MOSFET C–V Extraction Tool (BSIM4, ngspice)

Links

Language	GitHub Pages 🌐	GitHub 💻
🇺🇸 English

This tool automatically extracts gate capacitance (Cgg–Vg) characteristics
from BSIM4 NMOS/PMOS models using ngspice DC analysis.

Key features:

• Per-process-node analysis (e.g., 130nm)
• Independent extraction for NMOS and PMOS
• Temperature sweep (LT = −40°C, RT = 25°C, HT = 125°C)
• Model files separated under models/
• Automatic netlist generation → ngspice batch execution → Cgg extraction → PNG output

📁 Directory Structure

BSIM4_ANALYZER_CV/
│
├── models/
│   ├── nmos130.sp
│   └── pmos130.sp
│
├── template_cv.cir
├── run_cv.py
├── plot_cv.py
├── README.md
│
└── results/
    └── 130nm/
        ├── nmos_130nm_RT.cir
        ├── nmos_130nm_RT.log
        ├── nmos_130nm_RT.png
        ├── pmos_130nm_RT.cir
        ├── pmos_130nm_RT.log
        ├── pmos_130nm_RT.png
        └── … (LT / HT also produced)

🧩 1. template_cv.cir (Analysis Template)

A voltage sweep template used by Python .format() to embed settings.

NMOS (standard)

• Connections: S = D = B = 0V, gate swept
• Sweep: 0V → VDD

PMOS (real device behavior)

• Connections: D = S = B = VDD, gate swept
• Sweep: VDD → 0V
  → Matches physical ON/OFF behavior
  (OFF at Vg = VDD → ON at Vg = 0)

Output: Cgg only

.print dc V(g) @m1[cgg]

❗ Why Only Cgg Is Extracted (Important)

BSIM4 internal capacitances behave as:

• Cgg = ∂Qg / ∂Vg, a physically meaningful total gate capacitance
• Cgs, Cgd, Cgb are partitioned charges (model-dependent), not physical CV values

Problems with partitioned capacitances:

• Cgs + Cgd + Cgb ≠ Cgg
• Cgs/Cgd depend heavily on the charge-partition algorithm
• They may contain unstable values at low Vg

→ Therefore, this tool extracts Cgg only, ensuring physical interpretability.

🚀 2. Netlist Auto-Generation & Batch Execution (run_cv.py)

Run:

python run_cv.py

This performs:

• Auto-generation of NMOS/PMOS × LT/RT/HT netlists
• Automatic batch execution of all 6 cases using ngspice

Update ngspice path if needed:

NGSPICE_CMD = r"C:\Program Files\Spice64\bin\ngspice.exe"

📊 3. Plotting Cgg–Vg (plot_cv.py)

Run:

python plot_cv.py

Processing steps:

1. Read V(g) and Cgg from .print dc
2. Remove index=0 (DC initial non-physical point)
3. Remove Cgg ≤ 0 (solver startup noise)
4. Save PNG to:

results/<node>/<basename>.png

🧪 Example Output Behavior

• NMOS: accumulation → depletion → inversion (U-shaped curve)
• PMOS: reversed sweep (VDD → 0), following real device biasing
• Units: F (farads), matplotlib handles scientific notation automatically

📦 Model Files (models/*.sp)

BSIM4 educational models:

• nmos130.sp
• pmos130.sp

(Not matched to any foundry process; intended for learning & analysis.)

🔧 Adding a New Process Node

Add a node definition inside run_cv.py:

"90nm": {
    "vdd": 1.0,
    "nmos_model_file": "nmos90.sp",
    "pmos_model_file": "pmos90.sp",
    "nmos_model_name": "nmos90",
    "pmos_model_name": "pmos90",
    "lch": "0.09u",
    "wch": "1u",
    "toxe": "1.8e-9",
},

✔ Environment

• Windows 11
• ngspice 41 (64-bit)
• Python 3.9+
• matplotlib 3.x

📘 Summary

This tool provides:

1. Correct terminal conditions for NMOS/PMOS and physically accurate sweep directions
2. Physically meaningful extraction of Cgg only
3. Fully automated batch processing for 6 conditions (NMOS/PMOS × 3 temperatures)
4. Clean separation of models / template / results for easy scaling
5. Complete automation from ngspice → log → PNG

Ideal for device physics education, compact modeling, and process comparison studies.

Possible extensions:

• dC/dV extraction
• Cox estimation
• CV-derived Vth
• Experimental-research mode for Cgs/Cgd/Cgb
• Multi-node comparison plots

■ NMOS C–V Characteristics (130nm, RT)

• Device: NMOS (130nm)
• Temperature: RT (Room Temperature)
• Behavior: Accumulation → Depletion → Inversion
• Extracted Parameter: Cgg vs. Vg

■ PMOS C–V Characteristics (130nm, RT)

• Device: PMOS (130nm)
• Temperature: RT (Room Temperature)
• Sweep Direction: VDD → 0 V (realistic PMOS biasing)
• Extracted Parameter: Cgg vs. Vg

📄 Hybrid License

This project uses a Hybrid License:

👤 Author

📌 Item	Details
Name	Shinichi Samizo
GitHub