BTeV Trigger and Data Acquisition System

1
BTeV Trigger andData Acquisition System
Presentation to the DOE Fermi Group, January 11,
2001
2
Overview

• The challenge for the BTeV trigger and data
  acquisition system is to analyzeparticle tracks
  and interaction vertices for every interaction
  that occurs in the BTeV detector (15 million per
  second), and to select interactions with Bs.
• The trigger performs this task using 3 stages,
  referred to as Levels 1, 2 and 3L1 looks at
  every interaction, and rejects 99 of
  uninteresting dataL2 uses L1 results and
  performs a more refined analysis for data
  selectionL3 performs a complete analysis
  that uses all of the data for an interaction
• The data acquisition system saves all of the data
  in memory for as long as isnecessary to analyze
  each interaction ( 1 millisecond on average),
  and movesdata to processing units and data
  storage for selected interactions.
• The key ingredients that make it possible to meet
  this challenge
• BTeV pixel detector with its exceptional pattern
  recognition capabilities
• rapid development in technology FPGAs, DSPs,
  microprocessors
• good ideas

FPGAs field programmable gate arrays DSPs
digital signal processors
Erik Gottschalk f
3
Trigger/DAQ
Erik Gottschalk f
4
Vertex Trigger

• The vertex trigger is the primary physics trigger
  for BTeV. The trigger looks forparticles that do
  not come from the main interaction, but appear to
  come from aB-decay vertex close to (of order 1
  10 mm) the main interaction vertex.
• The first stage of the trigger, L1, reconstructs
  tracks and vertices for every beamcrossing (7.6
  million beam crossings per second), and every
  interaction (averageof 2 interactions per beam
  crossing). L1 handles 15 million
  interactions/second!
• With new data arriving at a rate of 7.6 MHz, the
  L1 trigger must produce atrigger decision
  (accept or reject) every 132 ns.
• To accomplish this the L1 trigger uses
  parallelism pipelining
• data for each beam crossing is subdivided and
  processed in parallel
• processor farms use parallelism to handle
  multiple crossings simultaneously
• data can remain in the processing pipeline for
  as long as a millisecond, but trigger decisions
  occur at most every 132 ns

Erik Gottschalk f
5
Simulated Pixel Data (compressed along beam
direction by 10)

• pixel data for a beam crossing (an interaction
  with B-decay vertices) input to L1
• data are subdivided and processed in parallel

Erik Gottschalk f
6
Track segments found by the L1 vertex trigger
A key feature of the L1 vertex trigger is that
partial track reconstruction issufficient at
this stage of the trigger. Therefore, we find
tracks as the enterthe pixel detector (close to
vertex), and as they exit (momentum measurement).
entering track segment exiting track segment
Erik Gottschalk f
7
L2 and L3 Vertex Trigger

• Unlike the L1 trigger, which is built with field
  programmable gate arrays (FPGAs)and digital
  signal processors (DSPs), the L2 and L3 triggers
  are implemented as afarm of commercial
  processors (e.g. INTEL processors running LINUX).
• L2 trigger (data for one beam crossing are sent
  to a single processor)
• starts with tracks and vertices found by the L1
  trigger
• requests data from other detectors (such as the
  forward tracking system)
• improves the track and vertex reconstruction
• imposes more selection criteria to reject 80
  90 of crossings that pass L1
• L3 trigger (L3 begins when all data for a beam
  crossing are sent to a processor)
• further refinement of selection criteria
• reject 50 75 of crossings that pass L2
• reduce the amount of data that will be recorded
• Output 4000 crossings per second, with an
  average size of 50 Kbytes 200 Mb/sec

Erik Gottschalk f
8
Final Remarks

• The BTeV trigger and data acquisition system is
  an ambitious project.
• This type of trigger has never been proposed in
  HEP.
• We benefit from new technologies that help us
  meet the challenge
• pixel detectors
• rapid development of FPGAs, DSPs,
  microprocessors

Erik Gottschalk f