1. L1 Trigger Strategy Ted Liu, Lawrence Berkeley Lab
(for the trigger group)
L1 Trigger Requirements and Trigger Lines
L1 Trigger performance
Background Study
Improving Performance: near-term strategy
Future Prospect: long-term strategy
summary
T. Liu, Oct. 99, L1 Strategy

2. L1 Performance Overview
Still at 700 Hz after
recent lumi&current increase:
HER: 450 -> 600 mA
LER: 800 -> 1000mA
Pessimistic extrapolation
to design current:
HER +450*(750/600)
LER +150*(2000/1000)
Cosmics
Bhabha+Physics
TOTAL Hz
We are in good shape, but
can do better --> room for improvements
Near-term concern
T. Liu, Oct. 99, L1 Strategy
Long-term concern

3. The Game is to Reduce rate as much as possible
Bunch X: 238 MHz
L1: < 2 KHz
L3: < 100 Hz
OPR: ~ 50 Hz?
OFFLINE: ~ 2 Hz
The Game is to Reduce rate as much as possible
at each stage, and make life much easier downstream !
T. Liu, Oct. 99, L1 Strategy

4. Trigger Line Definition/Convention
DCT Objects:
B: short track reaching SL5, Pt ~120 MeV
A: long track reaching SL10, Pt ~180 MeV
A’: high Pt track reaching SL10, Pt > 800 MeV
EMT Objects:
M: minimal ionizing cluster, E >100 MeV
G: intermediate energy cluster, E >300 MeV
E: high energy cluster, E >700 MeV
Y: backward barrel cluster, E >1 GeV
IFT Objects:
1U: 1 high P muon in backward endcap o
B*,M*...* means back-to-back objects(>120 )
D2 = 2B&1A, D2* = B*&1A
D2*+ = B*&1A&1A’ , A+ = A&A’
T. Liu, Oct. 99, L1 Strategy

5. Trigger Lines for Hadronic Events
run9304
Requirements:
B Physics
Pure DCT
CP Channels: high eff.
Non-CP channels:
high + precise eff.
DCT + EMT
Continuum
High & precise eff.
for background sub.
Pure EMT
Strategy for eff. Measurements:
orthogonal lines
redundant lines
record all trigger DAQ data
prescaled lines
T. Liu, Oct. 99, L1 Strategy

6. Use orthogonal DCT-EMT lines to study eff.
From Su Dong
Total eff for hadron skim: (EMT OR DCT) > 99.3%
T. Liu, Oct. 99, L1 Strategy

7. Trigger Lines for non-physics Events
run9304
Trigger Lines for non-physics Events
Requirements:
Luminosity, calibration:
2-prong ee,
High + precise eff.
Pure DCT
Calibration, beam monitor:
1-prong ee,
DCT + EMT
 only
Pure EMT
Background monitoring
Background&Trigger studies
T. Liu, Oct. 99, L1 Strategy

8. Trigger Lines for  and 2 Events
Requirements:
 Physics
Pure DCT
Rare & Asy.: high eff.
Branch Ratios:
precise eff.
DCT + EMT
2 Physics
precise eff.
Pure EMT
Shortcomings:
All unprescaled lines have high Pt
cut (800MeV) and with back-to-back
requirement. No unprescaled orthogonal lines.
All high eff. Lines have to be prescaled.
high background rate, will try to improve…
add 2BM+1AM, lower Pt cut?
T. Liu, Oct. 99, L1 Strategy

9. Need good understanding of background --> next
L1 Strategy to improve performance
Better Combination of trigger lines:
DCT+EMT phi matching
Give up some orthogonal lines
...
Improve trigger objects:
Short-term: PTD ---> DOCAD
Long-term: DOCAD --> DOCAZD?
Need good understanding of
background --> next
T. Liu, Oct. 99, L1 Strategy

10. Single beam L1 pass thr. events Z distribution LER HER LER HER
From S. Petrak
T. Liu, Oct. 99, L1 Strategy

11. Many tracks from beampipe: off Z --> off IP in x-y
LER
HER
LER
HER
Single beam
L1 pass thr.
events
x-z view
T. Liu, Oct. 99, L1 Strategy
Many tracks from beampipe: off Z --> off IP in x-y

12. DCT Trigger Hardware GLT A’ PT Discriminator PTD (x8) DOCAD
Drift Chamber
DCT Trigger Hardware
Binary Track Linker (BLT x1)
Coarse data
for all
supercell
hits
Trigger data
24 Gbits/s
A, B 16
GLT 16 A’
Fine position
data for
segments
found for
axial SLs
PT Discriminator
PTD (x8)
New idea
T. Liu, Oct. 99, L1 Strategy
DOCAD
Track Segment Finder (x24)
(only firmware change)

13. Drift Chamber Trigger ( DCT )
The heart of DCT is the Track Segment Finder
TSF continuously live image processor
A novel method: using both occupancy
and drift-time information, to find track
segments continuously with:
time resolution of ~ 100 ns,
event-time jitter window ~ 100 ns
spatial resolution ~ 1 mm
used for track Pt Discrimination
(1) send segment patterns downstream for
L1 trigger decision making
Upon a L1 accept:
(2) send segment patterns to the DAQ
system for use in Level 3 ...
Track Segment Finder
T. Liu, Oct. 99, L1 Strategy
Use 24 TSF modules

14. T. Liu, Oct. 99, L1 Strategy

15. T. Liu, Oct. 99, L1 Strategy

16. (baseline design: simple algo.)
Current PTD algorithm
New DOCAD algorithm
(baseline design: simple algo.)
A10 seed x x
800MeV
Pt threshold (800MeV)
850MeV
SL7
900MeV
SL7
1 GeV
SL4
SL4
SL1
SL1 x x IP IP
4 hits on one track with ~1mm,
should be able to tell whether the
track is from IP
Tracks (above threshold) coming from IP
should leave all the hits in one of the slices
Tracks not coming from IP will most likely
leave hits in different slices
T. Liu, Oct. 99, L1 Strategy

17. LER single beam run: background
Apply DOCAD algorithm to TSF data
Hadron skim: signal
Effect on Signal:
PTD on
Use hadron skim (~10% bkg)
~ 4% loss at turn on
DOCAD on
Effect on Background:
Use single beam run
L1 pass thr events
Max Pt in the event
40% rejection on A’ line
with simple algorithm
PTD on
Background rejection: 40%
The goal here is not to find the
best algorithm possible, but one
which is easy to be implemented
and with good performance
DOCAD on
T. Liu, Oct. 99, L1 Strategy
LER single beam run: background

18. 4% rejected signal events are picked up by other lines
Hadron skim
Only one event out
of 10**4 is lost
T. Liu, Oct. 99, L1 Strategy
New DOCAD algorithm is SAFE for signal events

19. Some rejected bkg events triggered other lines
LER
T. Liu, Oct. 99, L1 Strategy
Some rejected bkg evts will still contribute to overall rate

20. Fully rejected bkg evts
Some rejected bkg evts triggered other lines as well
Run 9304
L1 pass thr only
Fully rejected bkg evts
L1 rate reduction~11%
Total L1 rate reduction depends on other trigger lines
T. Liu, Oct. 99, L1 Strategy
Room for improvement --> better line combination

21. DOCAD algorithm: possible improvement
Base line design:
Require:
3/3 in the outer slices
2/3 in the inner slices
If we require:
3/3 in ALL slices
70% bkg rejection!
But 20% signal loss
Due to cell ineff.
More careful TSF LUT calibration
is necessary
T. Liu, Oct. 99, L1 Strategy

22. = Require 3/3 in ALL slices DOCAD algorithm: possible improvement  1
TSF can determine segment
location with 1 mm resolution
only for 4/4 patterns
But 20% signal loss
The resolution is degraded
for 3/4 patterns:
Due to cell ineff.  1 2 x x x x x = x x + x x x x
3/4
4/4
4/4
1 <  < 2
calibrating 3/4 patterns should help
70% bkg rejection!
using the  information should help
(baseline design ignored )
T. Liu, Oct. 99, L1 Strategy
Currently 3/4 pattern LUT is derived from 4/4 patterns

23. The Drift Chamber ( DCH )
L1 Long term strategy:
cut in Z using TSF fine position info from stereo superlayers?
Cell size:
1.2 cm x 1.8 cm
The Drift Chamber ( DCH )
10 superlayers of 4 layers each
Axial and stereo alternate
T. Liu, Oct. 99, L1 Strategy

24. Stereo SLs --> Z Axial SLs --> DOCA DOCAZD Tough but possible !
L1 long-term strategy:
Can we cut in Z using TSF position info from stereo SLs?
Stereo SLs --> Z
Axial SLs --> DOCA
DOCAZD
Close to IP in z
Cosmic ray events
In other words:
move some L3 DCH algorithm
upstream to L1 (cut in doca&z)
Off IP in z
Tough but possible !
T. Liu, Oct. 99, L1 Strategy
What to expect? The answer is in the TSF data we already have

25. Stereo SLs --> Z Axial SLs --> DOCA DOCAZD
L1 long-term strategy:
Can we cut in Z using TSF position info from stereo SLs?
Stereo SLs --> Z
Axial SLs --> DOCA
DOCAZD
Current
“trigger
volume”
Possible improved “trigger volume”??
What to expect? The answer is in the TSF data we already have!
T. Liu, Oct. 99, L1 Strategy

26. Level 3 Filter Upon a L1 Accept ... DAQ
Drift Chamber
Upon a L1 Accept ...
DAQ
Using segments found by TSFs as
seeds, fast algorithm applied to DCH
data to further reject background...
reduce the rate from 2 kHz to 100 hz.
Level 3 Filter
DAQ
Fine position
info for
segments
found for all
superlayers
Trigger data
24 Gbits/s
Currently TSF provides fine position info
for all SuperLayers to L3, only axial SLs
info passed to PTD.
T. Liu, Oct. 99, L1 Strategy
Track Segment Finder (x24)

27. One possible scenario Level 3 Filter GLT DOCAZD ?? DAQ A,B Supercell
Drift Chamber
One possible scenario
DAQ
Level 3 Filter
DAQ
All track
segments
found
Trigger data
24 Gbits/s
A,B
Supercell
hits
GLT
DZ?
Fine phi info
for all(?) SLs
DOCAZD ??
T. Liu, Oct. 99, L1 Strategy
Track Segment Finder (x24)

28. L1 Strategy Summary Below and up to design luminosity
In good shape, with some trigger line improvements
(assume dataflow backplane build is successful & on time)
3 X 10**33 < luminosity < 1 X 10**34
Should be ok with DOCAD deployment
Overall: Anders Borgland
should be done with firmware change only
(LBL new posdoc)
Technical feasibility study in progress
VHDL: Chris LeClerc
Beyond 10**34
Possibility of cutting in Z using the unique capability of TSF
Need to build new hardware for sure
Can use existing TSF data to find out what to expect
T. Liu, Oct. 99, L1 Strategy
Careful TSF LUT calibration is essential !

29. LER
T. Liu, Oct. 99, L1 Strategy
HER

30. BABAR Trigger System ( L1 + L3 )
Design challenges
Data taking environment at PEP-II:
Beam crossings occur at 238 Mhz
Design goal:
May have severe beam background
Event time must be determined
High eff for ALL physics events
Keep L1 rate below 2 kHz
L1 Event time jitter < 1 s
IFR Trigger
Global L1
trigger
(GLT)
Trigger latency < 12 s
EMC-EMT path
Calorimeter
Trigger
Processor
(TPB)
Binary
Track
Linker(BLT)
L1: hardware-based
L3: software-based
DCH-DCT path
Track
Segment
Finder(TSF) Pt
Discriminator
(PTD)
T. Liu, Oct. 99, L1 Strategy

31. T. Liu, Oct. 99, L1 Strategy

32. T. Liu, Oct. 99, L1 Strategy

33. Who trigged events without any track?
HER
1U = 1 high P muon in
backward IFR endcap
T. Liu, Oct. 99, L1 Strategy

34. Run9304, L1 pass thr Trigger lines for events without any tracks
T. Liu, Oct. 99, L1 Strategy

35. Who trigged LER? (+ 50% cosmic)
T. Liu, Oct. 99, L1 Strategy
D2*+ = B* & 1A & 1A’

36. Who trigged HER? (+1/4 cosmic)
T. Liu, Oct. 99, L1 Strategy
D2*+ = B* & 1A & 1A’

37. Run9304, L1 pass thr Number of tracks and trigger lines for run9304
T. Liu, Oct. 99, L1 Strategy
Run9304, L1 pass thr

38. L1 Trigger Lines for Hadron Skim
T. Liu, Oct. 99, L1 Strategy

39. The Drift Chamber ( DCH )
10 superlayers of 4 layers each
Axial and stereo alternate
T. Liu, Oct. 99, L1 Strategy