Return Ratio and Blackman's Formula
Measure the return ratio of one controlled source, then get the exact closed-loop gain and any port impedance from it.
You Will Be Able To
- Measure the return ratio of a controlled source in a netlist.
- Predict a closed-loop gain from the return ratio, the ideal gain and the direct feedthrough.
- Find a port impedance from two return ratios with Blackman's formula.
- Explain why a return ratio and a two-port loop gain can differ.
- Lesson
- E10.5
- Time
- 35 min
- Runs on
- ngspice
- Check
- 5 questions
An engineer put 1 kΩ in series with a follower's output to protect it. His note divided the total resistance by 1 + T and promised 1.44 kΩ, but the bench read 2.29 kΩ.
- the run
- Z0/(1 + T)
- Series resistor RISO 20 Ωthe run = 1.307 kΩ · Z0/(1 + T) = 1.29 kΩ
With almost no series resistor the two agree. The loop divides the follower's own resistance by 1 + T.
- Series resistor RISO 2 kΩthe run = 3.287 kΩ · Z0/(1 + T) = 1.594 kΩ
At 2 kΩ the run reads 3.287 kΩ and the rule 1.594 kΩ. The loop does not shrink RISO.
Read the plot as a table
| Series resistor RISO | the run | Z0/(1 + T) |
|---|---|---|
| 20 Ω | 1.307 kΩ | 1.29 kΩ |
| 500 Ω | 1.787 kΩ | 1.364 kΩ |
| 1 kΩ | 2.287 kΩ | 1.441 kΩ |
| 1.5 kΩ | 2.787 kΩ | 1.517 kΩ |
| 2 kΩ | 3.287 kΩ | 1.594 kΩ |
The rule divides everything by one loop gain. The run does not.
One loop gain cannot describe every port, so this lesson measures two return ratios instead.
The Return Ratio of a Controlled Source
- The closed loop
G_M pushes gm·vc into node s. While the loop is closed, vc = vg − vs.
gm·vc
- Zero the input
Set vin to zero. The return ratio belongs to the loop alone.
- Break and drive
Cut the control of G_M from the circuit. Drive node c with a test voltage vt.
- Read what returns
The current gm·vt flows into Rs. The circuit returns vg − vs = −gm·Rs·vt.
gm·vt
Read the circuit as text
- v_in voltage source: + →
n1, − →0 - g port: terminal →
n1 - v_t voltage source: + →
n2, − →0 - c port: terminal →
n2 - G_M current source (g_m·v_c): + →
0, − →s - R_s resistor (8.381 kΩ): pin 1 →
s, pin 2 →0 - R_ISO resistor (1 kΩ): pin 1 →
n3, pin 2 →s - out port: terminal →
n3
Rs = 1/(gmb + gds). Nothing loads the out pin here.
A controlled source sets its output from a voltage or current elsewhere. A transistor's gm is one.
The return ratio T of a controlled source is minus the returned signal divided by the test signal. Here T = gm·Rs.
The minus sign makes T positive for negative feedback. No feedback network was named: T belongs to the source and the whole circuit around it.
The lab copies gm and Rs from the SKY130 follower and prints t_open = 5.511562e+00.
Closed-Loop Gain from T and the Feedthrough
Two more runs give the exact gain. A∞ is the gain with a perfect controlled source. A0 is the gain with the source switched off, the direct feedthrough.
The follower has A∞ = 1 and A0 = 0. The formula gives g_formula = 8.464270e-01 and the transistor shows gain = 8.464269e-01.
- Return ratio
Drive the gate of M2 with vt and read node d1. The lab prints
t_rr = 9.300043e+00. - A perfect M2
Give M2 unlimited gm. The loop then forces the signal at d1 to zero:
a_inf = 8.245674e+00. - M2 switched off
Hold the gate still. The input still reaches the output through M1 and RF:
a_0 = 2.401716e-01.feedthrough
Read the circuit as text
- R_D resistor (60 kΩ): pin 1 →
VDD, pin 2 →n1 - M1 NMOS transistor: gate →
n2, drain →n1, source →n3 - M2 PMOS transistor: gate →
n4, source →VDD, drain →n5 - v_t voltage source: + →
n4, − →0 - d1 port: terminal →
n1 - R_F resistor (18 kΩ): pin 1 →
n5, pin 2 →n3 - R_1 resistor (2 kΩ): pin 1 →
n3, pin 2 →0 - v_in voltage source: + →
n2, − →0 - out port: terminal →
n5
In the closed loop, the gate of M2 connects to d1.
| Quantity | Lab line |
|---|---|
| A∞·T/(1 + T) + A0/(1 + T) | 7.468 |
| Closed-loop gain in the run | 7.468 |
The formula is exact. The feedthrough term adds 0.023 to the gain.
Blackman's Formula for Any Port
A port is a pair of terminals where you measure an impedance. Blackman's formula needs three numbers at that port.
- I/O
- Measure
- Analysis
- The dead loop
Switch the controlled source off and measure the port. That impedance is Z0.
- Two return ratios
Measure T at the same controlled source twice: once with the port shorted, once with it open.
- Combine
Multiply Z0 by 1 + Tshort and divide by 1 + Topen. The result matches the direct run.
Read the diagram as text
- The port: out pin of the follower model
- Source off: Z_0 = 9.381 kΩ
- Port shorted: T_short = 0.588
- Port open: T_open = 5.512
- Blackman: Z = 2.287 kΩ
- Direct run: Z = 2.287 kΩ
- The port → Source off: measure Z
- The port → Port shorted: measure T
- The port → Port open: measure T
- Source off → Blackman: Z_0
- Port shorted → Blackman: times 1 + T short
- Port open → Blackman: over 1 + T open
- Blackman → Direct run: matches
The loop is never closed and probed at the port. Three simpler runs replace that.
| Port | Z0 | Tshort | Topen | Blackman | Direct run |
|---|---|---|---|---|---|
| Follower model, out | 9.381 kΩ | 0.5875 | 5.512 | 2.287 kΩ | 2.287 kΩ |
| Feedback pair, out | 16.87 kΩ | 0 | 9.300 | 1.638 kΩ | 1.638 kΩ |
Impedances in ohms. Both rows match the direct run to seven digits.
With the follower's out pin shorted, part of the test current flows through RISO. The loop still returns 0.588. That is why RISO is not divided by 1 + T.
Short the pair's output and M2's current goes to ground. Nothing returns, so Tshort = 0 and Z = Z0/(1 + T).
The rules of E10.1 are the two special cases. A shunt port has Tshort = 0 and divides. A series port has Topen = 0 and multiplies.
The transistor follower reads z_direct = 2.293491e+03, which is 0.28 % above the model. The model keeps only gm, gmb and gds.
When Return Ratio and Two-Port Loop Gain Differ
E10.2 analysed the same pair with the two-port method and found T = 9.48. The return ratio of M2 is 9.30.
| Method | Loop gain | Predicted gain | Predicted Rout |
|---|---|---|---|
| Two-port (E10.2) | 9.48 | 7.540 | 1.659 kΩ |
| Return ratio of M2 | 9.300 | 7.468 | 1.638 kΩ |
| The run | none | 7.468 | 1.638 kΩ |
The two-port row is 1.0 % high in gain and 1.3 % high in Rout.
The two-port method assumes signal flows forward only through the amplifier and back only through the network. Here M1 also drives the output through RF.
The return ratio makes no such assumption, so its predictions are exact. Its price is two extra runs for A∞ and A0.
Neither number is wrong. A loop gain is a property of a method, and each method must use its own companion terms.
Asked in interviews
How do you find the return ratio of a transistor?
Zero the independent sources. Cut the control of the gm source and drive it with a test signal. The return ratio is minus the signal that comes back, divided by the test signal.
State Blackman's impedance formula.
Z = Z0·(1 + Tshort)/(1 + Topen). Z0 is the port impedance with the controlled source off. The two return ratios are taken with the port shorted and with it open.
Why does a resistor in series with a follower's output not shrink by 1 + T?
The loop senses the voltage before the resistor. With the output pin shorted the loop still works, so Tshort is not zero. In the lab the result is 2.287 kΩ, where dividing by 1 + T gives 1.441 kΩ.
A return ratio and a two-port loop gain disagree. Which is right?
Both, within their own method. The return ratio with A∞ and A0 gives the exact gain. The two-port method ignores forward signal through the feedback network, so it is close but approximate.
Try It
Part 1 holds the SKY130 follower and its linear model. The alter lines copy gm and Rs from the operating point into GM and RS.
EC copies vg − vs onto node c while the loop is closed. Setting its gain to 0 cuts the control, and VT then drives it.
RP is the port switch. Runs 5 and 6 read the return ratio with the port open and shorted.
Part 2 is the pair of E10.2 and a copy. EK holds the copy's gate at the closed loop's DC level, so both sit at one operating point.
Predict before each change. Set RISO and RISOM to 100: how close do z_blackman and z_one_t come? Then set RF and RFB to 180k: does a_0 rise or fall?
Run the netlist to draw its plots here.
gm_mf = 6.576136e-04 rs_mf = 8.381147e+03 gain = 8.464269e-01 z_direct = 2.293491e+03 z_model = 2.287118e+03 z_0 = 9.381150e+03 t_open = 5.511562e+00 t_short = 5.875145e-01 z_blackman = 2.287118e+03 z_one_t = 1.440691e+03 g_formula = 8.464270e-01 g_closed = 7.468444e+00 rout_closed = 1.637840e+03 t_rr = 9.300043e+00 a_0 = 2.401716e-01 a_inf = 8.245674e+00 g_asym = 7.468444e+00 z0_pair = 1.686982e+04 z_pair = 1.637840e+03Check Yourself
Common Mistakes
Wrong Divide every port impedance by 1 + T.
That holds only when shorting the port stops the loop, and a series port or a pin behind a resistor does not.
Fix Find Tshort and Topen for the port you care about, then apply Blackman's formula.
Wrong Leave the input source at AC 1 while you measure a return ratio.
The input's own signal then adds to the returned signal, so the ratio is no longer a property of the loop.
Fix Zero every independent source first. In the netlist, give each one AC 0.
VINM g 0 DC 0 AC 1
alter EC gain=0
alter VT ac=1VINM g 0 DC 0 AC 0
alter EC gain=0
alter VT ac=1Wrong Cut the loop at a wire that carries current and call the result the return ratio.
Such a cut removes the loading that each side puts on the other, so the measured ratio belongs to a different circuit.
Fix Cut at the control of the controlled source. At low frequency a MOS gate is such a point.
Remember
- To measure a return ratio: zero the inputs, cut the control of one controlled source, drive it, and read what returns.
- The closed-loop gain is A∞·T/(1 + T) + A0/(1 + T). On the feedback pair it matches the run to seven digits.
- Blackman: Z = Z0·(1 + Tshort)/(1 + Topen). A port that stops the loop when shorted gives Z0/(1 + T).
- A resistor outside the sensing point is not reduced by the loop: 2.287 kΩ measured against 1.441 kΩ from the one-T rule.
- Return ratio (9.30) and two-port loop gain (9.48) differ on the same circuit. Each belongs to its own formula.