φ Meson Production in pp Reactions at 3.5 GeV with HADES Detector 


Marek Pałka 
Thesis Supervisor Prof. UJ dr hab. Piotr Salabura 
Faculty of Physics, Astronomy and Applied Computer Science 
of the Jagiellonian University 
Cracow, 2011 

Contents 
I7 
1 Introduction and Physics Motivations 8 
1.1 QuarkModel .................................. 8 

1.2 ChiralSymmetry ................................ 9 

1.3 Light Vector Mesons φ and ω . . . . . . . . . . . . . . . . . . . . . . . . . 13 
1.4 OZIRuleanditsViolation -ExperimentalData . . . . . . . . . . . . . . . 15 
2 Accelerator Area and HADES Spectrometer 19 
2.1 AcceleratorArea ................................ 19 

2.2 HADESSpectrometer ............................. 21 

2.2.1 MultiwireDriftChambers ....................... 23 

2.2.2 Superconducting Electromagnet .................... 25 

2.2.3 Ring Imaging CherenkovDetector. . . . . . . . . . . . . . . . . . . 26 
2.2.4 TOFDetector. . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 28 
2.2.5 TofinoDetector ............................. 30 

2.2.6 Pre-ShowerDetector .......................... 31 

2.2.7 StartandVetoDetector ........................ 32 

2.2.8 ForwardWallDetector ......................... 33 

2.2.9 ResistivePlateChamberDetector . . . . . . . . . . . . . . . . . . . 34 
2.2.10Target .................................. 35 

2.3 Trigger and Data Acquisition System of HADES spectrometer . . . . . . . 36 
2.3.1 FirstLevelTrigger ........................... 37 

2.3.2 SecondLevelTrigger .......................... 38 

1 

3 Data Analysis 41 
3.1 Introduction. . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 41 
3.2 Tracking inHADESSpectrometer. . . . . . . . . . . . . . . . . . . . . . . 43 
3.3 TimeofFlightRecalculation. . . . . . . . . . . . . . . . . . . . . . . . . . 44 
4 Event Selection and Particle Identification 48 
4.1 MomentumCuts ................................ 49 

4.2 Masscuts .................................... 50 

4.3 MDCEnergy LossCuts ............................ 52 

4.4 Particleidentification. . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 55 
4.5 Missing MassCut. . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 55 
4.6 VertexCuts ................................... 57 

5 φ Meson Yield Estimation 59 
5.1 EstimationofSystematicErrors........................ 63 

6 Comparison of Experimental and Simulated Data 65 
6.1 AngularDistributions.............................. 66 

6.2 MomentumDistributions. . . . . . . . . . . . . . . . . . . . . . . . . . . . 67 
6.3 MassDistributions ............................... 69 

7 φ Production Cross Section 71 
7.1 pp Elastic Scattering Cross Section -Normalization Factor . . . . . . . . . 72 
7.2 Production Cross Section for pp → ppφ at 3.5 GeV . . . . . . . . . . . . . 74 
8 Conclusions and Outlook 79 
II 80 
9 Upgrade of HADES Data Acquisition System 81 
9.1 TriggerReadoutBoard :TRBv2 . . . . . . . . . . . . . . . . . . . . . . . . 81 
9.2 Front-end and Readout Electronics of RPC Detector . . . . . . . . . . . . 91 
9.3 Front-end and Readout Electronics of TOF and Forward Wall Detector . . 95 
9.4 Front-end and Readout Electronics of START and VETO Detector . . . . 98 
9.5 ReadoutElectronicsofShowerDetector . . . . . . . . . . . . . . . . . . . 99 
9.6 ReadoutElectronicsofMDCDetector . . . . . . . . . . . . . . . . . . . . 101 
9.7 Front-end and Readout Electronics of RICH Detector . . . . . . . . . . . . 103 
9.8 HADESTriggerSystem ............................ 104 

9.8.1 HADESDetectorTriggerSignals . . . . . . . . . . . . . . . . . . . 104 
9.8.2 CentralTriggerSystem -CTS..................... 107 

9.8.3 TriggerDistribution .......................... 114 

9.9 HADESHUB .................................. 116 

9.10 HADESDAQ Summary ............................ 117 

10 Forward Wall Readout Electronics 120 
Abstract 

The High AcceptanceDi-ElectronSpectrometer(HADES) working atGSIDarmstadtis an unique apparatus aiming for systematical investigations of dilepton and strangeness production in elementary and heavy ion collisions in 1-4 AGeV energy range. Measurementsperformed with this spectrometerinheavyionandproton-nucleuscollisionsaimin investigationofhadronpropertiesinnuclearmatter.Inwide scopeof the HADESphysics program theinvestigations ofhadronproductionin elementarycollisions are alsoforeseen. In the framework of this dissertation the exclusive φ meson production in proton-proton reactions at3.5 GeV kinetic energyispresented.The ratio φ to ω production cross section has been obtained and discussed in connection to the OZI rule violation phenomena. 
Description of the analysis methods, whichleads to estimation of the pp → ppφ meson productioncross sectionsispresented.Thisisfollowedbyacomparisonof theextracted φ mesonproductioncharacteristicswith the simulationof thisreactionchannel.Inthe simulation an isotropic φ angulardistributionintheCenter ofMass(CM) frame of colliding nucleons and an uniform phase space population in the proton-meson Dalitz distribution are assumed. A good agreement of the measured angular, momentum and proton-φ invariant mass distributions with the simulation have been achieved. 
The extracted exclusive φ production cross section at 3.5 GeV kinetic energy equals: 
σ4π 
φ =1.05[µb]±0.2[µb](stat.)±0.13[µb](syst.) 
The exclusive ω production cross section obtained in this experiment allows to calculate the ratio of φ to ω production cross section which equals: 
Rφ/ω =0.0099±0.0018(stat.)±0.0012(syst.) 
The obtained ratiois close to the one measuredinpion-nucleon reaction which exceeds by factor ∼ 2-3 OZI rule predictions. 
In the second part, selected tasks of the HADES data acquisition (DAQ) system upgrade, which were in responsibility of the author of this thesis, are described. They concerned mainly following aspects: 
• 	
developing,testing and validating of thecentral trigger system(CTS), 

• 	
developing, testing and validating of the readout electronics for RPC, START, VETO, Forward Wall and TOF detectors. 


Thenewly developed HADESDAQsystem,asitispresently usedinongoing Au+Au campaign,withitstriggerdistribution scheme,front-endboardsand readout components are described as well. 
Part I 

Chapter 1 
Introduction and Physics Motivations 
1.1 Quark Model 
The first elementary particle found in experiment was electron, discovered by J.J.Thompson in 1897. 50 years later, in extended cosmic rays measurements physicists discovered several newparticlespecies.Itwasalso soonnoticed thatonecangroup them into families using specific symmetries. In 1964 Gell-Mann and Zweig proposed that all hadrons, particles interacting with strong forces, are built out of the elementary objects with a spin value of 12 (fermions), calledquarks(onlyquarksu,d, s wereproposedatthis time). It was proposed that quarks have electric and color charges, which are responsible for the electromagnetic and strong interactions, respectively. Later, the c, b, t heavy quarkswerediscoveredand aquarkfamilyhasbeencompleted.Table1.1 summarizesthe quark basic properties: masses, electromagnetic charges and quantum numbers. 
One of the main assumptions of thishypothesisis that only composite objects without anyspecificcolor(color singlets)canbeexperimentally observed.Thisputsconstraints on how the hadrons can be built. If each quark/anti-quark can have three colors/anticolors(red,green, blue/or corresponding anti-colors) then the hadrons are built either out of the threequarks(baryons) orquark and anti-quark combinations(mesons).The theory whichdescribes strong colorinteractionsis calledQuantumChromoDynamics(QCD). 
Theseinteractionsaremediatedby exchangeof massless fieldbosonscalledgluons. Each gluoncarriesthecolor/anti-colorcharge.Thismakespossible,incontrasttothe situation in electrodynamics, direct interactions between the gluons. 
1.2 Chiral Symmetry 
The Lagrangian, which describes the dynamics of a free fermion, has the following form: 
¯
L = ψ(iγµδµ −m)ψ 
According to the Noether theorem, if a given Lagrangian is invariant under certain global transformations of the fields ψ, then the corresponding charges and currents are conserved. For example, for the above Lagrangian one can consider two global transformations: 
τ 
 
θ 
2
• vector SU(2)V transformation: ΛV : ψ → ψe−iF
where: iτ -Pauli matrices, 
F
 
2
• axial SU(2)A transformation: ΛA : ψ → ψe−iγ5 τ θ 
where: γ5 = iγ0γ1γ2γ3,γ -Dirac matrices. 
Itcanbe shownthatthe firstpartof theLagrangian(mq =0) is invariant under the axial and the vector transformations. The second part is also invariant under the vectortransformationbut not underthe axial(forthe non-zero massfermions).The vector 
Table1.1:Quark masses and theirquantum numbers. 
Flavor  Mass [MeV/c2]  I  I3  S  C  B  T  Q e  
u  1.7 to 3.3 1 2  1 2 0  0  0  0  +2 3  
d  4.1 to 5.8 1 2 −1 2 0  0  0  0  −1 3  
s  ∼ 101  0  0  -1  0  0  0  +2 3  
c  ∼ 1270  0  0  0  1  0  0  −1 3  
b  ∼ 4190  0  0  0  0  -1  0  +2 3  
t  ∼ 172000  0  0  0  0  0  1  −1 3  

transformation correspondstoa rotationintheisospin space[1].Theinvarianceunder thistransformationmanifestsitselfinthemassdegeneracy ofhadron stateswith the same isospin(e.g. pions).Theaxialtransformationrotateshadron fields ψ into states of the opposite parity but same mass. The invariance of the Lagrangian under both transformations SU(2)V × SU(2)A is called chiral symmetry. The fact that it is not observed in nature means that the symmetry is broken. For example the ρ meson with(JPC)=1−− 
has mass mρ =0.77 GeV/c2, which is significantly lower than the mass of the a1 meson 1++ (
with JPC = ma1 =1.23 GeV/c2)). 

The masses of the quarks are generated in the Standard Model in two ways. One, like for the leptons, via coupling to the Higgs field in the electroweak interactions. This explicitly breaks the chiral symmetry of the Lagrangian and is responsible for the mass generation of heavy quarks c, b, t. The other mechanism, essential for the masses of light quarks u, d, s, is induced by a spontaneous chiral symmetry breaking. This mechanism is signalizedby the appearance of two,fourquark andgluon condensates(for a recent review see [2]). The quark condensates are non-vanishing structures of the quark-anti quark pairs createdinthevacuumby non-perturbativeQCD effects.Thedynamicalinteraction between these condensates and the quarks causes that the quarks acquire large effective mass. On the other hand early predictions based on the Nambu-Jona-Lasino model[3] indicated that at higher nuclear densities and/or temperatures the expectation value of twoquark condensatedecreases(see Fig.1.1,[4]).Furthermore,it has also beenproposed that under such conditions lowering of the masses of light vector mesons ρ, ω and φ can be used as a signal for the chiral symmetry restoration in dense nuclear matter. However, morecomplexcalculationsbased onQCD sumrulesindicatethattheconnectionbetween the meson masses and the quark condensates is much more evolved and is related to the integral of themesonspectralfunction.Therefore,theQCD sumrulesprovideconstraints on both the WIDTH and MASS of the meson at a given density, but do not answer the question aboutin-medium massesinthe unique way[2]. 
Independentlyfromtheoreticalconsiderations,fromthe experimentalpoint of viewthe fundamental question ”how does the hadron masses change in dense and hot medium?” remains open. To address this there are in principle two experimental methods followed in the world : 
a) Measurementsofthemesonmassdistribution(spectralfunction)insidenuclearmatter. Asdemonstratedby the NA60 collaborationthismethodis suitablefortheshortlived ρ meson[5], 
b) Measurements of the so-called ”transparency ratio” of nuclear matter for a given me-son. This method has been used to study properties of the ω and φ mesons in cold nuclear matter (nucleus). It allows to derive conclusions on the meson broadening (widthincrease)insidethe nuclear matter[6][7][8]. 
For thelatter one, measurements ofthe φ and ω productioninp−p reactions , whichis the subject of this thesis, is an important reference. Though, the masses of both mesons are similar, their internal quark structures differ from each other (see next section for more details), hence can help to disentangle effects related to u, d and s quarks. Indeed, the measured meson transparency indicate large absorption inside nuclear matter, which is equivalent to at least 10-fold increase of the meson natural width. 
Recently, the HADES collaboration also measured a φ to ω cross section ratio in Ar +KCl collisions at 1.756 AGeV [9]. It is an order ofmagnitude larger as compared to the one measured in pp collisions closetotheproductionthresholdbutitisin agreement with thermal modelpredictions(seeFig.1.2).The result mightindicatelarger absorption of the ω as compared to the φ meson.Ontheotherhand one should notethatinthiscase the φ mesoniscreated at sub-threshold energy(ǫ< 0). Thus, observed enhancement of the φ productionmayindicateimportanceofmulti-stepprocesseswithintermediate shortlived resonancesinvolved[10] or/and meson final-stateeffectsinthedensenuclearmatter [2]. To complete the picture, further measurements have been performed with p + Nb system at 3.5AGeV with the HADES spectrometer and data analysis is in progress. 
Rφ/ω 

10-1 
10-2 
10-3 

∈= Ec.m.- E ) [GeV]
thr(φ 
Figure1.2:Comparison of the φ to ω ratio as afunction of the excess energy ǫ.The result obtained with the HADES spectrometer in the Ar + KCl collisions at 1.756 AGeV is placedamong other experimental results -the NN → NNφ and the πN → Nφ reactions. Alsothermal model(THERMUS[11]) inrelationtothe HADESresultispresented. 
The High Acceptance Di-Electron Spectrometer (HADES) working at SIS energies (see fig. 1.1)aims for systematical investigations of properties of the light vector mesons in elementary and heavy ion collisions along both lines. The e+ e− decay channel was chosen as the best option to directly study properties of the light vector mesons. This is because the leptons don’t interact strongly and can reveal the behavior of the mesons inside the dense nuclear matter. But also the meson reconstruction via hadronic decays, as shown in this thesis, is possible. 
1.3 Light Vector Mesons φ and ω 
The light vector mesons ρ, ω, φ areformedfromthequarktriplet u,d, s(SU(3) flavor multiplet). The meson ground states composed by these quarks are listed in Table 1.2 giving their basic properties. 
Table 1.2: Mesons ground states with the different total angular momentum, strangeness andisospin content[12]. 
quark combination  I  I3  S  JP 0−  Mass [MeV/c2]  JP 1−  Mass [MeV/c2]  
|u ¯d >  1  1  0  π+  139  ρ+  775  
|d¯u >  1  -1  0  π−  139  ρ−  775  
(|d ¯d > −|u¯u >) 1 √2  1  0  0  π0  135  ρ0  775  
|u¯s >  1 2  1 2  1  K+  494  K∗+  892  
|d¯s >  1 2  −1 2  1  K0  497  K∗0  892  
|¯us >  1 2  −1 2  -1  K−  494  K∗−  892  
|¯ds >  1 2  1 2  -1  ¯K0  497  ¯K∗0  892  
(|u¯u > +|d ¯d > −2|s¯s) 1 √6  0  0  0  η8  φ8  
(|u¯u > +|d ¯d > +|s¯s) 1 √3  0  0  0  η0  φ0  

The vector mesons φ and ω are linear combinations of the singlet φ0 and octet φ8 states of the SU(3) flavour symmetry nonet. The φ0 and φ8 states have the same isospin andthehyper charge(Y = B + S)quantum numbers, therefore they can mix and form the observed ω and φ mesons: 
φ = φ0sinθv −φ8cosθv and ω = φ8sinθv + φ0cosθv 
From the following equations one can obtain meson masses: 

Mφ 2 = M02sin2θv + M82 cos 2θv −2M08sinθ cosθ, Mω 2 = M82sin2θv + M02 cos 2θv +2M08sinθ cosθ, M2 =0 =(M2 −M2)sinθ cosθ + M08(sin2θ −cos 2θ)
φω 08
the θv, which is a mixing angle, can be obtained: M2 −M2 
φ 8 
tg2θ = . 
M82 −M2 
ω 
Next using a Gell-Maan-Okubo empirical mass formula [13] [14] one can calculate that θv = 39o. It is also possible to calculate this value from the meson radiative decay widths[15][16] and one obtains θv =37o . 
On the other hand an “ideal” mixing angle: 
→ θid 
sinθvid = √ 1 v =35.3o 
3 
makes the φ and ω ideal mixing states: 1 
φ = |s¯ω = √ u> +|d ¯
s> (|u¯d>). 
2For thisideal mixing angle the φ mesonisbuilt only out of the strangequarksand the ω meson only from the u and d quarks. However, if the mixing angle would be the ideal one φ decayinto3pions shouldbe strongly suppressedbecauserespectivequarkdiagram for the decay displays disconnected quark lines . This is in fact the principle for the so calledOZI rulesdiscussedin section1.4(seeFig.2.2 there).Ontheotherhand,assuming a small deviation of θv from the ideal, one is able to explain the following branching ratio for the φ and the ω mesons[12]. 
φ → K+K−49% 
→ K0K¯0 34% 
→ π+π−π0 15% ω → π+π−π0 90% 
→ 
π+π− 1.7% 

→ 
π0γ 8.9% 


1.4 OZI Rule and its Violation -Experimental Data 
As it has already been mentioned, the branching ratio for the φ → π+π−π0 is suppressed andthe φ → K+K− channelis enhanced.Suppression ofthethreepion with respecttothe kaondecay channelis explainedby a ruleformulatedby Okubo-Zweig-Iizuka(OZI[17]). Thisrule saysthat aprocess,inwhich therearedisconnectedquarklines(fig.1.3a),isless probableto occurthantheprocess with connectedquarklines(fig.1.3b). 
(a) φ decay into the three pions -discon-(b)φ decay into two kaons. nected quark lines. 
Figure 1.3: OZI rule examples. 
TheOZI rule also allows to makepredictionsfor theproduction of the φ vector meson in proton or pion induced reactions. 
The production of φ meson, composed out of the pure ss¯state, in pp reactions is schematicallypresentedinFig.1.4.The φ mesonhasparity and spin JPC =1−− and that is why an odd number ofgluonsis neededintheproductiongraph.Atleastthreegluons 

d Φ 
d 

p 
Figure 1.4: φ production in pp reactions. 
are required to produce a color singlet state. Such process would be strongly suppressed and can proceed only due to a small admixture of the u, d quarks in the φ meson wave function. 
The cross section of the φ mesonproductionisgenerally normalized tothe ω production cross section. The ratio of this cross sections equals[18]: 
A + B → φX 
Rφ/ω =	= tan2(δθv)f =4.21· 10−3 · f, (1.1) 
A + B → ωX where f is a ratio of the available phase space for the ω and φ production, in a given reaction, and δθv is a deviation from the ideal mixing angle. The validity of this rule was studied since the 70’ties in various experiments: 
• 	
π protoninduced reactions [19],[20],[21],[22],[23],[24],[25],[26],[27], 

• 	
high energy proton -proton reactions (for the momentum of the beam particles above 10 GeV/c) [24],[28],[29],[30] 

• 	
proton -anti-proton annihilation [31],[32],[33] 


Resultsfromthesemeasurementsare summarizedin[34] and almostall of themshow indications of the OZI rule violation, being largest in proton-antiproton annihilation at rest.InFig.1.5 anexampleof such measurementsfromtheOBELIX experimentisshown. 
Asit canbe seen,theOZI ruleisstrongly violated.Itismostly evidentforthereactions where in addition to the φ and ω light particles(γ,π)are produced. The violation of the OZI ruledependsalsoonthefourmomentumtransfer(the smallermomentumtransfer the larger the effect). This phenomenon was explained by several scenarios: 

• 
the proton has a polarized ss¯contributioninits wavefunction[36], 

• 
two step kaon exchange contribute to the φ meson production process[37][38], 

• 
φ resonance production vi cryptoexotic baryon Bφuddss¯[39][40]. 


Toconfirmoneofthe scenariositwas suggestedtoperformmeasurementsinkinematic conditions as close aspossibletotheproton-antiproton annihilation at rest.Fortheproton -proton induced reactions the φ to ω ratio was available only for the higher energies. However, during last 10 years the DISTO and ANKE experiments measured φ and ω cross sections close to the production threshold of the φ meson. This allowed to estimate theenergy rangewheretheOZI ruleviolationislargest(see fig.1.6). 

Figure1.6:Ratio ofmatrix elementsforthe φ to ω productionis shownasafunctionofthe excess energy. The value predicted by the OZI rule is marked with red arrow and yellow bandwithFSI corrections[41].Starpointsarethedatafromthe ANKE collaboration[42], the squaresfromtheDISTO[43],whilethetrianglesand circles(> 10 GeV )are extracted from ref.[44],[45],[46],[47]. 
InFig.1.6 acompilationofproton-protondatais showntogetherwith thepredictions frommodel,inwhichthefinal stateproton-protoninteractions(FSI) aretakenintoaccount[48](yellow) andOZI rule(red arrow). As one can see, extracted ratio of matrix elements is in conflict with the OZI rules and largest deviation is observed at energies closetotheproductionthreshold.Themodel of A.Sibirtsev[41] isfoundtobeingood agreement with theexperimentaldatapointsathigh energies(fig.1.6yellowbar),but still it can not explain the data at energies close to the φ production threshold. The example of this model illustrates the importance of the intermediate energy to clarify the beam energy dependence of the φ production in pp reactions. 
Theresultsfromthe HADES pp experiment at3.5 GeV kinetic energy(ǫ =280 MeV ) will allowto fillthegapbetweenthedatapointsinhigh energy region(> 10 GeV )and energyjust abovethe φ meson production threshold. 
Chapter 2 
Accelerator Area and HADES Spectrometer 
2.1 Accelerator Area 
Before going into detailed description of the High Acceptance Di-Electron Spectrometer (HADES)experiment setup,theacceleratorareawherethespectrometerislocatedispresentedinthis section.Theacceleratorcomplex(fig.2.1,[49]),whichprovidesbeamsforthe HADES experiment, is located at the GSI Helmholtzzentrum f¨ur Schwerionenforschung GmbH facility in Darmstadt, Germany. 
Itcanbelogically dividedinto severaldifferentareas.Theaimof thefirstpart(UNILAC -UNIversal Linear Accelerator) is to extract ions and inject them into the synchrotron. This is realized in several steps. First, the ions are extracted out of a MEVVA or MUCIS ion source. In the next a Low Energy Beam Transport system (LEBT) together with a mass spectrometer for selecting appropriate isotopes is used and the beam istransportedtoa HighCurrentInjector(ger. Hochstrominjektor -HSI). After this the energy of ions is 1.4MeV/u at maximum. In front of the next stage the beam is stripped and a particular ion charge state is selected. Then, the Alvarez linear accelerator[50] increases the energy of the ions to 11.4MeV/u. In the next step the heavy ion synchrotron (ger. SchwerIonen-Synchrotron -SIS 18) is used to increase the momenta of ions to the required high energy. The whole facility canprovide beams ofprotons up to uranium and a designed maximum momentum for this system equals: 

p = qB0R0, 
where : B0R0 =18[T ∗ m]and this corresponds to: 1 −4.5 GeV/c. 

TheSIS18consists of24bending magnets and24lenses.The vacuuminthebeampipe is on thelevel of1.3· 10−7 Pa.Theaccelerationof theionsisrealizedintwocavities. Each ion at this point experiences a potential of 16kV in frequency range of 0.8−5.6 MHz. Afterwards, the beam is transferred to the experimental areas , for example to the FOPI, HADES, FRagment Separator (FRS) or to the Heavy Ion Storage Rings (ger. ESR -Experimentier Speicher Ring). The full process of the ion acceleration takes ∼ 1 − 3s, depending on the required ion kinetic energy. 

2.2 HADES Spectrometer

 -> RPC 
Figure 2.2: Artistic view of HADES spectrometer. Several different types of detectors are used:fordirect electronidentificationRICH and Pre-Shower,timeof flight measurements (TOF, TOFino) for particle identification MDC and superconducting magnet for the momentum reconstruction. The detector arrangement is according to a sixfold symmetry. The green line represents the beam axis. Detailed description of individual components can be found in the following sections. 
The HADES[51] spectrometer, presented in Fig. 2.2, is located at the GSI Helmholtzzentrum f¨ur Schwerionenforschung GmbH in Darmstadt, Germany. Itspurpose is to measure and reconstruct products of heavy ion, proton and pion collisions. Experiments are done with beams with energies up to a few AGeV and intensities of typically 108particles/s. The main focus in the detector design was put on electron-positron pairs which carry undisturbed information about the high density phase of the collision. The chosendi-leptonchannelhasrelatively smallbranching ratio.Forexampleforthe ω meson it is on the level of 10−5 and for the φ meson 10−4. Therefore, the experimental set-up hastobe ableto cope withhighinteraction rates and todiscriminate electronpairsfrom the overwhelming hadronic background. In more details these demands are related to: 
• 	
large acceptance -the spectrometer covers polar angles in the range of 18o to 85o and almost all azimuthal angles(besidesthe area ofdetectorframes and magnet coils). This allows to have 35% acceptance for lepton pairs, 

• 	
mass resolution -in order to separate different particle species e.g. ω from ρ, the invariant mass resolution δM of the di-electrons must be below 1%, 


M 
• 	
high interaction rates -the detector and its electronics has to cope with 2−4· 104 triggered interactions per second to provide sufficient statistics in the interesting decay channels, 

• 	
trigger -designed in such a way, that it allows to accept events with electrons and discards non relevant data, 

• 	
highgranularity-facilitatestheparticletrackinginahighmultiplicity environment. 


All specifications mentioned above define the HADES experimental setup. It consists of several different detector systems: 
• 	
Multiwire Drift Chambers(MDC), 

• 	
Superconducting electromagnet, 

• 	
Ring Imaging CHerenkov detector(RICH), 

• 	
Time Of Flight TOF and TOFINO, 

• 	
Pre-Shower, 

• 	
START and VETO, 

• 	
Resistive PlateChambers(RPC). 


This set of different detectors is versatile such that it allows to measure, besides the electronpairs,hadronicproductsofreactions.Inthisdissertation(chapter 4) K+ K− identificationwillbedescribed.Topresentthewholepictureof the HADESspectrometer functionality it is necessary to focus on its different components, described in the next sections. 
2.2.1 Multiwire Drift Chambers 
TheMultiWireDriftChamberdetector, showninFig.2.3,allowstomeasureaposition, direction, energy loss per length and together with the magnetic field of the superconducting magnet(see subsection 2.2.2) momentumof theparticle. 
2 (-20°) 
(a) Artistic view of MDC set-up together with (b) One MDC module consists out of six differa magnet. For the sake of the clarity a few MDC ently orientated anode wire planes and 7 cathode moduleswereremovedfromthepicture.MDCde-planes (forbettervisualizationthe cathodewire tector,whichis consistingoutof24modules,is planes are notshown). logicallydividedintofourplanes.Firsttwo(MDC I and II) are placed in front of the magnetic field, created by the superconducting magnet. The two others(MDCIII,IV) are putbehindthemagnetic field. 
Figure 2.3: The MDC detector 
In order to be able to measure all these parameters with desired resolutions (e.g. 1% mass resolution for the ω meson) the detector is divided into four layers. Each layer consists of 6 modules located in individual HADES sectors. One module covers almost 60o of azimuthal angle andpolar anglefrom18o till 85o. The size of the modules depends onthelayernumber.Thedimensionsof allplanesarelistedinTab.2.1.Tokeep the same cell granularity the distance between the wires also increases and translates to the cell sizes ranging from 5 ×5 mm to 15×10 mm for theinnermost to the outermost chambers, respectively. 


Table2.1:DimensionsforthedifferentMDCplanes(see fig.2.3b). 
Plane  a[mm]  b[mm]  c[mm]  
I  839  767  139  
II  1049  905  205  
III  2139  1804  310  
IV  2689  2224  345  

Theinclinationof thedetector,with respecttotheverticalposition,ischosenin such way, that the trajectory of the particles, which paths are the shortest from the target to theMDC, areperpendiculartothedetector(seeFig.2.3a).A chamber module consist of 6 field/sense(anode)planesand7 cathodeplanes(sensewireshavepotential of0 V and cathodes 2 kV ). Thewireorientationareestablishedin suchway(seeFig. 2.3b),that they are rotated by ±0o , ±20o , ±40o with respect to the x axis in Fig. 2.3b. For the two 0o planes, the wires are shifted between each other by half of the distance between the wires to determine, on which side of the sense wires the particle is passing by. 
To measure momentum a track deflection in a toroidal magnetic field is determined. Themagnetic fieldiscreatedby the superconducting magnetbetweenplanesI-II andIII
IV.Whentheparametersofthemagnetic field,particlechargeandthedeflectionangleare knownthenthe momentum of theparticle canbe calculated.With the achievedposition resolution of 120 µm per wire plane a typical momentum resolution amounts to 2 −3% forprotons andpions and 1−2% for electrons, depending on theparticle momentum[51]. 
On top of this basic tracking feature, MDC system is used to perform the particle energy loss (dE/dx) measurement. The measurement itself rely on the gas ionization in the chamber. HADES MDC chambers operate with gas mixture of He-Isobutan with the ratio of 3 : 2. The admixture of isobutan is necessary to absorb photons, created by an avalanche. The minimum energy needed for He ionization is 25 eV . When the particle is crossing the module, it creates so called primary electron-ion pairs. Electrons are accelerated near the anode wires and acquire enough energy to ionize an additional He atoms(secondaryionisation).Thisprocessleads todevelopment of the avalanche with electrons andionsdriftinginthe oppositedirections.The current createdbythe avalanche (mainlymoving ions) is then amplified and shaped in a front-end electronics based on a ASD8 ASIC[52].Oncethe signal crossesapredefinedlevel,alogicpulseiscreatedforthe drifttime measurement.In additionTimeOverThreshold(TOT) is created.The width of this signal is proportional to the signal amplitude and hence energy loss. 
The dependency of dE/dx is described by the Bethe-Bloch[53] formula: 
dE 2 Zz2 12mec2β2γ2Tmax δ 
−( ) = (4πNAremec 2)[ ln −β2 − ]where : 
dx Aβ22 I2 2
NA : Avogadro’s number =6.022× 1023 mol−1 , 
re : classical electron radius =2.817× 10−13 cm, 
I : mean excitation potential, 
me : electron mass, 
Z : atomic number of the absorber, 

A : atomic weight of the absorber, 

z : charge of the particle in e units, 


β = v : beta of the particle, γ2 =1 , 
c 1−β2
σ : density correction, 
Tmax : maximum energy deposit in a single collision. 
andis utilizedfor theparticleidentification.Combining measurementsfromindividual 24MDC wireplanes, a truncated meanis calculated.The obtained resolution of theMDC energy loss measurement is around ∼ 6−7% and it depends on the particle momentum [54]. 
2.2.2 Superconducting Electromagnet 
Asmentioned above,themagnetic fieldforthemomentummeasurementiscreatedby the superconducting toroidal magnet. The design of the Iron-Less Super conductive Electro-magnet(ILSE,[55]) accountsfor severalimportantrequirements: 
• 
it covers the HADES acceptance in azimuthal and polar angles, 

• 	
produces a transverse momentum kick δp/p ∼ 2−5%, 

• 	
magnetic field is negligible in the region of the Cherenkov detector and the time of flight wall detectors, 

• 	
it is build with light materials to reduce production of secondary particles. 



Figure2.4:SuperconductingmagnetILSE composed out ofsixcoilslocatedinthevacuum chambers. 
The magnet is built out of six coils placed in the vacuum chambers. During the operation of the magnet the temperature inside the coils is lowered to the 4.56 K. Therefore thetoroidal coilscanworkina superconducting mode.Themaximumallowed currentfor each coil is 3566 A. In total it gives for all coils NI = 484000 A and creates a magnetic field inside the coil B =3.77 T andbetweenthe coils(inthe air) 0.7 T . The direction of thecurrentis suchthatthetrajectoriesoftheparticleswithpositive(negative) charge werebenttothehigher(lower) polar angles. 
2.2.3 Ring Imaging Cherenkov Detector 
The RICH (Ring Imaging Cherenkov, see Fig. 2.5) detector plays a significant role in the lepton identification. The concept of the RICH operation is based on the Cherenkov effect: particles moving with a speed grater than the phase velocity of light in the given material emitphotons[56].Created electromagneticwavesareradiatedina specificpolar angle direction, which can be described by a formula: 
1 
cos(θ) = where: 
βn n − index of refraction V 
β = 
c 
mounting frame 
photodetector, 
with 6560 pads 
in each sector 

window
2

by the spherical UV-mirror, passing the CaF2 window and are focused on the position sensitive photodetector. 
The HADES RICH detector uses the advantage of threshold character of the Cherenkov effect. Particles created in the target, which is placed in the middle of the spherical RICH mirror, are traversing the C4F10 radiator gas volume shown in Fig. 2.5. 
The γthr = √1 valueforthe radiatorgasis18.2.Thisputs constraints onthe momen
1−β2 
tum of the particles, which can create the Cherenkov light. Therefore, only leptons with momentum pthreshold > 9, 3MeV/c are able to create Cherenkov photons. For example, forpions the required momentum is pthreshold > 2.55 GeV/c, which is not accessiblein the reactions under study with the HADES. This allows to reduce the hadronic background in the interesting di-lepton signal. 
Emitted light is deflected by the aluminized carbon mirror. The mirror material is selected due to its stiffness, small radiation length and a high reflectivity, which amounts to80%.Theshapeof thismirroris such thatit reflectsthelight and createsringsof radius of 5.5cm onaposition sensitivephotondetector.Asit canbee seeninFig.2.5,thephoton detectoris separatedfromtheradiatorgasvolumeby a CaF2 window. Such construction is necessary to achieve good transmission for the ultra-violet light, which dominates the photonspectrum.Theincomingphotonsimpingeonalightsensitivematerial CsI, which cover position sensitive cathode plane of the photon detector, and induce emission of photo-electrons. The photo-electrons are accelerated in the electric field created by the anode wires of a multiwireproportional chamber,ionize thedetectorgas andgenerate the electronavalanche.Createdionstravel tothecathodeand thisproducesa signal amplified by the front-end electronics. In average 8−15photons,depending onthepolarangle, are detected. 
2.2.4 TOF Detector 
TheTOFdetector(Fig. 2.6) coverspolar anglesfrom45◦ to 85◦ and it is built out of six sectors(Fig.2.6a).Each sectoriscomposed outof8 modules,each oneconsisting of8 scintillatorrods(BC408).The structureforone sectorispresentedinFig.2.6b.Thearea of the rodes cross sections varies from 2 × 2 cm for the four lower ones to 3× 3 cm for the upper one. Also the length of the rods changes from 147.5 cm to 236.5 cm. 
Thepurposeofthisdetectoristomeasureatimeofflight of theparticlescreatedinthe target. The particle crossing a scintillator rod creates light, which is transported to both endsof thedetector.Thenitisconverted toanelectric signalinaphotomultiplier(EMI 9133B).TheinformationfromtheTOFisused alsofora firstleveltrigger(LVL1).Itis based onahitmultiplicity,whichfor A+A collisionsisagood measureof centrality.The data from the TOF system is acquired with 200 ps time resolution, which corresponds to the position resolution of 3 − 4 cm. In the further analysis this has influence on the particle species separation based on the reconstructed velocity and momentum. 


Figure 2.6: TOF detector. 
Thetimeofflight,positionandtheenergyloss(dE/dx)correspondingtothe amplitude of the signal can be determined by means of the following formulas: 
• time of flight:  
1  L  
tTOF  =  2(tright + tleft)−  2Vg  where :  

Vg − is the group velocity of the light in the rod, 
tright,tleft − time measured on the left and right side of the detector, L − length of the rod, 
• position: 
1 x =(tright −tleft)Vg
2
• and energy loss of the particle: 
� L 
dE = kAleftArighteλla where: Aleft,Aright − istimeoverthreshold(amplitude) of the signal, λla − light attenuation length in the rod, k − constant. 
2.2.5 Tofino Detector 
As already mentioned, the TOF detector covers only some part of the acceptance of the HADES spectrometer.Theremainingpartiscoveredby theTOFino(Fig.2.7) detector. 

This detector has, because of funding problems, much smaller granularity -four scintillatorpaddlesper sector(intotal24forthe HADESspectrometer) arrangedlong wise to the beam axis. Like the previous TOF detector it is used to extract time of flight information: 
xPreShower 
ttof = t − where: Vg 
Vg − is a group velocity of the light in the paddle, 
t − time registered by the photomultiplier. 
xPreShower − hit position derived from the Pre-Shower 
The signal is readout from one side only, which makes it impossible to determine the hit position from the detector alone. This information is taken from the Pre-Shower detector(subsection2.2.6, see fig.2.7). Alsohereenergydepositcanbeextractedfrom the amplitude of the signal. 
The main disadvantage of this detector is its low resolution of 400 ps and low granularity. These facts were the main reason of replacement of this detector with a high granularity Resistive PlatesChamber(RPC) detector(operational after2008). 
2.2.6 Pre-Shower Detector 
The Pre-Shower detector (Fig. 2.8) is also divided into six sectors. Each sector is constructed with three chambers separated by two lead converters. One chamber is built with two cathode planes and one anode plane, with sense and field wires displaced by 
7.5mm.Onecathodeplaneis subdividedintopads(32 × 32) with individual read-out. The main goal of this detector is to distinguish electrons from hadrons by means of an electromagnetic shower produced inside lead converters. This method supplements TOF measurementin a region whereparticleshavehigher momenta andhence are moredifficult to separate by the TOF method. 
When an electron is passing the detector, it loses the energy in the lead converters and produces the electromagnetic shower. This happens because a high energy electron, when scattered in the electric field of atoms also emit electromagnetic radiation called bremsstrahlung (deceleration radiation). The bremsstrahlung is strongly dependent on themassof theparticle.Itisinverselyproportional tothe squared mass,henceimportant for the electrons. 
A basic idea of the operation of Pre-Shower detector is shown in Fig. 9.20. Emitted photonsareconvertedintoelectron-positronpairs .Thesepairsagainemitphotonswhich Figure 2.8: Pre-Shower detector. 

(a) Photograph of the Pre-Shower detector  (b) Cross section of the Pre-Shower sector.  
mounted and fully equipped with readout elec- Three chambers separated with lead converters  
tronics.  and the response of the detector to two differ 
ent types of particles (electron and proton) are  
schamatically shown.  

convertsandthisproducestheelectromagnetic shower.By comparing anintegrated charge signalfrom3×3pads signalinthepre(beforeconverter) andpost1 and/orpost2(behind theconverters)chambersonecandifferentiateelectronsfromhadrons -thesignalbecomes more significant in the post1/2 chambers for the electrons. 
2.2.7 Start and Veto Detector 
VETO and START detectors are symmetrically built diamond detectors. The diamonds arecreatedwith socalledChemicalVaporDepositiontechnique(CVD,[58]).Thiskind of material ensuresafastresponse, small secondaryparticlesproduction(duetoa small thickness of the detectors) and very good time resolution of 20 − 30 ps. The detectors consistof eight sectorsofdiamond aligned with respecttoeach other(Fig.2.9a).Both detectors are normally placed 75cm away from the target(Fig. 2.9b). 
When a particle crosses the diamond detectors it produces electron-hole pairs which movetowardsmetal electrodes,attached tothediamond andkeptatpotential of 250 V. 

(a) START/VETO detector composed of eight  (b) The Placement of the START and VETO de- 
small detections fields (4 inner and 4 outer re tectors according to the beam and the target.  
gions) with transistor amplifiers [59].  

Figure 2.9: START and VETO detectors. 
The signalfromtheSTARTdetectorinitiatesa firstlevel(LVL1) triggerdecisionif there was no signal in the corresponding segment of VETO detector. Such case indicate that there was a reaction in the target. 
These detectors are routinely used for heavy ion reactions. However, they appeared not suitable for the pp experiments because of too low efficiency. Only recently, new development based on mono-crystal diamond material resulted in successful operation. However, for the p + p experiment described in this thesis no START detector was used. 
2.2.8 Forward Wall Detector 
The Forward Wall detector presented in Fig. 2.10 is a newly installed detector. Its main purpose is to detect the spectator particles in deuteron proton reactions and event plane reconstruction in A+A collisions. It covers polar angles from 7, 1o to 0.13o. It consists of threedifferentsize scintillatorsandphotomultipliersmodules.Thesizeof modulesvaries from40mm×40mm for156 modules,80 mm×80 mm for88pieces to160 mm×160 mm for76 ontheborderof thedetector.Intotal thereare320 scintillatorandphotomultiplier modules(see Fig.2.10a). 
Having the time of flight with typical time resolution of σtof = 400ps and angle with respect to the beam, one can calculate the velocity. If the assumption is made for the 
MDC III/IV 
MDC I/II RICH 
beam 
target 
Pre-Shower 
Forward Wall 
START 
7,3m 
(a)  Front view of the Forward-Wall de (b) Sketch of the HADES spectrometer and the place 
tector without mounted detectors. Size of  ment of the the Forward Wall module.  
the modules increases when going from  
the inner to the outer area of the detec 

tor. 
Figure 2.10: Forward Wall detector. 
particletype(e.g.protonindeuteronproton reaction),itis alsopossibleto estimatethe momentum. 
2.2.9 Resistive Plate Chamber Detector 
TheResistive PlateChamber(RPC,[60],Fig.2.11a)detectorreplaced theTOFinodetectorin2008.Therefore,thisdetector was not usedduringthedatatakinginthe experiment described in this thesis. 
The requirements for the newly constructed detector are challenging. It has to cope with rates of the order of 1000particles andprovidesdoublehit ratesbelow10%.Inorderto 
cm2 
distinguishleptonsfrompionsithastohaveaverygoodtimeresolution. Asdemonstrated in seriesof experiments,thedetectorachieves70 ps time resolution [61], whichis enough to separate pions from electrons up to momentum of 400 MeV/c. The HADES RPC detector is, as most of the other detector systems, composed out of six sectors. Each sector has two partially overlapping layers of the individually shielded RPC cells, which is shown in Fig. 2.11a. One cell of the RPC consists of: 
• shielding of the cell(Fig.2.11b,1), 


(b) The construction of the individual RPC cell. 
Figure 2.11: RPC detector. 
• 
plasticpressure object(2), 

• 
aluminum electrodes(3), 

• 
glass plates(4), 

• 
kapton insulation(5). 


The cell is filled with admixture of SF6 and C2H2F4 gas. To the electrodes a high voltage(5 kV )is supplied and when a charged particle is crossing the cell it ionizes the gas. The electrons are accelerated in the electric field towards the anode. This causes furtherionizationand createsanelectronavalancheand ameasurableelectric signal.The signal is detected on both sides of the detector cell, by dedicated front-end electronics [62].Itallows,asincaseoftheTOFdetector(subsection 2.2.4),todeterminethehit positionbut withbetter(8 mm)resolution. The amplitude of the created signal is used to improve the time of flight measurments by walk corrections. 
2.2.10 Target 
Inthe HADESexperimentthetypeofthetargetdependsonthecurrentlyrealizedphysics program. It was either thin foils or a liquid form of hydrogen. For instance in a C +C at 1−2GeV a5 mm thick foil was used. The density of this material was 2.15 g/cm3 and interaction probability was 5%. 

In case of p +p reactions at 3.5 GeV describedinthisthesis,theprepared target was composed out ofliquidhydrogen(LH2, Fig. 2.12). It has been developed at Institute de Physique d’Orsay(IPN). 
Itwasbuiltoutof twovessels -oneinsidetheother.Theinnerone(5 cm long with a diameter of2.5 cm)washolding the LH2 liquid andit was operating at20K temperature and at normal atmospheric pressure. The outer one was providing a thermal isolation. The system was operated in the vacuum and the target was closed with a 100µm thick Mylar foil. The interaction probability with the Mylar foil was 0.05%. This was much lower in comparison with LH2 target interaction probability -0.7%, and hence did not produced significant background. 
2.3 	Trigger and Data Acquisition System of HADES spectrometer 
The HADES experiment has a significant amount of electronic channels, which is around 80000.Itisneitherpossiblenornecessarytotransportandstoreallthedataonthe storage devices. Therefore the experiment had two levels of the triggers for an event selection. Thesystemforthedatatakingisbrieflydescribedinthefollowing sections(statusduring the 2007 p +p year production beam time). The Forward Wall readout hardware will be presented in chapter 10 in more details as an example. This choice is based on the fact, that author of thisdissertationparticipated in thedevelopment of the readout electronics for the Forward Wall detector. 
2.3.1 First Level Trigger 
As mentioned in the introduction of this chapter, to reduce the amount of the data two levelsofthetriggering areusedin HADES. Thefirstleveltrigger(LVL1) isgenerated based onapredefined signal multiplicityfromtheTOF and theTOFinodetectorsand,if available,withSTART(no VETO)signal.TheTOF/TOFinomultiplicity conditionsused in the p + p experiment described in this thesis are presented in Tab.2.2. 
Table 2.2: Multiplicity conditions during pp at 3.5 GeV kinetic energy beam. 
Multiplicity type  Downscaling  
Mult.�2  256  
Mult.�3  1  

The main trigger used for the analysis of the pp → ppK+K− channel was Mult.�3. Theothertriggerwas usedforthenormalizationpurposes(pp elastic scattering) and was downscaled to avoid too large dead time. 
Time 
undef 100ns 100ns 
100ns 

trigger code
event tag lo
event tag hireserved data
trigger code
event tag lo
event tag hireserved data 
Figure 2.13: LVL1 trigger data transmission. It contains information about the trigger type, number(tag) and reference time. 
TheLVL1trigger signalwasgeneratedintheCentralTrigger Unit(CTU)andwas distributed via HADES trigger bus with differential signals presented in Fig. 2.13. The information which was contained in the LVL1 trigger data transfer was following: 
• 	
referencetime -sent onTSlineandit wasusedforsynchronizationof thedatafrom all readout systems, 

• 	
trigger code -valid on the rising edge of the reference time signal, mainly only two types of the code was used: take data and calibration, 

• 	
trigger tag -trigger sequential number saved into the headers of transferred events, allows event builder to merge all incoming data into one file without risk of mixing it. 


AfterreceivingLVL1triggerthedatawas storedinthelocal readout electronicsbuffers awaiting fora secondlevel triggerdecision(LVL2). 
2.3.2 Second Level Trigger 
After each LVL1 trigger decision the data from the Pre-Shower, TOF and RICH detector wasalsosimultaneouslyprocessed and senttotheMatchingUnit(MU,[63]) whereelectron signal was searched in real time to generate the LVL2 trigger. The schematic overview of the HADES trigger system is presented in Fig.2.14. 
The algorithms for the data preprocessing were realized in the electronic hardware, integrated locally with the detector specific data readout boards. In the Shower IPU (Image ProcessingUnit) thealgorithm[64] waschecking weatherelectromagnetic shower occurred in the acquired event. This was done by checking if the accumulated charge in the fired pads of clusters of the three consecutive Pre-Shower chambers was increasing. 
TheRICHIPU[65] was searchingfortheringpatterns.Duetothespecialdesignof theRICHdetector(padswithtunedsizes)thering radiusisalways8x8pads. Nevertheless it is still a challenging task to find a ring on top of an electronic noise and background originating from the hits of charged particles on the pads. 

Figure 2.14: Overview of the HADES trigger and readout system. The Central Trigger Unit(CTU) accepts the multiplicity triggers comingfrom theTOF andTOFinodetectors and converts them into the LVL1 trigger. After receiving the LVL1 trigger the Detector Trigger Units start the readout electronic and the experimental data is stored in LVL1 pipe. In parallel the pattern recognition algorithms are executed in the detector specific Image Processing Units(IPUs).Thematching unit(MU)combinestheinformationfrom theIPUsandbased ontheresults sendseitherpositiveornegativedecision(LVL2trigger). If the decision was positive the data is read out via the VME CPUs and transported to the EventBuilder(EB). 
Based on the information from the TOF and START detectors, the TOF IPU calculated particle velocity and by this mean it could distinguish electrons or positrons from slower particles. 
Finally, theMU was combining the correspondinginformationfrom theRICH and the Pre-Shower/TOF IPUs and performing spatial correlation of the hit position of the electron candidates found by these detectors. Based on this grounds, the data was discarded ortransportedfurthertotheEventBuilder(EB),whereitwascollected and sentfurther to record on the tapes. 
For the presented analysis of the pp → ppK+K− it is essential to mention, that part of thedata(LVL1 events)is storedindependent ontheLVL2 triggerdecision.Suchdownscaled events(down-scaling wasdefinedintheMU) allowtoperformhadronic analysis. 
Chapter 3 
Data Analysis 
3.1 Introduction 
Thedatapresentedinthisthesishavebeentakenduring3 weeksinApril andMay 2007. The experiment was done using the proton beam with a kinetic energy of 3.5 GeV .A total amount of 1.7· 109 events have been collected. Among these events 1.15· 109 were producedby the multiplicity3 trigger(M3).In order to be able toidentifyparticle tracks inside the HADES spectrometer it is necessary to combine information from different detector subsystems. The HADES collaboration has developed a standardized strategy for a such complex analysis. HYDRA(Hades System for Data Reduction and Analysis [66])isa softwarepackagebased ontheC++ROOTplatform[67].With thehelp of this program Data Summary Tapes(DST) filesforthefurtheranalysisarecreated,which includesacalibrateddatafromalldetector systems.Inmoredetail,theDSTs filescontain identified hits on RICH, TOF/TOFINO and Pre-Shower with corresponding parameters like coordinates, time of flights values, energy loss etc. and also reconstructed tracks in MDCs.Inthenext step of theanalysisaPAT(PostDST Analysis Tool[68]) is executed (see for the details chapter 4). In this part ofanalysis particle identification of a specific reaction decay channel is done and kinematic variables as the invariant masses, emission angles of reconstructed particles etc. are calculated In the last stage of the analysis the obtained results are compared with simulation data, which are processed in exactly the same way as the real data. The whole path of the experimental data analysis is shown in Fig. 3.1. 

The main purpose of this analysis was the identification of the pp → ppφ exclusive channel and the estimation of the φ meson production cross section. In order to obtain correctionsduetoacceptanceand reconstructionefficiencylosses,detailed simulationsare neededtodeterminethe effects ofthe apparatus(see chapter6). Simulated events are generated using aMonteCarlo eventgenerator(PLUTO[69]).Itproduces eventsfor a given reaction channel based on physics model, defined by the user. Particle properties anddecay channel aretakenfromtheParticleDataGroup[12] and are storedinsidethe PLUTO data base. The events from the PLUTO program are then used as an input to theDetectorDescription andSimulationTool calledGeant[70]. TheGeant simulation environment is used to define the HADES geometry and material budget. The simulated tracksofparticlesaretransported through the HADESspectrometerandgeneratedetector hits. These hits are created out of a realisticphysical models ofprocesses takingplace insidethedetectors. Next,thehitsaredigitized andpackedintotheevents,which represents the response of the HADES detectors including front-end and readout electronics. The following steps of data processing are the same as for the experimental data. 
Having both, experimental and simulated data, it is possible to calculate respective corrections due to reconstruction efficiencies and detector acceptance. Finally, the exclusive cross section of φ mesonproduction(chapter7) canbe extracted.Fortheinvestigation of theOZI ruleviolation(chapter1.4) ratio of the φ to ω meson cross section needs to be calculated.The respective pp → ppω exclusiveproduction mustbe analysed.This analysis has been performed by K. Teilab from University of Frankfurt and the respective cross section for ω production has been derived[71]. 
3.2 Tracking in HADES Spectrometer 
As mentioned before, the HYDRA package is used to reconstruct tracks and hits related to the particles measured inside HADES acceptance. 
First, theparticle trajectoryis reconstructedbased on theposition and thedrift times of fired wiresintheMDCdetector.Based onthispartially reconstructed tracks, socalled track segments, can be obtained. The track segments are created independently for the MDC I/II (inner) and MDC III/IV (outer) doublets. In next steps both tracklets are merged together and aligned with the TOF or the TOFINO/Pre-Shower hits. 
The mechanism of track segment matching is presented in Fig. 3.2. The inner MDC track segments are projected to the target area assuming straight line approximation. Segments with best match to the target are selected. For outer segments all combination of hits are considered but then both, inner and outer MDC track segments are projected and matched on a special plane -kick plane. The kick plane is a hyperplane, obtained from simulations, which approximates the place of the deflection of charged particles in the HADESmagneticfieldjustby suddenchange -a ”kick”-of thetrajectory[72]. 
Knowing the deflection of a particle in the magnetic field of known strength and its charge one is able to calculate its momentum. This is done in two steps. First the cubic splinemethod[51],[73]isapplied tocalculate firstapproximationof themomentum. Second, based on the previous result, a fourth order Runge Kutta algorithm of Nystrom [74]is used. Implementation of this method solves differential equations of motion in the known magnetic field. 
In order to identify electrons, the inner MDC track segments are matched with the rings reconstructed in the RICH detector and hits in the TOF or TOFINO/Pre-Shower detectors. 

For the hadron (proton, pion, kaons) identification the tracks are correlated with hitsintheTOF orTOFINO/Pre-Showerdetectorsonly. Havingparticlemomentumand correlated time of flight and/or energy loss in the detectors a particle identification can be performed. 
3.3 Time of Flight Recalculation 
Asit wasalready mentioned,therewasnoSTARTdetectorin HADES set-upduring p+p experiment. Consequently, there was no direct information about time of the reaction whichis mandatory forthe velocity(β = V/c)calculation and, in next steps of analysis, fortheparticleidentification.Therefore,itwasnecessary torecalculatethetimeof flight 
(tnew)using different approach. The method applied in this analysis was based on following information: 
• 	
measured stop times in the TOF and TOFino detector with respect to the LVL1 trigger time -denoted below as ttof, 

• 	
hypothetical time of flight calculated from the momentum and the path length of thereconstructed track assuming thatmassof theparticle(id) isknown -denoted below as tmom id. 


One should note that measured time of flights ttof are related to the calculated ones by simple relation: ttof = tmom + offset, where the offset is atime offset(time whenthe LVL1 trigger fired) changing eventby event,whichhowever,canbeeliminatedforagiven event using time differences between particles produced in the reaction. 
Let us illustrate the algorithm on example of the pp → ppφ reaction channel and the φ → K+K− decay branch. For the complete channel reconstruction only one proton and K+ , K− pair identifications are necessary. 
The time of flight for K+ , tnew K+ , is obtained from the following equations: 
ttof p −ttof K+ 
tΔ pK+ = 	(3.1) 
2 
tmom p + tmom K+ 
tmean pK+ = 	(3.2) 
2 tnew K+ = tmean pK+ −tΔ pK+ (3.3) 
Likewise for the 2nd particle: 
tnew K− = tmean pK− −tΔ pK− 	(3.4) 
The reference particle, which is chosen to be a proton, is calculated as hereafter: 
tmean pK+ + tmean pK− + tΔ pK+ + tΔ pK− 
tnew p = 	(3.5) 
2 Thetimeofflightrecalculationforthe φ meson exclusive analysisisdonefor allpossible mass association to the three reconstructed tracks and associated ttof for thegiven event. 
There are in fact only 2 combinations per event to be considered since a negative charge particleis always assumedtobe an antikaon.This assumptionis of courselater verifiedby means ofparticleidentification cutsdescribedbelow.For each combination, a correlation of the momentum and the velocity of theparticleisplotted and compared to the expected ones for the kaons and the protons. 

Toillustrateresultsof such recalculationand experimental resolution,eventsfromthe simulated reaction pp → φpp → K+K−pp are displayed in Fig. 3.3. The displayed results is for the correct mass associations. In order to evaluate resolution effects important for the particle separation, the mass of the particle has been calculated from the measured momentum and velocity : 
(1−β2)p2 
m = 
β2 
The results, mean andRMS values of the massdistributions, arepresented as afunction of momentuminFig.3.4forprotons andkaons.InFig.3.4aitis clearly visiblethat mean values of the mass distributions (fitted with a Gauss function) are in agreement with the expected masses of proton and kaon. Furthermore, the relative mass resolution is plotted as a function of the momentum in Fig. 3.4b . The resolution decreases for the increasing momentum.Abetterresolutionobtainedforthegiven momentumfortheprotons as compared to the kaons results from the time recalculation procedure which takes into account 3 and 2 time measurements, respectively. As one can expect a time error resulting fromthetimerecalculationwith moretracksis smaller(seeeq.3.5). 

(a)  Proton and kaon masses as a function of the  (b) Errors for the proton and kaon mass distribu 
momentum, with calculated σ. Red lines shows ex tion as a function of the momentum.  
pected masses of the particles  

Figure 3.4: Mass resolution as a function of the particle momentum resulting from the time reconstruction procedure. 
Chapter 4 
Event Selection and Particle Identification 
The φ meson is reconstructed in the channel pp → φpp → K+K−pp. This decay mode has been chosen because of its large branching ratio -49.2±0.6%. The particles, which are required to be detected for the exclusive channel, are at least two particles with a positive charge(proton and K+)and one with a negative charge(K−). Identification of theseparticlesand reconstructionof theirmomentaare sufficienttoretrievefull reaction kinematics(includingnotdetectedproton).Furthermore,the HADESacceptanceforthree prong events is higher by an order of magnitude as compared to reconstruction of fourprong events(twokaons and twoprotons). 
In the next sections of this chapter applied cuts are described in more detail. In the first step of the analysis events containing 2 or 3 particles with a positive charge and one particle with a negative charge are selected. In the next step, particle identification by means of time of flight and energy loss in MDC is performed. In order to reduce a particle misidentification and to reduce the background in the φ meson mass region, cuts on theparticle momenta applied.Thus, analysisprocedure consists offollowing conditions imposed on reconstructed tracks: 
• 	
momentum cuts -section 4.1, 

• 	
particle identification via conditions defined on the particle mass calculated from the time-of-flight and momentum reconstruction -section 4.2, 

• 
particle identification via cuts on MDC energy loss section 4.3. 


If oneoftheparticlesisnotfulfilling agivencut(e.g.theprotonmass cut) thenit is excluded from the hypothesis that it is a specific type of the particle. After all cuts, if therearemorethan3 remainingparticlecandidatesperevent,event selectionismadevia global χ2 test(section 4.4). Afterwardsfromtheeventonly onecombinationischosen with identified p, K+ and K− particles.Finally, whenparticles areidentified, a p, K+K− missing massis calculated and compared to theproton massby means of onedimensional condition to select the ppK+K− reaction channel(section4.5 and4.6). 
4.1 Momentum Cuts 
The momentum distributions retrieved from the simulation of the ppφ exclusive channel are presented in Fig.4.1. It shows that they can be utilized to partially separate protons from kaons. 
Figure4.1:Momentumdistributionfordifferentparticles -simulation.Theredlinesrepresent imposed momentum cuts. 
Inordertopreservemaximal statisticsfortherelevant reactionchannel aconservative conditions onthe momentadistributionshavebeenimposed(seeFig.4.1 redlines): 
350MeV/c < K+ ,K− < 1200MeV/c 
mommom 
800MeV/c < pmom < 2050MeV/c 
It is clear that for the momenta smaller than 1200 MeV/c proton and kaons must be distinguished by other means, as explained in the next section. 
4.2 Mass cuts 
As described in section 3.3, because of the lack of START detector the time of flight reconstructionwas mandatory.It wasbased onthehypothesisthat agivenevent consists fromone(ortwo)protonsand twokaons. Afterassigning K− identity to the track of the particle with a negative charge,two combinationsper event(3in case of4particlesper event) were considered. Assuming a specific particle mass association to the remaining trackscorresponding timeof flightwerecalculatedfromthemeasured momenta(tmom), asgivenby equations3.1-3.5in section3.3.Finally,fromthereconstructed timeofflights (tnew)particle masses were calculated and could be compared to the true ones by means of one dimensional conditions. 
Inordertodeterminewidthsofsuch conditionsadedicatedsimulationforthe ppK+K− channelisperformed.Toillustrate the method, the conditionfor theprotonidentification ispresentedindetailbelow.Fig.4.2a showsamassdistributionasafunctionof theproton momentum asderivedfrom the time reconstruction algorithm.Thisdistributionisdivided into 30 MeV/c momentum slices. Next, as shown in Fig.4.2b, the projection on the axis is fitted with a Gauss function. 
2500 
mass [MeV/c^2] 
2000 
1500 
1000 
20
150 
450 400 
counts [a.u.] 
50 
350 
300 40 
250 30 
200 
100
500 
10 
50 
0 
00
200 400 600 800 1000 1200 1400 1600 1800 2000 2200 0 200 400 600 800 1000 1200 1400 
p [MeV/c] proton mass [MeV/c2] 
(a)Proton mass distribution as a function of the (b) Reconstructed proton mass distribution for momentum for the ppK+K− channel. the 1000 MeV/c − 1030 MeV/c momentum region. 
Figure 4.2: Determination of the width of the proton mass distribution based on simulation. 
In this way, the mean value and σ of the mass distributionsforprotons andkaons as a Figure4.3:Distributions usedtodefine momentumdependentmass cuts.Theblackpoints with green error bars show the simulation results, black points with blue error bars represent the experiment data. 

function of the momentum(shown in Fig.4.3)is evaluated. For theparticle identification, a mass window with the width equal to 2σ is defined for both protons and kaons. 
Since the calculated widths are very important for the correct particle identification and later efficiency corrections they should be validated by a direct comparison with the ones derived from the experiment data. Unfortunately there is not enough statistics to obtain such comparison for the interesting ppK+K− channel. However, the reaction channel pp → ppπ+π− is much more abundant and could be analyzed in this context. Fortheparticleidentification selectionsontheMDC energyloss,describedinmoredetail in section 4.3, is applied by means of two dimensional conditions shown in Fig 4.4a. Furthermore, a cut on the pπ+π− missing massaround theprotonmass, showninFig.4.4 by redlines,hasbeenutilized toincreasethepurity of thechannel selection. Using these events the widths of the proton and pion mass distributions could be extracted from the experimentaldata(Fig.4.3blackpoints withblue errorbars) and comparedtothe simulated ones(Fig.4.3blackdotswithgreenerrorbars).Asitcanbe seenfromFig.4.3 both distributions are in agreement. 
3
10 
1600 
no cuts 
1400 
dEdx cut on all particles 
1200 1000
counts [a.u.] 
missing mass cut 
800 600 400 200 0
0 200 400 600 800 1000 1200 
missing proton mass [MeV/c2] 
(a)MDC dE/dx cuts individually applied for pi-(b)pπ+π− missingmass distributionforthe pp→ 
ons and protons. ppπ+π− event hypothesis in different stages of 
analysis.Thegreen lineisjustafterthe particle 
identification, blue one after the dE/dx cuts and 
red one after the missing mass cut. 
Figure 4.4: ppπ+π− selection for validation of proton and pion mass reconstruction algorithm 
4.3 MDC Energy Loss Cuts 
Particleidentificationby means of theMDC energy loss(MDC dE/dx)is very similar to theoneusedinthetimeofflight analysiswiththeexceptionthat noeventhypothesismust be used.Theidentificationisperformedbymeans oftwodimensional conditions, extracted from the simulation, which were applied on the momentum versus dE/dx distributions. 
In order to validate respective conditions the pp → p pπ+π− reaction channel is used for the extraction of appropriate momentum dependent cuts. In order to select the ppπ+π− reaction channelparticleidentificationby means ofthetime recalculation method wereapplied.Fig.4.5(leftpanel) showsobtained massdistributiontogetherwith applied mass windows to select protons and pions. Finally, a condition on the pπ+π− missing distribution around the proton mass was imposed for the final selection of the reaction channel. 
After the events with proton and two pions are uniquely identified, the MDC dE/dx versus momentum distribution is constructed and divided into 30 MeV/c momentum slices.Foreach slicea fitwiththeLandaufunctionismade(examplefortheprotonis shown in Fig.4.6). 
proton missing mass after best comb. selection 3missing mass after cuts on idividual masses 
Entries 
2.511442e+08 
Entries 
8.371473e+07

mass after best comb. selection 
x103 cuts on idividual masses  10 
cut on proton missing mass
800 masses after cut on proton missing mass 1600 
700 
1400 
counts [a.u.]
π-π+
600 
1200 
p 
500 
1000 
counts [a.u.]
800 
600 
400 
200 
0
400 200 0 200 400 600 800 1000 1200 0 200 400 600 800 1000 1200 
mass [meV/c^2] proton missing mass [MeV/c^2] 
(a) Mass distributions reconstructed from time (b) Proton missing mass cut. 
of flights with prominent peaks corresponding to 
protons and pions. 

Figure 4.5: ppπ+π− selection for validation of the MDC dE/dx distributions for protons and pions. 
(a) Proton dE/dx distribution as a function of (b) Energyloss distribution for the 400 MeV/c− the momentum. 440 MeV/c momentum region. 
Figure 4.6: Momentum dependentproton dE/dx distributions from simulation with Landau function fit. 
ThemostprobablevalueandRMSisextractedfromthe fitsforprotonsandpionsas afunctionof momentum(blackpointswithblueerrorbars)and are showninFig.4.7.A similarprocedurewasrepeatedforthe simulation(blackpointswithgreenerrorbars). As one can see, the obtained mean and RMS values agree with the simulated ones very well. The respective distributions for kaons were obtained from the simulations and are showninFig.4.7,too.Asonecan see,thediscriminationofkaonsfromprotonsandpions 
is possible only below momenta of 800 and 600 MeV/c, respectively. 
Figure 4.7: dE/dx distributions used for defining momentum dependent cuts. 
Figure 4.8: dE/dx cuts. Red lines are fitted polynomial functions to the most probable values of the dE/dx as a function of the momentum. The yellow line is a a cut for the kaons, green for the protons. 
SincethedE/dx distributions are not symmetric,the conditionsdefinedfortheparticle identification are defined in the following way: 
K+/− MDC dE/dx > most probable value −2.5RMSkaon 
p MDC dE/dx > most probable value −2.5RMSproton 
The graphical representation of these cuts imposed on the experimental distributions is shown in Fig. 4.8. The red lines describe the most probable dE/dx values for a given particle type. The green and yellow line represents the dE/dx cut used in the analysis, which separates respectively protons and kaons from the pions. 
4.4 Particle identification 
Afterapplying thecutsdescribedintheprevious sections,theparticleswerequalifiedfor a particular type -p, K+ , K−.In somecases(forexample4particlesperevent) there were more than one classified combination per event and one needed to determine the best pK+K− particle combination. In such case to identify the particles in the unique way, the MDC dE/dx measurements were used and a global event χ2 is calculated: 
(dE/dxIdeal i −dE/dxmeasured i)2 
χ2 
= 
combination I,II,III 
(dE/dx)2 
i=1 resolution 
The combination with the lowest χ2 value is used to identify the particles. At this point, when checking the simulation data, in ∼ 99% of the events the particles were properly identified. 
4.5 Missing Mass Cut 
In thepresentedppφ exclusive reaction channelitis only required todetect threeparticles. Thekinematics of thefourth one(proton) canbe obtainedfrom theknown momentum vectors of thebeam,thetarget and theidentified3particles(p, K+ , K−). 
Fig.4.9a shows the pK+K− missing mass distribution for the identified proton and kaon candidates. As one can see the background is very large and exceeds the signal visible as a small peak around the proton mass. The background originate mainly from misidentified pions from reactions of the type ppπ+π−π0, which has much larger cross 
1200 1000 
800 

1400 
1200 
600 
800 
400 
600 
200
400 
200 
0 
0 
850 900 950 1000 1050 
−
p K+ K missing mass [Mev/c2] p K+ K− missing mass [Mev/c2] 
(a) Background fit with fourth order polynomial  (b) Zoom ofexperimental data for the pK+K−  
function (blue line) to extract the pK+K−  miss- missing mass. To the experimental points a  
ing mass signal.  Gauss function is fitted (magenta line).  

Figure 4.9: Missing mass cut 
section(around 4 mb, to be compared with 1 µb for the ppK+K−). In the right panelof Fig.4.9 a zoom around the pK+K− missing mass peak is shown together with a Gauss function fit.Fig.4.9b shows,thatthemeanvalueof theexperimentaldatapointsis shifted relative to the proton mass by 8 MeV/c2 and it is located at 930.1 MeV/c2 (instead of 938 MeV/c2). This has recently been explained by a 1% uncertainty of the magnetic field value given by the current read-out, which corresponds to the absolute field value. A similar effect has also been observed for the other mesons (ω, K0) and it is under investigation. 
Following the obtained results, for the best signal to background ratio a cut for the ppK+K− selectionhasbeendefinedin the pK+K− missing mass around theprotonpeak: 
930−1.4σ<p missing mass < 930+1.4σ 
where σ amounts to 8.6MeV/c2 for the experimental data and 6.5MeV/c2 for the simulation. The number of the protons in the extracted peak (1.4σcut) is 183 ± 15.9 (statistical error) and should be larger or at least equal to the number of extracted φ meson particles. 
4.6 Vertex Cuts 
The vertex cutisbasedonthetrajectoryinformation carriedbytwokaons.Based onthisit ispossibletoextractthevertexposition -marked with V inFig.4.10a.The coordinates of thispoint and thedistance(R inFig.4.10b) to the actualbeamposition canbe calculated. Theposition of thebeamis extractedfrom all available experimentaldatafor adifferent timeperiodsoftheexperiment.Thisallowstocompensateforanyshiftscausedby changes of thebeamdirection[75].InFig.4.10b the the x and y position from one day of data taking is presented. 


Figure 4.10: Vertex cut. 
The geometrical cut is applied based on an assumption that for the kaons, which are the result of the φ meson decay, the vertex should be close to the actual beam position. The data from the experiment (black points with blue error bars Fig.4.11) and from simulation(greenlineFig.4.11) ispresentedinlogarithmic scaleinFig.4.11.Thecutis represented by the red line. As one can see above 8 mm the experimental data are higher thanthe simulation(whentheareasbelowthecutareinagreement).Thisindicatesan admixture of uncorrelated events. Therefore, the events are accepted when the distance between the kaons vertex and the beam position is smaller than 8 mm. 

distance to beam [mm] 
Figure4.11:Distance of thekaons vertex to thebeampositionfrom the experimentaldata (black points with blue error bars) and simulated (green line). The red line represents upper limit(8 mm) of the distance between K+ K− vertex and beam position for the accepted events. 
Chapter 5 φ Meson Yield Estimation 
180 
160 
counts/5 MeV/c2 
140 120 100 

80 60 40 20 0 
K+ K- invariant mass [MeV/c2] 
Figure 5.1: K+ K− invariant mass signal after all cuts described in the previous sections and particle identification. Different lines under the φ meson peak represent various fits for thebackgroundestimation.Theproperbackground estimationisdescribedin the next part of this chapter. Also the visible structure above 1100 MeV/c2 is explained later on. 
As a result of previously presented cuts applied to the data sample an experimental signal with φ meson peak, visible in Fig.5.1 at 1020 MeV/c2 K+K− invariant mass, is obtained. To extract the φ meson yield it is necessary to estimate the shape of the background. In Fig.5.1 there is no clear indication how it should look like. Green, blue and red lines are examples of fits obtained with different background functions. Fitting thebackground shapewithpolynomialfunctionswouldlead toalarge systematical error. 
In orderto avoid this and to estimatein aproper way thebackground underthe φ signal the side band method is employed. 

In the implementation of the side band analysis method, the events in the missing mass of pK+K− which are outside the proton mass were taken to simulate the behavior of thebackground(seeFig.5.2 -AandB range) and normalized tothebackground area around the φ peak: 
Minv K+K− < 1000 MeV/c2 and 
1040 <Minv K+K− < 1075 MeV/c2 
The conditions for the side band widths are: 
pK+K−(5.1) 
missing mass > 945 MeV/c2 and 915 MeV/c2 −60 MeV/c2 · W <pK+K−. (5.2) 
missing mass < 915 MeV/c2 
Where W isaparameterused todeterminehowaccurateand stableisthebackground as a function of the width of left band (Fig.5.2 -area A). It was varied from 1 to 6 (60 MeV/c2 to 360 MeV/c2) and under such conditions it was found that the shape under the φ peak changes only slightly. The background obtained by this method with W = 1 is plotted as black points with red error bars in Fig.5.3 and, as one can see, it nicely describes the background behavior around the φ meson peak. By subtracting the background from the K+ K− invariant mass the φ meson signal is established. 
180 
160 

counts/5 MeV/c2
140 120 100 

80 
60 
40 
20 0 20 
K+ K- invariant mass [MeV/c2] 
Figure 5.3: Experimental signal -black points, background estimated by the side band method(with W = 1)-red points with red error bars, extracted φ meson signal -black points with blue error bars(exp. signal minus background). 
The meanposition of thepeakpresentedinFig.5.3is µ =1020.5 MeV/c2±2 MeV/c2 andthe width σ =7.5MeV/c2±2MeV/c2.Thederived φ yieldequalsto183±35.2(stat.)± 16.8(syst.1).Wherethe firstpart of theerror(stat.)is statisticaland secondpart(syst.1) is systematical. It is calculated from the different background estimations under the φ peak. This is in agreement with the number of counts in the proton peak extracted from the pK+K− missing mass distribution, which equals to 190 ± 13.8 (stat.). In Tab.5.1 the number of counts in the proton peak extracted from the missing mass spectrum is compared to the φ yield from K+ K− invariant mass. This is done for various widths of the missing mass cut, where the cut width is changed in terms of the RMS width. As one can see, both yields are consistent. 
cut value [RMS]  pyield φyield  
2  1.05  
1.9  1.07  
1.8  1.02  
1.7  1.06  
1.6  1  
1.5  1  
1.4  1.04  
1.3  0.97  
1.2  0.92  
1.1  0.94  
1  1.03  

Table 5.1: Comparison of φ yield withyield in theprotonpeak from the pK+K− missing mass for different missing mass cuts. 
Above 1100 MeV/c2 intheinvariantmassspectrum(seeFig.5.3) onecan seeanother structure, which can not be described by the background created with side band method. The origin of this structure can be explained by simulation of the dominant background channel the pp → ppπ+π−π0. In Fig.5.4 the green histogram represents K+K− invariant massdistribution obtainedfrommisidentifiedpions originating fromthe pp → ppπ+π−π0 reaction channel. Simulated events were treated in the same way as experimental data the samecutsandparticleidentificationasfor φ meson exclusive channel analysis. In the same mass region as for the reconstructed K+ K− invariant mass(above1100 MeV/c2) a second structure appears. The blue line shows the invariant mass distribution for the misidentified pions but obtained with the side band method described above. As one can see the background obtained with the side band method describes the background shape around the φ meson mass region very well. On the other hand, like for the experimental K+ K− invariant mass, the structure above 1100 MeV/c2 is not reproduce. Hence, one canconcludethatthebackground estimationbased onthe sideband techniqueisagood 
70 
60 
50 
40 
30 
20 
10 
0 
misidentified inv. mass [MeV/c2] 
Figure5.4:Green curve:misidentified K+ K− invariantmassfromthe pp → ppπ+π−π0 reactionchannel with visible structureabove1100 MeV/c2.Theblueline:data constructed like in the side band method. This indicates that the three pions channel is a source of this K+ K− invariant mass shape. 
approach only for the 950 <MK+ K− < 1100MeV/c2 . 
5.1 Estimation of Systematic Errors 
As it was discussed in previous section, the systematic error was obtained by different background estimation. However, it is necessary to introduce an additional error due to thedifferent cut effectsonthe simulated and experimentaldata.Thisimposeanadditional systematic error for the determination of the final φ meson exclusive production cross section. To estimate the cuts influence it is necessary to compare the simulation and experimentalyields when varying these cuts.The cuts are changed in respect to theRMS values obtained on a given distributions (see section 4.2 and 4.5). The results of this investigations are presented in Tab.5.2. 
A relative change is calculated by comparing the fraction of the φ meson yield from the simulation and experiment when cuts are open to 2RMS to the one when the cuts are changed: 
φsim. diff cuts φsim. 2σ cuts 
RC = ÷ ·(5.3) 
φexp. diff cuts φexp. 2σ cuts 
·
counts [a.u.]



cut type  cut value [σ]  Relative Change RC  
K+ mass  1.8 1.6  0.98 1.05  
K− mass  1.8 1.6  1.02 1.09  
proton mass  1.8 1.6  1.04 1.21  
K+K−p missing mass  1.9 1.8 1.7 1.6 1.5 1.4 1.3 1.2 1.1 1  1.03 0.98 1.02 0.97 0.98 1.03 0.98 0.93 0.97 1.08  

Table 5.2: Comparison of simulated and experimental φ yields under different cuts conditions. 
The estimated systematical error(syst.2) is on the level of 7% and can be quadratically added to the φ meson yield systematic error, finally φyield = 183 ± 35.2(stat.)± 16.8(syst.1)±12.8(syst.2). 
Chapter 6 
Comparison of Experimental and Simulated Data 
In order toget the φ mesonexclusiveproductioncross section(seeChapter7) corrections for the spectrometer acceptance and reconstruction efficiency losses must be applied. Respective correction factors can be obtained only by means of dedicated simulation. For the simulation a certain reaction model must be assumed. Therefore it is important to compare the experimental and simulation results inside the HADES acceptance. If the model describes the measured distribution, extrapolation to the full phase space is justified. 
Inthenextsubsectionstheangular(sec.6.1),momentumof the φ meson(sec.6.2) and massdistributions(section6.3) ofthekaonsandprotonsarepresented.The simulation is done as illustrated in the section 3.1. In the event generator used for the φ meson simulation a simple production model was assumed. An isotropic φ angular distribution intheCenter ofMass(CM) frame and an uniformphase spacepopulationintheprotonmeson Dalitz distribution were assumed. 
In order to show theφ signaldistributionsitis necessary to extract thebackgroundfor each of investigated distributions. To do so, the background subtraction must be applied. Two different techniques described in next sections were investigated. 
6.1 Angular Distributions 
Infirst step,angulardistributionsintheCMframeareanalyzed.Theprotondistribution for theppK+K− reaction(shownin thebottompanelFig.6.1)is obtained asthedifference between two distributions obtained for: 
a) the events inside the window around the φ peak in the K+K− invariant mass distribution(Fig.6.1 upperpanel -blacksymbols), 
b) theeventscontainedinthe sidebands( Fig.5.2) withthenormalizationtothearea below the φ peak(Fig.6.1 upperpanel -blacksymbolswith red errorbars). 
350 
300 
250 
200 
150 
100 
50 
0 cos(θ) 
p cos( θ) 
cm 
cm 
experiment 

side band method background 
70 
60 

exp. p angular dist. 
50 
sim. p angular dist. 
40 
30 
20 
10 
0 
10 

1 0.8 0.6 0.4 0.2 0 0.2 0.4 0.6 0.8 1 
cos(θ) 
p cos( θ) cm 
cm 
Figure6.1: Angulardistributionofthedetected(blackpointswithblueerrorbars)and simulatedproton(greenhistogram) inthecenterofmassframe.Upperplot showsdistributions for the events inside the window around φ peak (black points) and for the eventsinsidethe sidebands(blackpointswithrederrorbars).Signal(blackpoints)is the difference. Bottom plot shows comparison of the signal to the simulation 
Similarprocedurehasbeen applied toderivethe φ angular distribution. The result is shown in Fig.6.2. 
As one can see both distributions are in agreement in the shape with the simulation, which wasnormalized tothe samearea.Thereforetheassumptionof theisotropicangular distribution is fairly correct. This conclusion is consistent with the results obtained by 
dσ dσ 
[a.u.] [a.u.]
dΩ dΩ 

the DISTO experiment but the excess energy for the φ meson production was 80 MeV [76]. 
dσ 
[a.u.] 
dΩ 
80 60 40 20 
0 

Φ cos( θ) 
cm 
Figure6.2: Angulardistributionof the φ meson reconstructedfrom the experiment(black pointswithblueerrorbars)and simulated(greenhistogram) datainthecenterofthe beam and target proton mass. 
6.2 Momentum Distributions 
In next step, the momentum distributions of the protons and φ meson are investigated. The distributions obtained from experiment and simulation are shown in Fig.6.4. In the right column the background was extracted with the side band method described above. In this case the simulated data and experimental one do not fully cover. In order to check the reason for this discrepancy the condition for the side band was changed to the following one: 
Minv. K+K− < 1000 MeV/c2 or 1040 MeV/c2 <Minv. K+ K− < 1070 MeV/c2)(6.1) 
Mmissing mass > 855 MeV/c2 (6.2) 
This ensures thatthebackgrounddistributions are createdonlyfrom the events, which arein adirect neighborhood of the φ mesonpeak(seeFig.5.3).Itisalso similartothearea wherethebackground extracted with the sideband methodforthe K+ K− invariantmass fully covers experimentalpoints(seeFig.5.3). This also confirmsthe observation made with the simulation presented in Fig.5.4 where one see that the side band background describesthedistributionof missidentifiedpionsinthe K+K− invariant mass only in the mass region given by equation 6.1. 

counts [a.u.] counts [a.u.] 

60 50 40 30 20 10 0 10 

1000 1200 1600 2000 
1400 1800 
2200 
Φ momentum [MeV/c] 
(a)φ meson momentum distribution. (b)φ meson momentum distribution. 
counts [a.u.] 
60 50 
40 
30 
20 10 
0 
10 


40 
30 
20 
10 
0 

10 
proton momentum [MeV/c] proton momentum [MeV/c] 
(c) Detected proton momentum distribution. (d) Detected proton momentum distribution. 
counts [a.u.] 

40 
30 
20 
10 
0 

50 
40 
30 
20 
10 
0 

800  1000  1200  1400  1600  1800  2000  2200  2400  2600  800  1000  1200  1400  1600  1800  2000  2200  2400  2600  
proton momentum [MeV/c]  proton momentum [MeV/c]  
(e) Missing proton momentum distribution.  (f)Missing proton momentum distribution.  


80 
70 60 50 
40 
30 
20 
10 
0 10 300 400 500 600 700 800 900 1000 1100 1200 

K+ momentum [MeV/c] 
(g)K+ momentum distribution. (h)K+ momentum distribution. 


0 
K
 momentum [MeV/c] 
(i)K− momentum distribution. (j)K− momentum distribution. 
Figure6.3:Comparisonof the simulated(greenhistograms)and experiment(blackpoints withblue errorbars) momentumdistributions.Theleft column represents thebackground extracted with conditions described in equation 5.1, the right corresponds to the method used for φ meson K+ K− invariant mass background estimation. As one can see the first method describes the channel reaction momentum distribution in a better way. 
With such conditions the simulated and the experimental data are found to be in a good agreement. One should mention that the φ meson and proton angular distribution presented above does not change when applying this modified condition. Therefore the conclusion about the agreement between the simulation and experimental data holds for all presented differential distributions. 
6.3 Mass Distributions 
As the last step, in Fig 6.4 we show comparison of the proton, and kaon mass distributions for the φ signal extracted with the side band method to the simulated ones. As once can see, obtained mass distributions agrees very well with the expected ones and confirms, in particular for the kaon masses, that the observed signal originates from the ppK+K− channel. Fig.6.4d shows pφ invariant mass distribution, which also agrees with 
the simulation confirming assumed phase uniform distribution. 
 � 


40 
50 
counts [a.u.] counts [a.u.] 

35 
ts [a.u.]
40
30 
30
25 20 
20 
15 
10 
co
10 
0 
5 
0 

10 
K+
proton mass [MeV/c2]  mass [MeV/c2] 
(a)Detected proton mass distribution. (b)K+ mass distribution. 
50 40 


50
counts [a.u.] 

30 
20 
10 
40 30 
20 10 0 
0 
10 
K- mass [MeV/c2] proton and Φ inv. mass [MeV/c2] 
(c)K− mass distribution. (d)p and φ invariant mass distribution. 
Figure6.4:Comparisonbetweenthe simulation(greenhistogram) and experiment(black points with blue error bars) data of mass distributions for a different particles. As one can seeinall casesthedataisinagreement(within2σ). 
The obtained results also ensure that the model used for the φ meson production agrees reasonable with the measured distributions and can be used for calculation of the acceptance and efficiency correction factor. The correction factor has been obtained with the PLUTO and Geant simulation programs and full reconstruction scheme will be presentedinthenext section(seechapter7). 
Chapter 7 
φ Production Cross Section 
To calculate the exclusive φ meson production cross section following formula has been used: 
σel.4π σel.acc 
σφ 
= 	= ,where: 
4π Nφ 4π Nφ 4π
Nel.4π Nel.acc 
Nφ 4π − Number of events corresponding to φ meson production in p + p reactions in the full phase space, 
σel.4π − cross section for the elastic scattering, Nel.4π − number of proton-proton elastic scattering events, σel.acc − cross section for the elastic scattering inside the HADES acceptance, 
Nel.acc − number of events with the elastic scattering reconstructed inside the HADES acceptance. 
In order to obtain the acceptance and reconstruction corrected number of counts for the φ meson(Nφ 4π )one has to consider: 
• 	
acceptanceand efficiency correctionfactorof the HADESspectrometerforthe pp → ppφ → ppK+K− exclusive channel -AccEff, 

49.2+0.6 

• 	
branching ratio br for the φ → K+K− decay -BR = −0.7[12], 

• 	
downscaling factor of the LVL1 trigger -dsc, 


CHAPTER 7. φ PRODUCTION CROSS SECTION 
The acceptance and efficiency corrections are extracted from the pp → ppφ → ppK+K− exclusive channel simulation. This is done by calculation of the ratio of the numberof simulated eventstothenumberof reconstructed eventsafterall cutsdescribed in chapter 4. The obtained correction factor is used to multiply obtained φ meson yield. The downscaling factor is related to the LVL1 events, which were stored independently from the LVL2 trigger decision, and it equals 2.87. Having all these factors one is able to determine the Nφ 4π: 
AccEff · dsc 
Nφ 4π = φexperimental yield 	(7.1) 
br In the next section the elastic scattering cross section correction factor σel.acc/Nel.acc is presented. 
7.1 	pp Elastic Scattering Cross Section -Normalization Factor 
To obtain the Nel.acc it is necessary to select experimental events containing pp elastic scattering. For this purpose during the experiment the dedicated M2 trigger was used. Theelastic scattering reactioncanbefullydescribedby twoindependentvariables[79]. These variables are alaboratory azimuthal angle(Φ) and alaboratorypolar angle(Θ). Using these variables one can derive the following equation for the elastic proton-proton events : 
|Φ1 −Φ2 | = π (7.2) cotΘ1 · cotΘ2 = γcm (7.3) 
where, the γcm is the Lorentz factor of the CM system in relation to the laboratory system. It is defined as: 
Ebeamproton + mproton 
γcm = 
2mproton 
After applying a two-dimensional cut around the expected values derived from the 
CHAPTER 7. φ PRODUCTION CROSS SECTION 
Eq. 7.2-7.3 a prominent peak is visible in Fig.7.1 and one can uniquely identify elastic proton-proton scattering events. 

Figure7.1:Angulardistributionofthe scatteredprotonsintwodimensionalrepresentation of |Φ1 −Φ2 |versus γcm. The peak originates from the pp elastic scattering. 
However, integrating the visible signal does not deliver Nel.acc ,yet because it is necessary to correct this number for the reconstruction efficiency. This is done with usage of simulation chain presented in section 3.1. The results of this analysis are presented in Fig.7.2. The ratio between the reconstructed number of the elastic events and the generated with PLUTO(withinthe HADES acceptance) givesthe HADESefficiencyforthe pp elastic scattering reaction channel[77]. 
As one can see in Fig.7.3 the PLUTO event generator (Fig.7.3 solid black line) is not able to describe properly the elastic scattering cross section at beam momentum of 
4.34 GeV/c, which corresponds to 3.5 GeV beam kinetic energy. Therefore, when determining the σel.acc itis necessary to rely onthe experimentaldataderivedby the[78] at beam momentum 4 GeV/c (Fig.7.3 green points) and 4.5 GeV/c (Fig.7.3 gray points). Themeanvalue(Fig.7.3bluepoints)of thesecross sectionsistakenasanapproximation to the σel.acc at 4.34 GeV/c beam momentum. 
Finally, the correction factor σel.acc =267· 108 ±2%(stat.)±7%(syst.)[77]. 
Nel.acc 
CHAPTER 7. φ PRODUCTION CROSS SECTION 

Figure7.2:Comparisonof thepolarangledistribution simulated with PLUTO(lightred line) and reconstructed(dark redline) pp elastic pairs. 

Figure7.3:Differential elasticcross sections.Greenandgray measurements[78] areused to estimate σel.acc at 3.5 GeV beamkinetic energy(bluepoints). 
7.2 Production Cross Section for pp → ppφ at 3.5 GeV 
Withthecorrectionfactorsdescribedintheprevious sectionsthe σφ 4π exclusiveproduction cross section can be calculated and it equals: 
σ4π 
φ =1.05[µb]±0.2[µb](stat.)±0.13[µb](syst.) (7.4) 
CHAPTER 7. φ PRODUCTION CROSS SECTION 
10210 
1 
10-1

ε 
Figure 7.4: Exclusive ω and φ production cross section in pp reactions [80]. HADES result at 3.5 GeV kinetic energy is placed among already existing data points. The three lines represents a three body phase space for the production as a function of the energy normalized to the ω cross section at high energy (black line ), φ meson at low energy (red dashed line)and φ inhigh energy region(redline) cross sections.Obtained result at 280 MeV excess energy for the φ meson production is in-line with the data from ANKE and DISTO experiments. 
Fig.7.4 shows the world data on the φ and ω meson production cross sections in pp reactionsas afunction of theexcessenergy[42][43][81][82][83][84][85].Thethreelines represents a dependence of three body phase space as a function of the excess energy where the final state proton-proton interaction effect is taken into account as well. This is equivalent to the assumption of a constant matrix element for the meson productions and production governed by the phase space only. The black line is normalized to the ω production cross section at high energy. For the φ meson two lines are presented, which differby selected regionof normalization: reddashedlineshowsresult with normalization close to the production threshold while the solid presents result for the normalization fixed at high energy region. For the beam energy studied in this thesis the excess energy 
CHAPTER 7. φ PRODUCTION CROSS SECTION 
for the φ meson production equals 280 MeV . In Fig.7.4 HADES data for exclusive φ mesonproductioncross sectionisshownwith orangecircle.Theachieved resultisin-line with the results accomplished by ANKE and DISTO experiments close to the φ meson threshold production. 
In HADES experiment the ω meson cross section was obtained as well. Fig.7.5 shows the ω meson peak in the proton-proton missing mass distribution extracted from the analysis of pp → ωpp → ppπ+π−π0 reaction channel [71]. The ω meson cross section production was estimated to be 106.5±0.9(stat.)±7.9(syst.)[µb]. It is necessaryto underline that at this kinetic energy the angular distributions measured for the ω meson in the CM frame are only slightly anisotropic in contrast to the results obtained at lower energieswhere strong anisotropy wasobserved[81].Thiscanindicatedifferentproduction processes for the φ and ω mesons close to the production threshold. 

Figure7.5: ω mesonpeak at782 MeV/c2 extractedfrom pp → ωpp → ppπ+π−π0 reaction channel with HADES[71] 
Based on these result the φ to ω production cross section ratio Rφ/ω in pp reaction at 
3.5 GeV kinetic energy can be calculated and amounts to: 
Rφ/ω =0.0099±0.0018(stat.)±0.0012(syst.) 
In Fig.7.6 this ratio is shown together with already existing results from N −N (red points), π −N (pinkpoints) andAr +KCl (HADES-blackpoint)collisions as afunction 
CHAPTER 7. φ PRODUCTION CROSS SECTION 
of excess energyfor the φ mesonproductionin elementary collisions.The excess energyfor theheavyion collisionsis calculated assuming N −N reactionsofthe samenominalkinetic beam energy. It leads to a negative value indicating subthreshold production which can proceedthroughsecondaryprocessesor/anddirect N −N reactions utilizing nucleons with large fermi momentum inside colliding nuclei. As already discussed in the introduction, the ratio observed in heavy ion reactions is larger at least by one order of magnitude as compared to the N − N reactions, even the one observed at much higher energy of 3.5 GeV obtainedinthis work(greenpoint). This mightindicatethatindeed other reaction mechanism is responsible for the φ meson production in heavy ion reactions. Among many possible reactions channels, secondary pion-nucleon collisions seems to be an obvious candidate. However, also for this reaction channel observed Rφ/ω is larger, if one take into account that secondary pions are produced with low energies. 
Rφ/ω 
10-1 
10-2 
10-3 

∈= Ec.m.- E ) [GeV]
thr(φ 
CHAPTER 7. φ PRODUCTION CROSS SECTION 
It is also interesting to compare obtained Rφ/ω (green point) to the one for the πN reactions at high energy. Both results are similar and are factor ∼ 2-3 above predictions givenby theOZI rule(4.2× 10−3). For the latter one, phase space factor f in equation 
1.1 canbeneglecteddueto sufficiently high energy abovetheproductionthreshold what is reflected by almost constant Rφ/ω value for the π − N reactions. This excess above naive OZI rule predictions is clearly lower as compared to the one measured close to the production threshold in N −N reactions(∼ 7-8[42],[43]). 
Chapter 8 
Conclusions and Outlook 
The experimental results presented in this work, despite the low statistics, allows to estimate the φ meson exclusive production cross section σφ 4π =1.05[µb]±0.2[µb](stat.)± 0.13[µb](syst.). The agreement between simulation, where isotropic angular distribution in the CM frame and an uniform phase space population in the proton-meson Dalitz distribution are implemented, is observed. This agreement confirms that extracted in this work production cross section for the φ meson production (in context of acceptance and efficiency corrections) are reliable. From the same experimental data, but other analysis, the ω meson exclusive cross section production was acquired. Therefore the direct comparison between these two cross sections is possible and results in Rφ/ω =0.0099±0.0018(stat.)±0.0012(syst.). The extracted data points are supplementary to the data of the φ to ω ratio in Ar + KCl collisions at 1.756 AGeV -measured by HADES and other low-energy N − N data. In order to complete the picture on the Rφ/ω excess seen in the heavy ion collisions new data from p −A reaction channel at the same energy regime are required. The HADES collaboration measured p + Nb system at 3.5AGeV . The data analysis is currently in progress and its results will complete the picture of φ and ω meson production at this range of the kinetic energy. 
Part II 

Chapter 9 
Upgrade of HADES Data Acquisition System 
In the near future the measurements of the HADES spectrometer will be performed at aSIS-100,oftheupcomingFAIR[86] acceleratorcomplexatGSI. Afterthis,following years the HADES collaboration will continue its experimental program up to the kinetic beam energiesper nucleon of8 GeV . Also during experiments at SIS-18 with a heavy ion collisions(like Au+Au) with ahigh chargedparticlemultiplicity,thevolumeof thedata will exceed this for what the system was prepared. It will reach 400 MB/s for the whole detector system with accepted LVL1 trigger rate of 20 kHz and 100 kHz in peak. The above was the reason for an upgrade of the HADES detector and its trigger and readout systems[87]. 
9.1 Trigger Readout Board : TRBv2 
To build this new system it was necessary to take a decision what should be a structure of it and answer the questions -how the new readout electronics for each detector, what kind of themediaforthedatatransfer,protocols,central trigger systemand eventbuilder will looks like. The whole approach was constructed on an idea of one platform for all subsystems. Adapting the ”one platform” concept, the TRBv2 (Trigger and Readout Board version2) wasbuilt[88]. 
To accomplish the above requirements this board consists of: 
CHAPTER 9. UPGRADE OF HADES DATA ACQUISITION SYSTEM 

(a) TRBv2 board top view. Equipped with 128 (b) Main components of the TRBv2. Dashed 
time to digital converter channels, Virtex 4 FPAG, line for the connector placed in the back of the 
ETRAX-FS multi-processor, optical link, DSP, board. 
power supply, Ethernet connection. 

Figure 9.1: TRBv2. 
• 	
four32 channelTimetoDigitalConverter chips(HPTDCs), 

• 	
FPGAVirtex4LX40 -it managesall thedataflowontheboard anditisconnected to all main components on the board, 

• 	
ETRAX-FS multi-processor[91] -itisusedfortheDAQ andslow-controlfunctionality.Theprocessorrunsa standardLinux2.6kernelinthe128MBytesof memory and it is directly connected to the 100 Mbit/s Ethernet. The integrated three coprocessors (each 200 MHz) allow a high I/O bandwidth without the main CPU intervention, 

• 	
flash, where the ETRAX-FS processor boots from, 

• 	
Tiger-Shark DSP[93](600 MHz)-itcanbeusedfortheprepossessing of thedata, 

• 
AddOnconnector -isaveryhighdata-ratedigitalinterfaceconnector,which allows 


• 	optical link(2.5 Gb/s)-the FPGA sends data with the help of a TLK2501 Texas Instrumentdevice[92] to the opticalSFP(Smallform-factorpluggable transceiver), 
CHAPTER 9. UPGRADE OF HADES DATA ACQUISITION SYSTEM 
to connect add-on (AddOn) boards. This can expand the field of possible applicationsbyproviding additionalfront-end electronics, computationalpower(other FPGAs and DSPs) or particular interfaces for the given detector subsystems. The pins of the AddOn connector are linked to the Virtex 4, Etrax-FS and also to the power supply of +5 V , 
• 	
two 1 Gb SDRAM memory banks, connected to the ETRAX-FS and to the FPGA, 

• 	
48 V galvanically isolated power supply. 


The means for programming the HPTDC ASIC are generally the same as in the TRBv1(seeappendix10) butwith somedifferencesoftheboardarchitecture.Firstof all, the HPTDCs are in one JTAG chain and the access to the JTAG lines is through the FPGA registers. Additionally, the user has an easy-to-use configuration file. Using the existing Perl script it is possible to create the TDC configuration files, which are written automatically in the STAPL language. Next the user can use the Jam Player and load all required setting to the TDCs. Also with the JTAG interface the Virtex FPGA, board clock manager and chips on the AddOn boards can be programmed. 
For the usage of the TRBv2 several scenarios were necessary to apply: 
• 	
stand alone TDC readout board, 

• 	
independentfromthe HADESDAQ system setup(section9.2), 

• 	
temporaryCentralTriggerSystem -CTSduring2010 HADESbeamtests(together with General Purpose AddOn used for converting the signals standards e.g. from TTL to LVDS), 

• 	
first platform for the readout tests of the Shower and MDC AddOn boards (see section 9.5 and 9.6), 

• 	
integrated with the current HADES DAQ and trigger system as a TDC readout board(RPC,TOF,START, VETO,ForwardWalldetectors), 


In the stand alone mode the system is started with the rising edge of the reference time. The reference time is split to five signals. Four of them are connected to the 32nd 
CHAPTER 9. UPGRADE OF HADES DATA ACQUISITION SYSTEM 
channel of HPTDCs to synchronize the TDCs locally and in a bigger system with other data coming from the other readout boards. The fifth reference time initiates the readout ontheFPGA.Thedatataking schemefromthe HPTDCsisjustthe sameasinthecase of theTRBv1(seeappendix10).Whenthewholeeventiscollectedinthe VirtexFIFO, itistransported viaport B(15:0) and C(15:0) of the ETRAX-FSinterface,by using a customized protocol, to its internal memory. 

Figure 9.2: Data transfer protocol between FPGA and ETRAX-FS. 
Asitcanbe seeninFig.9.2,theFPGA signalizeonportBpin16(PB16 -DATA VALID,Fig.9.2) thatthereisanewdataonthebusandif the ETRAX-FSisnotbusy,it confirms(PB17 -ACK, Fig. 9.2) that the data was written to its internal memory. After this, transmission of next word can be executed. 
This also canbedone withusing theDMA architectureimplementedinI/Oprocessors ofthe ETRAX-FS(seeFig.9.3).Inthis situation,theFPGA sendsthedatainpackets. Each consist of 15 data words -this is caused by input DMA FIFO size, which is only 16 words deep. When the data transfer is started, the ETRAX-FS I/O processor rises the busy signal immediately. 

Figure 9.3: Data transfer protocol between FPGA and ETRAX-FS with DMA usage. 
The algorithm prevents in this way situation in which the Linux processor indicates with unacceptable delay that it is busy meanwhile the FPGA sends a new event. This can happen when ETRAX-FS can not transfer the data via the 100 Mbit Ethernet to the Event Builder or it can not copy the data from the DMA FIFO to the already filled up internalmemorybuffer.In this case a communicationbetweentheI/O and mainprocessor 
CHAPTER 9. UPGRADE OF HADES DATA ACQUISITION SYSTEM 
is needed, which can take not fixed time. If this is not the case, the busy is released and next data packet can be sent. 
Table 9.1: Data transfer in the DMA mode with different data load. 
32 bit words per event  LVL1 frequency [kHz]  Speed [MB/s]  
14  124  11  
22  82  9.2  
38  44  8.5  
64  27  8.1  
110  18  7.9  
170  10.5  7.8  
640  3  7.7  

Moredetailsabouttheimplementationof thealgorithmsinthe ETRAX-FSprocessor and theI/O co-processors canbefoundin[94]. 
Asit was mentionedbefore,itis required tohave access to theinternal registers of the FPGA e.g. setting the delay between the reference time and trigger sent from the FPGA to the TDCs. For such situation a read/write protocol was developed. Since there are 36 lines between the FPGA and the ETRAX-FS and 34 already are occupied by the data transfer, the protocol had to be implemented on the remaining 2 lines -PC17 : 16(see Fig. 9.4). 
Data ACKIf read 

Figure 9.4: Data transfer protocol to read/write registers from/to the FPGA and also from/to other devices. 
Line PC(17)is used toinformtheFPGA that a newbit of thedatatransfer(PC16) 
isonavailable.Inthe first80bitsthereis(seeFig.9.4): 
CHAPTER 9. UPGRADE OF HADES DATA ACQUISITION SYSTEM 
• 	
bits3 :0indicate,from whichdevice user wants to read/write.It canbedirectly the FPGA or via specially adapted interfaces: DSP, SDRAM or memory block inside the SFP, 

• 	
15 -modether/waccess,ifbit15 =1thenthisisread modeotherwisewrite, 

• 	
47 :16 -addresstoread/write, 

• 	
79 :48 -datatowrite, sentevenintheread mode. 


Ifthisisthe read mode, after80bits ofthedatatransfer,the81stpulse (PC(17)) from the ETRAX confirms, that the line PC(16) changed its direction and the FPGA can start to send data. The described stand alone mode is useful for small applications and test setups.Thismodeisalsoused totestanewlyproducedTRBv2 with aspecially designed AddOnTester(seeFig.9.5a).Thisboard consistof oneLattice ECP2MFPGA, which canbeaccessedfromtheETRAX-FS ontheTRBv2(the sameread/writeprotocol as between the VIRTEX FPGA and ETRAX-FS ( see Fig. 9.4). The TRBv2 and the Tester AddOn are attached to each other via the AddOn connector. The TRBv2 is not visibleonthisfigure -itislocated undertheTesterAddOnboard.TheTesterAddOnhas sixteenclockdrivers[95].OneclockdriveracceptsonedifferentialLVDS signalfromthe LatticeFPGA and splitsitto8differential outputswith additiveinputtooutputjitter less than 1 ps (see Fig. 9.5b). 
With the AddOn Tester board it is possible to check if all the measurement channels areworkingproperly intermsof atimeresolution.During such testthereferencetimeis senttotheTRBv2 and alsoperoneeventtwoout of eightsignalsare senttotheindividual HPTDC channels. The time difference between such signals has to be constant and any deviation of the time difference measurements shows the time resolution of the HPTDC togetherwithTRBv2 PCB(PrintedCircuitBoard) layout,connectionsand cables. 
Fig.9.6presentsmeasurementsfortheHighResolutionmodeofthe HPTDC(96 ps HPTDC bin size). The mean value of two channels resolution is on the level of RMS = 0.426× bin size =40.9 ps. 
For the START and VETO detector there is a necessity to accomplish much better results.ThiscanbedoneinVery HighResolutionmode(24 ps HPTDC bin size). In this 

CHAPTER 9. UPGRADE OF HADES DATA ACQUISITION SYSTEM 
Connector to HPTDC Connector to HPTDC 


A 
B 
(a) Tester AddOn board top view with visible  (b) Block diagram of the AddOn Tester board.  
on the photograph cables going to the TRBv2  One signal sent from Lattice FPGA to 8 differ 
being under  a test.  ent HPTDCs channels.  


Figure 9.6: High Resolution (HR) mode. On the left side the channel to channel time resolution for all 128 TRBv2 HPTDC inputs. On the right side an accumulated RMS of all channel to channel measurements. 
CHAPTER 9. UPGRADE OF HADES DATA ACQUISITION SYSTEM 
mode, in the HPTDC, only one fourth of the channel numbers is available -the rest is used to improve the performance of the HPTDC in terms of the time resolution. 

Figure9.7:VeryHighResolution(VHR) mode.Ontheleft sidechannel tochannel resolution for all 32 TRBv2 HPTDC inputs. On the right side accumulated RMS of all channel to channel measurements. 
In such conditions, the whole test system delivers RMS =0.74× bin size =17.8 ps. 
The achieved time resolutions of the TRBv2 for the HPTDC High Resolution mode and the Very High Resolution mode are comparable with this what the manual of this ASIC demonstrates(see Tab. 9.2). 
Table 9.2: Intrinsic HPTDC and the test system time resolution comparison. 
Mode  HPTDC [ps]  TRBv2 [ps]  
High resolution mode  34  40.9  
Very High resolution mode  17  17.8  

This is due to the fact that the PCB layout was prepared with great carefulness to guarantee the impedance matching and the decoupling of the transmission lines of the LVDS-timingsignals.Toaccomplishcorrespondingoutcomeit wasalsonecessary toapply corrections for time to digital conversion non-linearities. Accordingly, a code-density test wasapplied(signalsdeliveredfromTester AddOn).Such testisbased ondelivering tothe 
CHAPTER 9. UPGRADE OF HADES DATA ACQUISITION SYSTEM 
HPTDCsalargenumberofthehits(toget asignificantstatistics)from someuncorrelated source.Afterward,it canbe seen,whichbin(arawvalueof themeasurement,inHRmode it is enough to check 8 bits, in VHR mode 12 bits) of the given channel is getting less or more data. The plot in Fig. 9.8 shows an example of such corrections for one channel in VHR mode. 

The main purpose of the TRBv2 is to serve as main readout board for the timing detectorsand tobeintegrated with the HADESDAQ.Sinceall theotherdetectorspecific electronic readout boards were supposed to be exchanged, the idea of one platform was also implemented in area of the trigger, data and slow control transfer. There is also one common protocol for the data/control/trigger transport to connect all boards -TRBnet. Thewholedescriptionof theTRBnetcanbefoundintheJ.Michael Ph.D thesis[96]. 
When describing the whole process of the event readout, it is necessary to start from the source of the triggers -the Central Trigger System(CTS). The whole process starts, whenanappropriatelayoutofdetectortrigger signals(seechapter 9.8.2) formsatrigger for the whole HADES system. In such situation the reference time and the LVL1 trigger ontheTRBnetsideare sent separately.Thereisalsoaconstant need to sendspecial types 
CHAPTER 9. UPGRADE OF HADES DATA ACQUISITION SYSTEM 

of triggersforaspecific subsystemtoperform suitableactions(seesection 9.8.2).Inthis situationonlyLVL1 trigger(with special triggertypenumber) viatheTRBnetprotocolis transferred. The TRBnet packets are sent via optical links to specially designed HADES HUBs(section 9.9) andtothe finaldestination -inthe shownexampletotheTRBv2 (see Fig.9.9). 
When the TRBv2 receives either the reference time or LVL1 TRBnet packet it starts the HPTDC readout. Then, it saves the event to its internal FIFO and transports it furthertoaTRBnetbuffer(stillinsidethe VirtexFPGA).Whenthisprocessis finished the LVL1 busy release signal is sent to the CTS. All other readout boards are acting in the same manner and all these systems signalize on appropriate time to the CTS, that the first phase of the readout is finished. Next, the CTS can send either new LVL1 trigger(togetherwith thereferencetime) orareadout request. Afterthereadout request, 
CHAPTER 9. UPGRADE OF HADES DATA ACQUISITION SYSTEM 
thecollected eventis senttothe HADES HUB(seechapter 9.9) and viaGbitEthernet furthertotheeventbuilders(EBs).Asit canbe seentheread/writeaccesstotheinternal registers of the Virtex FPGA can be done by using the TRBnet protocol from the Slow Controlboard(theTRBv2 canbeadoptedtofunctionasa sourceoftheSlowControl requests). There is also an independent way for r/w operations -via 100M bit Ethernet and Etrax FS processor. This alternative route is useful particularly for the stand alone systems. 
9.2 	Front-end and Readout Electronics of RPC Detector 

ThedevelopmentoftheRPCdetectortookplaceinlastfewyears.Finally,itisinstalled andintegrated withthe HADESdetector[61].Tomakepossible suchfinaltobecomea true and undoubtful success a few years of testing the prototypes and commissioning had to be in the background of such detector development. In parallel to the building of the detector also front-end and readout electronics was design and prepared. There were several significant stepsduring thisprocess.To showhowthiswholeprocessevolved in time, few of the steps will be highlighted. The first test under the beam conditions 
CHAPTER 9. UPGRADE OF HADES DATA ACQUISITION SYSTEM 
were completed in 2005 year. In Fig. 9.10 there is a visible RPC detector with partially mounted on it front-end electronics. 

This electronics consist of a Mother Board (MBO), which serves as a distribution platform for the power and threshold settings for the attached to it Daughter Boards (DBO).On theDBOareplaced operational amplifiers anddiscriminators.The thresholds ontheDBO are setfromtheTRBv1/2board viaSPI(SerialPeripheralInterface[97], see Fig. 9.12). 
At this point of time (TRBv1 was used) it was possible to achieve 86 ps of time resolution(detectorplusall electronicboards).This schemeof theelectronicschaindesign waskeptinthe nextphases ofthedetector andits electronicsdevelopment and commission time(Fig. 9.11). 
In 2007 the TRBv2 was already existing and also one full sector of the RPC detector was equipped with the front-end electronics. The readout part was connected with currently existing triggeranddatadistributionsystem(see section2.3). 
The configurationinFig.9.13 allowed to check thetrackingperformance with connection to the whole HADES system: 95% efficiency over selected large detector area and 
CHAPTER 9. UPGRADE OF HADES DATA ACQUISITION SYSTEM 

AddOn 

(a)RPC detector test in 2007, configuration of the readout setup. 	(b) Pictures taken during this test. 
Figure 9.13: 2007 year test setup. 
average time resolution of 81 ps. The obtained results were more than satisfying and at this point of time it was decided to go for the mass production of all 6 sectors. Built a stand alone readout system independent from the HADES DAQ made the detector test process much easier to perform. It was also prepared to connected two RPC sectors at the same time (see Fig.9.14). In this setup eight TRBv2 boards were used as readout 
CHAPTER 9. UPGRADE OF HADES DATA ACQUISITION SYSTEM 
electronics. One TRBv2 connected with the GP AddOn was playing the role of the CTS and the newly built optical Test HUB AddOn was used as a trigger and busy signals distributor(optical connections). 

Ethernet connection - for the FPGAs register settings 2 Gbit/s optical connections -for the LVL1 and LVL2 trigger/busy distribution Trigger input from the RPC multiplicity logic to the CTS Cables to/from the RPC detector (timing signals and treshold settings) 
(a)The block diagram of the setup used for the final RPC tests. 
(b) The pictures taken during installation of this setup. 
After all evaluationphases oftheRPCdetector,in2010it wasfullyintegrated withthe HADES system. After this significant effort, there were further test/commission beams. Thatallowed cametothefollowing results(still with someplacefortheimprovements) for all six RPC modules : 
• 
average time resolution : 73 ps 

• 
average position resolution : 8 mm 


CHAPTER 9. UPGRADE OF HADES DATA ACQUISITION SYSTEM 
• 	average efficiency 94 % 
9.3 	Front-end and Readout Electronics of TOF and Forward Wall Detector 

For the Time of Flight and Forward Wall detectors it is necessary to use a specially designed AddOn board, called TOF AddOn. The TOF AddOn has 128 channels, which fully covers one sector of the TOF detector. The information attained from the Photo MultiplierTube(PMT) signalis: 
• 	
arrival time -time of the particle crossing the detector, 

• 	
generated charge -energy deposited (energy loss) by the particle in the TOF/Forward Wall detector. 


Inordertoperformmeasurementsmentioned abovetheTOF AddOnisequipped with NINO[98] ASICs,which aredesigned atCERNforthe ALICETOFdetector.Theinput 
CHAPTER 9. UPGRADE OF HADES DATA ACQUISITION SYSTEM 
signalis splittotwoandfedintotwoindividualinputsof the NINOchips(slowandfast NINO, Fig.9.16). 

On the first one the rising edge of the signal is discriminated (with thresholds set from ETRAX-FS fig.9.17). On the second the falling edge is created, but first the signal is shaped and discriminated in the way that the time distance between the rising and falling edge is proportional to the particle energy loss. The signal generated out of these twoedgesis senttothe HPTDCplaced onthe,attached with AddOnconnector,TRBv2. 
To do correct discrimination in the NINO ASIC, the chip itself provides several settings, which can be fixed with applying appropriate voltages: 
• 
commonthreshold value(coarse) 

• 
stretching of the output signal 

• 
setting the hysteresis 

• 
channel threshold value(fine) 


First three parameters from the list can be applied individually for the fast and slow part of the measured signal. All these voltages are set from the ETRAX-FS, which has direct access to the DAC SPI lines via port E and the addon connector. 
CHAPTER 9. UPGRADE OF HADES DATA ACQUISITION SYSTEM 

As it was mentioned before, one TOF AddOn is connected to all 128 channels of one TOFsector.Thisleavesnospaceforthemeasurement of thereferencetime,demandedfor synchronization with other detectors. That’s why every 32nd channel is logically OR-ed with the reference time to overcome thisproblem. Desired additional electronics isplaced directly on the TRBv2 and its placed on the specially left open pads. In this situation, the signal coming from the hit and reference time has to be distinguished. This can be doneeasilyby applying timewindowcutsonthemeasured timesof the signals,which are suspected tobe referencetimes(Fig.9.18). 
With such configurationof thehardwaretheTOFAddOnintroduces30 ps additional jitter to the signal. The power dissipation for the individual channel is on the level of 80 mW . The process of the integration of the TOF/Forward Wall electronics took place between 2010 and 2011 and was successfully commissioned and approved for the future HADES experiments. 
CHAPTER 9. UPGRADE OF HADES DATA ACQUISITION SYSTEM 
Hits 	Ref. time Trigger 

9.4 	Front-end and Readout Electronics of START and VETO Detector 
The board used for the readout of the START and VETO detectors is the TRBv2. Since there is only 8 channels to read from one detector and also the time resolution is an importantfactorintheprocessofdifferentparticlesspecies separation,the HPTDCsare used in very high resolution mode. Just to remind -in this mode the time resolution is on the level of 20 ps and 8 channels per one HPTDC are producing the results of the measurements. 

(a)Top and bottom view. (b) START/VETO front-end electronics block diagram. 
Figure 9.19: START and VETO front-end electronics. 
As it can be seen in Fig. 9.19b the signals from the detector are coming directly to 
CHAPTER 9. UPGRADE OF HADES DATA ACQUISITION SYSTEM 
the discriminator. The discrimination levels are linked directly with voltages set on the DAC chip. The values of these voltages, like in case of the RPC front-end electronics, are set via the SPI with the chain of Etrax FS and the Virtex FPGA registers. After the discriminationthe signalsare splitintotwogroups:onegroupistransportedtothecentral triggersystemboard(CTS) andthe second oneislinked withthe HPTDCinputs.One signal out of this eight is treated in a different way. The reference time, which is essential for different systems synchronization is logically OR-ed exactly with the 7th signal. The obtained results from the 2010 test and commissioning beam time time regarding the RMS time resolution was σ =32 ps -detector, front-end electronics and measurement all together. This was possible after the corrections of measurements nonlinearities, signal walk corrections and cross talk rejection. 
9.5 Readout Electronics of Shower Detector 
The development of the Shower readout electronics can be divided into two phases. In the firstphasetheShowerreadoutprototypewasbuiltproviding atestplatformusedfor given hardware configuration: 
• 	
VariableGain Amplifiers(VGA)togainthe signalscomingfromtheShowerdetector Front EndBoards(FEB, [99]), 

• 	
8-channel10-bitanalog todigital converters(ADC, AD9212)(intotal96 channels on the board), 

• 	
Lattice LFE2-70E-5F900C FPGA chip, 

• 	
threeDACsfortheVGAgain setting(onechipperoneShowerdetectorplane). 


In this first step the TRBv2 was used for the readout. The data was transported through the AddOn connector to the TRBv2 Virtex FPGA and further to the Etrax FS and withUserDatagramProtocol(UserDatagramProtocol)packetstotheEventBuilder (EB).Thisfirst stageconfirmedthattheproposedboard architectureconceptisadequate totheneedsand thenthe finalboard wasbuilt(Fig.9.20). 
CHAPTER 9. UPGRADE OF HADES DATA ACQUISITION SYSTEM 100 

Detec
tor
 s
i
gn
a
l
s 
Detec
tor
 s
i
gn
a
l
s 


The Shower AddOn readout board placed in the HADES DAQ system operates in three different modes: 
• 
data taking, 

• 
calibration, 

• 
maintenance. 


Inthedatataking mode,theShower AddOnafterreceiving theLVL1 trigger(orthe referencetime)startsthereadoutprocess.ItforwardsthetriggertotheFEBs.TheFEBs constantly convert a charge coming from the detector to the electric signal, which amplitude is proportional to the charge value. These analog values are saved in the local 
CHAPTER 9. UPGRADE OF HADES DATA ACQUISITION SYSTEM 101 

capacitor CR. After receiving the forwarded trigger the outputs of FEBs are multiplexed totheShowerAddOnDACs,digitized and read outby theFPGAs(Fig. 9.20 -firstand secondFPGA marked onthepicture).Afterthisreadout sequenceall CR are discharged. Acquired valuesarecompared withpreviously setthresholds(maintenancemode,accessing FPGAs via TRBnet). The thresholds allow to reject noise coming from the detector. Next, the data is transferred to the third FPGA, where it is combined and sent further viaGbit Ethernettothe EventBuilders(EBs). 
When theboard operatesin the calibration mode(after receiving aShower calibration trigger via TRBnet from the CTS), the Shower AddOn starts the process of injecting a defined charge(the charge alternates10 pC with 0 pC)to the channels of the FEBs. In thenext step,thereadoutprocesstakeplace.Baseonthecollecteddatathedetectorcan be calibrated. 
9.6 Readout Electronics of MDC Detector 
The MDC readout electronics upgrade can also be divided into two phases. In the first phase like in the case of the Shower AddOn, a MDC readout prototype was built and it served as aplatformto checkthepossibility of replacingthe oldMDC readout chain. Again theTRBv2(likeintheShower AddOnupgradeprocess)wasused asa supportforthe data readout. In this first stage all connections were electrical and this type of connection wasproventobea sourceof noise.Thiswasthereasonof changing themediaforthedata transfer from the electrical to the optical links. Taking into consideration a limited space inthedetectorarea(400 readoutboards)the PlasticOpticalFiber(POF)[100]linksand FiberOpticalTransceivers(FOT,FDL300TfromFirecomms) weredetermined as abest candidatestoperformthedatatransmission.Thechosen POF technology ischeaperand easier to handle than glass fibers. 
Based on all experiences gained during the previous system usage and the first phase of the development of the MDC readout electronics it was decided to: 
• 
keepthedaughterboardsresponsibleforthedetector signalgainanddiscrimination, 

• 
continue to use motherboards, on which the time measurementisperformed(TDCs 


CHAPTER 9. UPGRADE OF HADES DATA ACQUISITION SYSTEM 102 

are placed there, Fig. 9.21a), 
• replace the whole readout chain with two types of boards: 
– 
Optical EndPointBoard(OEPB), 

– 
MDC HUB. 



(a)MDC mother board with attached to it daughter boards. 

(b) Optical End Point Board. (c) MDC HUB. 
Measurementsdoneby theTDCsareacquiredby theLattice ECP2M20placed onthe OEPB(viathe connectors marked withthe reddashedline,Fig.9.21a andFig.9.21b). Afterward,the collected eventis transmitted with250 Mbit/s speed totheMDCHUBvia PlasticOpticalFiber(bluedashedline,Fig.9.21b and fig.9.21c).Thedataisreceivedby theMDC HUB,whichisequipped with32 optical transceivers.Tocopewith such amount of connectionsanddata,thisboardisequipped withfiveLattice ECP2M100FPGAs.Four of this FPGAs are used for all data transfer to/from group of 8 OEPBs. The fifth one is usedtocollectallthedataand senditviaGbitEthernet(fig.9.21cyellowdashedline) 
CHAPTER 9. UPGRADE OF HADES DATA ACQUISITION SYSTEM 103 

tothe EBs.TheTRBnetentityisimplementedinall abovementionedFPGAs(OEPB and MDC HUB) and besides the data transfer it manages also the transfer of the LVL1 trigger, data readout request and slow control data. All transmissions are done on the same optical connections. 
To prove the stability of such big system an irradiation test has been performed at GSI.DuringHADESp+Nbproductionbeamtimeat3.5GeV ofkinetic energy,theOEPB was placed near the beam dump. The purpose of such test was to see the reaction to the expositiontodifferent reactionproducts(pions,protons, andheavierfragments).By usingdedicatedSingle EvenDetection(SED) component,availableintheLatticeFPGA, it waspossibleto observe an average of2 errorsperhour at aparticle rate of106 particles . 
scm2) 
Though, changing the position of the electronics to the comparable of the final OEPB location caused absence of the errorsinducedby the radiation[101]. 
9.7 	Front-end and Readout Electronics of RICH Detector 
IncaseoftheRICHdetector,inthephaseofthe HADESDAQupgrade,boththefront-end electronics and digital readout were redesigned and rebuilt. The overview of the system is presented in Fig. 9.22c. 
Thedetectorsignalsarelinked with theanalogfront-end card APVASICinputs.The APVASIChas128input channels,which amplifyshapeand savetheincomingsignalsto the internal analog memory. After the arrival of the LVL1 trigger, the Analog to Digital Converter Module (ADCM) initiates the readout of the APV (by two 12 bit, 40 MHz ADCs), digitizes the data, forms the data event and sends it with TRBnet protocol, via theopticallinkstothe HADES HUB(section 9.9). 
In order to offer the described functionality ADCM module contains the following components: 
• 
theoptical connectiontothewhole HADESDAQ system(SFP), 

• 
the Lattice FPGA ECP2M100 for the ADC control, online calculations, data com


pression and TRBnet functionality, 
CHAPTER 9. UPGRADE OF HADES DATA ACQUISITION SYSTEM 104 


(a)ADC module. (b) APV front-end. 

(c) Block diagram of the RICHreadout and front-end electronics. 
Figure 9.22: New RICH detector electronics. 
• 
RISC micro-controller for slow control and power supply quality constant check, 

• 
SPI flash for the automatic hardware boot procedure after the power cycle. 


9.8 HADES Trigger System 
9.8.1 HADES Detector Trigger Signals 
To start,describedinprevious subsections,thedetectorelectronicreadoutsadefinedpatternofdetectortrigger signalsisused.This signalsarrangementindicates,thatinteresting physics appearedduringtheion,proton orpion collisions.Based on this theLVL1 trigger and reference time is sent from the CTS. In the HADES spectrometer the main trigger detector sources are: 
CHAPTER 9. UPGRADE OF HADES DATA ACQUISITION SYSTEM 105 

• 
RPC, 

• 
TOF, 

• 
START and VETO. 


TheTOFhardwareforthesesignals,producedand senttotheCTS,islocateddirectly on the TOF AddOn. It consists of three levels of the summing modules. In the first level eight signals are grouped and summed. Four results of the sum are again summed and thenoutofthisfourthe final signaliscreated(fig9.23).Thecomponentsarepickedin theway,thatawholerangeof operational amplifiers(inthe summingpart) isused.The output signal is a linear function of the active inputs -20 mV per one channel. One has tobearinmind,that onehitintheTOFdetectorcreatestwo signalsontheTOF AddOn (left andright side of PMT) and the corresponding signal is 40 mV high. 

For the RPC multiplicity trigger signal production a special board had to be created (see Fig. 9.24). 
The multiplicity signals are initially created directly on the DBO from the sum of the four detector signals. After this, these eight signals are summed on the MBO (just to remind -eight DBO is attached to one MBO) and transported to the noticeable LEMO connectors. There are two visible LEMO connectors, which are placed in perpendicular way totheboard.Leftoneisusedto sendfinal multiplicity triggersignal(a sumof all incoming signals)tothecentral triggersystemboard.Onthe second oneitispossibleto 
CHAPTER 9. UPGRADE OF HADES DATA ACQUISITION SYSTEM 106 


(a)Top view. (b)BottomviewwithvisibleAddOnconnector. 
Figure 9.24: Trigger multiplicity board for the RPC detector. 
observeinputsignals.The selecting oftheincomingsignalisdoneviaSPIbyprogramming amultiplexer.Themultiplicityboardisattached via AddOnconnector(fig.9.24b) tothe TRBv2.This allows to adjust remotely the thresholds(theDAC voltages) todiscriminate the RPC detector signals. The access to the DAC and multiplexer chain is through the EtraxFSprocessor -likeintheTOFAddOn(fig.9.17).All arriving RPCtriggersignals are going to the TRBv2 and are counted in the Virtex FPGA. Then the counter values aretransported with a specialtriggertype(0xE)along with collecteddatatotheevent builder. This allows to check the performance and behavior of the specific RPC detector areas. It is also possible to send a test signals from the individual TRBv2 to check, if a related RPC part detector and electronics function in a proper way. 
Thetrigger signalscreatedintheTOFandtheRPCdetectors(timing wallsignals)are summedtogetherontheboard,whichisthe sameasRPC multiplicityboard(the same length of the trigger signal cables is assured for this two systems). With such difference thattothisboard only a small addonisattached(instead of theTRBv2).Theconnectors which are marked in Fig. 9.25 provide: 
1. 
the RPC and TOF trigger multiplicity signals from the individual sectors linked with the CTS, 

2. 
three different multiplicity signals of the whole timing wall (the TOF and RPC detector together) -the multiplicity value corresponds to the particular threshold 


settings, 
CHAPTER 9. UPGRADE OF HADES DATA ACQUISITION SYSTEM 107 

3. 
SPI(programmingfromtheTRBv2)forthethresholdsregulation-individualinputs and aggregated timing wall multiplicity signal discrimination, 

4. 
SPI for selection, which input can be observed on the dedicated LEMO output (bottom side ofthe board) 



012345 012345 TOF sectors RPC sectors 
Figure 9.25: Central multiplicity unit. 
Incaseof theSTART and VETOdetectorsthetrigger signalsarelinkeddirectlyfrom the front-end electronic boards to the CTS (fig. 9.19b). The other important function, which the central multiplicity unit is responsible for, is compensation of the base line (groundlevel)shiftindifferentpartsof thedetector.Otherwise,itwouldbenotpossible to relay on the input signals voltage levels especially when the lowest one is 40 mV . 
9.8.2 Central Trigger System -CTS 
The backbone of the new DAQ architecture are the fast optical connections. In this circumstances also the CTS along the other readout systems required rebuilding and a newboardhasbeendesigned and constructed(Fig.9.26). 
The essential thing, which the CTS has to accomplish as piece of hardware, is to accept signalsincomingfromthe selecteddetectors:TOF(6inputs),RPC(6),START (8), VETO(8)and otherphysical triggersPT(8) and adjustthem, suchitispossibleto 

CHAPTER 9. UPGRADE OF HADES DATA ACQUISITION SYSTEM 108 

2 
1 
createthe firstleveltrigger signals(theLVL1andreferencetime). Thearchitectureof the algorithms running on the CTS has to take into account different conditions for the particularbeamtimes(heavyion orpioninducedbeams). Thisis assuredby applying adjustable VHDL components inside one of the FPGA designs (Lattice SCM (marked as 1 in Fig.9.26) mounted on the board. Exactly this FPGA, is used to make all the operationsontheincoming detectorsignals.The second one(Fig.9.26 -2) isadoptedfor the standard HADES DAQ operations (via two optical connections) with the TRBnet protocol: 
• 	
slow control channel -accessing the CTS read/write registers, 

• 	
sendingtheLVL1triggerand readout requeststothe HADESdetectorDAQsystems (formed in the first FPGA), 

• 	
transmitting data from the CTS to the event builders. 


SincetheTRBnetprotocolisimplemented only onthe secondFPGA,totransferdata 
CHAPTER 9. UPGRADE OF HADES DATA ACQUISITION SYSTEM 109 

from/tothe firstFPGA,aspeciallydedicatedprotocolisimplementedbetweenthesetwo chips.Onceallthedetectorspecifictriggersignalsarecreated and senttotheLatticeSCM FPGAplaced ontheCTS,abasicmanipulationsontheincomingsignalsare started.The first logical input component is used to create out of the rising edge of incoming signal a pulse. This is due to the fact that this information is sufficient for forming the final LVL1 trigger signal(and thereferencetimealong it).Itisalsoeasiertoperformalllater manipulationsonthiskind of the signal.Theaccuracy of theedgedetectionisonthelevel of1.25 ns.Thisis accomplishedby using theDDR architectural elements availablein this FPGA, where the input signals are sampled with 800 MHz. Since the internal logic runs with 200 MHz clock the signal, which is created after DDR component is a 4 bit vector (see Fig.9.27). Value of this vector represents the sub-clock precision. Consequently in later steps, actions taken on the signals can be done with 1.25 ns precision and one has to remember that all these operations are done on the 4 bit vectors (if not mentioned otherwise). 

After the pulse creation the signal can be delayed. This is required in the case when from the different detector systems the detector trigger signal travel different distances. ThisittrueforexamplefortheTOF/RPC signalsandtheSTARTdetector signals.There are two types of delay components in the FPGA. One group is used to delay the signal with 1.25 ns precision. As it can be seen in Fig. 9.28 first the signal is rewritten to the new12bitvector -inthepositiondefinedby thedelay value(intheshownexamplethe delay equals to5×1.25 ns).Inthenext stepsthisvectoris savedtothetemporary vectors A1,B2, C3 each after every clock cycle(5 ns).The finalsignaliscreatedout of thisthree 
CHAPTER 9. UPGRADE OF HADES DATA ACQUISITION SYSTEM 110 

vector signals by the OR operation. 

Unfortunately such precision can not be used when it is necessary to delay signals by larger times(up to 1.3 µs). This wouldtake too much hardware resources of the FPGA. Therefore,the secondgroup of thedelay componentshasgranularity of5 ns and allows to delay up to8 signalsinparallel.Tomakethispossible,togetherwith thelowusageof the resourcesof theFPGA,thearchitectural component -theFIFOisused(seeFig.9.29). 

Sometimes it is desired to suppress one of the types of the incoming trigger. In such 
CHAPTER 9. UPGRADE OF HADES DATA ACQUISITION SYSTEM 111 

circumstances it is possible to downscale a given trigger signal and transport only each 2downscalevalueth rising edgeof this signal(seeFig.9.30). 
00 


Figure 9.30: Downscaling the incoming trigger signal. 
Whenthe signalsarealigned toeach other(afterapplying acorrectdelay values),the next logical step is to widen the signals. This is done in order to get an overlap between the interesting signals. Again, like in the delay components out of the same reasons there are two types of this components. With the first type it is only possible to enlarge the signal with multiplication of5 ns.Themechanismof thisalgorithmisquite simple.When width value is larger than 0 then according to the position of the input rising edge the corresponding output rising and falling edges are created -a Look Up Table is used for thispurpose(LUT). Accordingly theoutputsignaliscreatedfromthesetwoedgesand in between(w −1)0× F vector signalsareinserted(seeFig.9.31). 

If thefrequency of the signals,which areusedintheprocessof finding thecoincidence is large the 5 ns granularity of the width enlargement is not enough. This is the case when START and VETO signals are used for the final trigger creation. In this situation 
CHAPTER 9. UPGRADE OF HADES DATA ACQUISITION SYSTEM 112 

a special component is used, which can enlarge the signal with a 1.25 ns step. This component consistsofROM whichisworking asaLUTandbased oninput signal creates immediately the output vector(seeFig.9.32). 

Figure 9.32: Setting the width with 1.25 ns resolution. 
It is necessary to underline, that for the sake of description simplicity of the above components, this document does not cover some special cases of the incoming signal configurations and how these signals are handled. 
Besides the mentioned functionalities (see Fig. 9.33 -full CTS trigger logic block diagram) the trigger logic can perform: 
• 	
enable/disable inputs and outputs, 

• 	
RPC/TOF sector-wise coincidences, 

• 	
enable anti-coincidence logic between the START and VETO signals, 

• 	
realize correlations between the START and VETO signals with all other trigger signals. 


Duringthe systemoperationitismandatorytohaveapossibilitytoobserveallsignals, which are undergoing all manipulations mentioned above. This allows to properly set parameterslikedelaysandwidthstomakeapropercoincidencebetweenadifferent signals. 
This is assured by using two multiplexers, each one forwards selected signal to the output connectorwired with theoscilloscope.Intotal thereare110internalFPGAsignals which can be observed: 
• 	
after the edge detection, 

• 	
after the delay, 


CHAPTER 9. UPGRADE OF HADES DATA ACQUISITION SYSTEM 113 

Figure9.33:CTStrigger logic. 

Multiplicyty logic 

Start part A 
0x8 input output (pulse)
INPUT_ENABLE(0xC3(0)) 

I 
CHAPTER 9. UPGRADE OF HADES DATA ACQUISITION SYSTEM 114 


• 
after the width enlargement, 

• 
START and VETO signals used for the coincidence signal creation. 


To see these signals in appropriate way the internal 4 bit vectors are changed back with output DDR components to the LVDS signal. 
OtherimportantfeaturethatCTSgrantsis a convenientjudgmentofthebeamquality settings. The CTS creates from the START, VETO, RPC or TOF trigger signals ten different histograms (see Fig. 9.34). Each histogram has 500 samples and each sample correspondstothenumberofhitsinagivenadjustabletime. Anexampleof usageof this histograms can be a correct beam focusing on the target. When the beam axis is moving during one spill it is visible on the acquired histograms from individual START detector channels. It is noticeable that maximum number of counts moves from one side of the detector(channel) tothe other. 
9.8.3 Trigger Distribution 
Intheprevious subsection,themeanshowtocreatetheLVL1trigger,areindetailed man-ner presented. Aim of this section is to clarify how the CTS cooperates and is connected with other detector readout systems. 
CHAPTER 9. UPGRADE OF HADES DATA ACQUISITION SYSTEM 115 


The CTS when accepting the given configuration of trigger coincidence signal has to take into account the busy (e.g. readout electronics is during the process of retrieving the data from the front-end) of the whole system. If all subsystems are ready to take a new event, then in such situation the LVL1 trigger can be sent via TRBnet protocol to the readout electronics. Inside the trigger data packet, transported via the optical links, there is a trigger number and trigger type. The trigger number allows to detect if any mismatch of the events collected on the tapes occurred. The trigger type informs what kind of action should be taken by the readout electronics: 
• 
type 0x1 -normal data readout, 

• 
0x9 -MDC calibration, if enabled the CTS is sending this type of trigger, 

• 
0xA -Shower calibration, sent from the CTS on user request, 

• 
0xB -Showerpedestals,if enabled theCTSis sending thistypeof thetrigger,when 


CHAPTER 9. UPGRADE OF HADES DATA ACQUISITION SYSTEM 116 

there is a spill brake, 
• 	
0xC -RICH calibration, sent from the CTS on the user request, 

• 	
0xE-statusinformation,sent each second,thesubsystemsshould send actualstatus of its internal state machines, registers etc. 


Inparallel to theLVL1 digital trigger theCTS forms an analog reference time trigger. This trigger is sent with copper cables as PECL/LVDS 100 ns signal. This signal allows to synchronize data from the different detectors. 
After each trigger the CTS waits for response from all subsystems which tells that full event was collected inside internal memories of readout electronics. After this it can sent the next LVL1 trigger and in parallel to it asks with readout request to send the datatothe EventBuilders.Inthereadout requestpackettheCTSinsertsthe EBnumber (corresponds todefinedbeforeIPaddress), to which thehardware responsibleforforming the final event(HADES HUB,Shower AddOn,MDC HUB,RICH readout electronics) should send current event. In the new HADES DAQ system it is possible to have up to 16 EB and events are sent with round robin algorithm to these computers. How many eventsand which EBisusedisdefinedjustby writing thecorresponding valuestothe CTS registers. 
9.9 HADES HUB 
Toconnect all subsystemsintooneoptical network the HADES HUBboard wasprepared (see Fig.9.36). The main purpose ofthis board is to transport all of the internal HADES DAQ system data such as: 
• 	
LVL1 trigger, 

• 	
readout request, 

• 	
read/writeaccesstotheregistersfrom/todifferent subsystems(slowcontrol), 

• 	
collect incoming detector data and send it to the EBs via the Gbit Ethernet. 


CHAPTER 9. UPGRADE OF HADES DATA ACQUISITION SYSTEM 117 


To cope with such amount of the data the board is equipped with 20×2.5 Gbit/s optical connections controlled by two Lattice ECP2M100 FPGAs. In both FPGAs the TRBnet components are located and in the first one, to send the data to EBs, the Gbit Ethernet controller is implemented. 
9.10 HADES DAQ Summary 
During the HADES tests and commission beam times in the late 2010 the whole updated system was able to take for example in U + U collisions at 1.2GeV a 500 MB/s with a 55kHz of acceptedtriggers.Howbigtransformationitisonecan seewhencomparing with theoldsystemperformance -only ∼ 2kHz forCa+Ca system.The whole overview of the HADES DAQ structure can be seen in Fig. 9.37. The multiplicity signals, created by the detectorsastheresultofthebeamtarget reaction, are senttotheCTS(shownasdarkblue lines).TheCTSlogicmakescorrelationsamong thesesignalswith thepredefined settings (via TRBnet-orange lines) and generates the LVL1 trigger transferred via TRBnet. In parallelthereferencetimeis sentfromtheCTS toall readoutboards(greenlines).When the readout electronics receive the reference time, they start to acquire data from the front-ends or like in the case of the TRBv2 from the HPTDCs. 
When all events are collected in the internal memories of readout boards and busy 
CHAPTER 9. UPGRADE OF HADES DATA ACQUISITION SYSTEM 118 


Figure9.37:Structureof the HADESDAQsystem.Thereare several typesof connections forproper systemoperation.ThedesignedsystemisabletoacceptLVL1/readoutrequests at ∼ 55 kHz for heavy ion beam and up to 100 kHz for light particle reactions. 
release signals are sent from all subsystems to the CTS, the CTS transfers a readout request or the nextLVL1 trigger(together with the reference time).When the readout 
CHAPTER 9. UPGRADE OF HADES DATA ACQUISITION SYSTEM 119 

request reaches to the readout electronics the data is transferred directly to the Event Builders(EBs,violetlines).If thereisnodirect connectiontotheGbit Ethernetthedata is transferred via TRBnet to the HUBs and then further to the EBs. 
Chapter 10 
Forward Wall Readout Electronics 
As the Forward Wall was newly build detector therefore it was also needed to build electronicsfor thedata acquisition.For thispurpose aTRBv1(TriggerReadoutBoard version 1) was used. 

The TRBv1 has 4 input connectors (Fig. 10.1,1) (80 pins), which application is to 
CHAPTER 10. FORWARD WALL READOUT ELECTRONICS 
provide31LVDStiminginput signals(LVDS -Low-VoltageDifferentialSignaling) and alsospecialsignals(describedinthenextpart of this section).Ithasfour32 channelHigh Performance TDCs ASICs [89](HPTDC,fig. 10.1,2). Each 32nd channel is connected to theexternallygenerated(see2.3.1) referencetime signal.Thisisnecessary to synchronize thedatawith all other subsystems.The HPTDCASICisflexiblesuchit allowstomeasure the arrival time of the rising or the falling edge (or both), arrange different HPTDC bin widths(from780ps to 25ps), disable/enable individual channels and prepare set of parameterslike(see fig.10.2): 
• 	
trigger latency -fixing the point in the past from which measurement data should beaccepted(itisadvisedby the HPTDC manual nottoexceed25 µs), 

• 	
matching window -time period from which data is saved, 

• 	
reject latency -time after which the data is discarded to prevent overflows in the internal buffers of the HPTDCs. 


Old 
Hits Trigger data 

Figure 10.2: Windows settings for HPTDC chip. 
Allthese settingsare setviaindividualJTAG(JointTestActionGroup) interface. First of all this configuration data has to be prepared. This was done in several steps: 
• 	
modifying therequested HPTDCparametersinaheadprepared standard setup file with all default settings, 

• 	
running the previously written C++ program to create the configuration files in StandardTestand Programming Language(STAPL[90]), 


CHAPTER 10. FORWARD WALL READOUT ELECTRONICS 
• 	starting the Jam Player [90] adopted to the on-board Etrax MCM computer chip [91] with Linux running on it. The running Jam Player interpreter drives Etrax port pins, which are corresponding to the standard JTAG lines TDI,TDO,TCLK,TCK,TMS. 
The sameinterfaceisused toprogramtheSpartan2EFPGA(5)(FieldProgrammable Gate Array). This chip is used as a central entity for the data flow management and it is connected to the LVL1(fig. 10.2,3) and LVL2(4) data buses, FIFO(6), two memories(7), Etrax MCM(8), and HPTDCs. 
When the LVL1 trigger and the reference time signal arrives, first the data is selected and saved in the HPTDCs. Next, the FPGA begin the readout by sending the token to the TDC chain. The HPTDCs, one after the other, send the data to the FPGA. 

After this stage the FPGA builds an event, which is sent further to the FIFO. The stored eventswaitforthe secondlevel triggerdecisionarrival totheFPGA.If thedecision 
CHAPTER 10. FORWARD WALL READOUT ELECTRONICS 
is positive they are transported to the memories, if not it is readout from the FIFO without saving them. Afterwards, the FPGA signalizes to the Etrax, that it can start the readout of the collected events. Then, the data is encapsulated by the Etrax with the standard HADESheaderandfetch via UserDatagram Protocol(UDP),over100MBit Ethernet(fig.10.2,9),tothe eventbuilder(EB). 
With this schemeit waspossibletorunwith80kHzacceptedLVL1 triggersand20kHz LVL2 triggers(forfourteen32bitdata wordsper event), which correspondsto1.2MB/s of the data transfer. 
Itis asloimportantto mention,thattheTRBv1 uses agalvanicly isolated48Vpower supply(fig. 10.2,10), which makes much more simple power distribution. This avoids ground loops and gives the possibility to mount the TRBv1 directly on the detector frame. 
Bibliography 
[1] P.Koch,Z. Pchys.C57(1993) 283. 
[2] S. Leupold, V. Metag, U. Mosel, Hadrons in strongly interacting matter, Int.J.Mod.Phys.E19, 147(2010). 
[3] Y. Nambu and G. Jona-Lasinio, Dynamical Model of Elementary Particles Based on an Analogy withSuperconductivity.II, Phys.Rev.124,246(1961). 
[4] Nambu AndJona-LasinioModel, Phys.Lett.B249(1990),386-390. 
[5] NA60 Collaboration, (R. Arnaldi et al.). May 2006. 5pp., Phys.Rev.Lett.96:162302,2006. 
[6] MKaskulov,E HernandezandEOset, Eur. Phys.J. A31,245(2007). 
[7] PM¨uhlich andUMosel, Nucl. Phys. A773,156(2006). 
[8] TIshikawaetal, Phys.Lett.B608,215(2005). 
[9] DielectronproductioninAr+KClcollisionsat1.76AGeV.,ByHADESCollaboration 
(G.Agakishievetal.).Mar2011.16pp, Phys.Rev.C84,014902(2011). 
[10] M. Abdel-Bary etal.(COSY-TOFCollab.), Phys.Lett.B647,351(2007). 
[11] 47,21(2011).S.WheatonandJ.Cleymans,Comput. Phys.Commun.180,84(2009). 
A. I. Titov, B. K¨mpfer, and B. L. Reznik, Phys. Rev. C. 
[12] http://pdg.lbl.gov/. 
[13] M. Gell-Mann, ”The Eightfold Way: A Theory of Strong Interaction Symmetry,” CaliforniaInstitute ofTechnology SynchrotronLaboratory ReportCTSL-20(1961), unpublished.; S. Okubo, Note on Unitary Symmetry in Strong Interactions, Prog. Theor. Phys.27(1962). 
[14] C.Casoetal.,RevofParticle Properties Eur. Phys.J.C3(1998) 1. 
[15] ParticleDataGroup, Phys.Rev.D54(1996) 25. 
[16] Y.NambuandJ.Sakurai,Phys.Rev.Lett.8(1962) 79. 
[17] S.Okubo, Phys.Lett.B5(1965) 165;I.Iizuka, Prog.Theor. Phys.Suppl21(1966) 37; G.Zweig,Cernreportno.8419/th412(1996); S.Okubo Phys.Rev.D16(1977) 2336. 
[18] H.J.Lipkin, Phys.Lett.60B(1976) 371. 
[19] Davies D. et al. Phys. Rev. D. 1970. V. 2. P. 506. 
[20] Danburg J. et al. Phys. Rev. D. 1970. V. 2. P. 2564. 
[21] Abolins M. et al. Phys. Rev. Lett. 1963. V. 11. P. 381. 
[22] Ayres D. et al. Phys. Rev. Lett. 1974. V. 32. P. 1463. 
[23] Cohen D. et al. Phys. Rev. Lett. 1977. V. 38. P. 269. 
[24] Baldi R. et al. Phys. Lett. B. 1977. V. 68. P. 381. 
[25] Woodworth P. et al. Phys. Lett. B. 1976. V. 65. P. 89. 
[26] DorofeevV.etal.(Lepton-FCollaboration) Proc. NAN’95Conf.,Moscow,1995. 
[27] Aguilar-Benitez M. et al. (LEBC-EHS Collaboration) Z. Phys. C. 1989. V. 44. P. 
531. 
[28] Blobel V. et al. Phys. Lett. B. 1975. V. 59. P. 88. 
[29] GolovkinS.etal.(SPHINXCollaboration) Proc. NAN’95Conf.,Moscow,1995. 
[30] Aguilar-Benitez M. et al. (LEBC-EHS Collaboration) Z. Phys. C. 1991. V. 50. P. 
405. 
[31] Cooper A. et al. Nucl. Phys. B. 1978. V. 146. P. 1. 
[32] Donald R. et al. Phys. Lett. B. 1976. V. 61. P. 210. 
[33] Chen C. et al. Nucl. Phys. B. 1977. V. 130. P. 269. 
[34] V. P. Nomokonov, M. G. Sapozhnikov Experimental tests of the Okubo Zweig Iizuka rule in hadron interactions, Phys.Part.Nucl. 34 (2003) 94-123; Fiz.Elem.Chast.Atom.Yadra 34(2003) 189-238. 
[35] Bertin A.etal.(OBELIXCollaboration) Phys.Lett.B.1996. V.388.P.450. 
[36] Ellis J. et al. Phys. Lett. B. 1995. V. 353. P. 319. 
[37] Ulf-G.Meißneretal.,Phys.Lett.B408,381(1997). 
[38] M.P.LocherandYangLu,Z. Phys.A351,83(1994) ;D.BuzatuandF.M.Lev, Phys.Rev.C51,R2893O.Gortchakovetal.,Z. Phys.A353,447(1996). 
[39] L.G.Landsberg, Phys. Usp.37,1043(1994). 
[40] L.G.Landsberg, Phys.Rep.320,223(1999). 
[41] A.Sibirtsev, Nucl. Phys. A604,455(1996). 
[42] M. Hartmannetal., Nucl. Phys.A755,459(2005). 
[43] F.Balestraetal., Phys.Rev.Lett.81,4572(1998); F.Balestraetal., Phys.Rev.C 63, 024004(2001). 
[44] V.Blobel etal.,Nucl.Phys.B69,237(1974). 
[45] M.W. Arentonetal., Phys.Rev.D25,2241(1982). 
[46] S.V.Golovkinetal.,Z. Phys.A359,435(1997). 
[47] Landolt-Bornstein, NewSeries,Springer,I/12(1998). 
[48] A. Sibirtsev, J. Haidenbauer and U. -G. Meißner, Aspects of φ-meson production in proton-proton collisions, Volume 27, Number 3, 263-268, DOI: 10.1140/epja/i200510269-9. 
[49] “http:www.gsi.de/beschleuniger/Beschleuniger e.html” 
[50] Alvarez, L. et al. Berkeley Proton Linear Accelerator, UCRL-236(rev.); OSTI ID: 4375045, October 13, 1953. 
[51] G. Agakichiev et al., ”The High-Acceptance Dielectron Spectrometer HADES”, Eur.Phys.J. A41:243-277, 2009 
[52] F.M. Newcomer,IEEETrans. Nucl.Sci.40(4)(1993) 630. 
[53] L. Landau , On the Energy Loss of Fast Particles by Ionization, J. Phys. USSR 8 (1944)201 
[54] J. Markert , A. Schmah , C. M¨untz , J. Stroth, Energy loss measurement with the HADES drift chambers using the Time-over-Threshold method. 
[55] Thomas Bretz. Magnetfeldeigenschaften des Spektrometers HADES. Master thesis, Technische Universit¨at M¨unchen, 1999. 
[56] Cherenkov,Pavel A.(1934). Visibleemissionofcleanliquidsby actionof γ radiation. Doklady Akademii Nauk SSSR 2: 451. Reprinted in Selected Papers of Soviet Physicists, Usp.Fiz. Nauk93(1967) 385. 
[57] B¨ohmer, M.: Das Auslesesystem f¨ur den Ringabbildenden Cherenkovdetektor im HADES Spektrometer. Diplomarbeit, Physik-Department der Technischen Universit¨at M¨unchen, Institut E12, 1999 
[58] E.Berdermannetal.,Nucl.Phys.(Proc.Suppl.) B78,533-539(1999). 
[59] W.Koenig et al.,DiamondsastimingdetectorsforMIPS:The HADESproton-beam monitor and start detectors, INTRUMENTS-METHODS-02, GSI Scientific Report 2008. 
[60] R.Santonico,R.Cardarelli, Nucl.Instr.andMeth.A187(1981) 377. 
[61] A.Blanco: RPC HADES-TOF wall cosmic ray test performance, Nucl. Instr. and Meth. A(2010), doi:10.1016/j.nima.2010.08.068 
[62] D. Belve, et. al., Performances of the Front-End Electronics for the HADES RPC TOF wall on a 12C beam. 
[63] M. Traxler, Real-Time Dilepton Selection for the HADES Spectrometer Inaugural-Dissertation zur Erlangung des Doktorgrades der Naturwissenschaften der JustusLiebig-Universit¨at Gießen, 2001. 
[64] Petri, M.: Entwicklung eines kombinierten Auslese-und Triggersystems zum Nachweis von Elektron/Positronen-Signaturen in einem elektro-magnetischen Schauerdetektor, PhD. thesis, Universit¨at Gießen, 2001. 
[65] Lehnert, J.: Echtzeit-Mustererkennung zum Elektronennachweis mit einem RICH-Detektor in relativistischen Schwerionenkollisionen, PhD thesis, Universit¨at Gießen, 2000. 
[66] M.Sanchez,Momentumreconstructionandpionproductionanalysisinthe HADES spectrometer at GSI, Ph.D. thesis, University of Santiago de Compostela, 2003. 
[67] R. Brun and F. Rademakers, ROOT, an object-oriented data analysis framework, Nucl.Instr.Meth. A389(1997) 81. 
[68] W.Przygoda http://hades-wiki.gsi.de/cgi-bin/view/Homepages/PAT. 
[69] http://www-hades.gsi.de/computing/pluto/html/PlutoIndex.html. 
[70] http://web-docs.gsi.de/holzmann/hadsimul.html;http://wwwasdoc.web.cern.ch/wwwasdoc/geant 
[71] K.Teilab,Theproductionof η and ω mesonsin3.5GeVp+pinteractionin HADES, Ph.D. thesis. 
[72] J. Markert, Untersuchung zum Ansprechverhalten der VieldrahtDriftkammern niedriger Massenbelegung des HADES Experiments, Ph.D. thesis. 
[73] H. Wind, Nucl. Instr. Method 115(1974) 431. 
[74] W.H. Press, Numerical Recipes, 3rd Edition, Cambridge University Press(2007) 
[75] http://hades-wiki.gsi.de/cgi-bin/view/SimAna/Apr07GlobVertexPos. 
[76] DISTOColl.(F.Balestraet.al), Phys.Rev.C63,024004(2001). 
[77] A. Rustamov http://hades-wiki.gsi.de/cgi-bin/view/SimAna/Apr07Normalization. 
[78] R. C. Kammerud et al. Large angle proton proton elastic scattering at intermediate momenta. Phys. Rev., D4:1309–1324, 1971. 
[79] A. Rustamov. Exclusive η meson reconstruction in a proton-proton collisions at 2.2 gev withthehades spectrometer andhighresolutiontracking.Disseration.Technische Universitat Darmstadt, 2006. 41, 45, 81. 
[80] IrakliKeshelashviliPh.D.thesisKaonPairProductioninppCollisionsatthe ANKE Spectrometer. 
[81] TOFCollaboration:S. Abd El-Samad etal., Phys.Lett.B522,16(2001). 
[82] F. Hibouetal., Phys.Rev.Lett.83,492(1999). 
[83] A. Baldini et al., in Numerical Data and Functional Relationships in Science and Technology,TotalCross–SectionsforReactionsof High EnergyParticles, NewSeries, Group 1,Vol.12(Springer–Verlag,Berlin, Heidelberg,1998). 
[84] R.Baldi etal., Phys.Lett.B68,381(1977). 
[85] V.Blobel etal.,Phys.Lett.B59,88(1975). 
[86] http://www.gsi.de/portrait/fair.html. 
[87] M. Palka et al., The New Data Acquisition System for the HADES Experiment, Nuclear Science Symposium Conference Record, 2008, NSS 08, IEEE, p. 1398-1404. 
[88] I. Fr¨ohlich et al., A General Purpose Trigger and Readout Board for HADES and FAIR-Experiments. IEEE Trans.Nucl.Sci. 55(2008) 59-66. 
[89] HPTDC, Jorgen Christiansen, Digital Microelectronics Group, CERN 
[90] Altera, http://www.altera.com/support/devices/tools/jam/tls-jam.html. 
[91] ETRAX, AXIS Communications, Sweden 
[92] http://focus.ti.com/docs/prod/folders/print/tlk2501.html. 
[93] http://www.analog.com/en/ 
[94] R.Trebacz : Investigation of dielectron production in quasi-free p − n scattering at 
1.25 GeV with HADES, PhD. thesis, Jagiellonian University, 2011 
[95] http://focus.ti.com/docs/prod/folders/print/cdclvd110.html?CMP=AFCconv SF SEP 
[96] J.Michael PhD thesis in progress, Johann Wolfgang Goethe-University, Frankfurt, 2011 
[97] http://en.wikipedia.org/wiki/Serial Peripheral Interface Bus 
[98] F. Anghinolfi, P. Jarron, F. Krummenacher, E. Usenko and M. C. S. Williams, “NINO: An ultrafast low-power front-end amplifier discriminator for the time-offlight detector in the ALICE experiment,” IEEE Trans. Nucl. Sci. 51, 1974(2004). 
[99] A.Balanda et al., “Development of afastpad readoutsystemforthe HADESshower detector,” Nucl. Instrum. Meth. A 417, 360(1998). 
[100] http://www.firecomms.com/products-optolock.html. 
[101] A. Tarantola et al., The Upgrade of the Multiwire Drift Chamber Readout of the HADES Experiment at GSI. Nuclear Science Symposium Conference Record, 2008, NSS 08, IEEE, p. 2146-2149. 
[102] M.Raymondet al., “TheCMSTracker APV250.25 µmCMOSReadoutChip,”, Proceedings of the6th workshop on electroicsforLHC experiments, CERN/LHCC/2000041, 130-134.