Converter and Bridge Models

Converter and Bridge Models

Every converter chooses its own bridge model. The network connection and external bridge ports stay the same, so a case can change fidelity without rewiring the system, and different units in one network may use different models.

units:
  vsc_1:
    bridge: {model: switching}
  vsc_2:
    bridge: {model: pwm_averaging}

When bridge.model is omitted, the file default is pwm_averaging.

Model comparison

Model	Electrical bridge	Digital timing	Typical cost
switching	Ideal switch states at exact carrier-comparison instants.	ADC, controller interrupts, computation completion and PWM register loads.	Highest.
pwm_averaging	Active duty ratios applied continuously between register loads; no carrier ripple.	Same sampled controller and PWM-register schedule as switching.	Lower than switching.
averaging	Ideal controlled voltage source driven by continuous controller equations.	No ADC, controller-interrupt, computation or PWM-register schedule.	Usually lowest for control-envelope studies.

The models answer different questions. Similar low-frequency trajectories do not make them numerically identical.

Switching

switching evaluates the carrier comparison and changes the ideal bridge state at the exact edge time. Every edge becomes an integration boundary, so solver stages before and after it use the correct bridge state.

Use switching when the study needs:

• switching ripple or harmonics;
• carrier phase and synchronization effects;
• exact device-state transitions;
• fast protection behavior tied to switching or register-load instants.

The fixed step is a maximum step: PESLite shortens an interval when an event, output point, controller interrupt or PWM edge occurs first. A smaller simulation.solver.dt improves the continuous-state integration between those boundaries, at additional cost.

PWM-period averaging

pwm_averaging removes the carrier waveform from the electrical bridge, but it does not remove the digital control system. The controller still samples through the ADC, computes on its interrupt grid and writes a shadow duty register. The active register loads at the configured carrier instants, and the bridge holds that active duty until the next load.

This model therefore retains:

• controller sampling and computation delay;
• active and shadow PWM-register semantics;
• single or double PWM updates;
• asynchronous or synchronous timer placement;
• startup, protection and event timing.

Use it when digital timing matters but switching ripple does not. It is generally the best default for controller and system-level studies with sampled control.

Ideal averaging

averaging represents the bridge as an ideal controlled voltage source. Measurements and controller outputs are evaluated at solver stages, and controller states join the continuous plant state vector.

There is no ADC window, controller-interrupt grid, computation wait, carrier, or active/shadow PWM register in this mode. Sampled-only settings may remain in a shared file and are silently ignored, which allows the same case to be run under any of the three presets.

Use ideal averaging for:

• control-envelope and electromechanical dynamics;
• long simulations where switching detail is outside the study bandwidth;
• adaptive integration with error tolerances;
• fast parameter sweeps where continuous control is the intended model.

Do not use it as a substitute for switching when conclusions depend on ripple, carrier timing or sampled-data delay. Continuous feedback can also create algebraic loops; PESLite detects connected feedback variables and solves them internally at each derivative evaluation.

Command presets

The three command-line flags are configuration presets rather than simple display options:

Preset	Bridge model	Solver preset
--switching	switching for every unit	fixed-step RK4
--pwm-averaging	pwm_averaging for every unit	fixed-step RK4
--averaging	averaging for every unit	adaptive DP45

An explicit --set has higher priority. For example, a run can select the averaging preset while overriding its solver choice:

peslite case.pes --averaging \
  --set simulation.solver.type=fixed \
  --set simulation.solver.method=rk4

Without a preset, each unit keeps the model written in the file, so mixed-model systems are supported.

Accuracy and comparison practice

Compare models on quantities within their common bandwidth and use identical events, references, initial values and output times. A useful workflow is:

1. establish the control behavior with ideal averaging;
2. repeat with PWM-period averaging to include sampled control and register timing;
3. use switching for the intervals where ripple and edge behavior must be verified;
4. reduce solver step or tolerances until the conclusion is insensitive to further refinement.

The output interval controls saved snapshots, not the integrator step. A visually smooth CSV does not by itself demonstrate numerical convergence.

Selection summary

Question	Recommended starting model
Switching harmonics, ripple or carrier interaction	switching
Sampled controller and PWM delay without ripple	pwm_averaging
Continuous control and long-duration dynamics	averaging
Mixed-detail network	Select the model independently for each unit

For initial setup and command priority, see Getting Started.