0.0.

1 Rosenstark’s method
Consider a LTI electric circuit, containing an amplifier with gain R, which sets a constraint
between the values of two variables y ∗ and u∗ according to the relation:

u∗ = Ry ∗

y* u*
R

u y
LTI

We denote as (system theory notation) :

• u: input of the circuit

• y: output of the circuit

We can consider the two variables that are the input and output of the amplifier y ∗ and u∗ as
an output and an input of the system respectively and imposing the constraint externally:

u* y*
R

u* y*

u LTI y

Since the
 system is LTI we can find a matrix
 A of transfer functions between the input vector
u y
u = ∗ and the output vector y = ∗ :
u y

y = Au

1
     
y a11 a12 u
=
y∗ a21 a22 u∗
If we had dynamical elements, the transfer function matrix would still exist in the Laplace
domain.
Let’s now rewrite the system of equations that constitute the matrix equation:
(
y = a11 u + a12 u∗
y ∗ = a21 u + a22 u∗

If we add the constitutive relation of the amplifier we get:

∗

 y = a11 u + a12 u (1)

y ∗ = a21 u + a22 u∗ (2)
 ∗
u = Ry ∗

(3)

Let’s represent the first one using the algebra of blocks. We start giving the diagram the shape
it will have at the end, even if it’s not complete yet

y = a11 u + a12 u∗ (1)

5
u u*
a12 y

u
a11

The second one becomes:

y ∗ = a21 u + a22 u∗ (2)

a22 u*

u
a21 y*

And the third one becomes:

u∗ = Ry ∗ (3)

u*

y* u*
R

2
If we connect the three components we get the following architecture:

a22 u*

u u
a21 y*
R u* u*
a12 y

u
a11

We derive now the transfer function between u and y in this architecture.

Start by finding a transfer function for the feedback:

u∗ = Ry ∗ = R(a22 u∗ + a21 u)

u∗ − Ra22 u∗ = Ra21 u
u∗ (1 − a22 R) = Ra21 u
Ra21
u∗ = u
1 − a22 R
Then, we can apply the rules for the cascade and the parallel to get:
 
∗ Ra21 Ra21 a12
y = a12 u + a11 u = a12 u + a11 u = + a11 u
1 − a22 R 1 − a22 R

So, the transfer function between the output and the input, also called the feedback gain is:
y Ra21 a12
AF B = = + a11
u 1 − a22 R
We can rewrite it as follows:
y Ra21 a12 + a11 − a11 a22 R
AF B = =
u 1 − a22 R
a21 a12
Ra22 + a11 − a11 a22 R
a22
=
1 − a22 R
a21 a12
( − a11 )Ra22 + a11
a22
=
1 − a22 R
a21 a12
−Ra22 (a11 − ) + a11
a22
=
1 − a22 R
We can introduce the loop gain as T = −a22 R and obtain the following:
a21 a12
(a11 − )T + a11
a22
AF B =
1+T

3
We can now define other two numerical constants:
a21 a12
A∞ = lim AF B = a11 −
T →+∞ a22
A0 = lim AF B = a11
T →0
Doing so, the expression for the feedback gain becomes:
A∞ T + A0 T 1
AF B = = A∞ + A0
1+T 1+T 1+T
| {z } | {z }
f eedback f eedf orward

The reason for this distinction into feedback and feedforward is that the first part (feedback)
depends on the blocks on the feedback loop, while the second (feedforward) just depends on the
feedforward branch (the lower one).
Rosenstark’s method provides a practical procedure to estimate values of these three parame-
ters: A∞ , A0 and T .

COMPUTING A0

By definition, it is:
A0 = lim AF B = a11
T →0
Let’s think about the meaning of the coefficient a11 .
We defined it as the transfer function between u and u in the matrix of transfer functions of
the LTI circuit without the amplifier R:

u* y*   
y
 
a11 a12 u
=
y∗ a21 a22 u∗

u LTI y y = a11 u + a12 u∗

We can obtain the value of a11 by setting u∗ to zero and computing it as:
y 
a11 = 
u u∗ =0
Since in the circuit including the amplifier R we have the relation:

u∗ = Ry ∗

we can set u∗ = 0 by simply setting R = 0 in the circuit with the amplifier:

u∗ = Ry ∗ = 0y ∗ = 0 ∀y ∗

4
Therefore:

u* y*
0
y 
A0 = a11 = 
u* y* u R=0

u LTI y

So, we can find the value of a11 as the transfer function between the input and the output of
the circuit when the gain of the amplifier is set to zero.

COMPUTING A∞

If we take the limit for R → +∞ of the overall transfer function AF B we get:

a21 a12
lim AF B = a11 − = A∞
R→+∞ a22
Notice that, when we set T → +∞ the gain of the feedback is finite and so will be the output
of the circuit for finite inputs. Since the output of the circuit depends in general linearly on the
output of the amplifier u∗ , such variable cannot be infinite. But, since the gain is infinite, the
only way for this to happen is for the input of the amplifier to be an infinitesimal (y ∗ → 0 ).
This consideration is very useful for the practical computations.
So, the parameter A∞ can be evaluated as the transfer function between the output and the
input of the circuit when the gain of the amplifier is infinite:

u*
∞ y*

a21 a12 y 
u* y* A∞ = a11 − = 
a22 u R→+∞

u LTI y

5
COMPUTING T

By definition, it is:
T = −a22 R
Since the value of R is known, we just need to compute the value of a22 .
This can be done considering the nature of A22 in the transfer function matrix:

u* y*   
y
 
a11 a12 u
=
y∗ a21 a22 u∗

u LTI y y ∗ = a21 u + a22 u∗

We can see that a22 is the transfer function between y ∗ (the input of the amplifier) and u∗ (the
output of the amplifier) when u is set to zero and in absence of the amplifier :

y ∗ 

a22 = ∗ 
u no R
u=0

Once a22 is found we can compute the loop gain as:

T = −Ra22

PRELIMINARY CONSIDERATIONS: R → +∞

We now wish to find a practical procedure to apply to a circuit in order to determine the
three constants that appear in the formula for the gain of the circuit.
In order to do this let’s study the properties of the block R to identify which elements in the
circuit are suitable candidates for R.
The block R takes one variable as input and gives another variable as output.
The input and outputs can be either one of the following combinations, according to the type
of characteristic equation that describes the element R:

input (y ∗ ) output (u∗ )

voltage voltage
voltage current
current current
current voltage

6
Example: Ohm’s law

7
We can see a resistor as an example of block R with the current as input and the voltage as out-
put:

v
v = Ri (Ohm’s law)
i ∗
y =i (Input)

8
∗
u =v (Output)
R
We can see it as a block that samples the current and gives back a voltage R times bigger:

i v
R
A resistor can also be seen as a two-port element with two short-circuited terminals that performs current
sampling (via a short-circuit, which doesn’t affect the circuit) and gives back a proportional voltage with
proportionality factor R (through a current-controlled voltage source):

i i

R
0 i Ri v

v=Ri

So, a resistor is nothing but a transresistance amplifier with input and output located in the same place.


We can think of the amplifying element as a two-port element that performs a current or voltage
sampling at its input and gives back a voltage or current at its output that is proportional to
its input.
The easiest kinds of amplifying blocks are controlled sources, which can be seen as follows:

• CURRENT-CONTROLLED SOURCE: two port element that performs current sampling

through a short-circuit(which doesn’t affect the circuit)
• VOLTAGE-CONTROLLED SOURCE: two port element that performs voltage sampling
through an open-circuit(which doesn’t affect the circuit)

7
If we don’t consider the type of output we have:

0 i

v v αv 0 i αi

Let’s now represent on a plane the input-output characteristic of such elements.

We have:

u*

*
αy
=
u*

y*

Where y ∗ is the controlling quantity of the controlled source and u∗ is its controlled quantity.
If we now let the slope go to infinity we get that the line coincides with the u∗ axis, which
means that:

• The input quantity is zero

• the output quantity is independent from the input quantity

If we consider the previous two cases of voltage and current sampling we have that, when we
let α → +∞:

• current-controlled source:

– the current sampling is given by a short-circuit ⇒ vin = 0

– the input is zero ⇒ iin = 0
– the output is independent from the input ⇒ no constraints on the output

8
 • voltage-controlled source:

– the voltage sampling is given by an open-circuit ⇒ iin = 0

– the input is zero ⇒ vin = 0
– the output is independent from the input ⇒ no constraints on the output

So, in both cases the input port is subjected to the constraints:

(
vin = 0
⇒ describe a nullator
iin = 0

And the output isn’t subjected to any constraint, which describes a norator.

7 7
The symbols for the two components are:

NULLATOR NORATOR

v v

i i

( (
v=0 no constraint on v
i=0 no constraint on i

Therefore, in the limit for α → +∞ the two-port elements both become:

This two-port element is called an operational amplifier.

9
The operational amplifier is a two-port element with the following characteristics:

i+ 
v+− = 0

iout i+ = 0
v+- 

i− = 0
i- no constraints on iout and vout
vout

iout
The lower terminal is often connected to the reference potential and, in this case, the symbol
of the ground is sometimes dropped:

PRELIMINARY CONSIDERATIONS: R = 0

We now want to see how we can set R = 0 in the practical case of a controlled source.
This corresponds to leaving the input untouched and setting to zero the coefficients of the con-
trolled sources.
Let’s not consider the type of controlling quantity but just the type of controlled quantity.
This gives rise to two possible types of topologies, which are (we denote by y ∗ a general con-
trolling quantity):

y*
? αy* y*
? αy*

If we set the coefficients α to zero, we have that:

• the voltage source (fixed voltage, allows every current) becomes a short-circuit (fixed zero
voltage, allows every current)
• the current source (fixed current, allows every voltage) becomes an open-circuit (fixed
zero current, allows every voltage)

10
So, we have the previous two topologies become respectively:

y*
? αy* y*
? αy*

y*
? y*
?

So, setting R = 0 is equivalent to substituting:

• a controlled current source with an open circuit

• a controlled voltage source with a short circuit

PRELIMINARY CONSIDERATIONS: choice of the controlled source

It is strongly recommended to choose as a controlled source for the application of Rosenstark’s

method a controlled source that has one terminal connected to the reference potential, since
this makes the computations much easier.

PRACTICAL PROCEDURE FOR A0

We defined A0 as:
y 
A0 =
u R=0


In order to set the gain R to zero we need to set the gain α of the controlled source to zero.
This is equivalent to substituting:

• a controlled current source with an open circuit

• a controlled voltage source with a short circuit

Because of this process of ”killing” the controlled source it is also referred to as the dead tran-
sistor gain.

11
 A0
1. • if the controlled source is a current source, substitute it with an open circuit
• if the controlled source is a voltage source, substitute it with a short circuit

2. find the transfer function between the output of the circuit (y) and the input of the
circuit (u) as y/u

The value of the constant is A0 = y/u

PRACTICAL PROCEDURE FOR A∞

We defined A∞ as:
y 
A0 = 
u R→+∞
Letting R → +∞ is equivalent to substituting the amplifying two-port controlled source with
an operational amplifier.
Once we perform the substitution, we just have to evaluate the transfer function between the
input and the output y/u.

A∞
1. • if the controlled source is current-controlled, open the branch in which the
controlling current flows and insert the input port (left port) of the op-amp
• if the controlled source is voltage-controlled, connect the input port of the
op-amp in parallel with the two nodes across which is defined the controlling
voltage

2. substitute the controlled source with the output port of the op-amp

3. find the transfer function between the output of the circuit (y) and the input of the
circuit (u) as y/u

The value of the constant is A∞ = y/u

PRACTICAL PROCEDURE FOR T

We said that it is possible to compute T in two steps. First compute:

y ∗ 

a22 = ∗ 
u no R
u=0

Once a22 is found compute the loop gain as:

T = −Ra22

12
The first step is equivalent to:

• removing the controlled source (no R); the input part (short or open circuit) is left un-
touched since it doesn’t affect the circuit

• setting the input of the circuit to zero (u = 0)

• evaluating the transfer function between what was the output of the controlled source
(u∗ ) and what would be its input (y ∗ ) as: y ∗ /u∗

To evaluate the transfer function between the output of the controlled source two-port element
and its input we put a test generator of the same kind of the removed controlled source in the
same place in which the controlled source was.
In order to make the procedure independent from our reasoning we call ys the output quantity
of the controlled source and us its input quantity.
Moreover, we will call k the constant a22 .

T
1. • if the controlled source is a current source, replace it with a test current source
of value ys and same orientation
• if the controlled source is a voltage source, replace it with a test voltage source
of value ys and same orientation

2. set the input voltage or current to zero (short-circuit a voltage generator, open a
current generator)

3. find the transfer function k between the new test source (ys ) and the input of the
removed controlled source (us ) as k = us /ys (using a test source of the same kind
of the removed controlled source)

The value of the loop gain is T = −kA, where A is the gain of the removed controlled
source.

13