Lumi System DAQ upgrade

After involving the fast beam monitoring (HERA request,June 96) the VME BUS has realy reached its saturation (500 additional events are readed and processed per second). The problem is not only a direct exeeding the VME bandwidth, but too complicated and long arbitration sequensies appeare in the BUS. As a matter of fact the seven CPU masters are standing now in one crate(4*FIC + 2*MACVEE + SPARC) and making asynchronious VME access with using DMA and reading big data from slow responsed units (scalers). Currently for the CPU with last priority approximatly an every hundred's VME access is finished with "Bus Error" condition. This means that these often VME BUS conflicts are happened due to arbitration errors in case of simultenious bus access from highest and lowest priority masters. The biggest asynchronious data flow comes from FADC crate where both DAQ and LUMI subtasks are making independent readout. Unfortunatly this independent readout can not be arranged due to differ type of events are needed for each subtask. One of the possible solution removing these simultenious access conflicts could be VME BUS separation for LUMI and DAQ subtasks. For this purpose a new system configuration was tested during 96/97 shut down. The basic idea was to move the LUMI FADC and other connected with lumi task moduls in to the separate standard and FADC VME crates with delegating the all needed communication between LUMI and DAQ brunches to VSB FIC - DPM connection. The stability tests with specialy written for new configuration software and with 2 metre VSB cable connection after some adjustment bring to an assured result. A complete
  • Lumi DAQ Upgrade was not performed during short 96/97 shut down basicaly due to very late delivery time for new FADC VME crate (~ begining of May 97). All other solutions were suggested (with the same or less profit, but more expensive) require much more both hardware and software changes, that would be dangerous to do for such complicated system. It is preferable also an idea of the second FADC crate due to current one is completly full and for future HERA-LUMI upgrade some spare slots will be realy usefull.

    Here is list of problems what we have now and what should be solved after upgrade:

    - Increased probability of the data corruption that were observed several times during 96 data taking.

    - A small instability in the one unit can cause a big overal instability with difficult procedure of fixing a real problem

    - Current LUMI subbranch dead time on 20% higher due to this bandwidth saturation

    - Absolutely impossible any future readout extension

    - Increased probability of wrong interrupt generation

    - Total computer power limitation due to slow data exchange and often BUS ERROR handling.