Experimental Physics and
Industrial Control System

[email protected] (Tim Mooney) · Thu, 12 Mar 1998 10:27:54 -0600

re...

> The essence of the mechanism needed is a state transition from "ready" to
> "busy" synchronously as the request is posted so that subsequent state
> queries cannot temporarily see a "ready" and be misled. Multi-record
> solutions tend to produce race windows and get out of synchronization.

The transition to "busy" need not be prompt if your completion detector looks for a
*transition* from "busy" to "ready" instead of looking only at the level of the
signal.  The wait record is great for this because it can be instructed to execute
its output link only on a transition.  (The full list of possibilities is "Every
Time", "On Change", "When Zero", "When Non-zero", "Transition To Zero", and
"Transition To Non-zero".)

We've been using the wait record in a primitive completion-detector database that
logically OR's all relevant busy signals, and signals completion when the result
makes a transition to zero.  This simple-minded completion detector fails when one
busy signal makes 0-1-0 transitions before another signal has made its 0-1
transition, but that's just because it uses a logical OR instead of counting 1-0
transitions as you plan to do.  (We don't count transitions because we want to
support scans of compound motions programmed by users at run time, so we can't
know in advance how many actual devices are being scanned.  So far, the benefits
have been well worth the problems, but it would be nice to minimize the
problems, and we're looking at ways to do this.)

Tim Mooney

Experimental Physics and Industrial Control System

Experimental Physics and
Industrial Control System