svt september 22, 2005alberto annovi1 svt & cpr upgrades svt what is the svtupgrade why it is...
TRANSCRIPT
September 22, 2005 Alberto Annovi 1
SVTSVT & CPR upgrades
• SVT•What is the SVTupgrade
•Why it is important
•Many upgrade steps at Fermilab: no shutdown what has been done what has to be done
•CPR•What is the CPR upgrade
•Performance results
Short description
September 22, 2005 Alberto Annovi 2
SVTSVT UPGRADE STATUS
July 21 2005: first AM++ & AMSRW
Inside SVT
The SVT team at work in Trigger room (90% italian)
P. Giannetti for the SVT Upgrade group
September 22, 2005 Alberto Annovi 3
SVT
TsukubaChicagoFermilab
Chicago
Pisa1st Pulsar:
Sequencer & Road Warrior
Associative Memory 512 kpatterns
2nd Pulsar:Hit Buffer
3nd Pulsar:Track Fitter
September 22, 2005 Alberto Annovi 4
SVTLarge Impact on Upgrade from Italy (~20 names)
AMchip03 project: Sartori, Tripiccione, Pedron, Schifano (FE)
validation, test: Annovi, Dell’Orso, Giovacchini (PI)
AM++ - LAMB Bardi, Giannetti (PI)
RW Belforte, Carosi, Catastini, Spinella (TS-PI-SI)
AMSRW Bitossi, Giannetti, Piendibene, Spinella (PI)
Software svtvme: Annovi, Giovacchini (PI)
monitoring-configuration: Carosi, Simoni, Volpi (PI)simulation: Annovi, Cerri, Simoni, Torre(PI-LBL-SI)
spymon: Di Ruzza, Rescigno(RO)
Tests: Annovi, Giovacchini, TorreInstallazione: Annovi, Dell’Orso, Giannetti, Giovacchini,
Piendibene (PI), Torre (SI), Di Ruzza, Rescigno (RO) USA: Adelman (TF++), Bellinger (software), Chappa (HB),
Furic(HB), Maruyama (HB), Tang (mezz.), Shochet(mezz.)
Project Leader: Alberto Annovi
September 22, 2005 Alberto Annovi 5
SVTTEMPI DI REALIZZAZIONE• Nuova AM-board: inizio estate 2004 (Pisa) durante estate 2004: test con FPGA (Pisa)
• Progetto prototipo AM-chip: luglio 2004 (Ferrara-Pisa)consegna chip ~2 mesi – disponibile ad ottobre.
• Nuova LAMB: montare nuovo AM-chip a ottobre 2004 (Pisa)
• test del chip + scheda: ottobre – dicembre 2004 (Pisa-Ferrara)
• produzione: inizio 2005 (Pisa-Ferrara)• installazione: estate 2005 (Pisa-Ferrara)• Altri DAQ/Trigger upgrade: previsti nel 2006
Road Warrior: . . . (~60 k$ Fermilab)
messa in opera entro fine 2003
2003 Schedule
September 22, 2005 Alberto Annovi 6
SVTReal data
Upgrade steps
Board/software Validation (step # 0)• Boards will be first tested in the test stand• Then they will be tested parasitically
with real data• Need few slots in a test stand placed near SVT
Hardware installation in 3 steps: 1. AMS/RW and AM++ inside SVT (July 29)
2. TF++ installed in ex-Road Warrior Pulsars 128 kpat
3. Install finally HB and 2nd AM++ board 512 kpat
Study of performances – firmware optimization (4th step - 06)
Other hardware improvements?? (5th step – after 06.. may be)
We are here
Installation & commissioning parasitically without interrupting operations.
September 22, 2005 Alberto Annovi 7
SVT
RW+TF: F>50s = 21%
September 22, 2005 Alberto Annovi 8
SVTSVT timing w/TF++
TF++ Run w/TF last week
(Similar luminosities ~85)
As expected, core of timing stays the same, but significantly reduced tails (look at RMS)
RMS=18
RMS=10.7
September 22, 2005 Alberto Annovi 9
SVTJuly: AM++ installation (1st step)
• AMS/RW + one AM++ board have been installed with minimal impact
AMB
AMB
AMS
MERGER
HF
HF
HF
HB
TF
AMS/RW
AM++
MERGER
HF
HF
HF
HB
TF
• Compatible with old boards (can work with 32k patterns)
• Minimal re-cabling and reconfiguration
Gain: 2 sec reduction on average SVT processing time@Lum=[78-25]*1030 due to RW moved before the HB.
Total gain so far: ~8 sec at 115*1030
September 22, 2005 Alberto Annovi 10
SVTimpact on L1 bandwidth
September 22, 2005 Alberto Annovi 11
SVTimpact on L1 bandwidth
0
2
4
6
8000 10000 12000 14000 16000 18000 20000 22000 24000 26000
Series1
Series2
Dead
tim
e %
L1 accept rate (Hz)
Before SVTUpgrade
Lumi 20-50E30
current SVTUpgrade
Lumi 90E30
17KHz 23KHz +6kHz (+35%) L1A bandwidth @ double inst. lumi.
Now using SVT up to lumi
130E30 (cm-2s-1)
Lumi 45E30
September 22, 2005 Alberto Annovi 12
SVTConclusion
• Parasitic tests of all boards before installation
• AMS/RW, AM++ and TF have been installed before the shutdown (July- August)
• Complete the upgrade (November 2005)
• Optimize Track trigger : XFT-SVT-L2 processors
fill L1 rate as much as possible (2006)
• May be after 2006….GigaFitter (R&D proposed to Murst.)? More AM??
DONE
To be DONE
September 22, 2005 Alberto Annovi 13
SVT
Michele Giunta 13
CPR UPGRADE_PEOPLE
INFN Pisa (7 tecnici per un totale di 4 MU tra assembl. e montaggio) S. Lami A. Menzione R. Paoletti A. Scribano N. Turini M. Giunta
Univ. Roma 1 ( 1 tecnico per un totale di circa 1 MU tra assembl. e montaggio) M. Iori
INFN Pavia G. Introzzi
Univ. Trieste & Udine (3 tecnici per un tot. di 2 MU tra assembl. e montaggio) A. Penzo G. Pauletta
September 22, 2005 Alberto Annovi 14
SVT
Michele Giunta 14
Position in the CDF detector is betweenSolenoid coil and calorimeter
September 22, 2005 Alberto Annovi 15
SVT
Michele Giunta 15
Installation 1
September 22, 2005 Alberto Annovi 16
SVT
Michele Giunta 16
CPR2 detector :: photos
A
DC
B
125 mm
125 mm
September 22, 2005 Alberto Annovi 17
SVTRandom Tracks vs W electrons
September 22, 2005 Alberto Annovi 18
SVTMuon Candidates from Stream A
Sent to Jun Suh for Installation in Simulation
September 22, 2005 Alberto Annovi 19
SVT
Sample of Z->ee with 1 track in crack.
Track Pt
September 22, 2005 Alberto Annovi 20
SVT
SVTBackup slides
backup
slides
September 22, 2005 Alberto Annovi 21
SVTNovember: HB + AM++ enlargement (3rd step)
Status after new HB
installation
AM++
AM++
AMS/RW
HF
HF
HF
AM++
AM++
AMS/RW
HF
HF
HF
newHB
newTF
MERGER
MERGER
newHB
newTF
AMBUS slots
Final configuration
512 kpat/wedge
newTF
newTF
MERGER
HF
HF
HF
newHB
MERGER
HF
HF
HF
AMS/RW
AM++
AMS/RW
AM++
newHB
empty
September 22, 2005 Alberto Annovi 22
SVTPreshower Occupancy
Reference based
on old detector
New detector, far fewer hits
from minimum bias particles
September 22, 2005 Alberto Annovi 23
SVT
TF++ EFFECT
0
1
2
3
4
5
6
7
8
9
10
15000 17000 19000 21000 23000 25000 27000 29000 31000
L1A RATE Hz
DEA
D
TIM
E
%
Green-white-black TF++
Lum=120-55
Brown-light Blue
oldTF Lum=130-100
Brown-light Blue
oldTF Lum=100-63Yellow
oldTF Lum=118-100
Pink TF++
Lum=80-45
Pink TF++
Lum=45-28
TF++ impact on L1 bandwidth
September 22, 2005 Alberto Annovi 24
SVTAugust: Track Fitter installation (2nd step)
After RW has been replaced by AMS/RW.Just reprogram the 12 RW boards with TF firmware
2- 3 sec gain at Lum=[91-81] *1030
<T>=25.1 sec <T>=22.8 sec
September 22, 2005 Alberto Annovi 25
SVT L1 accept rate is bottle neck
•L1A rate ~ 20 kHz (vs 50kHz design)•L1A rate limited by L2 exec timeRates at @ L=100E30 (current luminosity):~15kHz high Pt L1A~15kHz TTT L1A (now PS = 2)Rates growing quadratically with luminosity
Siliconreadout
SVTprocessing
L2 decisionprocessing
Level 2 pipeline structure:
Optimized but slower than
design
Upgrade in progress
Upgrade in progress
SVT upgrade has large impact on maximum L1A rate!
September 22, 2005 Alberto Annovi 26
SVTSVT: Silicon Vertex Trigger
XFT tracksSVX hits
Finding tracks in the silicon
September 22, 2005 Alberto Annovi 27
SVT
September 22, 2005 Alberto Annovi 28
SVT
Level 2• Asynchronous 3 Stage Pipeline• 20 s Latency• 300 Hz accept rate
CDF DAQ & TriggerDetector
Raw Data
Level 1
pipeline:
42 clock
cycles L1Accept
Level 2Trigger
Level 2
buffer: 4
events L2Accept
DAQ buffers
L3 Farm
Level 1•7.6 MHz Synchromous Pipeline•5544 ns Latency•50 KHz accept rate
To Mass Storage (50~100 Hz)
7.6 MHz Crossing rate
SVT here
Level 1Trigger
•50 kHz accept rate
• 20 s average Latency
~20 kHz actual
~35 s actual
Tails are important
September 22, 2005 Alberto Annovi 29
SVTTF++ impact on L1 bandwidth
TF++ EFFECT
0
1
2
3
4
5
6
7
8
9
10
15000 17000 19000 21000 23000 25000 27000 29000
L1A RATE
DEA
D
TIM
E
Blue-red TF++
Lum=100-55
Brown-light Blue
oldTF Lum=130-100
Brown-light Blue
oldTF Lum=100-63
Yellow
oldTF Lum=118-100
White TF++
Lum=80-45
September 22, 2005 Alberto Annovi 30
SVTYear chip boards devel. Total
2003? 120 kE10 kE (test b.) 5 kE 135 kE
2004 Ferrara 10 kE (protot.) 30 kE 40 kE
2005 53 kE 100 kE (produc.) 153 k
+40 kE +40 kE +50 kE +50
kE+60 kE
(MURST)
Pisa
Le Pulsar sono pagate dagli USA totale secondo upgrade 310k$
September 22, 2005 Alberto Annovi 31
SVTAM++ board (Pisa)
• Standard cell AMchip prototype working >40MHz– Production received: evaluating yield
• LAMB and AM++ vme board:– 3nd prototype build and tested– Need final test w/ board full of AMchips
• Building a 4th AM++ vme board prototype – fix minor issues– Production available by June
• ON SCHEDULE
September 22, 2005 Alberto Annovi 32
SVTAMS/RW status (Pisa)
• AMS and RW firmware implemented and tested
• next step implement SVT firmware tools
• ON SCHEDULE
Pisa had the most risky responsibilities, but we are now in very good shape
September 22, 2005 Alberto Annovi 33
SVTNon italian responsabilities
• Pulsars – Pulsar production arrived– Large RAM mezzanine production done – Small RAM mezzanine prototype under test– ON SCHEDULE
• TF
– First firmware written
– Standalone tests starting– ON SCHEDULE
• HB
– Firmware writing just started
– BEHIND SCHEDULE
– More man power (firmware engineer joined)
September 22, 2005 Alberto Annovi 34
SVTSoftware and integration
• SOFTWARE
– Work in progress
– Critical part for test and installation
– BEHIND SCHEDULE
– project reorganized after internal review
• 4 main blocks identified
• coordinators assigned: R. Carosi, A. Annovi, A. Cerri, M. Rescigno
– Additional man power (w.r.t. baseline) mostly Pisa people
– CAN RECOVER
• INTEGRATION
– Vertical slice parasitic test begun
with AM++ and AMS/RW
– Add boards as they become available
Real data
September 22, 2005 Alberto Annovi 35
SVTInstallation plan• Fermilab goal was to install during the shutdown
• Shutdown postponed to October (at least)
• Plan to install 128kpatterns per wedge by July/August
• Installing also new TF (as soon as it is ready)
• Take advantage, as soon as possible, of
• better AM resolution
• RW function moved before HB
• faster TF
September 22, 2005 Alberto Annovi 36
SVT2006:better Timing better algorithm (4rd step)
ScenA: Two_TRK2_Oppq_Dphi135_Sumpt5.5
x2
0.1
0.6
Increasing luminosity Low SVT
Background rejection
x6
x3
September 22, 2005 Alberto Annovi 37
SVTAlso Scenario C shows exactly same problem
ScenC: Two_TRK2.5_Oppq_Dphi135_Sumpt6.5
x2
x6
x3
September 22, 2005 Alberto Annovi 38
SVTFirst very good idea: veto events with a large number of interactions
MORE SVT studies:1) Limited # fits per road2) Adjust Chi**2 cut3) Adjust Hit Finder THRs4) Use new XFT infos (z?)5) Adjust L2 processor
algorithms
September 22, 2005 Alberto Annovi 39
SVTHOW to fill the available L1 bandwidth (step 4)??
A lot of discussion!!
TF++ EFFECT
0
1
2
3
4
5
6
7
8
9
10
15000 17000 19000 21000 23000 25000 27000 29000 31000
L1A RATE Hz
DEA
D
TIM
E
%
Green-white-black TF++
Lum=120-55
Brown-light Blue
oldTF Lum=130-100
Brown-light Blue
oldTF Lum=100-63Yellow
oldTF Lum=118-100
Pink TF++
Lum=80-45
Pink TF++
Lum=45-28
25kHz
10 % dead timeNew Trigger Table
September 22, 2005 Alberto Annovi 40
SVTSVT before July
SVT00Tracking
SVT01Tracking
SVT074 RoadWarrior
SVT02Tracking
SVT06Final MRG
GB
SVT03Tracking
SVT05Tracking
SVT04Tracking
SVT088 Road Warrior
AMB
AMB
AMS
MERGER
HF
HF
HF
HB
TF
AMB
AMB
AMS
MERGER
HF
HF
HF
HB
TF
SVT tracking crate
(2 wedges)
RW
RW
8
8
cables
Need extra crate
September 22, 2005 Alberto Annovi 41
SVTFinal configuration
SVT00Tracking
SVT01Tracking
SVT074 Hit buffer
SVT02Tracking
SVT062 Hit buffer
SVT03Tracking
SVT05Tracking
SVT04Tracking
SVT09 8 Track
fitter
SVT08Final MRGGB 4 TF
AM++
AM++
AMS/RW
HF
HF
HF
AM++
AM++
AMS/RW
HF
HF
HF
newHB
SVTtracking
crate(2 wedges)
newTF
MERGER
MERGER
newHB
newTF
12
cables
AMBUS slots
toLVL2
1