Beruflich Dokumente
Kultur Dokumente
D.K. Sharma
EE Department
IIT Bombay
1 Introduction
Modern IC technology has made it possible to put millions of transistors on a
single integrated circuit. This capability has been used to great effect in digital
circuits such as microprocessors, memories and various Application Specific Inte-
grated Circuits (ASICs). Indeed, such is the complexity of integrated circuits being
designed today, that one often talks of putting complete systems on a chip, rather
than designing a few function blocks. However full systems cannot, in general, be
exclusively digital. Analog function blocks are required at least at the input and
the output stage in order to interface with the world outside the chips. The analog
blocks at the input amplify signals and convert them to digital form. Once these
signals have been converted to digital form, complex digital functions can operate
on these and produce the desired results in digital form. These results then need
to be converted back to the analog form, in order to provide outputs which can be
used directly by the external world.
It is clear, therefore, that analog design has to complement digital IC design
in order to take full advantage of the power made available by semiconductor
technology. However, the skill required in designing analog circuits is much higher
than that needed for digital circuits. The requirements of stability, low noise
and low thermal drift are much more exacting in the case of analog circuits as
compared to digital circuits. Most digital circuits are inherently economic in power
consumption, because a device is either ‘ON’ (zero voltage) or ‘OFF’ (zero current).
This is not so in the case of analog circuits. Therefore, special care needs to be
taken in order to make an analog circuit economical in its use of power.
Analog design is, therefore, a challenging task. Unfortunately, it has not re-
ceived the same attention as digital design over the past few years. This is because
digital products such as microprocessors and memories have driven the Very Large
Scale Integrated Circuits (VLSI) markets up to now. As mixed signal products
such as communication chips come to center stage in the electronic market, the
need for analog and mixed signal design will be acutely felt. This course addresses
this need.
In this particular module, we shall see how simple building blocks used in
analog circuits are analyzed and designed.
1
2 Single Transistor Amplifier
2.1 DC gain
Consider the single transistor amplifier with
a current source load as shown in figure 1. A
I
d small signal input (vi ) is applied to its gate
v and the output is taken from the drain. The
o drain current of the n channel MOSFET is de-
v V
d pendent on its gate voltage (Vg ) as well as on
i Vg
its drain voltage (Vd ). Therefore, the change
in the drain current can be written as
∂Id ∂Id
Figure 1: A Single MOS transis- dId = dVg + dVd (1)
∂Vg ∂Vd
tor Amplifier
∂Id
By definition, ∂V g
= the transconductance (gm )
∂Id
of the transistor and ∂V d
= go , its output con-
ductance. Also, following small signal conventions, we can write differentials of Vg ,
Vd and Id as small case vg , vd , and id . We also notice that vg is the same as the
input signal vi and vd is the output of the amplifier, vo . Also, since the transistor
is fed by a current source, it does not permit any variation in the drain current.
Therefore, dId = 0. So,
0 = g m vi + g o vo (2)
Hence, we can write for the voltage gain (Ao ) of the stage,
vo gm
Ao = =− = −gm ro (3)
vi go
Where ro , the output resistance of the stage is defined to be 1/go .
gm and go depend on the transistor characteristics. Assuming the transistor to
be in saturation, we can use an approximate model and write:
K
Id ' (Vg − VTo − αVs )2 (4)
2α
where, K is the conductivity factor given by:
W W
K = K0 ≡ µCox (5)
L L
VTo is the threshold voltage without substrate bias, and α is a parameter which
accounts for the effect of substrate bias Vs . α is weakly dependent on Vd and its
value is close to 1. W and L are transistor width and length respectively. Vs is the
substrate bias. We define
Cgd
G vo
D
vi Cg gm vi ro Co
S S
Figure 2: AC equivalent Circuit
4
This describes the frequency response of a system with one dominant pole. The
bandwidth is given by 1/ro Ctot . The Gain Bandwidth product is given by
1 gm
GBW = gm ro · = (22)
ro Ctot Ctot
We see that the gain bandwidth product (or the cutoff frequency) is independent
of ro .
Ao
A - 3db
o Gain (db)
0 db
BW GBW
Frequency
The maximum possible cutoff frequency occurs for an unloaded amplifier, where
the capacitive load is entirely due to the device capacitance itself. we can take the
output node capacitance to be proportional to the device width. Then,
Ctot = χW (23)
K 0 VGT
GBWmax = (25)
αL
s
1 2K 0 Id
= (26)
χ αL
2Id
= (27)
χW VGT
The relationships derived above have been summarized in the table below.
Using expressions for DC gain and GBW max , we can express their product as:
s s
2L K 0 VGT 1 2K 0 W L 1 2K 0 Id
Ao · GBWmax = 0 · = 0 ·
λ VGT αχL λ αId χ αW L
5
Free Design Variables:
Parameters W, L, VGT q
W, L, Id L, VGT , Id
0
K W VGT 2K W Id
0 2Id
gm αL αL VGT
2
λ0 K 0 W VGT λ0 I d λ0 I d
go 2L 2
q L L
2L 1 2K 0 W L 2L
Ao λ0 VGT qλ αId λ VGT
0 0
K 0 W VGT 2K 0 W Id 1 2Id
GBW αLCtot αL Ctot VGT Ctot
q
K 0 VGT 1 2K 0 Id K 0 VGT
GBW max αχL χ αW L αχL
I
d
V
d2 v out
V Vg2
ref M2
V
d1
Vg1
v in M1
to the gate of the lower transistor. The gate of the upper transistor is held at
6
a fixed voltage Vref , such that both transistors operate in saturation mode. The
lower transistor acts as a common source amplifier, whereas the upper transistor
works in the common gate configuration. We shall use subscripts 1 and 2 to
denote the lower and upper transistors respectively. To analyze the cascode stage
and its circuit performance, we treat the two series connected transistors as a
single compound transistor with the gate and source of M1 and drain of M2 acting
as the corresponding terminals of the equivalent transistor. We then evaluate
its equivalent gm and go values, so that we can use the expressions for circuit
performance derived in the previous section. Since the two transistors are in series,
their drain currents are equal. We define gmeq and goeq in a fashion analogous to
the single transistor case, so that
Thus,
∂Id
gmeq = ∂Vg1
with dVd2 = 0 (30)
∂Id
goeq = ∂Vd2
with dVg1 = 0 (31)
Making use of these, we can write for the small signal change in drain current
id
so vd1 = −
gm2 + go2
Notice that the source of the upper transistor is not at the substrate potential.
Therefore, as the drain voltage of the lower transistor changes, the drain current
of the upper transistor changes due to two distinct reasons:
1. The gate source voltage of the transistor changes. As the drain voltage of
the lower transistor goes up, the value of the gate-source voltage of the upper
transistor reduces, reducing its current.
2. As the source voltage of the upper transistor changes, its threshold voltage
changes. This changes its effective over-voltage, and hence the current. If
the drain voltage of the lower transistor rises, the threshold voltage of the
upper transistor becomes higher, reducing its current.
7
It is assumed that gm2 is defined so as to account for both of these. We can
eliminate vd1 from equation(34) to obtain
go1
id = gm1 vg1 − id
gm2 + go2
and therefore, !
go1
id 1+ = gm1 vg1
gm2 + go2
So,
id gm2 + go2
gmeq = = gm1 (36)
vg1 go1 + go2 + gm2
Generally gm values are much higher than go values. In this case, we can see that
gmeq ' gm1 . The slight reduction in gm value from gm1 is due to feedback through
the drain voltage variation of the lower transistor.
8
3.3 DC Gain
Having evaluated gmeq and goeq , we can write for the DC gain of the stage:
This simplifies to
gm1 (gm2 + go2 )
Ao = −
g01 g02
We can write this as:
gm1 gm2
Ao = − · (1 + ) (43)
g01 g02
We can define the individual DC gains of the two transistors (in common source
and common gate configurations respectively) as
gm1 gm2
A01 ≡ − and A02 ≡ 1 +
g01 g02
Then,
Ao = −A01 · A02
Thus the DC gain is equal to the product of the DC gain of the two transistors.
3.4 AC behaviour
Since gmeq for the cascode stage is about the same as that for a single transistor
stage, the Gain Bandwidth product also remains unchanged. Because of a higher
output resistance, the bandwidth is reduced and the DC gain is higher for a cascode
stage.
The figure below gives the AC equivalent circuit of the cascode amplifier. As
ro2
Cdg1
G vx vo
vi ro1 gm2 vx Co
Cg1 gm1 vi
S
Figure 5: AC equivalent Circuit of the Cascode Amplifier
we shall see presently, vx is quite small. This permits us to ignore the effect of the
drain capacitance of the lower transistor and the gate capacitance of the upper
one. If we wish, we can always replace ro1 by a parallel combination of ro1 and the
9
capacitance seen at the drain of the lower transistor.
Because the DC gain is higher while the gain-bandwidth product remains un-
changed, it is possible to design amplifiers which go beyond the technological limit
discussed earlier (eq. 28). (The technological limit placed a bound on the product
of DC gain and gain-bandwidth product for a single transistor amplifier). However,
a limitation still exists on the achievable DC gain and gain-bandwidth product from
the cascode configuration. Assuming the two transistors to be identical, the DC
gain is the square of the one transistor DC gain, while the gain-bandwidth product
is unchanged. Thus we can see that for the cascode configuration, the product of
the square root of the DC gain and the gain-bandwidth product is bounded by the
technological constraint.
10
therefore, that the load also should be a current source made from a cascode pair.
This implies that the stage will use four transistors in series as shown in figure
6 below . (This is often called the telescopic cascode). This, however, limits the
Vbiasp1
Vbiasp2
vout
Vbiasn
vin
range of output voltages for which all transistors will operate in their saturation
regions, as required for proper circuit operation.
M3
Vbiasp1
M2 Vbiasp2
vin M1
vout
Load
the n channel M1 and the p channel M2 form the cascode pair. M3 provides the
necessary bias current. The output resistances of M1 and M3 are in parallel. How-
ever, the node where the drains of M1 and M3 are connected has its voltage swing
reduced by the cascode action provided by M2. Therefore, we can show that the
output resistance in this configuration is:
rout = (1 + gm2 ro2 )(ro1 ||ro3 ) + ro2
11
This is lower than the output resistance of the telescopic cascode stage, because
of the paralleling of ro1 and ro3 . However, it is much higher than the single
transistor output resistance. Fig.7 does not show a load connected to the stage.
Clearly, the load used in this case should be cascoded as well, otherwise, its low
output resistance will come in parallel with the gain-multiplied output resistance
of the stage. The practical folded cascode circuit is shown in the figure below:
M3
Vbiasp1
M2 Vbiasp2
vout
Vbiasn2
M1
vin
Vbiasn1
I
d
V
d2 v out
V +
ref M2
-
Vg1 Vd1
v in M1
12
higher the DC gain of the cascode stage. This basic idea behind the cascode am-
plifier can be carried a step further. The gain of the output resistance amplifier
element can itself be increased by an ancillary amplifier. This is illustrated in
fig.9 The ancillary amplifier could be a cascode stage and we can boost its gain by
the same technique. We could extend this gain boosting to any number of boost
stages. The maximum obtainable DC gain is limited by thermal drift and noise
etc. In practice, the boost amplifier itself is rarely gain boosted further.
Vdd
vo 1 vo 2
vi 1 vi 2
One (not very good) way of implementing a difference amplifier is to use two
single ended amplifiers as shown in figure10. The difference of the two outputs
would be the output. Analogous to the definition of the common mode and dif-
ferential input voltage, we can define the common mode and differential output
voltages as:
vod ≡ vo1 − vo2 (51)
vo1 + vo2
vocm ≡ (52)
2
The common mode and differential gains can be defined as:
vod
Adif f ≡ (53)
vid
vocm
Acm ≡ (54)
vicm
13
For a good difference amplifier, the differential gain should be high and independent
of input common mode voltage, whereas the common mode gain should be as low
as possible. The common mode rejection ratio is defined as
Adif f
CM RR ≡ 20 log dB (55)
Acm
In the simple suggested circuit, transistor currents, and hence the differential gain,
will depend on the common mode voltage. This is not a desirable feature as we
would like a good difference amplifier to ignore the common mode voltage and to
amplify just the difference signal. A better difference amplifier can be implemented
if we add a current source to keep the total current constant. If the common mode
Vdd
vo 1 vo 2
vi 1 vi 2
Vs
Is
voltage appearing at the two inputs changes, it will only change the voltage at the
node where the two sources join (Vs ). However, the current remains unchanged
due to the current source - and therefore, the differential gain is unaffected by the
common mode voltage. This results in a high common mode rejection ratio.
But due to the current mirror action, I(M p2) = I(M p1). Since Mp1 and Mn1 are
in series, I(M p1) = I(M n1). Therefore,
14
Vdd
Mp1 Mp2
i out
vi 1 vi 2
Mn1 Mn2
Vs
Is
and
gm
GBW =
CL
where CL includes Cdg and Cd for Mn2 and Mp2, as well as the load capacitance.
15
Vdd
Mp3
Mp1 Mp2
i out
vout
vi 1 vi 2
Mn1 Mn2
Vs
Mn3
Vbias Mn4
R1 C1 R2 C2
16
C
A1 A2
Because the drain and gate of Mp1 are shorted, it is always in saturation. Mp2
has its source and gate at the same potentials as those of Mp1, and also has the
same drain current. Thus, its drain-source voltage must be equal to that of Mp1.
Therefore it is also always in saturation.
Mp3 has the same gate and drain potentials as Mp1 and Mp2. If we ensure
that it has the same drain potential as Mp1 and Mp2, it will also be guaranteed
17
to be in saturation. This can be ensured if
I(M p3) W/L(M p3)
= (60)
I(M p1) W/L(M p1)
R1 C1 R2 C2
gm11 v1 gm22 v2
The voltage gain of the first stage is given by gm (M n2) × (ro (M n2)||ro (M p2)).
The voltage gain of the second stage is gm (M p3)×(ro (M n3)||ro (M p3)). Therefore,
the overall voltage gain is:
• We select a ratio of the bias currents of stage 2 and stage 1. Stage 2 would
typically be operated at 5 times the tail current of the differential amplifier.
We should ensure that the slew rate limitation comes from the first stage
charging the Miller capacitor and not by the second stage charging the load
capacitor.
• The slew rate required from the op amp determines I(Mn4), the tail current.
Currents through Mn1, Mn2, Mp1 and MP2 are half this value. Current
through Mp3 is the assumed ratio (say 5) times the current through Mn4.
At this stage, We know the current through all the transistors.
18
• Since currents through Mn2,Mp2, Mp3 and Mn3 are known, their output
conductances can be calculated (go = Id /VA where VA is the Early voltage
= L/λ0 ). The overall DC gain is given by
gm (M n2)gm (M p3)
A=
(go (M n2)||go (M p2))(go (M p3)||go (M n3))
As gm for Mn2 and all go values are known, this determines the gm for MP3.
Once we know the gm as well as the current for Mp3, we can calculate its
geometry.
• Once the geometry of Mp3 is known, geometries of Mp1 and Mp2 are deter-
mined by eq 60.
• The geometries of Mn4 and Mn3 are determined from current mirror ratios
in relation to a given reference transistor for bias generation.
6 Example Design
Assume a technology where K’ for n channel transistors is 120 µA/V 2 and for p
channel transistors is 60 µA/V2
The threshold voltages for n and p channel transistors are +0.4 and -0.4V respec-
tively. We shall use channel lengths which are about 4 times the minimum channel
length. For this choice of L, assume that the Early Voltage VA = 20V for both p
and n channel transistors
2. We shall bias the second stage at 5 times the tail current of the differential
stage.
20
3. From the slew rate, I(Mn4) = 2 × 10−12 × 10−6
= 40µA
gm (Mn2)
2π × 50 × 106 =
2 × 10−12
19
6. To get a gm of 628 µ0 with a current of 20µA,
q
628 × 10−6 = 2 × 120 × 10−6 × (W/L) × 20 × 10−6
9. Since the geometry of Mp1 and Mp2 has to be in the current ratio with Mp3,
W/L of Mp1 and Mp2 should be ≈ 1.7.
20