The Series-Shunt Amplifier with Loading
Open a two-transistor feedback loop with its loading kept, and account for every step from the resistor ratio to the measured gain.
You Will Be Able To
- Find what a feedback network loads at the input and at the output of a series-shunt loop.
- Predict the closed-loop gain from the loaded open-loop gain and the feedback factor.
- Write a gain error budget and say which term is largest.
- Explain why a grounded body changes the gain that the resistor ratio promises.
- Lesson
- E10.2
- Time
- 35 min
- Runs on
- ngspice
- Check
- 5 questions
A two-transistor amplifier was built for a gain of 10, with RF = 18 kΩ and R1 = 2 kΩ. The run measures 7.47. Nothing is broken: three separate effects took the missing 25 %.
- Frequency 10 kHzclosed-loop gain = 7.468
The gain is flat at 7.47. The resistor ratio promised 10.
Read the plot as a table
| Frequency | closed-loop gain |
|---|---|
| 1 kHz | 7.468 |
| 10 kHz | 7.468 |
| 100 kHz | 7.468 |
| 1 MHz | 7.468 |
| 10 MHz | 7.468 |
| 100 MHz | 7.388 |
| 1 GHz | 3.673 |
| 10 GHz | 0.2704 |
No load capacitor is attached, so the gain stays flat far beyond any real use.
This lesson finds each effect and gives it a number.
The Pair and Its Network
- Two gain stages
M1 carries 19.7 µA and M2 carries 61.2 µA. Each stage inverts, so the output moves the same way as the input.
19.7 µA
- Sense and return
RF and R1 divide the output voltage and place a part of it on the source of M1. The input drives the gate. M1 amplifies the difference: a series-shunt loop.
- Loading at the input
Hold the output still. The source of M1 then sees R1 in parallel with RF: 1.8 kΩ of degeneration.
- Loading at the output
Disconnect the source of M1. The drain of M2 then sees RF in series with R1: a 20 kΩ load.
- The ideal feedback factor
What is left is a divider that loads nothing: β = R1/(R1 + RF) = 0.1.
Read the circuit as text
- R_D resistor (60 kΩ): pin 1 →
VDD, pin 2 →d1 - M1 NMOS transistor (4/0.5): gate →
n1, drain →d1, source →s1 - M2 PMOS transistor (20/0.5): gate →
d1, source →VDD, drain →out - R_F resistor (18 kΩ): pin 1 →
out, pin 2 →s1 - R_1 resistor (2 kΩ): pin 1 →
s1, pin 2 →0 - v_in voltage source (0.95 V DC): + →
n1, − →0 - out port: terminal →
out
s1= 0.162 Vd1= 0.618 Vout= 1.264 V
The body of M1 is at ground. That matters later in this lesson.
The ideal rule says the closed-loop gain is 1/β = 10. It assumes a very large loop gain and a network that loads nothing. Neither holds here.
A real network is made of resistors, and they carry current. At the input they sit in the source of M1 and lower its gain. At the output they load M2.
Open the Loop, Keep the Loading
To find the loop gain, we need the gain of the amplifier with the network's load on it but without its feedback. For a series-shunt loop there are two rules.
| Port | What you do to the other port | What the amplifier sees | Here |
|---|---|---|---|
| Input (series) | Short the output node | R1 ∥ RF in the source of M1 | 1.8 kΩ |
| Output (shunt) | Open the input connection | RF + R1 from the drain of M2 to ground | 20 kΩ |
Short a port that the network senses as a voltage. Open a port where it returns a voltage in series.
The lab builds this opened loop as a second copy of the two transistors. A voltage source holds the far end of RF at the DC output level. A current source replaces the DC current of M1 in R1. The copy then sits at the same operating point: both outputs are at 1.264 V.
With the loading in place, hand analysis gives each stage. A1 = gm1RD/(1 + (gm1 + gmb1)·1.8 kΩ) and A2 = gm2·(ro2 ∥ 20 kΩ).
| Quantity | By hand, from the operating point | From the AC run |
|---|---|---|
| First stage A1 | 9.678 | — |
| Second stage A2 | 9.305 | — |
| Open-loop gain A | 90.05 | 78.99 |
| Output resistance, loop open | ro2 ∥ 20 kΩ | 17.38 kΩ |
The hand value of A is 14 % high. It leaves out ro of M1, which lets the drain of M1 react to its source.
The Gain Error Budget
Now close the loop on paper. The closed-loop gain is A/(1 + A·β). One more effect enters first, and it is specific to this circuit.
The returned voltage arrives at the source of M1. The body of M1 is at ground, so a rising source also lowers VBS. The feedback therefore acts through gm1 + gmb1, while the input acts through gm1 alone. The effective feedback factor β′ is β·(gm1 + gmb1)/gm1 = 0.120, not 0.100.
| Step | Closed-loop gain | Lost at this step |
|---|---|---|
| Resistor ratio, 1/β | 10 | — |
| Body of M1 at ground, 1/β′ | 8.34 | 16.6 % |
| Finite loop gain, measured A = 79.0 | 7.540 | 9.5 % |
| The run | 7.468 | 0.9 % |
Each loss is a share of the gain left by the step before. The last 0.9 % is what the two-port picture leaves out, such as signal that passes forward through the network.
- Neutral
- Measure
- Analysis
- The body effect
The feedback acts through gm1 + gmb1. More loop gain cannot recover this step.
- The loop gain
The network loads both stages, so A is only 79.0. This step shrinks with more loop gain.
- The remainder
The two-port picture leaves out the last 0.9 %.
Read the diagram as text
- Resistor ratio: gain 10
- Body at ground: gain 8.34
- Finite loop gain: gain 7.54
- The run: gain 7.47
- Feedback factor: 0.120, not 0.100
- Loop gain: T = 9.48
- Forward signal: passes through the network
- Resistor ratio → Body at ground: 16.6 % lost
- Body at ground → Finite loop gain: 9.5 % lost
- Finite loop gain → The run: 0.9 % lost
- Feedback factor → Body at ground
- Loop gain → Finite loop gain
- Forward signal → The run
Three effects take the missing 25 %.
The loop gain is T = A·β′ = 79.0 × 0.120 = 9.48. A loop gain near 10 leaves about one part in ten as error. More loop gain shrinks that row. It does nothing for the body-effect row.
The output resistance confirms the loop gain. The opened loop shows 17.4 kΩ and the closed loop 1.64 kΩ. The ratio is 10.6, close to 1 + T = 10.5.
Asked in interviews
How do you find the loading of the feedback network in a series-shunt amplifier?
For the input side, short the output node and look into the network from the input: here R1 ∥ RF. For the output side, open the input connection and look in from the output: here RF + R1.
Why is the closed-loop gain below 1/β?
The gain is A/(1 + Aβ), which is 1/β times T/(1 + T). With a loop gain of 9.5 that factor is 0.905. Loading lowers A, so it lowers T and makes the error larger.
A feedback pair returns its signal to the source of an NMOS whose body is grounded. What changes?
The source drives the transistor through gm + gmb, and the gate through gm only. The feedback factor grows by (gm + gmb)/gm. Here the gain target falls from 10 to 8.34.
Try It
The first block is the closed loop. The second block is the opened loop: the same transistors, with R1A and RFA loading the source of M1 and RFB plus R1B loading M2.
EHOLD copies the output level of the closed loop, so RFA carries the same DC current as RF. FDC copies the DC current of M1 into R1B. Both copies carry DC only, because the closed loop is silent when the opened loop is driven.
The let lines repeat the hand analysis with gm, gmb and gds read from the operating point. Four AC runs then measure the closed gain, the open gain and both output resistances.
Predict before each change. Make RD and RDO 120k and lower both DC inputs to keep M1 near 10 µA: does the last row of the budget shrink? Then set RF to 22 kΩ in all three places: how close to 10 does g_closed come?
Run the netlist to draw its plots here.
vout_dc = 1.263782e+00 vout_open_dc = 1.263782e+00 id1 = 1.970451e-05 id2 = 6.121866e-05 a1 = 9.678187e+00 a2 = 9.304888e+00 a_hand = 9.005445e+01 beta_eff = 1.199738e-01 t_hand = 1.080417e+01 g_hand = 7.629034e+00 g_closed = 7.468444e+00 a_open = 7.898575e+01 g_two_port = 7.539527e+00 rout_closed = 1.637840e+03 rout_open = 1.737593e+04Check Yourself
Common Mistakes
Wrong Open the loop by removing the feedback network.
The network also loads the amplifier, so removing it gives an open-loop gain the closed circuit never has.
Fix Keep its loading: R1 ∥ RF at the source of M1 and RF + R1 at the output, and remove only the returned signal.
Wrong Lower the network resistance to make the feedback stronger.
The ratio sets β; a smaller network only loads the output stage and lowers the loop gain.
Fix Choose the ratio for the gain. Choose the level well above the output resistance the stage can drive.
Wrong Expect 1 + RF/R1 when the feedback returns to the source of an NMOS with a grounded body.
The source acts through gm + gmb, so the effective feedback factor here is 0.120 and the target is 8.34.
Fix Include gmb in the feedback factor, or use a device whose body is tied to its source.
Remember
- A feedback network loads both ports of the amplifier. Count that load before you compute the loop gain.
- Series-shunt: short the output to find the input-side load, and open the input connection to find the output-side load.
- The closed-loop gain is A/(1 + Aβ), which is 1/β times T/(1 + T). A loop gain of 9.5 costs 9.5 %.
- Feedback returned to the source of a grounded-body NMOS acts through gm + gmb. Here that moves the target from 10 to 8.34.
- The ratio of the open-loop and closed-loop output resistances gives 1 + T directly.