[evla-sw-discuss] Real Time Clock for MIB

[Barry Clark]

I'm under the impression that we can live with the on-chip timers, without
an external part.  The situation is not like the VLBA, where the station
may need to know the time to point the antenna, without being able to ask
what time it is via the internet.  I might reconsider if the on-chip 
timers can't do what I suggest below.

We would like a clock which can, independently of any CPU activity,
count time with reasonable resolution (perhaps of order one microsecond).
The clock overflow time should exceed one day.  The CPU must be able
to set the clock.  If possible, this should not be reset by system resets,
though power off and loss of LO are, I think, acceptable disruptions.

In addition, for scheduling within the RTOS, we need a timer that provides
an interupt reasonably frequently, say 100-1000 times per second.  vxWorks
certainly requires such a thing, and if something else does not, then
I think we will probably want to write the equivalent service ourselves.

In addition, it is desirable, but not required, to have a utility timer
to provide interupts at a programable interval different from the system
timer above.  (vxWorks use this for debugging facilities - it is very 
handy.)

And there must be a facility for the system heartbeat (19.2 Hz or whatever)
to cause an interupt.