Alarm Filtering / Patch for aiRecord

[Bernd Schoeneburg <bernd.schoeneburg@desy.de>]

At the codeathon we discussed a time-based filtering of alarms in the
ioc. Finally a design from Bob Dalesio and Eric Norum became consense. A
time constant "AFTC" controls the smoothing of alarm levels.  Short
peaks of value should not generate alarms, sometimes. Using this filter
can avoid it.

Dear Andrew, Bob, Eric, Ralph and all,
at the codeathon I worked on the implementation of the alarm filtering
in the aiRecord. The handling of the hysteresis inside the filter was
not correct. Now I found a solution to store the inner state of the
hysteresis without using an additional field. The code is more compact
now. The traditional HYST of the record - which is in egu-units and is
applied on the value - is used to make the alarm more "sticky" which is
currently that of the record (now: after filtering).
At the codeathon there was a discussion where (and to what value) to set
the lalm variable. Now, after discussions, I set it to the level of the
alarm after the filtering. This makes sense, I think.

Important: When AFTC is set to zero (default) the record behaves like
normal.

Please have a look on the changes. The patch file is based on the latest
cvs snapshot 16Apr09. Next records to work on will be "bi" and "calc".

- Bernd

Attachment: rec_DIFF_AlarmFilering.gz
Description: GNU Zip compressed data