Closed henrikt-ma closed 4 months ago
Would it make sense to replace the
ContinuousClock
source blocks with something more differentiable?
Would Modelica.Blocks.Sources.RealExpression(y=time)
be of rescue?
Would it make sense to replace the
ContinuousClock
source blocks with something more differentiable?Would
Modelica.Blocks.Sources.RealExpression(y=time)
be of rescue?
It would save System Modeler, but let's not worry about that; we'll fix that on the tool side, as decided (but still not noted) in the last MAP-Lib release preparation meeting.
Closing, as it was concluded that this is a tool issue. I'll open another issue (#4325) regarding the lack of excitation in the second order derivative.
This affects many models, but let's talk about
Modelica-4.1.0/regression/Tables/CombiTable1Ds/Test3
to make things concrete.After
Modelica.Blocks.Sources.ContinuousClock
was modified in https://github.com/modelica/ModelicaStandardLibrary/pull/4150, the model no longer builds in System Modeler, since the input to the table isn't consider sufficiently differentiable to compute the second order derivative of the output.Before the change to
ContinuousClock
, the input to the block was a piecewise smooth signal, and by not switching between theif
-equation during continuous-time integration, there was no lack of smoothness.Would it make sense to replace the
ContinuousClock
source blocks with something more differentiable?