Lab 3 · SG13G2Devices in weak inversion

Goal. Characterise the SG13G2 devices you will design with: gm/ID lookup tables, flicker noise against gate area, and the HBT against the MOSFET as an input device. Labs 4 and 5 reuse the tables. Goes with Chapters 5 and 6.

Run. With Docker and the container of the set-up:

python labs/spice/sg13_char.py

The script writes a .spiceinit that loads the PSP and resistor models, generates the netlists, runs them in the container with ngspice -b and parses the results (Chapter 30). It runs at the typical corner and 27 °C:

  1. gm/ID sweeps of sg13_lv_nmos and sg13_lv_pmos, W = 10 µm, L = 0.5, 1 and 4 µm, |VDS| = 0.6 V;
  2. the input-referred noise of both devices at |ID| = 1 µA for W/L = 10/1, 40/4 and 100/10 µm;
  3. IC and IB against VBE of the npn13G2 HBT at VCE = 0.6 V, and its input-referred noise at IC = 1 µA.

The results go to figsrc/data/sg13_gmid.csv, sg13_noise.csv, sg13_hbt.csv and sg13_hbt_noise.csv, the files that the Chapter 5 figures are drawn from. If your container reproduces them, your set-up is right.

Expected results.

Quantity Value Where in the text
Slope factor n; gm/ID in deep weak inversion 1.22; 31.7 V⁻¹ Section 5.1
Specific current Ispec per square, NMOS / PMOS 0.78 / 0.20 µA Section 5.1
Thermal noise at 1 µA in weak inversion 18.5 nV/√Hz Section 5.2
Flicker coefficient K, NMOS / PMOS 1.6 × 10⁻¹⁰ / 5.4 × 10⁻¹⁰ V²·µm² Section 5.3
Gate area per device for 1 µVrms unchopped, NMOS about 1800 µm² Section 5.3
HBT at 1 µA: gm/IC, β, input noise 37.7 V⁻¹, about 680, 15.1 nV/√Hz flat Section 5.4

In these models the NMOS has less flicker noise than the PMOS, the opposite of the usual rule, which is why Open-BioAFE uses NMOS input devices.

Questions.

  1. Build a lookup function gm_id(ic) from sg13_gmid.csv and use it to size an NMOS input pair for gm = 30 µS at 1 µA per device. What W/L comes out at L = 1 µm and at L = 4 µm, and what is the gate capacitance of each?
  2. Plot the measured input noise against 1/(W·L) at 1 Hz. Is it a straight line? Extract K yourself and compare with the table.
  3. At what source resistance does the HBT’s base shot noise make it noisier than the MOSFET? Compare with the electrode impedances of Chapter 3.
  4. Add the ss and ff corners and 85 °C to the script. How much do n, Ispec and K move? Which number in your design is most sensitive?
  5. From the PDK’s mismatch coefficients (AVT = 3.9 mV·µm for the NMOS), find the input-pair offset for the 1800 µm² devices, and say why chopping makes it irrelevant (Chapter 9).

Deliverable. gm/ID lookup tables for both devices at three lengths, flicker noise against gate area, and a one-paragraph choice of input device for the LNA of Lab 4.

All labs · Previous: Lab 2. Electrodes and interference · Next: Lab 4. The Harrison LNA