Speaker
Description
Summary
The Detector Control System in general is divided into three layers, Supervisory
Layer, Control Layer and Field Layer. The Supervisory Layer is the top-level with
the operator's user-interface. The Control Layer is a communication layer, and the
Field Layer is where the different hardware devices are found. All the components
in the system work in parallel, feeding the operator with useful information
concerning the status of the system, or responding to commands given at the top-
level. The DCS for the TPC Front-end electronics is designed to act upon errors as
close as possible to the source. Functional blocks are distributed over a larger
number of computing nodes and sub-nodes.
The system described is based on so called Front-End-Electronics-Servers
(FeeServers). A FeeServer abstracts the underlying Front-end electronics to a
certain degree and covers the following tasks:
- Interfacing hardware data sources and publishing data
- Receiving of commands for configuration and controlling the Front-end electronics
- Self-test and Watchdogs (consistency check and setting of parameters)
A dedicated communication software, the InterComLayer, handles the connection
between the lower Field Layer and the upper Supervisory Layer. The main tasks are
controlling the FeeServers in the Field Layer, sending configuration data to the
Front-end electronics and transporting monitoring data to the Supervisory Layer. The
InterComLayer implements three interfaces, a Front-End-Electronics client, a
Front-End-Device server and a client to the Configuration Database. These interfaces
are common abstraction layers within the ALICE experiment.
The FeeServer and other controlling software in the Field Layer runs on the DCS
board, an autonomous single-board computer which allows running complex controlling
software under the operating system Linux. Further custom hardware devices have been
developed covering specific tasks and serving as sub-nodes. The Readout Control
Unit (RCU) hosts basic controlling functionality for a set of FECs. Some tasks of
the DCS like the Monitoring and Safety Module are carried out by firmware modules
of the RCU.
In total the system consits of 216 embedded Linux nodes. Together with standard
computers in higher control layers the DCS board and the RCU board form a
distributed control system.
The architecture has several advantages:
- Distributed and module-based design with well-defined interfaces increases
structure and testability.
- Parallel systems increase bandwidth and reduce workload on each node.
- The system is independent of physical intervention. This is of high importance as
the system is unaccessible when it is in operative mode.
- Linux operating systems on the embedded computers provides flexibility and
standard tools.
- Software and Firmware that is easily reconfigurable.
- Low-level devices with intelligent error-handling decrease the possibility for
permanent failures.
The system is still under development. Several small scale control systems
consisting of the DCS board and applicable sub-nodes has already been used as stand
alone systems in major tests. The modularity makes it possible to test and review
each sub-system on it's own independently of the complete setup. Several
integration and beam-tests have been performed with satisfying results. The talk
will focus on experiences and results from the major integration test in May 2005.
