Professional Documents
Culture Documents
Duff 1973
Duff 1973
higher center frequencies. In fact, these results indicate [61 G. A. Rigby. and D. G. Lampard, “Integrated frequency-
selective amplifiers for radio frequencies,” IEEE J. Solid-
that reasonable performance can be obtained at center State Circuits, vol. SC-3, pp. 4174%, Dec. 1968.
frequencies of up to several megahertz. [71 D. 0. Pederson, Electronic Circutts (prelim. cd.). New
York: McGraw-Hill, 1965, ch. 10.
[81 B. A. Wooley, “The design optimization of integrated broad-
ACKNOWLEDGMENT band amplifiers,” Electron Res. Lab., Univ. California, Berk-
eley, Tech. Memo. ERL-M284, Sept. 1970.
The author wishes to thank Profs. D. O. Pederson [91 T. E. Idelman, F. S. Jenkins, W. J. McCalla, ,and D. O.
and R. G. Meyer of the University of California, Berk- Pederson, “SLIC-A simulator for linear integrated cir-
cuits,” IEEE J. Solid-State Circuits. vol. SC-6. ,..rm. 18S-203.
eley, for their help and guidance, and is grateful to Aug.’ l97l.
I. Getreu, B, Rosario, and W. Sansen for their many
helpful discussions. A special word of thanks is due to
both Mrs. D. McDaniel and B. Stafford of the Elec-
tronics Research Laboratory, Berkeley, for the help,
rendered in the IC fabrication, Finally, thanks are due William J. McCalla (S’67–M’67) was born in
to the Computer Center, University of California, Berk- Orange, Calif., on November 28, 1943. Here-
ceived the B. S., M. S., and Ph.D. degrees in
eley, for the generous support in terms of computer time. electmcal engineering from the University of
California, Berkeley, in 1967, 1968, and 1972,
REFERENCES respectively.
He has held appointments as a Research
[11 A. A. Gaash, “Synthesis of integrated selective amplifiers for
specified response and desensitivity,” Electron. Res, Lab., Assistant on the staff of the Electronics
Univ. California, Berkeley, Tech. Rep. 65-31, June 1965. Research Laboratory, University of Cali-
[21 A. A. Gaash, R. S. Pepper, and D. O. Pederso:, “Design of fornia, and as a Teaching Assistant within
integrable desensitized frequency selective amphtiers,” IEEE the Department of Electrical Engineering
J. Solio!XWe Circuits, vol. S01, pp. 2$-35, Sept. 1966. and Computer Sciences. He was employed by the Aeronutronics
[31 R. F. Adams, “Temperature sensitivity of integrated oscilla- Division of Philco-Ford, Irvine, Calif., during the summers of 1964,
tors,” Electron. Res. Lab., Univ. California, Berkeley, Tech. 1965, and 1966. During the summer of 1968, he was employed by
Memo. ERL-M247, May 1968. Motorola, Inc., Mesa, Ariz,, where he worked on computer-aided
[41 R. F. Adams and D. O. Pederson, “Temperature sensitivity simulation and optimization of integrated circuits. His principal
of frequency of integrated oscillators,” IEEE J. Solid-State
Circuits, vol. SC-3,pp.391-396, Dec. 1968. research interests have r~mained in this area. He is now a member
[51 W. J. Walsh, “Computer design of temperature desensitized of Technical Staff at Bell Laboratories, Holmdel, N. J.
integrated selective amplifiers,” Ph.D. dissertation, College Dr. McCalla is a member of Sigma Xi, Eta Kappa NLI, and Tau
of Engineering, univ. California, Berkeley, June 1968. Beta Pi
[(——
– )
simplifiedexpression is obtained, which is valid in the
d(8Q,)
neighborhood of the optimal value. k cos cot + h,’(aQ.)
dt
For operatingfrequenciessubstantiallybelow fT chal-ge
controltheory can be used to describethe time clepend-
ence of the transistorbase charge QT [2]:
+ h,’(8Qr)z + h,’(8Qr)3 = O .
1 (8)
W = ~
,=1
h,(3Q,)’ (4) [ (fundamental output power)’ 1 (lo)
M, = – 20 log,,
[
2
M2(f + o) ‘= – 20log,o
d-( g ~_~
2i,2h,’
)
(12)
-65-
2 135 mA
3.150mA
COMMON EMITTER
R. 2i12 RE=O
M,(j + m) = –20 log,, i-(
g>;.
)1
(13) IC=120, 135, !50 mA
v~:lov
% RL=75J2
~ .&) -
The bias dependent nonlinear Taylor coefficients de- f=2MHz
XN
finecl in (4) – (6) can be related to transistor parameters,
bias conditions, and loading conditions, The collector
hias current 1, is given as -55 -
therefore
L’
I d“
81
—Q qI.
. .— -45 1 I I 1 1
0 25 50 75 100 125 I o
dvb. nlcT Rg(OHMS)
az =
M,(f -+ ‘=) = –20 log,, IT ~ ‘—
2P0 aW~/aI,
4U,
“ (15)
.
2
1
‘w’~
kT O, t)n ~
[ cr31c
n
I. ax,
h, n+
I.’ 1 For this application, the bias point was located such
that d@/W. < 0. Inspection of (14) for the low-
. y$:$+~-y”
[ c c . c 1 frequency limit of .M2 shows that for the proper value of
Rg + RE, the denominator can go to zero since 8/3/dI,
The other coefficients can be computed in the same is negative at high current levels. This would indicate
manner.
considerably reduced distortion (a dip in the Mz char-
acteristics) when this condition is reached. This was
checked experimentally and theoretically with (11 ) and
the results shown in Fig. 2 which indicate a finite dip in
where /3 is the small signal transistor current gain, MZ versus Rg does occur. Thus, while affording some in-
sight, (14) is considered too idealized when the frequency
is not zero. ~
To study this further, (11) will be rewritten in a
_ aIc _ simpler form in (16) which enables more insight to be
i, — UT used.
aQ,
(16)
!
~ =R3=75Q
Vc-lov Ic-135n A
-65
-60 I
-.&_-
A
where
&i
~ -55(
NO,2095 .s:~:----- .--o-.<
ZN -=~=:=--%- -k- ---—-%
m-.% .-=_ ------ DATA
—__ A2757
-50 -B ,s. 1 TRANS-
%-4.- ISTORS
ICM MOOEL
-45~ I I ( I , ( ,, , $ c , , ,
2 3 4 567 S910 20 30 40
FREQUENCY (MHz)
M,ld,D = – 20 log,,
J-g
0
#-
/~
-90 -
g
/
/
+’ o /
/
/
/
n-/~0
/ /
/
I ./
/“
/
1
-85 ❑ DEVICE A
,~” ■ DEVICE S A2757
m
Equations (14) and (18) are very useful for de- ----
//0 A OEVICE NO 21013 TRANSISTORS
O DEVICE NO 20955
scribing second-order distortion. The dip in the Mz ● lCM’ MODEL
Ra + RE
-,o~
that causes the denominator of (14) to go
to zero. Equation (18) can then be used to evaluate
the magnitude of ~2 at its dip. To Mz in a minimize 2 3 4 56789IO 15 20 25 ?
circuit environment, the circuit or device is modified FREQUENCY (MHz]
in order to make the dip in Mz as large as possible and Fig. 4. JI, versus frequency for common emitter.
then Rg + RE is selected to operate as near the dip
as other circuit performance constraints allow.
An optimal collector bias current exists where dis- nonlinear coefficients, M2 and M3 are computed as
tortion becomes very small. From the denominator of function of frequency and compared with measured
(14) one can see that the term [ (1/1,) – l/n (dn/dIC) ] values as shown in Figs. 3 and 4. Fig. 5 gives the values
may become ~ero for the proper choice of IC, since dn/dI,. of P, n, f~, and their derivatives under various loading
>0. Hence, an optimal value of R. + RE of zero would conditions of interest, which were derived from the
be necessary for minimal distortion. From (18), dis- ICM model. This table gives some insight into how
tortion will be eliminated for Rg + RE = O,i.e., the base loading affects distortion through ( 18). For example,
is driven with a voltage source. For the transistor cir- fT/(afT/81c)increasesby a factor of 10 as R,. goes
cuit in this study, the condition described previously from 75 to 80. From (15) for the high frequency limit
was approached as 1, was increased. Data showing
of M2, one sees the @~Po term decreases by a factor
this decrease in the M2 dip as I, is increased are shown of ~. Hence, a high-frequency limit improvement
OUTWT RESISTANCE
W FIRST TRANSISTOR
RE=O, RL= 75 ,RL=75 RE=o,R~=8
RE=12 RE=12,
RL=8
~=lov ‘
A2757 +--1~=l35mA
B 76.78 76.16 80.42 79.74
Rg R
-277 -28! -218 -221 +
n 1.865 1.876 1.804 1.815 v
fT 5.96 X 108 5.33 x 108 1.719 XI09 1.2515 X109 & 047#F
X 109
~1,76 -1.62
X 109 -5.15XIOS -5.04x 108 RE 12$2
20nH
-0.339 -0.329 -3.34 -2.54
?
3.22 3.27 2.92 2.95 I J
1
LOW BETA
Fig. 7. Second transistor in Darlington stage,
I
RE’O, RL’33 ~=12, RL=20 RE=0,RL=20 RE=O,RL=B
I
E
- Sc
P 79.79 79.07 I 56.89 I 57.38 I
J
-227.65 -236.9 I -162.4 I -156,1 ]
n 1.815
n -75
5.37
f~ 1.28 XI09
J
& -878 X IO* -s.115 x 108 -8,767 X 108 -5.14 x 108
5oJa
-60
I
z
5on s
#
-55
-m
z~
R=on, p= 120
Fig. 6. Amplifier-simplified circuit.
-45
70 q
60;
R=Of2
MTA
-60 -
-55 -
..~
FREQUENCY[MHz]
12 5 10 15 20 30
of (14) should be minimized at 5 MHz. This can most FREQuENCY ( MHz)
easily be accomplished by the proper choice of Rg. (Note:
Fig. 10. Amplifier distortion for nonoptimal resistor values.
One would like to terminate the input of the transistor
in its optimal impedance at all frequencies.) Parameters
for the distortion equations have been calculated for The improvements in M2 resulting from an optimal
the case of an 8-0 load and zero emitter resistance in resistance of R = 80 Q are also illustrated in Fig. 9.
Fig. 5. With these parameters the denominator of (14) The addition of the optimal source resistance affects
vanishes when R = 55 tl. flhnce R~ = 5 Q for the circuit third-order distortion slightly. A maximum degrada-
in Fig. 7, a resistor R = 50 Q must be added in series tion in Ma of 3.8 dB occurs at 30 MHz.
with Rg.Equation (18) then predicts the magnitude It was observed that second-order distortion in the
of Mz at 5 MHz to be 69.2 dB. Hence, an improvement amplifier does not improve with a one-to-one cor-
of 30 dB is predicted at 5 MHz. This improvement is respondence with an improvement in the isolated Darl-
substantiated in Fig. 8 using (11) and the ICM model ington stage. It is presumed, therefore, that distortion
of an A2757 transistor. Fig. 8 also has measured data contributions from other sources in the total amplifier
for an actual Darlington pair with and without an are also important.
optimal resistor. Substantial improvement in Mz at 5 Fig. 10 shows how improvement in Mz for the am-
MHz is also evident. It should be noted that an optimal plifier changes when a nonoptimal source resistance is
resistor of R = 50 ~ is predicted for the model device used in the Darlington stage. M2 only varies 3 dB at
with a ~ of 80. The device used for the data has a 5 MHz for values of R, between 60 and 100 Q.
P = 120 and a different 6’P/dIc than the ICM model
device, hence the optimal resistor for this case was
CONCLUSIONS
measured to be R = 80 0.
Equation (18) indicates that the magnitude of M2 Second-order distortion is modeled accurately for the
(with optimal generator resistance) increases when /3 common-emitter configuration via the integral charge
decreases. This effect is demonstrated in the preceding
control model and the distortion equations for real loads.
example because the prediction for ~M2I with the lower Load dependent device parameters (i.e., n, /3, dn/dlc,
~ is greater than the data for the higher ~. When the d,B/MC) are computed from the ICM model for use in
/3 of a device is changed, a number of other device the distortion equations. Theory predicts and measure-
parameters may also change so the optimal resistance ment verifies that ilfz is very dependent upon Rq, Ra,
may be modified accordingly. However, if both the ~ and /3. It is shown that the addition of a single resistor
and optimal resistance decreases, (18) predicts an im- in the Darlington pair for the amplifier can reduce Mz
provement in &fz. with little effect on Ms. An optimal generator resistance
The distortion performance of the amplifier is shown is found that has a resonable value and provides maxi-
in Fig. 9 for the unmodified amplifier with R = O ~. mum improvement in M2.
IEEEJOURNAIJ
OF SOLID-STATE
CIRCUITS,
DECEMBER
1973 453
By adding the optimal resistance, the predicted and [51 S. M. Sze, Physics of Semiconductor Devices. New York:
Wiley-Interscience, p. 285, 1969.
experimental 30-dB improvements in A& at 5 MHz are [61 H. K. Gummel and H. C. Peon, “An integral charge-control
achieved. It is also shown that further improvement in model of bipolar t.raniistors, !) Bell Syst. Tech. J., ~0]. 49,
no. 5, p. 827, 1970,
Mz is possible when a, lower ~ transistor with the same
optimal resistance value is used. Collector bias current
may also be selected to reduce distortion.
The optimal value of resistance is shown to be non-
Donald G. Duff (S’69–M’70) was born in
critical, and amplifier performance remains much im- Plainfield, N. J., on May 24, 1948. He re-
proved even for a 20 percent variation from the optimal ceived the B ,S. degree in electrical engineer-
resistance. The worst M2 for the unimproved amplifier ing from the University of Arizona, Tucson,
in 1970, and the M.S. degree in electrical
is —52.5 dB at 5 MHz while the worst Mz for the im- engineering from Stanford University, Palo
proved amplifier is --65.5 dB at 30 MHz. Alto, Calif., in 1971.
From 1971 to 1973 he was employed with
Bell Laboratories in Holmdel, N. J,, where
ACKNOWLEDGMENT he worked on semiconductor device modeling
for computer aided circuit design, He is
The authors wish to thank T, L. Mader for his ex- presently at the University of California, Berkeley, studying towards
perimental contributicms, and J. J, Golembeski and D. L. the D, Eng. degree in electrical engineering.
Scharfetter for their encouragement and support.
REFERENCES
[11 S. Narayanan and H. C. Peon, “An analysis of distortion in Hin-Chiu Peon (M’72) was born in Hong
bipolar transistors using integral charge control model and Kong in 1939. He received the B.SC. and
volterra series,” IEEE Trans. Circuit Theory, vol. CT-20, M. SC. degrees in electrical engineering from
pp. 341-351, July 1973. Massachusetts Institute of Technology, Cam-
[21 H. C. Peon, “Modeling of bipolar transistors using integral bridge, in 1962, and the Ph.D. degree in
charge-control model with application to third-order distor- applied physics from Harvard University,
tion studies,” IEEE Trans. Electron Devices, vol. ED-19, Cambridge, Mass., in 1967.
pp. 719–731,June 1972. He joined Bell Laboratories, Murray Hill,
[31 L. I. Schiff, Quantum Mechanics. New York: McGraw-Hill,
1968, p. 152. N. J., in 1966 and has worked in the field
[41 Members of Technical Staff of Bell Laboratories, Transmis- of semiconductor physics and device analvsis.
sion Sustems for Communications. 4th cd.. Western Electric Dr. Peon is a member of Eta Kappa “Nu,
Co., Winston~Salem, N. C., 1970; p. 247.’ Tau Beta Pi, Sigma Xi, and the American Physical Society.