EMU Front-end Electronics T.Y. Ling March 5_ 1998

Document Sample
EMU Front-end Electronics T.Y. Ling March 5_ 1998 Powered By Docstoc
					                                   FED Software
• Current features
    – Software thread waits for VME Interrupt from DDU
          • This does the work during normal running
          • DDU sets VME Interrupt when it detects an error
          • Diagnostics are performed as needed, sent to log file
    – GUI gives access to counters & expert-level functions
          • Mostly for debugging, should not be needed by shifters
          – No manual tasks are required before or during a run
    – XML initialization (GbE Prescale, channel mask)
          • SOAP compatible in case we want to use it at LHC, but…
          • Really not needed at LHC; we can use preloaded flash settings
                – FED crate is ready after Hard Reset using only TTC commands
• Operation at MTCC was stable and reliable

Jason Gilmore         CMS Week Emu Meeting, 5 Dec 2006                      1
                            FED Software plans
• Need to test multi-FED Crate operation
     – We will have 4 FED Crates in the full system
     – Use a common crate control PC with 4 PCI slots
• VME Interrupts & DDU diagnostics
     – DDU sets Interrupt any time Reset is required
           • Either Sync error or a critical data error
           • Crate software starts diagnostic cycle after Interrupt
     – Where shall we save Interrupt & Diagnostic reports?
           • Currently goes to common Log4Cplus file
                – Can we have an Emu-specific log?
• Error state sent simultaneously to FMM
     – FMM/TTS may respond with Reset (too) quickly
     – May want to set FMM errors from Software layer
           • Allows time for diagnostic cycle to complete (~100 usec)
           • Also allows for global CSC view in Reset logic

Jason Gilmore        CMS Week Emu Meeting, 5 Dec 2006                   2
                              DDU/DCC at MTCC
• DDU & DCC performed well
    – Any Emu problems detected at FED were reported to FMM
    – Operator-error related problems occurred several times
    – Don’t kill the messenger; the real cause is elsewhere
• Problems detected at MTCC from several sources
    – ALCT Timeout at DMB, likely a P.C. setup/software issue
          • P.Crate(n) goes bad when P.Crate(n+1) is configured: Martin?
          – Usually happens within ~100 events of run start, not fixed?
    – TMB event mix-up with close L1As (< 8 BX)
          • TMB has poor overlapped-event handling, sent DAV but no data
          • Occurs when TMB is set for 7 BX wide coincidence window
          • Changed L1 trigger rules to avoid problems, needs a better fix
    – TF–to–DDU interface is not as mature, still minor problems
          • Rare DDU-to-Slink “Bad FED CRC” errors, still investigating
          • TF-to-DDU CRC always fails, problem with TF algorithm
                – Needs some more test iterations from UF
Jason Gilmore         CMS Week Emu Meeting, 5 Dec 2006                     3
                       DDU/DCC at MTCC, cont.
 • High-rate test (2 x LHC): problems resolved
      – Rare DDU errors, only at high data rate
            • DDU error after ~250k events
            • For comparison, DT error at ~1000 events
            • HCAL error within first few events
      – Reproduced problems on test bench at OSU
            • DDU memory management bug in InputFPGA, fixed
                – Occurred ~4 per million events (running at 2 x LHC rate)
            • DDU InputFPGA read control book-keeping bug, fixed
                – Occurred ~1 per 10 million events (at 10 x LHC rate)
 • Now error-free at 10 x LHC rate for 30 billion events
      – Realistic testing at even higher rates is possible, but
        limited by bursts/fluctuations
            • Require FMM feedback in trigger to prevent buffer overflows

Jason Gilmore       CMS Week Emu Meeting, 5 Dec 2006                         4

Shared By: