I think the behavior that Rolf was questioning is about the way ALH
stops annunciating the alarm state. To register as an alarm, the PV
satisfies the criteria that it transitions to the ALARM state the
prescribed number of times within the prescribed time.
To exit that alarm registration, would logically seem to require a
different condition, such as no (or fewer than the prescribed number)
transitions to ALARM state. If the PV simply continues the same actions
that got ALH to register an alarm, it doesn't seem logical that
de-registering an alarm should occur when the PV continues to do what
caused the original registration of an alarm. Moreover, the ongoing
toggling of the alarm state of the record will once again cause ALH to
register a new alarm, even though nothing about the record behavior has
Does someone rely on this behavior for some use case, or would it be
considered as incorrect behavior?
Rolf posed this question in the context of writing an ALH work-alike,
and the first order requirement is that it duplicate the behavior of
ALH. In this instance, I question whether that means duplicating what is
essentially a bug.
On 05/28/2017 02:52 AM, Andreas Lüdeke wrote:
> I have an alarm PV which needs no acknowledgement and is configured
with an input count filter:
$ALARMCOUNTFILTER 3 8
> The alarm PV changes from "in alarm" to "out of alarm" once/second
for a long time.
I see the following behaviour in alh:
6 seconds after the alarm starts toggling, the PV goes into alarm.
After another 6 seconds, the PV goes out of alarm.
The behaviour is as intended, if the PV changes once per second its state.
Because you've specified the filter "3 8":
it goes in the alarm state, if the alarm either stays for 8 seconds,
or if it enters an alarm state three times within 8 seconds.
This PV enters an alarm state every two seconds, so after six seconds it raised an alarm the third time.
If you would change the filter to for example
$ALARMCOUNTFILTER 5 8
then it would never go into alarm (if the PV sticks to its behaviour).
When it would stop toggling and stays instead into alarm, an alarm would be raised after 8 seconds.
If it would change its state more frequent, an alarm would be raised as well.
I hope that helps to clarify the ALARMCOUNTFILTER.
Cheers from Switzerland :-)
- Re: alh Andreas Lüdeke
- Navigate by Date:
RE: Configuring multiple ports with variable port numbers for StreamDevice Mark Rivers
Use case for ao record HOPR != DRVH or LOPR != DRVL? J. Lewis Muir
- Navigate by Thread:
Re: alh Andreas Lüdeke
epicsqt Siddons, David