A distributed hierarchical multiprocessor system for data acquisition and filtering in high energy physics experiments
M. Weymann, P. Farthouat, González, M, J.P. Dufey, M. Passaseo, P. Gällnö · CERN Document Server (European Organization for Nuclear Research) · 1992
OutputA tree of crates can be constructed by using this structure recursively.VME-out' architecture allows each processor to exploit the full bandwidth of its input channel.the local VIC 8250 via VSB and read the data in the remote crate via VME.This 'VSB-in are then accessible by other processors via VME.A processor located in another crate can use read data in parallel via their 'private' VSB channel into the dual ported memory.Those data From these units a system can be built in which one or several FIC's in one VME crate timeout generation.trigger signals in a VME crate; it provides hardware support for deadtime measurement and ported between VME, VSB and VMV.The third board, CORBO (also from CES) controls the or slave on VMV and VME and slave only on VSB.Its 512 KByte buffer memory is triple interconnects VME crates via the VMV cable bus, a precursor of the VIC bus.It can be master receive mailbox interrupts for interprocessor communication.The VIC 8250 (also from CES) VME and the microprocessor's local bus, and a DMA controller.Hardware FlFOs are used to VME (master/slave) and VSB (master only).It has a 4 MByte memory, dual-ported between troller, from CES, Geneva), is built around a Motorola 68020 or 68030 and has access to both Our system is built on 3 types of boards.The processor board, a FIC (Fast Intelligent Con 2. Hardware architecture in [1].recording and online monitoring, but this part is not covered here; a brief survey can be found achieved and large buffering space is available.A VAX 6310 is used for run control, data trol the read-out and to format the events data.A high degree of parallelism in the read-out is This paper describes a distributed multiprocessor system, implemented in VME, to con systems for development or test purposes is also a requested feature.namely a low intrinsic dead time and a high data throughput.Standalone operation of the sub acquisition system be easily and quickly adaptable while providing also good performances, ing detector configurations and development of new readout equipments require that the data implemented in various standards (CAMAC, FASTBUS, VME and VSB).Frequently chang tor devices (Silicon micro-strips, RICH, MWPC, Drift chambers ) with read-out electronics SPS accelerator, accommodates some 3-4 experiments each year which use a variety of detec The Omega spectrometer, a multiuser facility for fixed target expeiiments at the CERN 1. Introduction specific top layer and two lower layers that are common to all subsystems.VME crates linked via VICbus.It is implemented in three software layers: a subdetector July 1991 at the Omega spectrometer at CERN.The system is distibuted over several at the root of the tree.Based on this model a data acquisition system is in operation since as they pass successive levels of standard nodes until a complete event is finally recorded is presented.Event data, generated by front-end nodes, are combined and possibly Hltered A model of a data acquisition system with a tree-structured network of readout nodes