70 likes | 84 Views
Learn about integrating acquired subsystems effectively. Case study of Bob Dalesio from 11/29/98 regarding RFQ vacuum control using Labview, Modicon PLC, and EPICS. Insights on manpower usage and system complexities.
E N D
Integrating Acquired Subsystems Bob Dalesio 11/29/98
Vacuum Subsystem - RFQ, RFQ Windows • Controls the vacuum system for the RFQ and RFQ windows • Developed at Lawrence Livermore Laboratory • Los Alamos controls involved from the beginning • Livermore developed in Labview and Modicon PLC, Granville Phillips Ion Gauge Controllers - estimated 6 months to learn EPICS • By agreement, the PLC contained all vacuum interlocks and all automatic startup was done in Labview • Los Alamos converted over to EPICS
Vacuum System Manpower Usage • LLNL • 6 months to collect requirements and do project engineering • 3 months to implement the vacuum system • LANL • 5 weeks to develop the Modbus driver • 4 weeks to integrate the GPIB driver from SLAC • 1 week to develop IP drivers and carrier board support - based on K.Kasimer’s work • 3 weeks to develop screens (more extensive set - including interlock displays) • 3 weeks to develop the database • 3 days to test and integrate • new requirements for the RFQ IOC to provide an interlock to HPRF and monitor strain gauges were done in EPICS • Drivers were used for window vacuum, and injector vacuum - no additional work
High Power RF • Developed by Contental • No meetings with LANL controls engineers • Done using Allen-Bradley PLCs with PanelMate displays • Multiple master provided for control channels (PLC/EPICS) • Allen-Bradley driver modifications to support monitoring outputs and asynchronously processing records to read the new values
HPRF - notes • I do not have the time estimates… • The multiple master control of outputs significantly complicated the database design and the driver • Signals were displayed on the Panel Mate in one set of units, were in different units at the variables we were given - again - complication in the database and requirement for save/restore. • Signals in the Panel Mate were used as scratch pad registers between Panel Mate update cycles and EPICS would see these intermediate values. • Complex logic for HPRF control and interlocks were in the PLC - producing a 3 inch listing that was time consuming to take over • Control channels from EPICS did not work initially - they were not implemented in the PLC • Each change to the PLC took at least a week - no support for integration
RFQ cooling • Provided by Allied Signal • Implemented in EPICS • Delivered and tested by Allied • Not yet in the same directory structure • Integration - took the time to transfer some displays over to the operator account • Database implementation took 9 months - estimated 6 of them to train a new user and have him come up to speed • Industry Pac drivers took 3 months - not communicating here.
Subjective Conclusions • Poor project controls make it difficult to get accurate measures of time spent in each activity. • Communication with the provider in the earliest stages to agree on implementation strategy simplifies the entire task and especially the integration time. • It is not always reasonable to ask a vendor to implement in EPICS - the time to come up to speed should not exceed the time saved. • Do not put too much in a PLC - they are difficult to maintain for complex logic. • Require integration as part of the result from the supplier. • Even when selecting common hardware - coordinate the schedule for implementing the driver to avoid duplication of effort (and code to maintain)