Error
recovery, and experiment timing in GREAT TDR.
In the
event of failure during an experiment,
detected by the Event Collator, or the experimentalist.
Assumptions:
·
The
Master Timestamp counter is in the Metronome.
·
The
Master Timestamp counter is maintained during power failure. ( how long ? )
·
The
Master Timestamp counter can be read via the VME interface.
This
failure can be anything from a Crate power-cycle to a single module timestamp
counter being out of step with the Master Counter.
The timestamp counter of the failed
portion will need to be loaded with a current value at a pre-determined time.
The master counter in the Metronome
module is read ( T ) . A suitable time offset ( To ) is added, and the new value (Tn) is written to the counter in the
failed portion of the system.
Tn is finally written to a
comparison register in the Metronome. When the Master counter reaches Tn, it
issues a second pulse on the SYNC line delayed from the true sync pulse by 16
counts.
On recognition of the second sync
pulse those counters which need to re-load will.
The Event
collator is where an timestamp fault error can be detected. The response to the
discovery of an error is to mark the data from the faulty source ( disable it
), inform the Experimental control software,
and downstream equipment. The Experimental control software is
responsible for correcting the fault ( re-load counter/ power-cycle/etc. ), and
subsequently informing the Collator that the fault has been corrected. Then the
Faulty source will be re-enabled, and any downstream equipment can be informed
the source is back on line again.