Clarification of solver methods for clocked systems

[modelica-trac-importer]

Reported by otter on 21 Nov 2013 12:24 UTC Section 16.8.2 of the Modelica Specification 3.3 about solver methods needs some clarification.

Issues:

• The solver methods are only with respect to Real variables (e.g. (u(i) + u(i-1))/2 cannot be computed for a Boolean).

• The given solver methods are not suited for applications where the input is from a hardware device. In this case the formula can only be evaluated when the signal from the device is available, and this will introduce a delay. Therefore, the formula should be given for different cases.

• The meaning of "u" in the formulas is not clear. For example:

  Real u1 = sample(u1C, Clock(0.1));
  Real u2 = exp(u1);
  Real x;
  equation
  der(x) = u2;

  is "u" in the formula "u1" or "u2" (it is "u1").

---

Migrated-From: https://trac.modelica.org/Modelica/ticket/1359

[modelica-trac-importer]

Comment by otter on 21 Nov 2013 13:32 UTC Replying to [ticket:1359 otter]:

Section 16.8.2 of the Modelica Specification 3.3 about solver methods needs some clarification.

Issues:

• The meaning of "u" in the formulas is not clear. For example:

  Real u1 = sample(u1C, Clock(0.1));
  Real u2 = exp(u1);
  Real x;
  equation
  der(x) = u2;

  is "u" in the formula "u1" or "u2" (it is "u1").

Here is a more precise formulation, based on input from Sven Erik:

A solver method is to be applied on a subclock partition. Such a partition has explicit inputs u marked by sample(u), subSample(u), superSample(u), shiftSample(u) and backSample(u). The arguments of these are to be used as input signals u in the discretization formula. Therefore, in the above example the "u" in the formula is "u1C".

[modelica-trac-importer]

Comment by otter on 24 Nov 2013 15:37 UTC Fixed in r7150 according to the suggestion.