90 likes | 204 Views
Global Track Trigger Technical. James Ferrando Glasgow For the GTT group. Outline. Latency and deadtime in 2004 Component data sizes in 2004 Component data latency in 2004 Trigger significance Hardware modifications Software plans and studies Summary and Outlook.
E N D
Global Track TriggerTechnical James Ferrando Glasgow For the GTT group J.Ferrando
Outline • Latency and deadtime in 2004 • Component data sizes in 2004 • Component data latency in 2004 • Trigger significance • Hardware modifications • Software plans and studies • Summary and Outlook J.Ferrando
GTT latency and deadtime in 2004 • Latency at GSLT is: • stable for GFLT output rates seen, • within CTD-SLT latency envelope, and • has a lower mean latency than the CTD-SLT • Deadtime does not appear to be introduced by the GTT, but… • Further study needs, per run, component GFLT deadtime numbers excluding external vetos. • Data source naming convention: • CTD = CTD axial+stereo data • CTDZ = CTD z-bz-time data • MVD= MVD data • MVD0,1, and 2 = upper, lower, and wheels • STT = STT data J.Ferrando
GTT component data sizes in 2004 • Data size cut at component interface used to control latency. If size exceeds cut, no data sent for event. • CTD and CTDZ both cut at 10kB • No event data is lost. • CTDZ data size not shown. • MVD cuts at 2kB • not all data sent - cut defined during bad background running in 2002. • renewed efforts to include MVD hits into barrel and forward algorithms mean: • the cut will gradually be increased, and • the MVD group will have to control bad channel readout. • STT cut at 6kB • more problematic than MVD as peak of distribution is not at low end. J.Ferrando
GTT component data latency in 2004 • Data ready latency at GTT • CTD and CTDZ latencies driven by CTD side TP data transfer and not by the GTT interface transfer. • MVD is usually quick, but tails driven by handling other triggers on the VME readout machines are evident. • STT latency should improve after the shutdown when more STT side TP readout links are used. J.Ferrando
Trigger significance • GTT currently working as an improved CTD-SLT • CTD and CTDZ data used, but MVD hits are not yet integrated into algorithms. • Z-vertex from initial forward algorithm showed no difference between NC and GSLT pass through events - use of MVD hits should improve this. • GTT quantities available and being studied by Physics groups: • Primary Z-vertex, • background rejection, and • J/Ψ selection J.Ferrando
GTT hardware modifications CTD stereo-axial • Separate CTDZ (z-by-time) component interface installed. • CTDZ tests made using the STT component interface crate. • Run# < 49376: MVD+CTD • 49376 < Run# < 49859: MVD+CTD+STT • Run# > 49860: MVD+CTD+CTDZ CTD z-by-time GSLT interface STT J.Ferrando
Software plans and studies • Started: • Improvement of the primary vertex finding algorithm which is currently based on the CTD-SLT algorithm - M.Bell. • Fill MVLDAQ bank allowing tie in with absolute timing at interfaces - A.Polini. • Split MVD data processing into 2 threads: barrel and forward - C.Youngman. • Starting Jan 2005 when manpower available: • Inclusion of MVD hits into barrel and forward algorithms and other tracking improvements (split tracks, etc.) - V.Roberfroid, M.Sutton, B.Straub. Studies required: • Data size, cut setting, bad channels, effect on trigger result, etc. • MVD hit weighting w.r.t. CTD and STT hits. • Improvements to DQM needed - N.N. ! • Ongoing studies: • Finalize performance (GTENV+MVLDAQ) monitoring - M.Bell. • Z-vertex efficiency - M.Bell, P.Allfrey, B.Straub + Physics groups. • Background rejection - B.Straub + Physics groups. • J/Ψ finding - B.Dunne, J.Ferrando + Physics groups. J.Ferrando
Summary and outlook • GTT performing well • inside CTD-SLT GSLT latency envelope. • mean latency significantly better than CTD-SLT • improved trigger quantities w.r.t. CTD-SLT • Improvements expected • Including MVD hits will: • significantly improve Z-vertex and track resolutions. • Lead to a useable forward algorithm. J.Ferrando