Add new integral function block for determining approximate integral over time. #648
Conversation
Test Results 110 files ±0 110 suites ±0 48s ⏱️ ±0s Results for commit 054e1c9. ± Comparison against base commit b2b0d82. This pull request removes 2 tests. |
| </EventOutputs> | ||
| <InputVars> | ||
| <VarDeclaration Name="IN" Type="REAL" Comment="Input value" InitialValue="0.0"/> | ||
| <VarDeclaration Name="TM" Type="DINT" Comment="Time since last call in msec" InitialValue="0"/> |
There was a problem hiding this comment.
I would prefer TM to have type TIME to better indicate the meaning of this value. But I fully understand the implications this has on the integration as soon as people start to use values in nanoseconds...
| IF (1.0E38 - IN * DINT_TO_REAL(TM) < OUT OR -1.0E38 - IN * DINT_TO_REAL(TM) > OUT) THEN | ||
| OVERFLOW := TRUE; | ||
| ELSIF (TM > 0) THEN // Time Difference since last call must be positive. | ||
| OUT := OUT + IN * DINT_TO_REAL(TM) / 1000.0; |
There was a problem hiding this comment.
When testing this in the Interpreter I got the following case:
- send RESET
- set IN to 9.0E9, TM to 1 send REQ; OUT goes to 8999999.0
- set the input to 1.0, send REQ; OUT does not change
- set the input to -1.0, send REQ; OUT also does not change
this is caused by the variable resolution of the REAL datatype - a warning for the case where the integration does not work (i.e. OUT does not change with IN <> 0) could be nice-to-have
There was a problem hiding this comment.
you are totally right:
our colleges at CODESYS have same issue with same Function.
this comes from the internal representation of REAL.
maybe we should also make a LREAL Version, as well a Fixed-comma Version in DINT.
i will think about it, and in the MEantime make the warning as written above.
|
Seeing the comments and the implications for the correct operation of this Function Block, I don't think this should be written as a SFB in ST at all, but as a SiFB, which can do all the relevant checking and error signalling on the system level. |
|
@MartinMelikMerkumians i totally agree to you. and then step a big Moment back, go deep, take time, and make the improved Version with ANY_NUM, agree ? |
No description provided.