Too many orbits in DMA packets in GMT data
This manifests with the warnings of the kind:
WARNING StreamProcessor::process: expected DMA trailer word deadbeef, found 7604f. Orbits per packet count 0x15, > expected, (0x14) skipping packet.
Examples in: /eos/cms/store/cmst3/group/daql1scout/run3/hexdumps/scout_GMT_555555_000007_PASSTHROUGH.hex
, e.g the packet that ends at line 08f2d620
which has 22 orbits.
We notice that the s_orbits_seen
in the does increment by 20 i.e the expected 0x14. Dropped orbit counter does not increment. Our suspicions are some edge case in the fifo filler i.e how robust is using d_ctrl_d1.valid = '0' and d_ctrl_d2.valid = ‘1’
to mark the orbit boundaries.
I have placed the outputs of re-running scdaq on the pass_through captures here:
/eos/cms/store/cmst3/group/daql1scout/run3/hexdumps/scdaq_output_passthrough_gmt_555555_000007.log
/eos/cms/store/cmst3/group/daql1scout/run3/hexdumps/scdaq_output_passthrough_calo_666666_000000.log
The calo one has an extra debug output that prints the orbit number, to help find the dodgy packets.