Changes between Version 1 and Version 2 of Ticket #5452, comment 31


Ignore:
Timestamp:
2019-06-25T12:35:58Z (5 years ago)
Author:
Francesco Casella

Legend:

Unmodified
Added
Removed
Modified

  • Ticket #5452, comment 31

    v1 v2  
    77As I understand it, what makes this model different from the ones where omc works is that the current in each of the three conductors is split in two branches. How exactly that is done is irrelevant (the specific diode equations do not enter into the set of constraint equations), what is important is the balance of currents at the connection nodes.
    88
    9 We should investigate why {{{resolveLoops}}} does not figure this out, and extend that algorithm to also handle this case. Then, we should also consider whether to make {{{resolveLoops}}} active by default, as it may be essential in handling some models which are found out there. Unfortunately we can't think of trying it after index reduction has failed, because the failure is not structural (which could be caught by the backend), but rather numerical, and so only visible at runtime, se we'd need to make sure that resolveLoops always works fine.
     9We should investigate why {{{resolveLoops}}} does not figure this out, and extend that algorithm to also handle this case. Then, we should also consider whether to make {{{resolveLoops}}} active by default, as it may be essential in handling some models which are found out there.
     10
     11 Unfortunately we can't think of trying it after index reduction has failed, because the failure is not structural (which could be caught by the backend), but rather numerical, and so only visible at runtime, se we'd need to make sure that {{{resolveLoops}}} always works fine.