Widlar Some Circuit Design Technique PDF
Widlar Some Circuit Design Technique PDF
Abstract-This paper describes some novel circuit design tech- BIASING CIRCUITS
niques for linear integrated circuits. These techniques make use of
elements which can be constructed easily in monolithic form to avoid One of the most basic problems encountered in inte-
processing dilEculties or substantial yield losses in manufacturing. grated circuits is bias stabilization of a common emitter
Methods are shown which eliminate the need for the wide variety of amplifier. Conventional methods [l] usually require sub-
components and tight component tolerances usually required with stantial dc degeneration and a bypass capacitor to reduce
discrete designs, which substitute parts which can be made simply,
and which make use of special characteristics obtainable with mono-
th.e degeneration at the frequencies to be amplified. With
lithic construction. integrated circuits, the required bypass capacitors are
much too large to be practical. In the past, this problem
has been overcome using some sort of differential [2]
or emitter-coupled [3] amplifier connection. These solu-
HE COMPONENTS available in integrated circuit tions have been adequate in many instances, but they
processes which are in production today are limited suffer from a lack of versatility.
T
both in type and in range of values. This limitation The close matching of components and tight thermal
has been more troublesome with linear circuits than with coupling obtained i.n integrated circuits permit much
digital, since many more different kinds of parts are used more radical solutions. An example is given in Fig. l(a).
with conventional designs. Therefore, serious problems A current source ca,n be implemented by imposing the
have arisen in adapting discrete-component designs to ernitter-base voltage. of a diode-connected transistor op-
monolithic construction. erating at one collector current across the emitter-base
In many cases an attempt has been made to develop junction of a second transistor. If the two transistors are
new technology in order to integrate existing designs. This identical, their collector currents will be equal; hence, the
approach has met with varying degrees of success, but operating current of the current source can be deter-
has generally resulted in circuits which are difficult to mined from the resistor (R,) and the supply voltage (V).
produce in large volume. To realize a practical micro- Experiment has shown that this biasing scheme is stable
circuit with any certainty and within a reasonable period over a wide temperature range, giving collector current
of time, it becomes necessary to use existing production matches between the biasing and operating devices typ-
technology. Fortunately, with some specialized circuit ically better than five percent, even for power dissipa-
design techniques, it is possible in certain cases to achieve ti,ons in QZ above 100 mW.
performance with present technology which is equal to An extension of this idea is shown in Fig. l(b). A
or better than that obtainable with discrete-component transformer with a low resistance secondary can be in-
circuits. These techniques make it possible to avoid the serted between the biasing transistor (Q1) and the second
restrictions imposed by limited types of components, transistor (Q2). Then QZ is stably biased as an amplifier
poor tolerances, and restricted range of component values. without requiring any bypass elements, and the trans-
They make use of the inherent advantages of integrated former secondary is coupled to the amplifier without
circuits: close matching of active and passive devices disturbing the bias conditions.
over a wide temperature range, excellent thermal coupling A third and more subtle variation is given in Fig. l(c).
throughout the circuit, the economy of using a large If R, and R, as well as Q1 and Q2 are identical, the collector
number of active devices, the freedom of selection of currents of the two transistors will be equal, since their
active device geometries, and the availability of devices bases are driven from a common voltage point through
which have no exact discrete-element counterpart. equal resistances. The collector current of Q1 will be
Some of these techniques will be discussed; an attempt given by
will be made to indicate in a broad sense what can be I _ 1 - Vm
done rather than going into great detail on particulars. Cl - - 2+3, 0)
Rl ( 1>
Examples of practical circuits will be given to illustrate
where a single V,, and I, term is used since both tran-
important points.
sistors are identical. For VBE << V and I, << I,,
Manuscript received March 8, 1965:, revised July 19, 1965. I,, = I,, s $+. (2)
The author is with Fairchild Semiconductor, Mountain View, I
Calif. If Rz = $R,, then
586
1965 WIDLAR: LINEAR INTEGRATED CIRCUITS 587
Ein Ein
(a> (b) Cc)
Fig. 1. Biasing techniques applicable to integrated circuitr; which
take advantage of precise matching and close thermal coupling.
(a) Current source. (b) Transformer coupled amplifier. (c) RC-
coupled amplifier.
R3
10K
CONSTANT CURRENTSOURCE
E, g $+ (3)
The formation of current sources in the microampere
current range can be difficult with integrated circuits
which means that the amplifier will be biased at its because of the relatively large resistance values usually
optimum operating point, at one half the supply voltage, required. A circuit is shown in Fig. 3 which makes possible
independent of the supply voltage as well as temperature a current source with outputs in the tens of microamperes
and dependent only on how well the parts within the using resistances of only a few kilohms. It makes use of
integrated circuit match. the predictable difference of the emitter-base voltage
An evaluation of the effects of mismatch on bias of two transistors operating at different collector currents.
stability is given in the Appendix. Its operation can be described as follows: the collector
A simple amplifier using the biasing of Fig. l(c) is current of a transistor is given as a function of emitter-
illustrated in Fig. 2. An emitter degeneration resistor base voltage by
(RB) is employed in conjunction with R, to control gain
and raise input impedance without disturbing the balanced I, = I, exp. 9,
( >
biasing. A cascade connection of Qz with Q3 reduces
input capacitance while the emitter follower (QJ gives where Is is the so-called saturation current, q is the charge
a low output impedance. of an electron, k is Boltzmanns constant and T is the
It can be seen from the foregoing that the matching absolute temperature. This expression holds up to high
characteristics of a monolithic circuit have made possible currents where emitter contact and base spreading re-
biasing methods which are superior to those practically sistances become important and down to low currents
attainable with discrete designs. Excellent biasing sta- where collector leakage currents cause inaccuracy. It
bility is achieved over a wide temperature range without has been shown to be valid, within a percent or so, for
wasting any of the supply voltage across bias stabilization operation over at least six decades of collector current
networks. In addition, no bypass capacitors are required. with well-made silicon transistors [7]-[9]. This contrasts
The use of a circuit similar to that in Fig. l(c) as the with similar expressions for diode current and the emitter
second stage of differential input, single-ended output current of transistors which show substantial error over
amplifiers is described in [4]-[6]. three decades of current operation.
588 IEEE TRANSACTIONSON CIRCUIT THEORY DECEMBER
APPENDIX
and where
Ah
h
10 cfe2.h
(44 fc1