Project Possibilities

A. Student-proposed projects

B. Instructor-suggested projects

Overview of the components and structure of a modern, modular, numerical simulator: slides: animated, non-animated

B1: modelling.

B1.1: MNA_EqnEngine and STA_EqnEngine for pyMAPP, assuming pyModSpec models

• background materials:
  • A MATLAB python implementation is provided for reference: springMass1d_EqnEngine implementation.
  • see also the pyModSpec project, below.

---

B1.2: pyModSpec skeleton, with hand-coded implementations of R, C, L, indep. sources, controlled sources, diode, BJT and SH-MOS (without using automatic differentiation). Demonstrated with check_ModSpec.

• background materials:
  • A MATLAB implementation will be provided for reference.
  • CICC paper on ModSpec/MAPP.
  • Overview of devices and ModSpec (slides).
  • ModSpec concepts, step by step: animated slides, non-animated slides

---

B1.3: pyModSpec backend for VAPP, demonstrated to produce correct pyModSpec models from VA.

• background materials:
  • VAPP code will be provided.
  • VAPP poster and overview slides (animated).
  • VAPP guide.
    • Well posed device models for simulation document.

---

B1.4: ME/MEMS equation engine, demonstrated in DC, transient and AC.

• background materials:
  • SUGAR web page: see their user manual, especially.

---

B1.5: RRE equation engine (including devising a good RRE netlist format), demonstrated in DC and transient.

• background materials:
  • RRE slides (animated).
  • A MATLAB implementation will be provided for reference.

---

B1.6: Optics equation engine (including devising a good netlist format), demonstrated in DC and transient.

• background materials:
  • A MATLAB implementation will be provided for reference.

---

B1.7: Multi-physics Master Equation Engine.

• dependencies: EE + at least one other equation engine.
• background materials:
  • A MATLAB implementation will be provided for reference.

---

B1.8: Automatic differentiation for pyDAE and pyModSpec: implement pyvecvalder (and/or use an existing python AD package), demonstrate in pyDAE and pyModSpec.

• background materials:
  • A MATLAB implementation (vecvalder) will be provided for reference.
  • autograd.
  • other python AD packages.
  • Neidinger's intro paper on AD.

B2: analyses and algorithms.

B2.1: Timestep control for transient (LTE and NR failure/iteration based), demonstrated on "difficult" problems without using NR convergence aids.

• background materials:
  • Jian Yao's paper.
  • Local truncation error concepts: go through pages 45,80-90 of these slides.
  • Chua and Lin's chapter on transient and LTE.

---

B2.2: Implement homotopy in pyMAPP, demonstrate on folds and hysteresis.

• depends to an extent on transient timestep control.
• background materials:
  • A MATLAB implementation will be provided for reference (if needed).
  • slides explaining homotopy: animated, non-animated.

---

B2.3: Implement (non-oscillator) shooting in pyMAPP, demonstrate on DAEs.

• background materials:
  • A MATLAB implementation can be provided for reference (but should not be needed).
  • slides explaining shooting: animated, non-animated.

---

B2.4: Implement (non-oscillator) HB in pyMAPP, demonstrate on DAEs.

• background materials:
  • A MATLAB implementation will be provided for reference (if needed).
  • slides explaining HB: animated, non-animated.