Sei sulla pagina 1di 12

Two-Port vs.

Return-Ratio Analysis
Sergio Franco - May 21, 2014

The most common approaches [1] to negative-feedback circuit analysis are two-port analysis (TPA)
and return-ratio analysis (RRA). There are differences as well as similarities between the two,
oftentimes confusing, which Ill try to clarify by means of familiar circuit examples. The two
techniques are block-diagrammed in Figure 1 using subscripts TP and RR to distinguish between the
two-port and the return-ratio types.

Specifically, aTP and aRR are the open-loop gains, and TP and RR are the feedback factors. As
discussed in a previous blog [2], Figure 1a postulates unidirectional blocks, whereas Figure 1b is
more general because it accounts also for feedthrough around the error amplifier, as signified by the
gain block aft.

Figure 1. Negative-feedback block diagrams for (a) two-port (TP)


and (b) return-ratio (RR) analysis.

Two-Port Analysis (TPA)

Depending on whether sI and sO are voltages or currents, we have four possible topologies, as

depicted in Figure 2 by means of op amps. The first term of each hyphenated pair refers to the
manner in which the inputs are summed (series for voltages, parallel or shunt for currents), whereas
the second term refers to the manner in which the feedback network samples sO to produce the
feedback signal sF (parallel or shunt for voltages, series for currents). For each topology, the
closed-loop gain takes on the form

where

is the loop gain, and Aideal is the value of sO/sI in the limit TTP , in turn achieved by letting aTP .
Moreover, the feedback factor is

Figure 2. Using op amps to illustrate the four basic feedback topologies.

TPA seeks an expression for aTP that will take into account any interaction between the amplifier and
the feedback network, such as loading. This task is facilitated by the fact that negative feedback
transforms the open-loop resistance rpa presented by each port into the closed-loop resistance

with +1 in the series case, and 1 in the shunt case. If TTP is sufficiently large, we can regard R as
an open circuit in the series case, and as a short circuit in the shunt case.
As our first example, let us apply TPA to the current amplifier of Figure 3a, which has already
been discussed in a previous blog [2]. This circuit has

To find aTP, we rephrase the error amplifier as in Figure 3b. Recall [2] that the resistance seen by
the input source in Figure 3a is Ri = R2/(1 + av), and that seen by the load is Ro = R1(1 + av). For
large av we expect Ri to be small and Ro to be large. So, if we approximate Ro as an open circuit
(OC), the feedback network as seen from the amplifiers input port is simply the series combination
R2 + R1. Likewise if we approximate

Figure 3. (a) Shunt-series configuration terminated on a short-circuit load. (b) Circuit for
finding the open-loop parameters aTP, ria, and roa using TPA.

Ri as a short circuit (SC), the feedback network as seen from the amplifiers output port is simply the
parallel combination R2//R1. We thus have

indicating that the open-loop gain is

(Note that aTP av.) The loop gain is, after simplification,

Reconsider the example with av = 10 V/V and R1 = R2 = 10 k. Plugging into the above equations
gives

In spite of the OC and SC approximations, this compares quite favorably with Aexact = 1.909 A/A
found via direct analysis [2]. Just to make sure that this closeness is not accidental, lets check the
values of Ri and Ro. By inspection of Figure 3b, we have ria = R2 + R1 and roa = R2//R1. Applying
Equation (4), we get

thus confirming that Ri is much smaller and Ro is much larger than the other resistances in the
circuit.
Return-Ratio Analysis (RRA)

Return-Ratio Analysis (RRA)

This method, block-diagrammed in Figure 1b, calculates the closed-loop gain as

where TRR is the loop gain, and Aideal and aft are the values of sO/sI in the limits TRR and TRR 0,
respectively. These limits are in turn achieved by letting aRR and aRR 0 in Figure 1b. TRR is
found as the return ratio of the error amplifiers dependent source aRRsE, according the following
procedure: (a) set sI 0, (b) break the feedback loop right downstream of the dependent source
aRRsE, (c) subject the circuit downstream to a test signal sT of the same type and polarity as that of
the aRRsE source, (d) find the signal sR returned by the dependent source itself, and finally (e) obtain
the loop gain as the return ratio

As we proceed, well often find it convenient to express TRR as a product, similar to Equation (2),

where the feedback factor RR is found as

or, more simply, as RR = TRR/aRR.


Let us apply the procedure to the current amplifier of Figure 3a. This results in the circuit of
Figure 4a, where we have, by inspection, vR = avvD = av(vT), so

Consequently, aRR = av and RR = TRR/aRR = 1. To find the feedthrough gain, let av 0 as in Figure
4b. By inspection, iO = iI, so aft = iO/iI = 1 A/A. Consider again the example with av = 10 V/V and R1 =
R2 = 10 k, for which we now have

Compare Equation (14) with Equation (8) and observe the differences in the values of the Ts,
as, and s. Also, while ARR is exact, ATP is only approximate. Constrained to conform to the diagram
of Figure 1a, which postulates unidirectional blocks, TPA tries its best at approximating Aexact by
making TTP = 20 (compared to TRR = 10). You can easily verify that making TTP = 21 (instead of 20)
would yield ATP = Aexact for the present value of av. However, this would not work for lower values,
such as av = 1 V/V, where feedthrough becomes more relevant. The greatest difference occurs for av
= 0, where we have ARR = Aexact = 1 V/V but ATP = 0.

Figure 4. Circuits for finding (a) the loop gain TRR and (b) the feedthrough gain aft of the
current amplifier of Figure 3a.

Figure 5. (a) Shunt-shunt configuration and (b) Circuit to find the error gain aTP.

A More Complex Example

Let us apply the two methods to the I-V converter of Figure 5a, but using a more realistic op amp
model with a non-infinite input resistance ri and a non-zero output resistance ro. As we know, this
circuit has

Since this is a shunt-shunt topology, the feedback resistance appears as a grounded resistance to
both the input and the output ports, as depicted in Figure 5b. We have

indicating an open-loop gain of

(Note again that aTP av.) Moreover, the loop gain is

For RRA, refer to the circuits of Figure 6, which give, respectively,

so the loop and feedthrough gains simplify as

(Note the opposite polarities of aft and Aideal.) Let us compare the two approaches for a particular and
easy-to-visualize case, say av = 60 V/V and ri = ro = R = 10 k (yes, a substandard op amp to better
evidence the differences). Plugging these data into the pertinent equations, we get

Figure 6. Circuits for finding (a) the loop gain TRR and (b) the feedthrough gain aft of the IV converter of Fig. 5a.

Note the differences in the magnitudes, polarities, and dimensions of aTP and aRR, and of TP and RR.
There is also a slight difference between ATP and ARR (=Aexact).

Were the op amp to have ro = 0, then there would be no feedthrough as per Equation (18). In
this case we would get TTP = TRR = 30 and ATP = ARR = 9.677 V/mA. If the op amp also had ri = ,
then TTP = TRR = 60 and ATP = ARR = 9.836 V/mA. However, we would still have dramatic
differences, namely, aTP = 600 V/ma and TP = 0.1 mA/V, and aRR = 60 V/V and RR = 1. Yet, both
parameter sets would manage to deliver the same value for A in spite of their differences!
Two More Examples

Two More Examples

Lets conclude with the single-transistor circuits of Figure 7a and b. As evidenced by their common
small-signal model of Figure 7c, the error gain is now based on gm (in the case of op amps it was
based on av). Also, both circuits are of the series-input type. However, depending on whether we
take the output as the emitter voltage vo or as the collector current io, we have a shunt-output or a
series-output type, respectively. Both circuits are simple enough that we can analyze them directly
[3]. However, investigating them via both TPA and RRA will be far more illuminating.

Figure 7. (a) Series-shunt and (b) series-series circuits, and (c) their common small-signal
model. Assume gm = 40 mA/V, r = 2.5 k, ro = 40 k, and R = 1.0 k.

TPA of the series-shunt circuit of Figure 7a: To find Aideal, let gm as in Figure 8a. This results in
v 0, or vo vi, implying Aideal = 1.0 V/V (= 1/TP). With reference to Figure 8b, we have, by
inspection, vo = gm(R//ro)v, or aTP = vo/v= gm(R//ro). Plugging in the data as usual, we get

Figure 8. Circuits for finding (a) Aideal and (b) aTP for the series-shunt circuit of Figure 7a.

RRA of the series-shunt circuit of Figure 7a: To find TRR, refer to Figure 9a, where ir = gmv =
gm[(-it)(r//R//ro)]; to find aft refer to Figure 9b, where vo = vi(R//ro)/[r + (R//ro)]. So,

Plugging in the data as usual, we get

Figure 9. Circuits for finding (a) TRR and (b) aft for the series-shunt circuit of Figure 7a.

TPA of the series-series circuit of Figure 7b: To find Aideal, let gm as in Figure 10a. This yields
v 0, thus establishing a virtual short across r, so iR = vi/R. KCL at the supernode gives io = iR =
vi/R, so Aideal = io/vi = 1/R (= 1/TP). To find aTP, proceed with Figure 10b as usual. The results are

Plugging in the data gives

Figure 10. Circuits for finding (a) TRR and (b) aft for the series-series circuit of Figure 7b.

RRA of the series-series circuit of Figure 7b: To find TRR, refer to Figure 11a. This is the same as
Figure 9a, so we have the same TRR. To find aft, proceed with Figure 11b as usual. The results are

Plugging in the data gives

Clearly, there is less feedthrough in the series-series circuit than in the series-shunt one, so ATP
ARR.

Figure 11. Circuits for finding (a) TRR and (b) aft for the series-series circuit of Figure 7b.

Comparing TPA and RRA

The foregoing discussion has covered all four feedback topologies using simple circuits based on op
amps and transistors as gain elements (gain is av for op amps, gm for transistors). Comparing
procedures and results, we make the following observations:

RRA is more general than TPA because it does take into account feedthrough around the error
amplifier. As such, RRAs results are exact, whereas TPAs results are only approximate.

The difference between TPA and RRA is minimal for high loop gains, but is most obvious when the
loop gain drops to zero, where ARR aft but ATP 0.

TPA calculates the loop gain as the product TTP = aTPTP, RRA calculates it as the ratio TRR = vR/vT.

TPA uses a different two-port representation for each of the four feedback topologies, so in general
aTP, TP, and TTP vary from one topology to another.

By contrast, the loop gain TRR of a given circuit is independent of the topology, in turn determined
by the type and location of the input and output signals (however, aft is generally topology
dependent).

Any interaction between the error amplifier and the feedback network, such as loading, is handled
differently by the two analyses. TPA postulates TP = 1/Aideal, and then it finds aTP by manipulating
the amplifier circuit using the OC and SC approximations, so in general aTP av (or aTP gm).

By contrast, RRA effects no circuit manipulation, aside from breaking the loop for signal injection.
RRA implicitly assumes aRR = av (or aRR = gm), and it shifts the effects of the interaction between the
error amplifier and the feedback network to the feedback network itself, so in general RR 1/Aideal.

TTP and TRR may at times coincide, but this should not be assumed to be the norm. In particular,
one should never use TRR to calculate ATP, or TTP to calculate ARR. See, for instance, the error
incurred in trying to use Equation (3) in [2].

RRA has a more intuitive feel to it, and is also better suited to computer simulations or testing in
the lab (a subject that I plan to address in a future blog). On the other hand, TPA forces you to
dissect a circuit in ways that are more revealing of the interplay between the amplifier and the
feedback network.

Open Questions

Which method do you prefer? Should both methods be covered in college courses? If so, with equal
emphasis? Or should one be dropped altogether? If so, which one? Comments welcome!

References

[1] P. J. Hurst, A Comparison of Two Approaches to Feedback Circuit Analysis, IEEE Trans.
Education, vol. 35, No. 3, pp. 253-261, Aug. 1992.
[2]
http://www.edn.com/electronics-blogs/analog-bytes/4427143/Feedthrough-in-negative-feedback-circ
uits[3] http://online.sfsu.edu/sfranco/BookAnalog/AnalogJacket.pdf

Potrebbero piacerti anche