Insufficient precision when storing sine phases

[massimo ceraolo]

1) Load Modelica.Blocks.Sources.Sine or Modelica.Electrical.Sources.Sine
2) Click on the model icon and choose as phase 90 (they are degrees)
3) Click again on the icon: the phase is now 90.002.

This is because phases are stored in the model in radians as 6-digit strings corresponding to simple precision.

I think that more should be used. I understand that it is not optimal to use the full double-precision for all model numbers, since .mo readability would reduce. I would not like to see 1.00000000000001 just to say 1.
Probably a good compromise is to convert internal binary representation in decimal form using, say, 10-12 digits, and truncating final zeros, whenever present.

[Francesco Casella]

There is an option in Simulation Setup|Output to decide whether you want to store single- or double-precision results. What is the best choice largely depends on the application.

[massimo ceraolo]

This does not apply to this ticket

[Francesco Casella]

@ceraolo, sorry, I didn't read it carefully enough...

The trick is to use at least one more significant digit in the code than in the parameter window.

[massimo ceraolo]

To avoid too large loss of precision we should use many digits, with the consideration of readability I mentioned I the ticket description that may suggest some reduction from the maximum. So I would prefer at least 10 digits

With this tickets example OM writes (radians):
1.5708
While dymola writes
1.5707963267949

If I remember well double precision translates into around 16 meaningful decimal digits. It this is true, I can guess that Dymola's programmer chose 14 to give a very high precision, while still being sure that roundoff errors do not affect the last shown digit. Maybe this is the right choice?

[Francesco Casella]

Sounds good. In fact, when dealing with transcendental numbers, precision is never enough.

[massimo ceraolo]

the issue is not present anymore: it must have been solved some time ago.