From tigwiki

[edit] Background

By writing unassembled events, we can approach the maximum throughput in data rate in the DAQ, which is set by the limit at which data can be read through the VME backplane. This (for pio mode) is 3 MB/s. Theoretically, if we use multiple frontends reading multiple backplanes, we can increase this data rate by a factor of n. In order to do this, the frontend program has had to be modified. These changes are at the moment only available in tigcol3 for the sherekhan test daq.

[edit] Starting Clients

The frontend can be run in multiple mode by:

sherekhan@lxdaq17$ ./fetigcol -i 1

and

sherekhan@lxdaq18$ ./fetigcol -i 2

This -i switch sets the frontend index, allowing the same code to be used for each frontend, and also single and multiple frontend modes. The various collector cards have to be told which frontend should be responsible for them, and this is done using the Frontend Index key in the odb. See the odb settings images for more information.

[edit] ODB settings

In order to store odb settings for the different collectors and TIG10s, a settings tree has to be written to the odb. For a single frontend (without the -i switch) this is /Equipment/Trigger/settings. For the multiple frontend mode, each frontend generates it's own /Equipment/TriggerN/ tree, which records statistics. In addition, the settings subtree is stored in /Equipment/Trigger01/settings for multiple frontend mode.

The included images show the correct settings for a system in which frontend 1 is responsible for the master and slave 1, and frontend 2 is responsible for slave 2. In this particular example, ports 7 and 8 are enabled in slave 1 and 7 and 9 in slave 2.

The new odb tree entry /Equipment/Trigger01/settings/(master,slave1,slave2,etc)/Frontend Index tells the frontend program which VMIC will be treating that particular TIGC card.

With these setup parameters, the MIDAS status window now looks like the image included on the right. Note, each frontend reports it's statistics under TriggerN.