gmid_IHP130

This repository contains GM/ID extraction data and related interactive scripts to get started with gm/id-based design using the IHP130 PDK. It enables users to perform circuit simulations, extract key transistor parameters, and generate insightful plots quickly and efficiently.

---

Setup

1. Clone the Repository:

   git clone https://github.com/chennakeshavadasa/gmid_IHP130.git
   cd gmid_IHP130

2. Set up a virtual Environment

   python -m venv venv
   venv\Scripts\activate # On Windows
   # source venv/bin/activate # On Linux/macOS

3. Install Dependancies

   pip install -r requirements.txt
   

---

Overview

This script is designed to process and visualize gm/ID design data for various transistor types (e.g., HV_NMOS, LV_PMOS). It allows designers to analyze device behavior across lengths and optimize operating points using interactive plots and numerical optimization.

---

Key Steps

1. Data Parsing and Structuring

The raw .txt files contain electrical characteristics for multiple transistor lengths. The get_data() function extracts and cleans this data for further analysis.

def get_data(data: List, key: str) -> pd.DataFrame:
    ...
    for i, file_path in enumerate(data):
        with open(f'{gmid_data}/{key}/{file_path}', 'r') as file:
            ...
            filtered_entry = [val for i, val in enumerate(entry) if i < 3 or i % 2 == 0]
            ...

All transistor lengths are labeled using the labels list (e.g., '0.13u', '0.25u', ...).

---

2. Feature Engineering

Once the raw data is loaded, additional performance metrics are computed:

• Vov (Vgs - Vth) – Overdrive voltage
• gm/ID – Transconductance efficiency
• gm/gds – Maximum Voltage Gain
• id/W – Current Density
• ft – Transit frequency
• Capacitance ratios – Cgd/Cgg and Cgs/Cgg

df['Vov'] = df['Vgs'] - df['Vth']
df['gm/id'] = df['gm'] / df['id_val']
df['gm/gds'] = df['gm'] / df['gds']
df['ft'] = df['gm'] / (2 * np.pi * df['Cgg'])

---

3. Interactive Plotting with Plotly

The plot_interactive() function creates responsive graphs across transistor lengths for various electrical metrics:

plot_interactive(df, 'Vgs', 'gm/id', 'Vgs', 'gm/id', f'{df["Device"].unique()[0]} gm/id versus Vgs')
plot_interactive(df, 'gm/id', 'gm/gds', 'gm/id', 'gm/gds', f'{df["Device"].unique()[0]} gm/gds versus gm/id')
plot_interactive(df, 'gm/id', 'id/W', 'gm/id', 'id/W', f'{df["Device"].unique()[0]} id/W versus gm/id')

Each trace represents a different transistor length, color-coded for clarity. Log-scaling and SI formatting are included for better readability.

---

4. Operating Point Optimization

Using SciPy's optimization, the script finds the device operating point closest to the target design metrics:

• Target gm/ID (e.g., 15)
• Target gm/gds (e.g., ≥ 100)
• Weighted error minimization

def objective_function(x):
    error_A = w1 * (x[0] - gm_id_target) ** 2
    error_B = w2 * (x[1] - gm_gds_target) ** 2
    return error_A + error_B

Optimization result:

result = minimize(objective_function, x0, method='L-BFGS-B', bounds=bounds)

The closest matching data row is retrieved:

df['distance'] = np.sqrt(
    w1 * (df['gm/id'] - optimized_gm_id) ** 2 +
    w2 * (df['gm/gds'] - optimized_gm_gds) ** 2
)
closest_row = df.loc[df['distance'].idxmin()]

Final output includes key performance metrics:

print("\nClosest Matching Row:\n", closest_row[['Device', 'Length', 'gm/id', 'gm/gds', 'id/W', 'ft']])

---

Plots

• LV-NMOS (1.65V Device)

  

• LV-PMOS (1.65V Device)

  

• HV-NMOS (3.3V Device)

  

• HV-PMOS (3.3V Device)

  

License

Apache License 2.0