We have performed a successful test of the integration of independent DAQ systems based on a correlation
scalar counted from a common clock signal. The principles of the integration method are described in the
following AIDA elog entry: https://elog.ph.ed.ac.uk/AIDA/29
The configuration of the setup we used in the RIKEN test is described by the attached figure:
DAQcorrelation_diagram.pdf. The chosen solution was to use a 25 MHz clock generated by the MACB modules of
AIDA, which was distributed to the other two systems by a SIS36/38xx module in the BRIKEN VME crate (a
slightly modified versio of the method proposed by PCS in AIDA Elog#29).
The correlation can be monitored online using the MIDAS DataXfer and DataSpy toolkits
(http://npg.dl.ac.uk/MIDAS/DataAcq/Data.html). Each system operates a DataRelay program that sends a data
stream consisting only of IDs and time-stamps to a common DataSink (this requires a DataRelay code that
does
some filtering of the raw data). The scheme is shown in the attached diagram (DAQsoftware.pdf).
The DataPeek_Merge and SyncCheck codes use C++11 for parallelization of tasks, which is a helpful feature
to
improve the efficiency of finding coincidences between the correlation pulses. Thus, the code will not
run
in PCs with any Linux version. For example, it requires version 7 of ScientificLinux, which installed in
the
PC being used for BRIKEN DAQ control that will stay at RIBF for the time being.
++ SYNCHRONIZATION RESULTS ++
The figure correlation_25Hz.png shows a correlation plot for the BRIKEN+RIBF+AIDA running with
correlations
done by a pulser at 25 Hz. A large fraction of the pulsers appear in the double-coincidence plots (top
row;
same plots, except the 'Partial' one is automatically cleared every ~10seconds).
The bottom plots show the pair-wise correlations between BRIKEN and RIBF or AIDA. BRIKEN was defined as
the
'master' in this test, but current version of the program is flexible to select any stream as the master
one.
The second figure (correlation_random.png) shows the same test, but using a non-periodic pulser to trigger
the correlation scalar. We observed the same level of synchronization (peaks look broader because of the
zoom level). |