Re: Advice sought on fanning data to several PVs

[mooney@aps.anl.gov (Tim Mooney)]

re...

> We use GPIB to access a DVM whose voltage translates into PEP-II total Ring current. Up to now we 
> have used only one number, along with the Epics timestamp from the associated PV. We have provided
> adequate device support to access the single number of interest.
> 
> This DVM is now to be multiplexed for two rings, delivering four doubles on each read: Beam 
> Current, Time of reading (ms since some arbitrary 0), Other Beam current, Other time of reading.
> We can fetch this data with modified device support (into an internal ascii buffer), but now must 
> decide how to get it into four seperate data.

I guess I'd write device support to connect four AI records and one BO record to the DVM.
The BO record would trigger each reading and would be the only record actually to talk
to the DVM.  Device support would hold the four-part result and dole out the appropriate
numbers to the appropriate AI records when they ask for them.  The AI records would have
very simple (almost "soft") device support and would really be connected more to the array
of results stored in device support than to the DVM.

devXxSkeletonGpib.c appears to have all the hooks required to support this application.
If you're based on that code, the BO would invoke a GPIBREAD operation and would require
a custom conversion routine to take the result and stash it where the AI device support
could find it.  The AI's would invoke GPIBSOFT operations, which would just grab the stashed
data and return.  The doc on all this skeleton code is at
http://www.aps.anl.gov/asd/controls/epics/EpicsDocumentation/HardwareManuals/GPIB/gpib.960325.html

Tim