To maintain precision in continuous-time simulation, you should update local data (continuous or discrete) only during physical events at major time steps.

In Stateflow charts, physical events cause state transitions. Therefore, write to local data only in actions that execute during transitions, as follows:

• State exit actions, which execute before leaving the state at the beginning of the transition

• Transition actions, which execute during the transition

• State entry actions, which execute after entering the new state at the end of the transition

• Condition actions on a transition, but only if the transition directly reaches a state

  Consider this Stateflow chart.

  

  In this example, the action {n++} executes even when conditions c2 and c3 are false. In this case, n gets updated in a minor time step because there is no state transition.

Do not write to local continuous data in during actions because these actions execute in minor time steps.

This rule applies to continuous-time charts because you cannot call functions during minor time steps. You can call Simulink functions in state entry or exit actions and transition actions. However, if you try to call Simulink functions in state during actions or transition conditions, an error message appears when you simulate your model.

For more information, see Using Simulink Functions in Stateflow Charts.

A Simulink model reads continuous-time derivatives during minor time steps. The only part of a Stateflow chart that executes during minor time steps is the during action. Therefore, you should compute derivatives in during actions to give your Simulink model the most current calculation.

This restriction ensures smooth outputs in a major time step because it prevents a Stateflow chart from using values that may no longer be valid in the current minor time step. Instead, a Stateflow chart always computes outputs from local discrete data, local continuous data, and chart inputs.

This restriction prevents mode changes from occurring between major time steps. When placed in during actions, conditions that affect control flow should be governed by discrete variables because they do not change between major time steps.

The presence of input events makes Stateflow charts behave like a triggered subsystem and therefore unable to simulate in continuous-time. For example, the following model generates an error if the Stateflow chart uses a continuous update method.

To model the equivalent of an input event, pass the input signal through a Hit Crossing block as an input to the continuous chart, as in this example.

When a mode change occurs during continuous-time simulation, the entry action of the destination state indicates to the Simulink model that a state transition occurred. If inner transitions are taken, the entry action is never executed.

Do not use event-based temporal logic. Use only absolute-time temporal logic for continuous-time simulation. See Operators for Absolute-Time Temporal Logic for details.

Event-based temporal logic has no meaning because there is no concept of a tick during a continuous-time simulation.

In continuous-time simulation, the during action of a state updates the outputs. A chart with no state produces no output. To simulate the behavior of a stateless chart in continuous-time, create a single state which calls a graphical function in its during action.

To implement change detection, Stateflow software buffers variables in a way that affects the behavior of charts between a minor time step and the next major time step.