1. PANDA EMC Trigger and Data Acquisition Development
Experimental Physics Center
Institute of High Energy Physics
Beijing
II. Physics Institute
Justus-Liebig-University
Giessen
Qiang Wang
December 7th, 2009
XXXI PANDA Collaboration Meeting

2. Outline Motivation PANDA physics programs EMC Introduction Simulation
EMC Readout Electronics Introduction
EMC DAQ schematics
Algorithms development
Future work

3. Motivation
The Compute Node V2 is ready for use, lots of simulation results on EMC are also available now, so it is the time for algorithms development;
Without real algorithms running on The Compute Node, we can not sure if it is powerful enough (new ideas come from application);
Software
Plug-in module PC

4. PANDA physics maps (3) Charmonium hybrids (1) Charmonium channel 1
Background
Background
>>not to loss photons (low Eth, Bump splitting)
(2) Charmonium channel 2
(4) Measurement of the time-like EM form factors of the proton
Background
Angular distribution
a suppression better than 108
>> efficient and clean electron identification and accurate measurement of the final state photons originated from decay

5. Interaction of Particles with Different Detectors
Calorimeter:
Charged particle: leptons( ),
mesons( ) and baryon( )
Neutral particle: photons

6. Requirements
Accurate energy and spatial resolution of reconstructed photon;
Good electron/hadron (shape parameters) and electron/photon separation (correlation with Track Detector );
Events channels have similar final state particles compared with background events, no direct trigger criteria. But some physics parameters need to be calculated on DAQ stage (e.g. Cluster Etot, Gravity center position, Shape parameters of cluster) and events selection should be done with the correlation of additional sub-detectors information.

7. PANDA Detector
Forward endcup
Barrel EMC
Employ PWOII Crystal

8. EMC geometry
3600
11360
592

9. Energy range Vs spatial resolution requirements
backward endcap EMC: 10(20) MeV- 0.7 GeV;
barrel EMC: 10(20) MeV- 7.3 GeV;
forward endcap EMC: 10(20) MeV GeV;
Forward endcup
From PROTO60 results
Barrel EMC
Backward endcup
To identify overlapping photons (from Pi0 decay with small opening angles), it is mandatory to efficiently split cluster into
individual photons which requires the central
hits of the involved photons are separated by two crystals.
Backward endcap: 10°
Barrel EMC:2°
Forward endcap: 1°

10. Simulation Events
GeV
Energy Range: 1~2GeV

11. Simulation Events
GeV
Energy Range: 1~2GeV

12. Simulation Events GeV Energy Range: 1~2GeV
Bump overlapping when Pi0 decay to 2 Gamma
Energy Range: 1~2GeV

13. PROTO60 results on cluster size
For E=1.414GeV photon, most of the energy deposits in a 5*5 crystals array, about 80% energy deposited in central crystal

14. Hit Rates and Absorbed Energy Dose in Single Crystals
Figure 1: Hit rate in the barrel part from the DPM background generator at pp=14 GeV/c.
60KHz
Figure 2: Integrated single crystal rate for the barrel
section for an energy threshold of E>3MeV using DPM
at 15 GeV/c incident beam momentum.
Figure 3: Integrated single crystal rate for the forward endcap for an energy threshold of E>3MeV using DPM at 15 GeV/c incident beam momentum.

15. ADC Module (PANDA_EMC_Tdr P.98)
Main Feature:
40~80 MHz ADCs;
2 overlapping 12bits ADC to provide dynamic range of 12000;
One ADC module provide 120 channels
Barrel EMC needs 4*11360 channels[2(Range overlapping)*2(APD)]
Functions:
time adjustment and distribution of the global clock signal;
noise calibration;
common mode noise suppression;
pedestal subtraction;
autonomous hit detection;
conversion of ADC data and linearization of
the full data range;
transporting the hit information together with the time stamp to the data multiplexer;
slow control

16. Data Multiplexer (PANDA_EMC_Tdr P.99)
Interfaces:
1 bidirectional 1 Gbit/s optical link to/from the Time Distribution System;
10 bidirectional 1 Gbit/s optical links to/from the front-end electronics (the digitizer modules);
2 bidirectional 2 Gbit/s copper links to/from the backplane to the neighboring multiplexers;
2 bidirectional 2 Gbit/s links to the DAQ system;
1 Ethernet link to a general purpose network for configuration and slow control;
1Gb/s from/to Time Distribution System
2x2Gb/s Optical Link to DAQ
Data Multiplexer
10x1Gb/s
Ethernet Link
Functions:
The data multiplexer performs advanced data processing by extracting the signal amplitude and time;
combining single hits into clusters, and sorting the clusters in a time-ordered sequence.
2x2Gb/s from/to backplane to neighbor Multiplexers

17. EMC Readout Electronics Introduction
Assumptions:
Run: 30 minutes ;
Bust: 2uS with 500nS gap;
+TDC information to separate events in the same Bust(20 bits needed with 25ps time resolution);
Burst ID
Bust1.Event1
Event i
Bust2.Event1
Bust i
TDC Info.
Energy
Crystal ID
Bust1.Event N
40Events/Burst*540Byte/Event*~22kByte/Burst
Total Data Rate:8.8GByte/s
EventSize=90(Multiplicity)*(20bit(Time info)+14bit(ADC)+14bit(DetectorID))=540B
ADC Module
Data Multiplexer
FEE
FEE
Read Out
Module
FEE
FEE
FEE
2X Optical Link
FEE
FEE
30 Crystal
+60 APD
60 Pre. Amp.
120 Ch/Module
10X1Gb/s
Up to 150 Crystal/Module for
Low Hit Rate
60 Crystal/Module for High
Hit Rate situation
For range overlapping,2*ADC/Pre. Amp
BarrelEMC:11360 Crystal
Up to 5 ADC Mod/Data Multiplexer

18. Readout Schematic 1 Compute Compute Node Node Compute Compute Node
First Stage
Second Stage
Read Out
Module
Compute
Node
Compute
Node
EMC info.
Compute
Node
Compute
Node
Compute
Node
Read Out
Module
Compute
Node
Compute
Node
Compute
Node
Regional Data sent to Compute Node via different channels, Compute Node need to collect information belong to the same sub-event for high level processing
Compute
Node
Compute
Node
Read Out
Module
Average 842Mbps/Module
76 Modules for Barrel EMC
19 Compute Node
other detectors info.
SODA
SODA
SODA
Tasks
1. Signal Feature extraction
(Time, Amplitude);
2. Data Zero compression (?);
1. Regional Clustering;
2. Correlation of regional
information;
3. Cluster Properties
extraction;
4. Pattern recognition;
1. Correlation;
2. Physical parameters
calculation;
3. Event building;
Read Out
Module
Ass.150 Crystal/Module
76 Modules for Barrel EMC is needed

19. Impact of region size on cluster finding efficiency
Region i, j
Region i+1, j
Data Concentrated
on one CN
Phi*Theta: 10Cx*6Cx
Phi*Theta: 20Cx*24Cx
80% need regional correlation
33% need regional correlation
Simplified calculation: a 5*5 cluster distributed in the region
Computing power of one FPGA also needs to be evaluated

20. Data Concentrator on One CN
FPGA1
Agent1
Agent2
Agent3
Agent4
FPGA3
FPGA4
FPGA2
Switch
FPGA0
Cross link of four FPGAs on the same board will be used;
2. An IP liked protocol for RocketIO based serial link is need to concentrate data automatically;
3. Concentration at shelf level is also possible;

21. Readout Schematic 2 Compute Compute Node Node Compute Compute Node
Second Stage
First Stage
Read Out
Module
Compute
Node
Compute
Node
EMC info.
Compute
Node
Compute
Node
Compute
Node
Read Out
Module
Compute
Node
Compute
Node
Compute
Node
Backplane Connection
Backplane connection provide
way to collect data belong to the
same sub-event
Compute
Node
Compute
Node
Read Out
Module
Other Detectors info.
150 Crystal/Module
SODA
SODA
SODA
1. Signal Feature extraction
(Time, Amplitude);
2. Data collection (?);
3. Data Zero compression (?);
1. Clustering;
2. Cluster Properties
extraction;
3. Pattern recognition;
1. Correlation;
2. Physical parameters
calculation;
3. Event building;
Tasks

22. What need to be done on the First Stage
Compute
Node
Compute
Node
EMC info.
Full event info.
Compute
Node
Compute
Node
Compute
Node
Compute
Node
Compute
Node
Compute
Node
Compute
Node
Compute
Node
Other Detectors info.
Event selection should be done by correlation
with other sub-detectors
Tasks( using Schematic1):
Regional cluster finding;
Correlation of regional information;
Feature extraction (e.g. Etotal, Gravity Position, Cluster Size, Bump Overlapping detection);
Clusters selection (e.g. Emax>Eth1, Etotal>Eth2……);
Sub-event format for high level correlation;

23. Reco. Cluster Finder Algorithm
For(i=0; i<nHits; i++)
For(j=0; j<i; j++) {
if(isNeighbor(a[i], a[j]))
TagHits(a[i]);
if(isInTwoCluster())
MergeTwoCluster();
else
CreateNewCluster(); }
Phi
Theta
Advantages
Able to escape losing hits when Cluster have holes;
No limitation for data transmitting order;
Shortcomings
Not a time-effective algorithm;
Need large mount of memory to store intermediate results;

24. Seeds finding (Ehit>Eth2 or Neighbors of Ehit>Eth attaching
Cluster Finder Algorithm (Preliminary)
Hits:48
simple lineal map function
Hits:
Two dimension map
hits filtering
6 Clusters left
Seeds finding (Ehit>Eth2 or
local Emax Finding)
Neighbors of Ehit>Eth attaching
Cal. Etot, ClusterSize,
Gravity Position, ……

25. Future work
Computing power of one FPGA needs to be evaluated with the algorithms( Cluster Finder, LinearPos, Esum);
Serial transmission protocol for Data concentration needs to be considered;
More simulation in PANDA Root framework and adaptive algorithm development;

26. Thanks !

27. Backup slides

28. Cluster Finder Algorithm from ALICE TCP
Step 1: Simulation data read from .root files and reformed in predefined forms;
Step 2: For data belong to the same event will be processed in column, neighbor hits in the same column will merged to as one strip;
Step 3: Column based merge will be done, neighbor strips will be merged to one cluster;
Step 4: Merged cluster above the Eth will be processed to get parameters like Total hits, Etot, Size, Gravity Position … …
Phi
Theta
Advantages
Processing data column by column, only the formal column data need to be stored;
2. Only need small RAM and FIFO, and fabric resources, algorithm is easy to be implemented in our platform;
Shortcomings
Data need to be transmitted in restrict spatial order;
May lose hits when cluster have holes;
Unable to do bump splitting;

29. Cluster Finder Algorithm(1)
Calculation Results:
|EventID:0 ||Entries:21 |
|EventID:0 ||Time: ||Row:22 ||Etot: ||Y1:106 ||H1:1 ||PwPrdY: ||Index:1 ||Appendix:0 |
|EventID:0 ||Time: ||Row:23 ||Etot: ||Y1:145 ||H1:1 ||PwPrdY: ||Index:2 ||Appendix:0 |
|EventID:0 ||Time: ||Row:28 ||Etot: ||Y1:118 ||H1:1 ||PwPrdY: ||Index:3 ||Appendix:0 |
|EventID:0 ||Time: ||Row:29 ||Etot: ||Y1:117 ||H1:2 ||PwPrdY: ||Index:4 ||Appendix:0 |
|EventID:0 ||Time: ||Row:30 ||Etot: ||Y1:117 ||H1:3 ||PwPrdY: ||Index:5 ||Appendix:0 |
|EventID:0 ||Time: ||Row:31 ||Etot: ||Y1:116 ||H1:2 ||PwPrdY: ||Index:6 ||Appendix:0 |
|EventID:0 ||Time: ||Row:37 ||Etot: ||Y1:135 ||H1:1 ||PwPrdY: ||Index:7 ||Appendix:0 |
|EventID:0 ||Time: ||Row:40 ||Etot: ||Y1:126 ||H1:1 ||PwPrdY: ||Index:8 ||Appendix:0 |
|EventID:0 ||Time: ||Row:41 ||Etot: ||Y1:124 ||H1:1 ||PwPrdY: ||Index:9 ||Appendix:0 |
|EventID:0 ||Time: ||Row:41 ||Etot: ||Y1:126 ||H1:2 ||PwPrdY: ||Index:10 ||Appendix:0 |
|EventID:0 ||Time: ||Row:41 ||Etot: ||Y1:129 ||H1:1 ||PwPrdY: ||Index:11 ||Appendix:0 |
|EventID:0 ||Time: ||Row:42 ||Etot: ||Y1:125 ||H1:2 ||PwPrdY: ||Index:12 ||Appendix:0 |
|EventID:0 ||Time: ||Row:43 ||Etot: ||Y1:124 ||H1:4 ||PwPrdY: ||Index:13 ||Appendix:0 |
|EventID:0 ||Time: ||Row:43 ||Etot: ||Y1:130 ||H1:1 ||PwPrdY: ||Index:14 ||Appendix:0 |
|EventID:0 ||Time: ||Row:44 ||Etot: ||Y1:124 ||H1:3 ||PwPrdY: ||Index:15 ||Appendix:0 |
|EventID:0 ||Time: ||Row:45 ||Etot: ||Y1:125 ||H1:1 ||PwPrdY: ||Index:16 ||Appendix:0 |
|EventID:0 ||Time: ||Row:45 ||Etot: ||Y1:127 ||H1:1 ||PwPrdY: ||Index:17 ||Appendix:0 |
|EventID:0 ||Time: ||Row:46 ||Etot: ||Y1:122 ||H1:1 ||PwPrdY: ||Index:18 ||Appendix:0 |
|EventID:0 ||Time: ||Row:46 ||Etot: ||Y1:130 ||H1:1 ||PwPrdY: ||Index:19 ||Appendix:0 |
|EventID:0 ||Time: ||Row:47 ||Etot: ||Y1:131 ||H1:1 ||PwPrdY: ||Index:20 ||Appendix:0 |
|EventID:0 ||Time: ||Row:50 ||Etot: ||Y1:57 ||H1:1 ||PwPrdY: ||Index:21 ||Appendix:0 |
|EventID:0 || Time: ||Hits:1 ||Etot: ||LeftCx:23 TopCx:145 ||Width:1 Height:1 ||PwPrdX ,PwPrdY ||X: ,Y: ||fAppendix:0 |
|EventID:0 || Time: ||Hits:8 ||Etot: ||LeftCx:28 TopCx:116 ||Width:4 Height:3 ||PwPrdX ,PwPrdY ||X: ,Y: ||fAppendix:0 |
|EventID:0 || Time: ||Hits:14 ||Etot: ||LeftCx:40 TopCx:124 ||Width:6 Height:4 ||PwPrdX ,PwPrdY ||X: ,Y: ||fAppendix:0 |
|EventID:0 || Time: ||Hits:1 ||Etot: ||LeftCx:46 TopCx:122 ||Width:1 Height:1 ||PwPrdX ,PwPrdY ||X: ,Y: ||fAppendix:0 |
|EventID:0 || Time: ||Hits:1 ||Etot: ||LeftCx:46 TopCx:130 ||Width:1 Height:1 ||PwPrdX ,PwPrdY ||X: ,Y: ||fAppendix:0 |

30. K0L EventID:27

32. Event Size Caculation (TDC Info. +Crystal ID +Energy) * Multiplicity
20 bits
14 bits
14 bits 90
540Byte/Events

33. Simulation Events
Data generated: K0L Event 0, Hits:32

34. Simulation Events
Data generated: K0L Event 2, Hits:48

35. Gravity Position Calculation