Lab 6
The Two-Stage OTA
Verify the course OTA of Chapter 12 with the double-injection probe over corners and temperature, see how the Miller capacitor and the nulling resistor set the phase margin, and why a naive loop break misleads.
Contents
| Lab 6 at a glance | |
|---|---|
| Goal | Verify the course OTA of Chapter 12 with the double-injection probe over corners and temperature, see how the Miller capacitor and the nulling resistor set the phase margin, and why a naive loop break misleads. |
| Reproduces | Chapter 11 (Figures 11.3–11.5 and 11.8) and Chapter 12 (Sections 12.7–12.8) |
| Time | 3–4 hours (the corner run takes a few minutes, longer in SKY130) |
| Files | labs/lab6: ota2s.sch and ota2s.sym (the OTA), ota2s_tb, comp_sweep_tb, break_tb (.sch) per PDK |
Hand calculation
Chapter 12 chose \(g_{m2}\) = 157 µS, \(C_c\) = 1 pF, \(I_5\) = 20 µA and \(I_7\) = 60 µA. Predict the GBW, \(g_{m2}/(2\pi C_c)\), and the two slew-rate limits, \(I_5/C_c\) and \(I_7/(C_L + C_c)\). Then use Eq. (11.9) to find the overshoot of the unity-gain follower for a phase margin of 60°.
Steps
- The OTA.
ota2s.schis the OTA of Chapter 12 with the same six ports as Lab 5’s. Its compensation network uses the values'Cc'and'Rz': they are parameters, and every testbench sets them with a.paramline in itsNGSPICEblock, so they can be swept without editing the OTA. - Verification. Run
../../sim.sh ota2s_tb.sch. Copiesxavandxaicarry the two halves of the double-injection probe of Figure 11.3 (a series voltage source and a shunt current source at the output); theNGSPICEblock combines them with Eq. (11.7) into the loop gain and prints its DC value, \(f_u\) and the phase margin. Copyxttakes a 0.3 V step for the slew rate (20–80 % of the step). - Corners. Run
../../corners.py ota2s_tb.sch --temps=-40,27,125: 15 runs, one table. Find the worst corner for each specification and compare with Section 12.8. - Compensation. Run
../../sim.sh comp_sweep_tb.sch. Withalterparamandresetinside aforeachloop it repeats the loop-gain and step measurements for eight values of \(C_c\) with \(R_z = 1/g_{m6}\), and then for six values of \(R_z\) with \(C_c\) = 1 pF, and prints a line per value. Plot the overshoot against the phase margin (Figure 11.5b) and the phase margin against \(R_z\) (Figure 11.8b). - A naive break.
break_tb.schis the gain-of-2 amplifier of Figure 11.4 three times: with the probe, with an LC break at the OTA output and with an LC break at the inverting input. Run it and compare the three loop gains.
Checkpoint
Table 18.7. Lab 6 results (CL = 2 pF, tt, 27 °C unless stated). Corner ranges: tt, ss, ff, sf and fs at −40, 27 and 125 °C.
| Quantity | SG13G2 | SKY130 |
|---|---|---|
| Loop gain: DC value, \(f_u\), phase margin | 62.2 dB, 21.5 MHz, 67.0° | 75.7 dB, 21.7 MHz, 72.8° |
| Slew rate rising / falling, power, offset | 17.0 / 15.6 V/µs, 111.6 µW, −0.13 mV | 16.5 / 14.6 V/µs, 161.7 µW, −0.09 mV |
| Corner ranges: DC gain, \(f_u\), phase margin | 60.9–63.1 dB, 15.9–27.4 MHz, 65.9–68.6° | 73.1–76.6 dB, 17.2–26.2 MHz, 69.4–77.7° |
| Corner ranges: falling slew rate, power | 14.4–16.4 V/µs, 110–113 µW | 13.6–15.3 V/µs, 160–163 µW |
| \(R_z = 1/g_{m6}\): phase margin and overshoot at \(C_c\) = 0.33 / 1 / 2.2 pF | 24.2°, 50.6 % / 60.0°, 6.8 % / 76.2°, 0.0 % | 33.1°, 37.7 % / 63.4°, 5.4 % / 77.2°, 0.0 % |
| \(C_c\) = 1 pF: phase margin at \(R_z\) ≈ 0 / 1 kΩ / 2 kΩ / 5 kΩ | 52.7 / 59.7 / 66.4 / 83.3° | 53.8 / 61.4 / 68.6 / 88.4° |
| Gain-of-2 amplifier, DC loop gain and phase margin: probe / break at output / break at input | 45.6 dB, 76.0° / 56.2 dB, 74.7° / 45.6 dB, 80.1° | 54.5 dB, 80.9° / 69.6 dB, 79.6° / 54.5 dB, 82.0° |
Questions
- Both slew rates are below the 20 V/µs of your hand calculation, the falling one by 22 % in SG13G2. Which current limits the falling edge, and which capacitance does it have to charge besides \(C_L\)?
- The break at the output overestimates the loop gain by 11 dB (SG13G2) and 15 dB (SKY130). Which impedance does it get wrong, and why is the break at the input right at DC but wrong in phase at high frequency?
- The SG13G2 phase margin rises by 31° when \(R_z\) goes from 0 to 5 kΩ. Why not use 10 kΩ?
Extension
Redesign the OTA for a GBW of 40 MHz at the same phase margin, as Section 12.9 asks. Before you simulate, use Eqs. (12.1)–(12.5) to predict how much current the change costs; then edit ota2s.sch, rerun ota2s_tb and the corner sweep, and compare.