MIMOSA Analysis Framework (MAF) used in Test beam

MIMOSA Analysis Framework (MAF) used in Test beam

MIMOSA Analysis Framework (MAF) used in Test beam and what a test beam analysis software should be able to do (personnal point of view) Short overview of MAF / test beam Shopping list Alignment issues Data analysis issues Auguste Besson (IPHC-Strasbourg) Strasbourg telescope Telescope 8 reference planes silicon microstrips, 256 x 50 m Trigger: 2 scintillitators planes (2x4 mm2 and 7x7 mm2) Spatial resolution ~2 m/plane Resolution in the DUT plane ~ 1 m Cooling system DUT rotation possible Data taking Online monitoring Rate

~ ~ ~ ~ 50 000 evts / hour (M9, 8k pixels) 2500 evts / h (M5, 500k pixels) 15-20 Runs / week 100s Go / week. Off line analysis first results available in few hours Acquisition and monitorin PC EUDET, Geneva, January 25th 2007 Auguste Besson Analysis PC 2 General structure of MAF (not optimal) Originally the program developped by RD42 collaboration (for strips) ~ 6 different authors Structure not optimized but take advantage of all the mistakes already made ! 1) 2) 3) 4)

5) 6) Generate eta function of the telescope planes Alignment of telescope planes (from 2 fixed planes) Reconstruction (hits in DUT and tracks selection) Generate eta function of DUT Alignment of DUT Analysis (eff, noise, S/N, etc.) Store: Alignment parameters of the telescope planes Eta functions parameter of telescope planes Alignment parameters for DUT Eta function parameters for DUT ? EUDET, Geneva, January 25th 2007 Auguste Besson 3 eta functions for optimised resolution Impact position given by CoG inside a 20x20 mm2 pixel Impact position given by eta function inside the pixel 1/ CoG is not the optimal way to obtain a position = bias introduced by non linearities of charge collection vs distance [diode-impact] eta Value (output) Eta function

CoG Value (input) CoG(5x5) Residual: = 2.30 mm 2/ eta method flatens the distributions (probability density is expected to be flat) Eta(3x3) Residual: = 1.84 mm 3/ improves resolution significantly EUDET, Geneva, January 25th 2007 (see back up slides for details) Auguste Besson 4 What a test beam analysis software should be able to do ? (1) The goal is not only to determine standard performances (efficiency, S/N, resolution, etc.) but also and mainly to determine and understand quickly any unexpected behaviour of the DUT (and there will be since were dealing with prototypes !) Absolutly crucial On line monitoring other software. rough alignment trigger rates, beam profile/ flux setting rough baseline, pedestal, noise, hit selection of DUT detect anomalies quickly Event display / scanning jump to a particular event and analyse it in details (whole matrix, noise, pedestal, etc.)

Filters: Map / mask of bad pixels, lines, etc. to tag / exclude it easily from analysis. Flexibility = Complete datacard/input system all possible parameters even radiation dose, Temperature, etc. so that this parameters are set in only one place and are accessible in analysis. keep track of all the information in a not ambiguous way. change the telescope configuration easily (positions, angles, number of planes, etc.) All DUT have different formats, different numbers of sub-structures which could be analysed separatly or not. adapt reco easily < any DUT particularities. e.g. add some common mode correction at the reconstruction level in a particular region of a given matrix for events taken between 2:12 AM and 3:47 AM. EUDET, Geneva, January 25th 2007 Auguste Besson 5 What a test beam analysis software should be able to do ? (2) No crucial but very convenient for users Complete analysis fast enough compared to data taking time to react quickly Most users will want to compare different runs between them some tool/framework to compare different runs easily: e.g. : 5 consecutives runs @ 5 different temperatures study Noise vs Temp. Most users will need some calibration/ Noise runs same framework to analyse these runs. necessary to compute a fake hit rate Most users will want to optimize their analysis perform different hits algorithms / sets of cuts in the same reconstruction to compare performances EUDET, Geneva, January 25th 2007

Auguste Besson 6 The dream of the user Datacards options Define matrices Define several regions (A,B,etc.) For each region define: region A Noise: algo A, B, etc. 1 Matrix region B Pedestal algo A,B, etc. hit selection algorithm x,y,z S/N thresholds cluster size (1, 2x2, 3x3, 5x5, 7x7, etc.) simulated output chain (ADC, etc.) filters (hot pixels, dead lines, etc.) Mask 1 root branch per region example M9: 4 outputs, 2 submatrices, study temperature gradient effect EUDET, Geneva, January 25th 2007 Auguste Besson 7 Alignment issues (1) In principle Alignment of the telescope done when necessary In practice Alignment of all planes is done for each run.

We are dealing with microns, so any change in temperature, any access in the test beam area could modify the alignment. 2 alignments: telescope alignment: relatively stable DUT alignment: done for each run and possibly for each submatrix individually ! to get the best alignment: assume the algorithm needs to minimize distances between extrapolated tracks and all the associated hits in the DUT. BUT: you dont know a priori which hits can be associated to a given track. To know this you need to know the correct parameters of the alignment need some maximum track-hit distance cut (large before alignment) some tracks dont go through the DUT itself (if the DUT is smaller than the track acceptance for instance) need to know the alignment to select the good tracks (=good acceptance of telescope) may be necessary to do it for each submatrix individually EUDET, Geneva, January 25th 2007 Auguste Besson 8 Alignment issues (2) Recursivity is not avoidable New alignement for each configuration change (temperature, prototype, etc.) Track quality monitoring is mandatory control chi2, check that track selection doesnt introduce a bias, etc. track hit distance, chi2 EUDET, Geneva, January 25th 2007 Auguste Besson

distance track-hit To make the optimised alignement, need to already know it pretty well angles in the fit are very useful (study hits with large incident angle for instance) : 5 parameters (and not only 2) if position z is correctly known. Alignment of the telescope before the reconstruction 2/NdF 9 Is resolution homogenous ? horizontal Temperature monitoring coupled to the anaysis software vertical Residual (mm) How Mimosa breathes in CERN Offset ~ 4-5 mm EUDET, Geneva, January 25th 2007 Event number (k) Auguste Besson 10 Acceptance Remarks on triggers Use of a large trigger to see where the DUT actually is

All track extrapolation position in the DUT plane which where not associated to a hit in the DUT Trigger acceptance (2 x 4 mm2) DUT acceptance having an adapted trigger acceptance compared to the DUT is the easiest way to reduce data taking time. EUDET, Geneva, January 25th 2007 Auguste Besson 11 Data issues The raw data can be very different from one DUT to another. The user should only have to specify: headers, trailers, data encoding (binary, hexa, etc.), nbits of ADC, number of submatrices, etc. Somehow need to reduce the amount of data via reconstruction Select hits and tracks objects keep only small amount of data Example: Mimosa 9: 4 matrices (with different pitchs) to get few 1000s events in a given matrix ~10-20 Go runs Somehow need to be able to monitor everything during the run: Example 1: pedestal and noise of a given number of pixels versus event number. huge amount of data if it is done for all the pixels. Example 2: some common mode in a given region.

Study inefficiency: Assume you reconstruct a track but have not corresponding hit in the DUT. You want to know what happened access to the DUT signal AFTER alignment is done ! Software analysis should allow to do: Some standard reconstruction Some monitoring reconstruction = Event display/scan Some 2nd access to raw data after reconstruction EUDET, Geneva, January 25th 2007 Auguste Besson 12 Some examples of monitoring Event display Efficiency vs fake rate Noise: vs regions, time, etc. Resolution: vs method, impact in the pixel, etc. S/N: in seed, neighbours, etc. Clusters: size, charge sharing, etc. Double hit separation (time stamping?)

Matrix uniformity Radiation, temperature, read-out frequency, etc. Incident angle Digitisation, ADC, etc. Edges effect EUDET, Geneva, January 25th 2007 Auguste Besson 13 Conclusion The software will have to deal with many different configurations Its not as simple as a select hits and tracks software. Users will want to study everything and in particular, things you didnt foresee. The architecture has to be carefully discussed at the beginning (very bad experience with software made of 100s patchs) User input is crucial EUDET, Geneva, January 25th 2007 Auguste Besson 14 back up Eta function principle 1/ Compute Center of gravity position from 3x3 cluster charge (Qi) information v N entries 10 9

Vdig U CoG Qi ui i 1 9 i Udig 9 8 7 UCoG - UDig CoG Dig (mm) 6 Q i 1 u UCoG - UDig UCoG - UDig 5 4 3 2 1 0

2/ Plot Center of Gravity distance from the center of If there was the pixel UCoG - UDig -pitch/2 no bias, this should be a flat distribution UCoG - UDig CoG Dig (mm) Integral(UCoG - UDig) 80 70 60 50 U 10 f 40 30 20 CoG U Dig dx pitch / 2

EUDET, Geneva, January 25th 2007 +pitch/2 90 3/ Integrate this distribution to get the f eta distribution function: x 0 0 UCoG - UDig -pitch/2 Auguste Besson 0 +pitch/2 16 4/ Normalize by the Number of event (N entries), multiply by the pitch and shift it of (pitch/2) xx U U dx CoG Dig CoG Dig

pitch pitch// 22 pitch pitch pitch / 2 pitch//22 U CoG U Dig dx Dig CoG pitch//22 pitch Integral(UCoG - UDig) 90 80 70 60 50 40 30 20 10 0 UCoG - UDig -pitch/2 +pitch/2 0

90 +pitch/2 out 0 80 70 60 50 40 N entries 5/ Get a flat distribution of all hits in the pixel 30 20 10 0 UCoG - UDig -pitch/2 0 in -pitch/2 +pitch/2 This width is proportionnal to the number of entries in the yellow bin. Consequences/issues: Needs to generate these eta functions

The bin size versus available statistics (N entries) The CoG can be outside the range [- pitch/2 ; + pitch/2] (happens ~1/1000) Number of different CoG values (from ADC so from different charge values) can not be lower than the number of bin. example: 2 bits ADC ~ 17 values; 3 bits ADC ~ 89 values; 4 bits ADC ~ 400 values Low statistics in the corners of the phase space . assume no correlation between U and V directions (not completly true) EUDET, Geneva, January 25th 2007 Auguste Besson 17 Objectifs Tester les capteurs avec des m.i.p. Caractrisation des capteurs via: Reconstruction du passage de la particule grce un tlescope Alignement vis vis du chip tester Rapport signal/bruit Efficacit de dtection Charge collecte (pixel sige et amas) Rsolution spatiale Diffrents paramtres:

Temprature Chips irradis (X, e-, n) Exploration de diffrents procds de fabrication Explorations des paramtres gomtriques Pitch; surface de la diode de collection de charge paisseur de la couche pitaxiale tudes complmentaires: Critres de slection/efficacit Angle dincidence Pouvoir de sparation des impacts Cartographie des matrices (uniformit des caractristiques) Uniformit entre les prototypes Effets de la digitisation, etc. La caractrisation prcise des performances des capteurs passe ncessairement par des tests en faisceau EUDET, Geneva, January 25th 2007 Auguste Besson 18

Recently Viewed Presentations

  • Latest on European funding VSNW Annual Conference 12

    Latest on European funding VSNW Annual Conference 12

    Arial MS Pゴシック Calibri BlissLight Osaka ヒラギノ明朝 ProN W3 Century Gothic Times New Roman ヒラギノ角ゴ ProN W3 ヒラギノ明朝 ProN W6 Verdana Wingdings Office Theme 1_Office Theme Microsoft Excel Sheet Latest on European funding Workshop aims Slide 3 Accessing ERDF European...
  • CHAPTER 6: THE JUDICIAL BRANCH - Juarez AP GOV

    CHAPTER 6: THE JUDICIAL BRANCH - Juarez AP GOV

    Note Taking with a purpose. 42% gone in 20 minutes. 56% gone in an hour. 66% gone in a day. 75% gone in a week. 80% gone in a month. Summary of info is more important than the info
  • Estuaries - Loudoun County Public Schools

    Estuaries - Loudoun County Public Schools

    Estuaries come in all shapes and sizes and go by different names. Bays, lagoons, ... The average American Spends 10 recreational days on the coast of an estuary each year. Health of Estuaries ... "Over fishing"- fewer crab and fish...
  • CSCI 203 Introduction to Computer Science I

    CSCI 203 Introduction to Computer Science I

    A bit twist. The same as above, converting to upper case, however, now we want to map each lower case letter to a random upper case letter. Example: 'abc123' 'SGC123', 'hello' 'SGEBU', '123 456' '123 456'
  • Comparison of Golding&#x27;s Lord of the Flies and Freud&#x27;s Theory ...

    Comparison of Golding's Lord of the Flies and Freud's Theory ...

    Comparison of Golding's Lord of the Flies and Freud's Theory of Personality English I Mayfield Who was Sigmund Freud? Sigmund Freud was an Austrian neurologist and the pioneer for psychoanalysis, a system of psychological theory and therapy investigating the interaction...
  • Chapter 12, Race And Ethnic Relations

    Chapter 12, Race And Ethnic Relations

    racial formation The process by which groups come to be defined as a race. minority group Any distinct group in society that shares common group characteristics and is forced to occupy low status because of prejudice and discrimination. dominant group...
  • Social Psychology

    Social Psychology

    Social Situations and Aggression Aggressive Objects as Cues Imitation and Aggression Imitation and Aggression Violence in the Media: TV, Movies, and Video Games Effects on Children What About Adults? The Numbing Effect of TV Violence How to Reduce Aggression Does...
  • The H-R Diagram

    The H-R Diagram

    Distance (in pc) = 1/(parallax angle (in arcsec)) Must use the right units here: define a new distance unit, for convenience… a parsec. 1 parsec = distance of a star with a parallax of 1 arc second. Typical parallax shift...