Lecture 29 : Module 02

IIT Kanpur July 201
20 followers
Oct 19, 2023
Welcome back, okay. So, now it is time to see what a real transistor differential amplifier
would look like. So, till now we have been dealing with this structure, right and this
was going, this current was going into some resistance and let us say I, for this purpose
of this discussion, I keep this resistance, the drain some arbitrary resistance Rd, right.
Then let us say this is V1, this is V2, this is Vx. So, this is the small signal equivalent
of the structure that we want. It is time to replace the small signal equivalent with
its transistor level equivalent. So, let us do that.
So, let us start with the gm on the left that is gm1. So, clearly it is a transistor. So,
we have to apply V1 here, but clearly we cannot apply simply V1, right. We have to apply V1
on top of some bias and we know several ways of applying the bias. So, I will not show
them explicitly. I will just assume we know how to apply the bias. Let us call this Vb
and this is V1, okay. So, this has to be Vx, this is known as Vx.
What do we need to do to the drain? In the circuit on the left, in the incremental model,
the drain is shorted, right. So, in this case, drain has to be higher voltage and shorted
which means that it has to be shorted to Vdd, okay. Now, let us switch focus, for the same
one let us switch focus to the gm on the right. So, this will be again another transistor
whose other input will be incrementally V2, but again it has to be some voltage Vb2. So,
let me call it Vb1, let me call that Vb2 and this is V2 and we have a resistance connected
at its drain and let us call this, I mean another side of the drain is connected to Vdd, okay.
And what should we do with the sources? As far as the two structures are concerned,
sources are shorted, so let us short them, okay. However, if we short them, what is going to
happen? Firstly, in a common source amplifier, the source is shorted to ground, right, but if
we ground each of them, right, if we short them and ground each of them, will this work? This
will not work because what is the reason? The reason is in the incremental model,
then this becomes a ground. If in the incremental model this becomes a ground, what is the, this
current will be gm1 times V1, this current will be gm2 times V2 and both of them will go to the
ground. We are not going to get the difference of V1 minus V2, right. So, why is the difference of
V1 minus V2 getting generated? That difference was getting generated because Vx was a function
of both V1 and V2, right. So, if I go back, yeah, so you see Vx was like gm1 Vi plus gm2 Vf by gm1
plus gm2, right. So, Vx in this case has to be gm1 V1 plus gm2 V2 by gm1 plus gm2. This is almost
like a weighted average of V1 and V2, weighted with the value of its own gm, right.
So, clearly Vx equal to ground is not possible, but note that more importantly, what is there
looking downwards from the source? It is an open circuit, right, it is an open circuit of resistance
infinite, ok. So, which means that looking downwards from the source, I need to see an
incremental open circuit and what contraption do you know that gives you an incremental open
circuit? You guessed it right, this is a current source, right. So, let us put a current source
and let us call this I naught, ok. So, this is m1, this is m2, this is the voltage Vx and this
becomes the transistor level schematic of our differential amplifier, ok.
So, now it is time to investigate what are these voltages, I mean what are the effects of this Vb1,
Vb2, sizes of m1, m2 and so on, correct. We note that we have already done, we have already done
the incremental analysis before because this transistor came from the incremental analysis,
correct. So, again just to reiterate, what will be the current, what will be the current through
this structure, incremental current? Assuming the, assuming some current flows, assuming some
quiescent current flows into m1 and m2, we do not know how much right now. Let us say the quiescent
current here is I1, quiescent current here is I2 and of course, I1 plus I2 has to be equal to I naught
and this sets up some gm on m1 and some gm on m2. So, what is the quiescent current through gm1,
not the quiescent current, I am sorry, what is the incremental current through gm1? This is gm1
times V1 minus Vx. What is the current flowing into m2? Note that I have reversed the direction.
So, the incremental current flowing into m2 will be gm2 Vx minus V2 and then we equate these two
and you will get what we got in the previous module, right. So, now time has come to delve
slightly deeper into this structure. So, let us first figure out,
let us first figure out what is going to happen if the transistors m1 and m2 are not identical.
So, if transistors m1 and m2 are not identical, then what can you comment on the quiescent
current I1 and quiescent current I2? So, clearly the quiescent current I1 is half
mu n C ox W over L of 1 VB1 minus Vx whole square, quiescent current of 2 will be
half mu n C ox W by L2 VB2 minus Vx square. Note that I have made an important distinction
here, important assumption here and that assumption is that both m1 and m2 are in saturation, right.
So, for m1 it is quite obvious, I mean as long as we can keep VB1 to be
one threshold voltage, I mean as long as we can keep VB1, difference between VB1 and VDD
to be less than one threshold voltage, then we are good. It is not particularly evident in case of
m2. However, let us say the transistor m2 had gone into linear, what can I do? I can reduce the
value of Rd to increase the quiescent voltage at the drain of m2 and that would be good, right.
So, we are now mature enough to know what to do to get the transistor back into saturation
if it has gone into linear. So, I will make the assumption here to start off
that our transistors are in saturation. So, I will make another critical assumption,
another assumption. Assumption is since these two, I mean this looks to be a symmetric circuit,
right. I mean as far as the operation of this circuit is concerned or at least the physical
nature of this circuit is concerned, it looks like there is nothing to distinguish
if you apply an input to m1 or you apply an input to m2, right. I can, the circuit will
essentially behave identically, just that the directions of the incremental current will change,
right. And that is why in general, even though in general you can choose any values,
different values of Vb1, Vb2, W by L of m1 and m2 also need not be the same.
But in most cases, you will see that we tend to use same values of Vb1 and Vb2
and same values of W by L for m1 and m2 just because it gives us that symmetry
in terms of I can choose which terminal to apply the input to without really thinking
what is going to, without really thinking of the effect of asymmetry, ok.
If we do not do that, if we do not do that, what we need to do? We need to ensure that
both, both the, I mean the sizing of m1 and m2 is such that and the sizing of and the voltages
of V1 and Vb1 and Vb2 are such that we have transistors biased in saturation, ok. So,
to make the point, let me ask you, let us assume that this, our transistors are biased
in saturation and Vb1 and Vb2 are different, W by L of m1 and m2 are different. Can you tell
me what is going to happen if I increase the volt? Let us say I do not have the incremental
inputs, let us say I do not have the incremental inputs and let us say I increase Vb1 from Vb1 to
plus delta V and Vb2 to also Vb2 to plus delta V. Now, tell me what is going to happen?
Will there be any, what will happen to the, what will happen to the quiescent currents
I1 and I2? Any comment on what is going to happen on the quiescent current of I1 and I2?
So, see the easy way to, easy way to, the most obvious way to think about this is to split this
current source I naught into 2 sub current sources I1 and I2, right. So, let us say I
split this current source into 2 sub current sources, one having current of I1, the other
having current of I2. So, if this were the case, if this were the case, can you comment on
whether I will be having any current, any quiescent current through this wire connecting
the 2 current sources? Note that, note that this choices of I1 and I2 are not arbitrary,
right. They have been chosen in such a way that, in such a way that the quiescent,
they will support the quiescent current of I1 and I2 through the transistors M1 and M2, right.
So, in other words, what is the current that is coming through this? This is I1. What is the
current that is going out? That is also I1. What is the current coming through this? This is I2.
What is the current going out? This is I2, which means that for all practical purposes,
I can say that this, this line does not exist, right, because there was no current
through, through this, through this wire, right. So, now tell me what is going to happen if I
increase the gate voltages of M1 and M2, both of them by same amount delta V.
So, what is constant? Constant is the current sources I1 and I2. So, they are not going to
change, right. If they are not going to change, then what do you think is going to happen?
If they are not going to change, then I can as well say that this current is not changing.
So, these currents will not change, this I1 and I2 will also not change, right. If they do not
change, I have applied an increment of delta V at the gate. However, the current I1 remains fixed,
then what is the consequence? What is the only thing that can happen? The only thing that can
happen is Vx of this terminal will also go up by delta V, this terminal will also go up by delta V,
right. So, regardless of whether these two terminals are connected or not,
right, the sources of both M1 and M2 will go up by delta V in order to keep, in order to keep your
current constant, ok. And I mean, note that we did this, I did this heuristic hand waving
analysis, we can also do it using root force method, right.
What we can simply do is, let us assume this delta V is a incremental input, right,
it is an incremental change, right. We can also do assuming that if delta V is small,
with in quotes small, right, what will be the change in I1 or let us say if delta V is small
and we start off with the assumption that the currents I1 and I2 will change, right.
If the currents I1 and I2, the quiescent current through M1 and M2 actually changes,
which means that quiescent current through M1 has to increase and M2 has to decrease or vice versa,
right, it cannot be both. Because, I mean, where will that, if both currents increase,
then which essentially means that the current source, since we have a current source at the,
connected at Vx, that will not allow for the increase of current on both M1 and M2,
which means either the current through M1 has to increase or the current through M2 has to
increase and the other has to decrease. So, if we assume, if we assume the current
through M1 increases by delta I, right, this direction, which means the current through M2
will also decrease by delta I, which essentially means that the current in the opposite direction
of delta I, correct. So, what is delta I? Delta I is equivalent gm, which is the gm1, gm2 by gm1
plus gm2 times V1 minus V2. What is V1 minus V2? Delta V minus delta V, which is equal to 0,
right. So, which essentially means that moment to, if you apply an incremental input on both
sides simultaneously in the same direction, right, there will not be any change in the
quiescent currents, right. Another way of, another way of saying the same thing is that
if we apply a common signal, if we apply a common increment to gates of M1 and M2,
then there is no change in current and in more technical terms, in using more jargon,
what we call this is, this differential amplifier rejects common mode excitation, right.
So, this is a very critical property of, crucial property of differential amplifier,
which has immense use, right. So, because now if our differential amplifier is able to give us gain
which is equivalent to difference of two voltages, but is also insensitive to any
common mode excitation, which means that I do not have to set VB1 and VB2 very accurately.
VB1 and VB2 can be slightly different as well, right and still our quiescent points
would not get disturbed too much, correct. So, another piece of jargon is,
so this is called often ACM, that is the ACM is equal to 0. So, with ideal I naught at Vx,
right. Often also Vx is also called the tail transistor, tail node, right. So, Vx is also
called, often Vx is also called the tail transistor or the tail node in this case,
tail node. Sometimes you will hear me talking about a tail transistor or tail node, which
essentially means that anything that is connected to that node Vx, ok. So, if ACM is 0,
what about if we apply a differential excitation? Let us say we apply if V1, if we apply
Vi by 2 or let me say V1 equal to Vi by 2 and Vi, V2 equal to minus Vi by 2, right.
Basically, I am increasing one side, I am increasing the gate voltage of M1 and reducing
the gate voltage of M2, that is all I am doing. So, what will be the incremental current? Delta
i will be gm times V1 minus V2, that is gm times Vi. Where will this current flow? This current
delta i is flowing into Rd and it will create a voltage, right. So, maybe I should draw this
stuff again, right. So, where will this delta i flow? Note that even if I am sketching it
using transistors, the analysis that I am doing is a small signal. By now, we should be comfortable
without replacing the transistor with a small signal equivalent and we will be able to
visualize what the incremental model would look like even without replacing the transistor with
its incremental equivalent, right. So, if delta i is flowing upwards and it is flowing into Rd,
what will be the incremental voltage at this node V0? Incremental voltage V0 will be delta i
times Rd which is gm times Rd times Vi which means the gain V0 over Vi is gm times Rd.
This is often called the differential gain or AD, right. So, this is called differential gain,
ok. So, why is this split, why do I bother about this differential gain and common mode gain?
The reason we bother about differential gain and common mode gain is because of the following
reason. So, let us say I apply an arbitrary input V1, arbitrary input V2 here, right.
So, note that I can express V1 as V1 plus V2 by 2 plus V1 minus V2 by 2, right. It is simple
algebra, I broke it up and also similarly I can express V2 as V1 plus V2 by 2 plus or rather
minus V1 minus V2 by 2, correct. So, why is this useful? This is useful because in the context of
differential amplifier, this is my common mode input, right. This is the common mode input
and this is the differential input, ok. So, since in the incremental sense our circuit is
linear, we can use superposition, right. In the incremental sense our circuit is linear
which means we can use superposition which means we can in principle apply
a common mode input here that is let me call this Vicm, Vicm and let me call this Vidiff.
So, we can apply a signal Vicm and Vicm on both sides and we can apply another signal
Vidiff plus Vidiff and minus Vidiff on both sides, right. We can do both simultaneously
or we can do both individually and do our analysis, right. So, that is one of the reasons why
you will see that breaking the input into common mode and differential
mode often helps a lot, ok. So, let us stop here. In the next class,
we will delve deep into some more detail into the properties of a differential amplifier.
Note that we have till now neglected the effects of channel length modulation. We will see what
is the effect of channel length modulation. Again, we cannot assume we have an ideal current
source all the time, right. So, we have to see what are the non-ideal effects that come in
if we use a real current source, right or by real current source we mean,
essentially mean that a current source is essentially a transistor, right.
So, a transistor will have its own GDS or own RDS. So, we will see what happens to this common
mode gain, what happens to the differential gain, right in the presence of, in the presence of
these non-idealities. We will also see how to, how to, how to size the transistors M 1 and M 2
in order to get certain quiescent, certain quiescent gain and certain quiescent, rather
what are the, to get certain quiescent current through them and so on, right, ok. So, let us stop.
