3.8 High-level Representation (SMI)

The State Management Interface (SMI) [133] describes the various subsystems of the experiment in terms of a set of objects.

Each object has a predefined set of states in which it can be, and for each state a set of allowable actions that can be requested of it. The state of each object is determined either by the state of other objects or, for elementary objects, by the state set in its associated Elementary Process. Similarly, actions requested of an object are either passed on to other objects or to an Elementary Process.

The definition of possible states, allowed actions, and the relationship between objects is made in a dedicated SMI language. In non-elementary objects it allows conditions to be specified which will result in an automatic state change or set of actions. For example, the state of a higher-level object can be determined by the states of lower-level objects, or commands can be issued when a state change occurs.

A group of related SMI objects forms an SMI domain, which is implemented in a single process. Communication between an SMI domain and other domains, or with the Elementary Processes or the user interfaces, is effected using the DELPHI Information Management (DIM) [134] system. This system allows SMI states to be directly viewable by the DELPHI User Interface (see section 3.8.2). It replaces SMI's native communication system [135], providing greater reliability, since it does not require all states to be held by a central server.


3.8.1 Use of SMI in DELPHI

SMI provides the primary high-level control and reporting mechanism for both the Slow Controls and Data Acquisition systems. Each detector partition is mapped onto an SMI domain, which contains an object for each Elementary Process, which oversees a single well-defined subsystem. Some of the possible states of these objects and the actions that can be performed on them are listed in table 3.3. The states of all Elementary Objects in an SMI domain are combined into summary objects SC and, where relevant, LEP_RELATED.

Table 3.3: SMI states (a) and commands (b) of the standard Elementary Process. These correspond to the states and commands of the associated object in the detector partition's SMI domain. For high voltage subsystems, the states STANDBY, CHANGING_LO, ERROR_LO, and OFF indicate that the volts are no higher than the standby level. The states of the LEP_RELATED object (ALLOW_BEAM_CT_DUMP_INJ, DISALLOW_CHANGE, etc.) are set according to the states of the high voltage objects.
(a) SMI state Condition
OFF All channels are off.
HELD_OFF Same as OFF, except that an explicit RELEASE command is required before any control is possible.
ON All are on and OK.
STANDBY All CAEN channels are at their intermediate level.
RUN Same as ON, except that an explicit RELEASE command is required before any control is possible.
CHANGING At least one channel is ramping up or down and at least one is above its STANDBY level.
CHANGING_LO At least one CAEN channel is ramping up or down, and all are at or below their STANDBY levels.
ERROR At least one channel is in error (e.g. reading outside limits or tripped CAEN channel) and at least one other CAEN channel is above its STANDBY level.
ERROR_LO At least one channel is in error, and all CAEN channels are at or below their STANDBY level.
NOT_READY No channels are ramping or tripped, but they are not all at the same stable state (e.g. some on and some off).
NO_CONTROL No communications with the G64, or between G64 and CAEN.
DEAD The Elementary Process is not running.


(b) SMI Action
command  
START Default settings from the Slow Controls Configuration Database are downloaded to the G64, and control channels are switched on.
STANDBY Equivalent to START but sets intermediate values.
REPAIR Equivalent to START/ STANDBY but only for CAEN channels that have tripped.
STOP Control channels are switched off.
MONITOR Performs an additional read of all channel values.
HOLD If the state is ON or OFF, inhibits further commands (either from SMI or HIPE). The SMI state goes to RUN or HELD_OFF.
RELEASE Cancels the HOLD command.
ABORTIT Tells the Elementary Process to commit suicide.


The SC object gives the detector partition's overall status. Its states are summarized in table 3.4a.


Table 3.4: SMI states (a) and commands (b) of a typical detector partition summary object (SC).
(a) SMI state Condition
READY Everything is on and can take data.
NOT_READY One or more subsystems is not ready to take data (e.g. at standby level).
ALARM Unsafe condition (e.g. gas leak).
AL_CANCEL A previous alarm condition, now gone, requires explicit clearing.
ERROR One or more Elementary Objects is in ERROR.
CHANGING High voltages are ramping up or down.
NO_CONTROL No communications with one or more G64s or CAENs.
EP_DEAD One or more Elementary Processes are not running.
DEAD SMI domain is not running.


(b) SMI command Action
Prepare_For_Run All subsystems are STARTed to prepare for datataking.
Respond_To_Background Lowers the voltages of subsystems which are sensitive to `dirty' beam conditions.
Prepare_For_Dump Lowers the voltages of subsystems which are sensitive to conditions following physics beam.
Prepare_For_Injection Lowers the voltages of subsystems which are sensitive to the beam conditions which occur during LEP injection.
Prepare_For_Coarse_Tuning Sets the voltages of subsystems which are sensitive to the beam conditions which occur during LEP coarse tuning.
Switch_On_Fastbus Ensures that all Fastbus crate power supplies are switched on in preparation for datataking.
Prepare_For_Shutdown Switches off all subsystems.
Clear_Cancelled_Alarms Clears a previous alarm condition, now gone, subsequently allowing apparatus to be brought back into operation.
Set_Central Switches the detector partition to central operator control.
Set_Local Switches to local control.


Since LEP activity (such as injection or coarse tuning) can produce a significant number of stray particles in the detector, it is advisable to reduce the high voltages of the more sensitive partitions during this time. The LEP_RELATED object shows the state of these high voltages, and can be used to ensure that they are all lowered before giving LEP the go-ahead for the operation.

Conditions in the ancillary gas and GSS systems (see section 3.9) relevant to each detector partition are relayed to that partition's SMI domain and can be used to switch off voltages when a serious condition is indicated. They can also contribute to the partition's SC summary state, giving the possibility of an ALARM state.

The summary states for each detector partition are relayed to a central SMI domain, which composes overall SC and LEP_RELATED summary states for DELPHI.

The main commands used for the control of each detector partition are summarized in table 3.4b. These commands can be received by the partition from one of two sources: during data taking, they normally come from the central operator via the central SMI domain; during setting up, they (and other commands designed for the control of specific detector partitions) are issued by detector experts from a local SMI display. A switch from local to central control and vice versa is provided so that central switch-on commands can be inhibited during the intervention of detector experts.


3.8.2 SMI Display

Operator display and control is provided by the DELPHI user interface (DUI) [136] to SMI. DUI is a general-purpose X-Windows/Motif graphical user interface. It is used to show information as varied as the Slow Controls statuses and the LEP collimator positions. DUI interfaces naturally to SMI due to the latter's use of the DIM system.

The SMI display program may be used to inspect and, when necessary, issue commands to individual detector partitions by local operators, or to the whole of DELPHI by the central operator.

The central Slow Controls SMI display, shown in figure 3.5,

Figure 3.5: A typical screen from the SMI display. The LEP-sensitive high voltages are currently lowered (hence the states indicating that any LEP activity (beam coarse tuning, dump, or injection) is permitted). The local SMI for the Outer Detector (OD), obtained by selecting the OD SC button, is shown at the top right. The command menu for all detector partitions, obtained by selecting the top CMD button, is shown bottom right with the Prepare_For_Run command (see table 3.4b) selected.
\includegraphics[width=\textwidth]{smi-display-94}
allows the operator to see the summary states of the central SMI, of each of the detector partitions, and of various ancillary systems. Details of the component states of each partition or ancillary system can also be presented from this display; an example is also shown in figure 3.5. The display can be used to send commands to all or part of DELPHI, or (where authorized) to an individual object within a detector partition. The available global commands are similar to those for an individual detector partition (table 3.4b); an example menu is also shown in figure 3.5.

A help facility is also available from the display to give advice to the operator on the diagnosis and cure of problems occurring in each detector partition.


3.8.3 Big Brother

After several years of operator experience, it was decided to further automate the system in order to provide the fastest-possible response to changes in the state of the LEP machine, Slow Controls, and Data Acquisition. This became possible once a reliable determination of the LEP machine condition was available to the DELPHI SMI [137]. It fitted smoothly into the existing software since the Data Acquisition controls are also based on SMI.

This system, dubbed Big Brother [138], was implemented entirely in SMI during the 1994 run. It introduces the following automatic actions based on correlations between the three hitherto independent systems.

In order to maximize the amount of usable data taken, the run is only paused for Slow Controls problems that would seriously impair the subsequent data reconstruction or analysis. Similar conditions are used at the start of a fill to determine when to start datataking. There are a number of conditions which, from the point of view of the Slow Controls, are considered errors (and yield, for example, an ERROR state) but for which datataking remains nonetheless profitable. In order to distinguish these cases, a set of RUN_RELATED SMI objects, in parallel to the ones described in section 3.8.1, are defined for the relevant detector partitions (ID, TPC, OD, HPC, STIC, and Trigger Partition). These can differ from the normal SMI states at the Elementary Process level (to distinguish, for example, a high voltage trip of a single wire from that of an entire sector) or at the detector SMI level (for example, temperature warnings, while requiring expert intervention, rarely affect or are affected by the state of the Data Acquisition, and so should not cause a pause of the run). The detector RUN_RELATED states can be excluded from the global RUN_RELATED state by the operator. This allows the run to be resumed if the error condition is determined to be less serious than the SMI state indicates.

As well as speeding detector operations (particularly when one of the operators is temporarily absent from the control room -- for example performing checks in the cavern), with a consequent improvement in the overall datataking efficiency, this system has the additional advantage of standardizing the conditions that determine whether data is is taken.

Prerecorded, digitized, audio messages are used to keep the shift crew aware of Big Brother's actions, to request confirmation of high voltage increases, and to alert sleepy operators to serious problems such as safety alarms, high LEP backgrounds, or Slow Controls or Data Acquisition errors. Inevitably the controlling process has been called `Big Sister'.

Tim Adye 2002-11-06