collector capacitance C = 1 pF, and CJE = 2 CJC . (a) Find the SPICE parameters and (b) find the small signal capacitance C and fT when IC = 0.1 mA. SOLUTION: (a) Assume that VT = .025 V. Then gm0 = transconductance under test conditions = 40 mA/V. Since CA << CJC , then CJC C = 1 pF, and CJE = 2 CJC = 2 pF. Also we find that
T

=2

318 MHz = 2 Gr/s

1pF)

Units were included in this case to indicate that our result should be always be in ns, since we are dealing with pF and Gr/s. The SPICE parameters are then: TF = 0.2NS, CJE = 2PF and CJC = 1 PF.

(b) For current IC = 0.1 mA, we will have gm = 40 we find that CD gM

0.1 = 4 mA/V. Then from equation (12.84), 0.425ns

4ma V

for which we will have C = 1.7 pF + 2 pF = 3.7 pF 4

corresponding to fT = 135 MHz. **********************************************************************************

137

(3.7

1)

gm

0.851Gr s

Then, using equation (12.813), we find 1 1 (2pF F 2Gr s 40mA V

0.425ns

1.7pF and

We have to be a little wary of this equation, because as the current I C 0, g m case we could end up with a nonsense value for F. In the limit, gm gm T C C JE C JC C

1 g m (C JE

 

 

gm

C JE

C JC

gm

(12.812)
F

C JC)

(12.813)

0, and C D

0,

in which

12.9 THE GUMMELPOON MODEL AND SPICE PARAMETERS In simulation of BJT circuits, the BJT functions as a currenttocurrent transducer (CCT). The emitter and base defects may be included with F. The parameter R is sufficient to define the BJT in the inverted mode. The complete flow of currents in the forwardactive mode of the BJT is shown by figure 12.91.
# ! "

Figure 12.91

Current flow in a onedimensional npn transistor, showing emitter and base defects.

At the level of analysis where hand calculations are the norm, we assume that IC / IB = constant = F. Over a wide range of currents, which is what we expect with real circuits, and which is the way in which we desire the circuit simulation software to work, this assumption is not particularly good. There are two currentdriven effects that degrade the ratio IC / IB , one at low levels of current, and one at highlevels of current. These effects are indicated by figure 12.92. Note that at lowlevels, the ratio IC / IB , which we also call F , rolls off, because IB has a smaller slope than IC . At highlevels, IC / IB rolls off because the slope on IC is smaller than that for IB . It appears that our favorite assumpconstant, is valid only within some median range. tion, I C I B
% $

Figure 12.92

Logarithmic plots of IC and IB vs VBE , showing lowlevel and highlevel knees.

But not to worry. We merely include these effects under the base current IB and the base charge QB that we identified under the linkingcurrent analysis in section 12.4. This approach is called the GummelPoon model [12.91],

138

"87 4 )1 ' 5 32(&    4 )1 ' 5 62(&  7 B A9 @ E C 2D)    ) ' 0(&

developed at Bell labs in the early 1970s. Other than the efficiencies and C , we have seen that the operating characteristics of the BJT depend almost exclusively on the electrical and physical profile of the base region and on the injection currents into and across this region. The linkingcurrent analysis, developed in section 12.4, tie these together in the form IC where I S [ e VBE

QB

IC

where IS is saturation current for the base region. The linkingcurrent analysis of section 12.4 shows that IS is the exactly the same for either the forward or the reverse direction. Lowlevel currents are dominated by recombination in the junction spacecharge regions. In the forwardactive mode this current will be primarily that associated with the emitter junction. In the more comprehensive model we also include the collector junction. Recombination current in the SCR does not link across the base region, and therefore becomes a contribution to the base current IB , for each of the two junctions. These two exponential terms must have emission factors nE and nC , respectively, which we expect to be of value approximately 2, consistent with recombination in the SCR.. The more comprehensive form for IB , with terms including recombination in each SCR, is then IB

where nE and nC are the emission factors for recombination in the emitter and collector junctions, respectively. ISE and ISC are the saturation coefficients for these terms. All of these coefficients are SPICE parameters, as identified by table 12.91. Highlevel current effects and the Early effect are associated with basecharge QB . QB represents the components of base charge which are not part of the driftdiffusion process for chargecontrol analysis between emitter and collector junctions. It includes (1) the depletion charge for the two junctions and (2) excess charges associated with highlevel current effects. With these factors included, QB then takes the form

     

IS

e VBE

VT

In the reverseactive mode, for which V BC

0, V BE IC

I SE e VBE

IS

e VBC

I S e VBE

0, I C it is of similar and opposite form I S e VBC

n EVT

VT

139

Note that, in the forward active mode, where V BE

0, V BC

and

VT

IS

e VBC

VT

1]

(12.91)

q 2AD B
WB

n2 i QB

(12.92)

p(x)dx
0

(12.93)

0, I C is of the form of an ideal diode current, i.e.

IF

IR

I SC e VBC

n CV T

QB0 is the zerobias value of QB , i.e. the value when VBE and VBC are = 0. QBE and QBC are the uncovered depletion charge regions of the BE and BC junctions, respectively, that encroach the base region. QKF and QKR are the excess charges in the base as result of highlevel injection. The highlevel excess charges QKF and QKR begin to become significant at levels of current defined by IKF and IKR . These are the knee current levels, for forward and reverse modes of operation, respectively, at which the linking current IC begins to rolloff. It is also convenient to identify the zerobias value of IS as IS0 . With this simplification equation (12.93) is of the form q 2AD Bn 2 i QB

where

qB

QB Q BO

qB is the normalized form of QB . Inasmuch as it contains the the complete electrical behavior of the contributions to the base charge, it is selfsufficient as a relationship which will let us analyze the highlevel collector current effects as a function of bias:

q1

q2 qB

q1 represents the terms associated with only with the doping NB . Note that the uncovered (depletion) charges on the base side of the BE and BC junctions are associated only with NB . These terms should dominate at low current levels. The term q represents excess charges associated with the highlevel injection, which therefore involves IS . These are expected to dominate at high current levels. Equation (12.98) is quadratic in qB and can be written as

q2 B which has a (positive) solution of

q Bq 1

q2

Note that in the limits of lowlevel and highlevel currents, equation (12.910) gives:
  

qB qB


q1


if q 2

q2 4 1 q2 4 1

(low (high

In terms of the normalized use of qB , equation (12.91) is of the form

140

 

q 2 if q 2

qB

(q 1 2) 2

level injection) (12.911) level injection)

q1 2

q2

qB

VT

V BE VB

V BC VA

1 I SO V q B I KF e BE

IS

q 2D BAn 2 Q BO i Q BO Q B

I SO qB

I SO V e BC I KR

VT

Q BO 1

VT

V BE VB

QB

Q BO

Q BE

Q BC

Q KF

V BC VA

IS V e BE I KF

Q KR IS V e BC I KR
VT

(12.95) 1

(12.96)

(12.97)

(12.98)

(12.99)

(12.910)

We can see that equation (12.912), along with (12.911), gives a behavior for IC that is sufficient to describe a rolloff at highlevel injection. For example, suppose the BJT is in forwardactive mode and at a state of highlev

el current, for which q B

q2 .

Then equation (12.912) gives

This is the appropriate IC behavior at highlevel injection that we noted with figure 12.92. The basic GummelPoon model for the BJT consists of equations (12.95), (12.98), and (12.912). The CCT ratio IC / IB is a realistic representation of the device current gain response, illustrated by figure 12.93.

Figure 12.93 Plot of IC / IB vs VBE . The maximum value is F, which occurs at approximately VBE = 0.65V in this case. The GummelPoon model has a total of some 21 parameters associated with the chargecontrol analysis. These include a few higherorder parameters needed to model variations in F which we have not detailed. In addition to the chargecontrol parameters there are also 5 parameters for builtin resistances, and 4 parameters which detail specifics of the pn junction, such as grading coefficients and the builtin potentials. 6 parameters for a substrate junction JS are also included, for those situations in which the substrate junction inadvertently may be forward biased. In addition there are noise parameters and bandgap parameters, for a total parameter count of approximately 40. A complete parameter list for the GummelPoon model is given by table 12.91. Note that some of these parameters may not have the same keywords from one version of SPICE to another.

141

I KFI S e VBE

2V T

IC

IS


I S I KF(e

VBE V T

(e VBE

VT

1) 1) (12.913) 1

IC

VT

I SO V q B e BE

e VBC

VT

(12.912)

TABLE 12.91 GummelPoon Model Parameters

Symbol IS

Description Forward carrier saturation current Maximum forward current gain Maximum reverse current gain Forward emission coefficient Reverse emission coefficient BE junction saturation current BC junction saturation current Highlevel injection forward knee current Highlevel injection reverse knee current BE junction emission coefficient BC junction emission coefficient Forward Early voltage Reverse Early voltage Collector resistance Emitter resistance Base resistance Minimum highcurrent base resistance 50 pct drop point for RB forward transit time reverse transit time Coefficient for bias dependence of F Voltage coefficient for bias dependence of

Keyword Alt keywordsDefault IS BF BR NF NR ISE ISC IKF IKR NE NC VAF VAR RC RE RB RBM IRB TF TR XTF VTF ITF PTF CJE VJE MJE CJC VJC MJC CJS VJS MJS XCJC FC XTB XTI EG KF AF JS 1016 100 1 1 1 0 0 1.5 2 0 0 0 0 0 0 0 0 0 0.75 0.33 0 0.75 0.33 0 0.75 0 1 0.5 0 3 1.12 0 1

Units A

nF nR ISE ISC IKF IKR nE nC VA VB RC RE RB RBM IRB

JRB

A s s V A deg F V F V F V

XTF VTF ITF PTF CJE mE CJC mC CJS mS XCJC FC XTB XTI EG KF AF

Highcurrent coefficient for behavior of F Excess phase factor at t = F Zerobias BE junction depletion capacitance BE junction builtin potential BE junction grading coefficient Zerobias BC junction depletion capacitance BC junction builtin potential BC junction grading coefficient Zerobias substrate depletion capacitance Substrate junction builtin potential Substrate junction exponent factor Basecollector capacitance fraction Coefficient for depletion capacitance F and R temperature coefficient Saturation current temperature coefficient Bandgap energy Flicker noise coefficient Flicker noise exponent

PE ME PC MC

TB PT

eV

142

C2 C4 JBF JBR NLE NLC VA/VAR VB/VBR

A A A A

V V

12.13 ANALYSIS OF THE BJT: EXERCISES 121. Find emitter defect E , base recombination defect B , and F for the following BJT. Let E = 10 ns, = 20 ns, C = 20 ns, and assume that both the emitter and base are shallow, with WE = 1.0 m and WB = 1.2 m. Assume that the transistor is of type npn. mobility
5

NE

NB
1
 

NC
2
 

units
#/cm3 cm2/Vs cm2/Vs

1017
 

1016

1015

122. Evaluate the transistor of exercise 121, for E, B and F, if it is of type pnp. Why does this transistor have so much smaller F than the npn version? 123. Evaluate the collector defect C and the reversecurrent gain factor, R for both of the types (npn and pnp) identified by exercises 121 and 122. Note that you can use the base defect, B , as already determined. 124. The dopings and mobilities for a symmetricallydoped transistor are shown. The transistor has a shallow emitter, of depth WE = 2 m , and thickness WB = 2 m of the base region. The recombination times are all equal, E = B = C = 20 ns. (a) Assuming an npn transistor, determine E , B , and C. (b) Why is the collector defect smaller than the emitter defect? (c) Determine F and R and comment on their relative magnitudes. mobility

NE
1 1017 1


NB
1016 1
 

125. How is the (metallurgical) base width WB affected by the transistor biases? Assume the same parameters (dopings, dimensions, mobilities) as given by exercise 121 and determine the shrink of the basewidth WB as result of the encroachment of the junction depletion widths. Assume forward bias VBE = 0.6 V and a CE bias VCE = 4 V. Note that VCB = VCE VBE ( = reversebias on BC junction). (a) Determine the builtin potential BE and BC for each of the junctions and (b) the encroachment widths WBC = NC /(NC + NB )WBC , where WBC is the depletion width of the BC junction. A similar form for WBE results for the BE junction. (c) What reduced basewidth WB results? 126. Determine the base transit time F for exercise 121, using the basewidth and the diffusion coefficient for the base as given. The chargecontrol model assumes that F = B /F, and that R = B /R . Let F = 50, and determine . Note that it is nowhere near that of B used in exercise 121. In fact, SPICE uses it as an idealized charge control parameter. Assume R = 1.25 and determine R . The parameters F , R , F, and R are four of the parameters that SPICE needs in order to use the transistor in circuit simulation. 127. When carrying a current IC = 0.5 mA an npn transistor with F = 80, R = 0.5, VA = 100V, C = 1 pF has unitygain bandwidth of fT = 318 MHz. Assuming that the junctions have CJE 2CJC and that CJC >> CA , (a) determine SPICE parameters CJE , CJC , F = TF and R = TR. (b) determine the smallsignal capacitances C C and CA. (c) What values do C and fT have if the current is reduced to 0.1 mA?
! "    

143

  

  

   

NC
1017

units
#/cm3 cm2/Vs cm2/Vs

  

128. The transistor used in the circuit shown has SPICE parameters BF = 100, VA = 120, BR = 0.5, TF = 0.25 ns, CJE = 2CJC = 1 pF. (a) Determine the smallsignal conductances gm , g , go , g . (b) Determine the diffusion capacitances CD and CA (c) From (a) and (b) determne the smallsignal circuit capacitances C and C and cutoff frequency fT. (d) If the current (IC ) is reduced by a factor of 2.5, what is the circuit capacitance C and what is fT ?
 

Figure P12.8

Exercise for relating SPICE BJT parameters to the smallsignal parameters.

144

# !  $" 

    ' & ""%