Lab 7

Noise and Monte Carlo

Measure the input-referred noise of both course OTAs and compare it with Eq. (13.8); measure the random offset of 200 copies of the 5T OTA and of a version with a four times larger input pair.

Contents
Lab 7 at a glance  
Goal Measure the input-referred noise of both course OTAs and compare it with Eq. (13.8); measure the random offset of 200 copies of the 5T OTA and of a version with a four times larger input pair.
Reproduces Chapter 13 (Sections 13.4 and 13.6–13.8, Figure 13.6a, Worked example 13.2)
Time 2–3 hours (each Monte Carlo testbench holds 1200 transistors; SKY130 takes a few minutes)
Files labs/lab7: ota5t, ota5t_4x, ota2s (.sch and .sym), noise_tb, mc_offset_tb, mc_offset4x_tb (.sch) per PDK; offset_stats.py

Hand calculation

Evaluate the thermal part of Eq. (13.8) for the 5T OTA with \(g_{m1}\) = 320 µS, \(g_{m3}/g_{m1}\) = 0.63 and the γ values of Section 13.4 (SG13G2: γ1 = 0.77, γ3 = 1.44). Then evaluate Eq. (13.14) for its random offset with the Pelgrom constants of Table 13.3 and the device sizes of Table 10.1. Predict what quadrupling the area of the input pair does to the pair’s term, and to the total.

Steps

  1. Noise. Run ../../sim.sh noise_tb.sch. Both OTAs sit in open loop with their DC loops closed by 1 GH inductors, as in Lab 5, and a noise analysis from 1 Hz to 100 MHz refers each output noise back to its input. The testbench prints the input-referred density at 1 kHz, 100 kHz and 1 MHz; setplot noise1 and plot inoise_spectrum loglog show Figure 13.6(a) for the 5T OTA.
  2. Offset. mc_offset_tb.sch holds 200 copies of the 5T OTA, each a unity-gain buffer with its own 10 µA reference, all driven by one input at \(V_{ic}\), with the mismatch models and a fixed seed. Run it (one operating point gives 200 offsets), then python3 ../offset_stats.py.
  3. Four times the area. mc_offset4x_tb.sch is the same with ota5t_4x, whose input pair has twice the width and twice the length. Run it and offset_stats.py again.

Checkpoint

Table 18.8. Lab 7 results (tt, 27 °C; Monte Carlo with the seeds in the testbenches).

Quantity SG13G2 SKY130
5T OTA, input-referred noise at 1 kHz / 100 kHz / 1 MHz 261 / 29.2 / 15.5 nV/√Hz 323 / 48.5 / 23.8 nV/√Hz
Two-stage OTA, input-referred noise at 1 kHz / 100 kHz / 1 MHz 428 / 47.7 / 25.2 nV/√Hz 370 / 56.0 / 31.3 nV/√Hz
5T OTA offset, 200 copies: mean, σ −1.09 mV, 1.65 mV −4.48 mV, 2.28 mV
The same with a 4× area input pair: mean, σ −1.01 mV, 0.90 mV −3.72 mV, 1.57 mV

With a different seed your σ will differ by a few per cent: 200 samples give σ to about ±5 % (Section 13.8), which is also the difference between these values and those of Worked example 13.2.

Questions

  • Quadrupling the input-pair area halved the SG13G2 spread only to 0.55 of its value and the SKY130 spread to 0.69. Which device sets the remaining offset in each PDK, and what would you change next?
  • The two-stage OTA is noisier than the 5T OTA at every frequency in SG13G2, though its input pair is four times larger. Use Eq. (13.8) to explain why: look at the current per side and at the length of its PMOS mirror.
  • The mean offset of the 200 copies equals the systematic offset of Lab 5 within a fraction of the spread. Why is that a useful check on a Monte Carlo run?

Extension

Make a third variant of the 5T OTA with the pair at \(g_m/I_D\) = 20 and the mirror at 6 V⁻¹ at the same \(g_{m1}\) (Figure 13.6b), size it with the tables as in Lab 2, and measure its noise and offset. Then run mc_offset_tb.sch with ota2s instead of ota5t and compare its σ with the 1.33 mV and 4.08 mV of Section 13.7.


Back to top

SkillSurf · The Art of Analog IC Design, draft v0.1 (November 2026). Every source is listed on the References page.

© 2025–2026 SkillSurf. Text, figures and 3D models: CC BY 4.0; code: MIT. The SkillSurf name and logo are not covered.

This course was written with the help of AI tools and may contain mistakes. If you find one, please tell us through SkillSurf or email skillsurfedu@gmail.com, and we will correct it.

This site uses Just the Docs, a documentation theme for Jekyll.