Diode-Connected and Triode Loads, and the Inverter as an Amplifier
Put three transistor loads on one driver at the same current. Measure the gain each gives and what each costs.
You Will Be Able To
- Predict the gain of a diode-loaded stage from two transconductances and the body effect.
- Explain why a triode load gives more gain over a smaller input range.
- Find the gain of a self-biased inverter from the gm and ro of both devices.
- Measure how the bias current and gain of an inverter amplifier follow its supply.
- Lesson
- E9.4
- Time
- 35 min
- Runs on
- ngspice
- Check
- 5 questions
An inverter with a feedback resistor made a cheap amplifier that drew 10.1 µA at 1.8 V. With the supply 10 % off, it drew 28.5 µA at 1.98 V and 2.3 µA at 1.62 V.
Read the plot as a table
| Supply | bias current |
|---|---|
| 1.5 V | 735.4 nA |
| 1.6 V | 1.939 µA |
| 1.7 V | 4.714 µA |
| 1.8 V | 10.1 µA |
| 1.9 V | 18.86 µA |
| 2 V | 31.27 µA |
| 2.1 V | 47.29 µA |
Twelve times more current from one end of the supply range to the other.
Nothing but the supply sets this current. Each simple load in this lesson has a cost like this one.
Gain as a Ratio of Transconductances
- Device
- Load
- Neutral
- Structure
- Analysis
- Diode load
The gain is gm1/(gm2 + gmb2). Sizes set this ratio, and it is low: 1.51.
- Triode load
The load acts as a resistor of 127 kΩ. The gain is 20.3, but that resistance follows the supply.
- Inverter
The input drives both gates, so gmn and gmp add. They work into 800 kΩ: a gain of 231.9.
Read the diagram as text
- Driver M1: NMOS 10/1, about 10 µA
- Diode load: gain 1.51
- Triode load: gain 20.3
- Inverter: gain 231.9
- A ratio of two transconductances: driver over load
- A triode resistance: 127 kΩ
- Two transconductances that add: into 800 kΩ
- Driver M1 → Diode load: stage A
- Driver M1 → Triode load: stage C
- Driver M1 → Inverter: stage E
- Diode load → A ratio of two transconductances: set by sizes
- Triode load → A triode resistance: set by the supply
- Inverter → Two transconductances that add: both devices amplify
Each load has a cost: low gain, a narrow input range, or a current set by the supply.
- A diode-connected load
The gate of M2 is tied to its drain. Such a device is always in saturation. Seen from its source it looks like a resistor of 1/(gm2 + gmb2).
- Bias
With 0.65 V on its gate, M1 draws 9.7 µA through M2. The output rests at 0.859 V.
ID = 9.7 µA
- Signal
M1 turns vin into a current gm1·vin. That current flows in the small resistance of M2, so the gain is about gm1/(gm2 + gmb2).
gm1·vin
Read the circuit as text
- M2 NMOS transistor (W/L = 2/1): gate →
VDD, drain →VDD, source →out - M1 NMOS transistor (W/L = 10/1): gate →
n1, drain →out, source →0 - v_in voltage source (0.65 V): + →
n1, − →0 - out port: terminal →
out
out= 0.859 V
Both transistors carry the same current.
| Load | gm1 | gm of load | gmb of load | gm1/gm2 | With gmb, gds | AC run |
|---|---|---|---|---|---|---|
| NMOS diode 2/1, body at ground | 168.1 µS | 94.8 µS | 15.0 µS | 1.77 | 1.509 | 1.512 |
| PMOS diode 4/1, own well | 164.3 µS | 69.9 µS | none | 2.35 | 2.310 | 2.319 |
Both at 0.65 V of input. The last two columns agree within 0.4 %.
The body effect costs stage A 15 % of its gain. The output is the source of M2, and its body stays at ground.
Stage B of the lab uses a PMOS diode load. It sits in its own well, tied to its source, so it has no gmb term. Its gain is the ratio of an NMOS gm to a PMOS gm, so the two mobilities enter as well as the sizes. The lab measures 2.32 for stage B.
Why It Is Linear and Low
- A: diode load
- C: triode load
- D: inverter
- Input voltage 650 mVA: diode load = 859.4 mV · C: triode load = 1.216 V · D: inverter = 1.426 V
At the bias point the slopes are 1.51, 20.4 and 177, in the order A, C, D.
- Input voltage 700 mVA: diode load = 776.7 mV · C: triode load = 53.7 mV · D: inverter = 22.67 mV
Only 50 mV later the triode stage sits at 54 mV and the inverter at 23 mV. Both have lost their gain.
- Input voltage 850 mVA: diode load = 474.8 mV · C: triode load = 17.8 mV · D: inverter = 751.4 µV
The diode stage still amplifies. Its slope is 1.98 at 0.75 V and 2.11 at 0.85 V, while its current rises from 39 µA toward 124 µA.
Read the plot as a table
| Input voltage | A: diode load | C: triode load | D: inverter |
|---|---|---|---|
| 400 mV | 1.121 V | 1.799 V | 1.8 V |
| 500 mV | 1.033 V | 1.783 V | 1.798 V |
| 600 mV | 927.9 mV | 1.63 V | 1.774 V |
| 650 mV | 859.4 mV | 1.216 V | 1.426 V |
| 700 mV | 776.7 mV | 53.7 mV | 22.67 mV |
| 800 mV | 580.1 mV | 22.77 mV | 2.232 mV |
| 850 mV | 474.8 mV | 17.8 mV | 751.4 µV |
| 900 mV | 371 mV | 14.68 mV | 261.6 µV |
| 1 V | 235.1 mV | 10.98 mV | 30.87 µV |
More gain means a smaller input range.
In the square law, gm = √(2·µCox·(W/L)·ID). Both devices of stage A carry the same current, so the current cancels in the ratio. The gain is √((W/L)1/(W/L)2) = 2.24, less the body effect.
The run agrees once M1 is in strong inversion. From 0.75 V to 0.9 V the gain stays between 1.98 and 2.11. At the 10 µA bias M1 is in moderate inversion, and the gain is 1.51.
The gain is low because it grows only with the square root of the size ratio. A load a quarter as wide (0.5/1) raises the gain from 1.51 to 3.28. The output then rests 0.19 V lower, because the narrow load needs more gate voltage.
The Triode Load
Stage C uses a PMOS with its gate at ground. It is always on and works in the triode region, as a resistor made from a transistor.
| Stage C at 0.65 V | Value |
|---|---|
| Current | 10.0 µA |
| Drop across the load | 0.584 V |
| Small-signal resistance ron of the load | 127 kΩ |
| gm1·(ron ‖ ro1) | 19.8 |
| Gain from the AC run | 20.3 |
The formula uses the operating-point values and lands 2.4 % under the run.
A poly resistor of 127 kΩ would need about 390 squares of the film in E9.3. This load is one small transistor.
The cost shows in the transfer plot. The load resistance depends on VDD − |Vth| and on the process. The gain also collapses once M1 leaves saturation, 50 mV above the bias point.
The Inverter as an Amplifier
- Self-bias
No DC current flows in RF, so the input rests at the output voltage. The stage settles where its transfer curve crosses Vout = Vin: 0.653 V and 10.1 µA.
I = 10.1 µA
- Both devices amplify
A rising input makes M1 pull down harder and M2 pull up less. Both changes move the output the same way, so the transconductances add.
(gmn + gmp)·vin
Read the circuit as text
- M2 PMOS transistor (W/L = 13/1): gate →
in, source →VDD, drain →out - M1 NMOS transistor (W/L = 10/1): gate →
in, drain →out, source →0 - R_F resistor (100 MΩ): pin 1 →
out, pin 2 →in - C_C capacitor (1 µF): pin 1 →
in, pin 2 →n1 - v_in voltage source: + →
n1, − →0 - out port: terminal →
out
in= 0.653 Vout= 0.653 V
The same two transistors as a logic inverter, held at the steep point of the curve.
In E3.3 the PMOS is a current source with a fixed gate, and only the NMOS amplifies. Here the input drives both gates.
| Stage E at 1.8 V | Value |
|---|---|
| gmn | 173.2 µS |
| gmp | 116.8 µS |
| ron ‖ rop ‖ RF | 800 kΩ |
| (gmn + gmp)·(ron ‖ rop ‖ RF) | 231.9 |
| Gain from the AC run | 231.9 |
The PMOS adds 67 % to the transconductance for no extra current.
Supply Sensitivity
Read the plot as a table
| Supply | gain |
|---|---|
| 1.5 V | 190.3 |
| 1.6 V | 214.9 |
| 1.7 V | 230.9 |
| 1.8 V | 231.9 |
| 1.9 V | 220.3 |
| 2 V | 202.7 |
| 2.1 V | 184.2 |
The gain moves far less than the current in the first figure.
| Supply | Bias current | Gain, AC run |
|---|---|---|
| 1.62 V | 2.33 µA | 219.0 |
| 1.80 V | 10.10 µA | 231.9 |
| 1.98 V | 28.50 µA | 206.4 |
A supply change of ±10 %.
The supply is the only thing that sets the two gate drives, so the current follows it steeply. With more current gm rises and ro falls, so the gain changes little.
The current sets the power, the bandwidth and the noise, and here all three move with the supply. The mirror load of E3.3 takes its current from a reference.
Asked in interviews
What is the gain of a common-source stage with a diode-connected load?
About −gm1/gm2, a ratio of two transconductances. With equal currents the square law gives √((W/L)1/(W/L)2). A mobility ratio joins it if the devices differ in type. An NMOS load with its body at ground loses more: the gain is gm1/(gm2 + gmb2).
Why is that gain low?
It grows with the square root of the size ratio. A higher ratio also needs a larger gate voltage on the load, which uses up the output range. Here a load four times narrower raised the gain from 1.51 to 3.28.
Why does an inverter have more gain than a current-source-loaded stage at the same current?
The input drives both gates, so both devices amplify. The gain is (gmn + gmp)·(ron ‖ rop). Here the PMOS adds 67 % to the transconductance.
What is the weak point of the inverter amplifier?
Its bias current is set by the supply alone. Here ±10 % of supply moves the current from 2.3 µA to 28.5 µA. Supply noise also reaches the output, because the PMOS source is the supply.
Try It
Part 1 sweeps the input and reads each stage at 0.65 V. It also finds where the inverter's output equals its input.
Part 2 builds each gain from the operating-point values: gm, gmb and gds. Part 4 measures the same gains with an AC run.
Part 3 sweeps the supply of the self-biased inverter. No input source is needed: RF finds the bias at every supply value.
Predict before each change. Tie the body of XM2A to oa: where does gain_a land? Set RF to 1meg: what happens to gain_180?
Run the netlist to draw its plots here.
id_a = 9.714265e-06 id_b = 9.432985e-06 id_c = 1.000640e-05 vo_a = 8.594430e-01 vo_b = 5.505675e-01 vo_c = 1.215707e+00 vo_d = 1.426403e+00 v_trip = 6.533675e-01 i_trip = 1.009704e-05 gm1 = 1.680568e-04 gm2_n = 9.483418e-05 gmb2_n = 1.495895e-05 ratio_a = 1.772112e+00 formula_a = 1.508621e+00 gm1_b = 1.642973e-04 gm2_p = 6.991455e-05 ratio_b = 2.349973e+00 formula_b = 2.310358e+00 r_on = 1.271655e+05 formula_c = 1.983443e+01 i_162 = 2.334870e-06 i_180 = 1.009936e-05 i_198 = 2.849787e-05 gmn = 1.731723e-04 gmp = 1.168213e-04 r_out = 7.998393e+05 formula_e = 2.319483e+02 gain_a = 1.512321e+00 gain_b = 2.318505e+00 gain_c = 2.031848e+01 gain_180 = 2.318911e+02 gain_162 = 2.190223e+02 gain_198 = 2.063695e+02Check Yourself
Common Mistakes
Wrong Write the gain of an NMOS diode-loaded stage as gm1/gm2.
The source of the load moves while its body stays at ground, so gmb2 adds to gm2.
Fix Use gm1/(gm2 + gmb2). In the lab the plain ratio is 1.77 and the measured gain is 1.51.
Wrong Raise the bias current of a diode-loaded stage to get more gain.
Both transconductances rise together, so their ratio stays the same in strong inversion.
Fix Change the size ratio. In the run the gain stays between 1.98 and 2.11 while the current triples.
Wrong Drive a self-biased inverter amplifier straight from a DC source.
The source then sets the input level, and a few millivolts of error times the gain pushes the output to a rail.
Fix Couple the signal through a capacitor. In the lab, 0.650 V on the input leaves the output at 1.43 V, not 0.653 V.
VIN ie 0 DC 0.65 AC 1
RF oe ie 100megVAC src 0 DC 0 AC 1
CC src ie 1u
RF oe ie 100megRemember
- A diode-loaded stage has a gain of gm1/(gm2 + gmb2): a ratio, set by sizes, low and linear.
- An NMOS diode load with its body at ground loses gain to the body effect: 1.77 became 1.51 here.
- A triode load is a small resistor-like transistor: gain 20.3 here, over a narrow input range.
- A self-biased inverter has a gain of (gmn + gmp)·(ron ‖ rop): 231.9 here.
- The supply alone sets the inverter's current: 2.3 µA to 28.5 µA over ±10 %.