Engineering 1620: High Frequency Effects in BJT Circuits – an Introduction

Especially for the Friday before Spring Break

I have prepared these notes because on the day before a major vacation break some peo-
ple find it necessary to leave early for travel connections. As this material is only partial-
ly covered in Sedra and Smith, I felt it worthwhile to summarize it for you. I hope it will
also help you understand the calculations for Lab 5, a cascode amplifier for video signals.

Over earlier classes in the week, I developed a number of formulae first for the capaci-
tances in the junctions of a transistor and then for their effect on the gain, input imped-
ance and output impedance of a generalized common emitter amplifier. Let me begin by
recounting the capacitances of a transistor. (I gave you a handout with the full derivation
of these results and have added that as an Appendix to this set of notes.)

The small-signal model in its simplest form has two capacitances: the base-emitter ca-
pacitance is commonly called C while the collector-base capacitance is variously COB
or C . (The first nomenclature is common on data sheets while the latter is more com-
mon in papers. The CAD/SPICE usage is “cmu.” There is a slight but difficult to deter-
mine difference between the two when there is parasitic series resistance in the base. We
do not observe this distinction.)
COB
b COB is the capacitance of a reverse biased
c
junction and often has the dependence on
voltage of an abrupt-junction. For at least one
g m vbe value of collector-base voltage, it is straight-
r
C forwardly shown on any datasheet. The
transconductance and the dynamic resistance
of the base-emitter junction are needed first
e before one can tease the value of C out of
the datasheet.
qI
The Q-point determines the remaining model parameters in the usual way: g m  C and
kT

r  1  h fe  re 
1  h fe  kT where I am using h interchangeably with  as the DC
fe
qI E
current gain of the transistor. In class we showed that because of the base-emitter capaci-
h fe tance, the current gain becomes frequency dependent.
For a grounded-collector circuit, the gain becomes
i h fe
  s  c  . This is a single pole
ib 1  sr  C  C 
in dB
function that looks like this sketch when  is
graphed on a log-log Bode plot. Here we have de-
fined two frequencies: f  at which the gain is down –
0dB
f fT
3dB from its DC value and fT at which the magnitude of the gain is unity. If h fe is large,
fT
then f   . The datasheet gives fT as a function of the quiescent current because
1  h fe
that frequency is not sensitive to the DC current gain. (It also helps in advertising the
properties of a transistor to advertise the larger number!) The dependence of fT on cur-
rent comes from r (or equivalently re ) being inversely proportional to current. Part of
the C capacitance is due to the depletion layer and is roughly independent of current.
The other part of C is due to charge in transit from emitter to collector and when that is
dominant, reC   , the transit time. The typical curve for a 2N2222A device is shown
below next to a generalized common emitter circuit.
VCC

ZC

Q1

vout
vbg
ZE

ztr

The capacitance from collector to base complicates the simple calculation of gain and
input impedance. To simplify the problem, we take advantage of the Miller effect (named
after John W. Miller who published it in 1920). Miller’s theorem points out the equiva-
lence for input impedance, output impedance and gain of the two block diagrams below
as long as the gain of the amplifier is known with the feedback capacitor in place. (The
proof of the theorem simply equates the current in the feedback capacitor to the two cur-
rents through the capacitors of the second configuration.) Because the feedback capacitor
connects between output and input and the output voltage is often bigger than the input,
the current in the capacitor is generally greater than it
CF would be if the capacitor were across the input to
ground. The theorem points out that this is equiva-
-A lent to a larger, possibly frequency-dependent ca-
pacitor across the input and a marginally bigger one
across the output. What often makes this theorem
useful is that the low frequency gain is known and
that gain is constant enough to use over most of the
useful frequency range of the amplifier.
1  1/ A CF -
-A We also developed a set of formulas for the input
impedance and gain of the generalized common

1  A CF 2
emitter amplifier shown at the top right. The idea was to use Miller’s theorem to move
the collector base capacitance to two places: the (1+1/A)C component simply became
part of ZC and contributes to calculating the gain. The (1+A)COB element moved to the
left and ended in parallel with the input. This transform left only C to complicate life.
The input impedance with only that parasitic capacitance was:

ztr 
1  h fe   re  Z E   sr C Z E
1  sr C 1  sr C

The first term is our old result for input impedance at low frequency but now there is a
decrease in impedance with increasing frequency from an extra single pole at
1 f
f   T , a nice but not surprising result. The second term assures that even
2 r C 1  h fe
when the base capacitance shorts the base-emitter junction, there will still be ZE left as
part of ztr. This term is not important until roughly fT and one would not usually try to
use a device close to its maximum frequency limit. Notice that the first term has the form
rC
of a resistor R  1  h fe   re  Z E  ] in parallel with a capacitor of value C  e  .
 re  Z E 
The resistance value is the low-frequency result we derived a couple of weeks ago. The
capacitance is proportional to but generally smaller than C . We will use this result in an
analysis example.

 Z C 1
The gain formula has a similar form: G    where Z  is
 re  Z E  1  sre Z E C /  re  Z E  C

ZC in parallel with Miller’s second capacitor C 1  1/ A  . [I am playing fast and loose

with exact results here. Actually as A becomes frequency dependent one has to be care-
ful to include the effect of that change on the input capacitance. We will see this more
clearly in MOS circuits later.] The first factor is the low frequency gain and the second is
a new pole generally somewhat above fT . In other words, C does not have a big effect
on the Gain except through a decrease in input impedance that causes loading of the input
signal source.

Now let us look at an example of these effects. The circuit on the left below is one we
used as a low frequency example some time ago. It has a quiescent point around 6 mA, a
current gain about h fe = 120 typical, and therefore re = 4.2 ohms and r  500 ohms.
From the graph above, fT  250 MHz. The datasheet value of COB is 7 pf. From this,
1 1
C   COB   7 1012  146 pf.
2 f T re 6.28* 4.2* 2.5 10 8

3
 VCC

1K

VCC
VBIAS
Q2

21K 1K 21K

CIN
Q1 Q1
2N222A 2N2222A
CIN

25 Ω 25 Ω
The CBYPASS
mid
vin
2.7K ban vin 2.7K
16 μ d 16 μ
75 Ω gain 75 Ω
is
.99 1 103
 34
4.3  25
.
By employing Miller’s theorem, we can draw the small signal model as:

CIN
rπ = gm vbe
21 K 500 ohm Cπ =
146 pf

vin 2.7 K 1K
vout
Cμ (1+34)
= 245 pf RE Cμ (1+1/34)
25 ohm = 7.2 pf

RBB = 2.5 K ztr Z C

4
For Zin we have (neglecting CIN because it provides only a low-pass cutoff):

21 K 121*29.2
= 3.5 K
2.7 K 245 pf 146  4.3
 21 pf
25  4.3

ztr
RBB = 2.5 K

The resulting system is 1.42 K in parallel with 266 pf. The input impedance is dominated
by the Miller capacitance (245 pf of 266 pf total) and its magnitude will start to decrease
with single pole behavior at 421 KHz. By contrast the effect of C on the gain is to in-
troduce a high-frequency low-pass pole at about 23 MHz from the 1K resistor and 7 pf
capacitance in parallel. [If there is appreciable source loading at this frequency, it is even
possible that this second pole will be even higher in frequency but that is a later topic.]

An amplifier that has constant gain to some high frequency but has so low an input im-
pedance as to load the signal source well below its gain cutoff frequency is a poor design
because one cannot use the gain for the full range of the amplifier’s potential usefulness.
The circuit on the right above is called a cascode amplifier and it attempts to solve this
problem with a second transistor. The tandem arrangement of a common emitter stage,
Q1, with a common base stage, Q2, is called a cascode connection. (And no, this is not a
spelling error.) The voltage gain of the common emitter Q1 is very low, fractional in this
case, because ZC for that stage is the input impedance of the common base stage, Q2.
(That input impedance is zin 2  re 2 || C 2 . The capacitance of Q2 is not important at a few
 re 0.99  4.2
megahertz so the voltage gain of Q1 is G     .15 ) There is no
 re  Z E  4.2  25
longer a direct capacitance between input and output. The output load no longer affects
the input impedance. For that reason, cascode circuits are sometimes said to be ‘unilat-
eral.’ This time the input impedance is the same 1.42 K resistive part but the capacitor is
only 21+7 = 28 pf and the capacitive part becomes a factor in loading the input only
above 4.5 MHz. That is a full order of magnitude improvement in pole placement.

5
 Appendix: The Effect of Cπ on Input Impedance and Gain
of a BJT CE Circuit

c
b

ib
Cπ rπ gmvbe
vbe
vbg ZC vout
e
ie ZE veg

ztr

Basic Equations:
veg
KCL at the emitter terminal, e: 0  ib  g m vbe 
ZE
r ib
Ohm's law across base-emitter: vbe 
1  sC r
KVL from input across base-emitter and emitter to ground: vbg  veb  vbg

vbg
Definition of ztr : ztr 
ib
Connection between hybrid-pi transconductance model and current controlled h-
model: g m r   0 where  0 is the low-frequency current gain.

Steps to solve:

veg  g m r   1   0  sC r 
 ib  1    ib  
ZE  1  sC r   1  sC r 

 1   0  sC r  r ib
veg  Z E   ib and vbe 
 1  sC r  1  sC r

6
  r  Z E 1   0  sC r  
vbg     ib (Used: r  1   0  re )
 1  sC r 

Input impedance is:

ztr 
1  0  re  Z E   sC r Z E
1  sC r

ztr 
1  0  re  Z E   sC r Z E
1  sC r 1  sC r

Gain is the ratio:

v Z i Z g v  Z C g m r ib  Z C g m r ib  ZC 0
H  s   out  C C  C m be   
vbg vbg vbg vbg 1  sC r  vbg 1  sC r  ztr 1  sC r 

 ZC 0  ZC 0 1
H s   
1   0  re  Z E   sC r Z E 1  0  re  Z E  1   sC re Z E /  re  Z E 
Summary:

The input impedance is lowered by C beginning at the beta cutoff frequency. It is also
asymptotic to ZE for frequencies above fT. In this equation the first term is the low fre-
quency impedance with a new pole at the beta cutoff frequency. The second term takes
care of the behavior that makes ZE the impedance at very high frequency.

ztr 
1   0  re  Z E   sC r Z E
1  sC r 1  sC r

Notice that the first term has the form of a resistance in parallel with a capacitance. As
you did in the first lab, we can manipulate that into the form of a resistance with an
equivalent capacitance in parallel by multiplying and dividing the frequency in the de-
nominator by the resistance from the numerator.

ztr 
1   0  re  Z E  if C  
C re
and R  1   0  re  Z E 
1  sC R  re  Z E 

7
The gain is little affected except there is a new pole at some frequency above fT. The first
factor is the low frequency gain and the second is a new high frequency pole.

 ZC 0 1
H s  
1   0  re  Z E  1   sC re Z E /  re  Z E 