HOME       UP       PREV       NEXT (Event Driven Simulation)  


Simulation of real-world systems generally requires quantisation in time and spatial domains.

There are two main forms of simulation modelling:

Finite-element simulation is used for analogue and fluid-flow systems.

It is rarely used in SoC design (just for low-level electrical propagation and cross-talk modelling).

Finite-element difference equations:
   tnow += deltaT;
   for (n in ...) i[n] = (v[n-1]-v[n])/R;
   for (n in ...) v[n] += (i[n]-i[n+1])*deltaT/C;
Basic finite-element simulation uses fixed spatial grid (element size is deltaL) and fixed time step (deltaT seconds).

Each grid point holds a vector of instantatious local properties, such as voltage, temperature, stress, pressure, magnetic flux.

Physical quantities are divided over the grid. Three examples:

  1. Sound wave in wire: C=deltaL*mass-per-unit-length, R=deltaL*elasticity-per-unit-length
  2. Heat wave in wire: C=deltaL*heat-capacity-per-unit-length, R=deltaL*thermal-conductance-per-unit-length
  3. Electrical wave in wire: C=deltaL*capacitance-per-unit-length, R=deltaL*resistance-per-unit-length

Larger modelling errors with larger deltaT and deltaL, but faster simulation. Keep them less than 1/10th wavelength for good accuracy. Generally use a 2D or 3D grid for fluid modelling: 1D ok for electronics.

Typically want to model both resistance and inductance for electrical system.

When modelling inductance instead of resistance, then need a `+=' in the i[n] equation.

When non-linear components are present (e.g. diodes and FETs), SPICE simulator adjusts deltaT dynamically depending on point in the curve.

23: (C) 2008-12, DJ Greaves, University of Cambridge, Computer Laboratory.