MEG DCH Analysis

22
17 February 2010 DCH Analysis 1/22 MEG DCH Analysis MEG Review Meeting 17 February 2010 W. Molz For the DCH Analysis Working Gro

description

MEG DCH Analysis. W. Molzon For the DCH Analysis Working Group. MEG Review Meeting 17 February 2010. Outline. Goals of DC analysis Overview of calibrations and analysis Low level performance: show some results, still improving resolutions Efficiency R f resolution Z resolution - PowerPoint PPT Presentation

Transcript of MEG DCH Analysis

Page 1: MEG DCH Analysis

17 February 2010 DCH Analysis 1/22

MEG DCH Analysis

MEG Review Meeting17 February 2010

W. MolzonFor the DCH Analysis Working Group

Page 2: MEG DCH Analysis

17 February 2010 DCH Analysis 2/22

Outline

•Goals of DC analysis

•Overview of calibrations and analysis

•Low level performance: show some results, still improving resolutions– Efficiency– R resolution– Z resolution

•High level resolutions: show some results, still improving resolutions and our measurements of the resolutions– Momentum– Track angle at target– Position at target

•Demonstrated performance vs. proposal performance vs. current MC

•Prospects for improvement– Hardware– Software

Page 3: MEG DCH Analysis

17 February 2010 DCH Analysis 3/22

Goals of DCh Analysis

•Optimize performance of spectrometer– Best low level resolutions: R-,Z, efficiency, noise rejection – Best high level resolutions: positron energy, trajectory

•Determine hardware limitations and possible improvements– Noise, alignment, stability

•Characterize performance for purpose of physics analysis– PDFs for likelihood analysis

•Optimize power of physics analysis – Selection criteria vs. efficiency

Page 4: MEG DCH Analysis

17 February 2010 DCH Analysis 4/22

Positron Spectrometer Impact on MEG Performance

• Select on positron energy within interval near 52.8 MeV– For fixed →e acceptance, BG/S proportional to p (MEG prediction RMS=180

keV/c)

• Select on enear – For fixed acceptance, BG/S proportional to x (MEG prediction RMS = 8x8

mrad2) – photon position resolution ~ 4 mm RMS ~6 mrad both and – Track fittingangle uncertainty 4-5 (7) mrad each – Position of stopping target: uncertainty 0.5 mm ~6 mrad

•Project to target and timing counter and correct te for propagation delay– For fixed acceptance, BG/S proportional to t (MEG prediction RMS = 64 ps,

~2 cm)– Path-length projection to target has negligible uncertainty– Uncertainty in path-length projection to timing counter dominated by scattering

and E loss after spectrometer– Improvements needed to incorporate position at timing counter and material

between spectrometer and timing counter into trajectory fit

• For all effects, tails in resolution function loss of acceptance proportional to integral in tail, small increase in background because source of background is uniform

Page 5: MEG DCH Analysis

17 February 2010 DCH Analysis 5/22

DC Performance 2009 vs. 2008

•Significantly improved performance this year

Hit map 2008

2009

Hit in plane near track projection

Hit in plane assigned to

track

Page 6: MEG DCH Analysis

17 February 2010 DCH Analysis 6/22

Significant Improvements in Tracking Analysis

• Incorporated use of TIC time for track time– Alternative to use of track time deduced from DCH itself– Necessary last year due to inefficient chambers– Track time from DC itself now much improved performance

•Much better algorithms for selecting/removing appropriate hits on track– Significantly improves resolution and efficiency– Re-optimize this year for better quality data

•Better understanding of merging of multi-turn tracks•Developed technique for measuring resolutions using two-turn tracks

– Fit each turn of a two turn track– Project each turn to common point of closest approach to spectrometer axis

between two turns – one projected forward, one backward– Measure difference in ,, R, z, p and infer resolution in these quantities

• Improved fit to Michel edge to extract momentum resolution •Better understanding of chamber performance, contributions to resolution•Work done on cross-checks of calibration•Work on cross-check of alignment using cosmic-ray muons•Better understanding of relating measurable resolutions to kinematic

resolution

Page 7: MEG DCH Analysis

17 February 2010 DCH Analysis 7/22

DCH Alignment •Primary alignment of chambers from optical survey•Correct chamber displacements by minimizing mean residuals to fitted

tracks using Michel data

•Residual chamber rotations after optical survey are negligible: no corrections

•Mean residuals reduced from ~100 m to 10-20 m•Compare to typical resolutions:

– Position resolutions R ~200 m; Z ~1000 m; – Chamber-to-chamber scattering deviation ~500 m

•CR data recorded with field off for cross-check of alignment– Different drift performance without magnetic field – Possibility of getting higher momentum tracks with less scattering– Remove possibility of correlated DC shifts being missed due to momentum fit

Plot of rotation diagnostic

Optical survey Optical survey

after software alignment RMS residual = 10 m

after software alignment RMS residual = 18 m

Page 8: MEG DCH Analysis

17 February 2010 DCH Analysis 8/22

Quality of Baseline Prediction• Charge on anodes and pads used for Z measurement• Baseline subtracted by measuring level early in waveform, subtracting average value

– Shown to be superior to linear, quadratic extrapolation• Bin to bin pedestal fluctuations larger in 2009 data vs. 2008 data

• Precision of prediction of baseline in signal integration affects position resolution

– Histogram difference in predicted baseline and average baseline in 50 ns signal region

– Measured on pedestal runs simultaneous with recent MEG data

• Contribution to Z resolution– Depends on both

the precision of the baseline and the size of measured charge

– Increased HV in 2009 gives 40-50% increase in mean hit charge wrt 2008

– Error in Z due to baseline fluctuationcalculated for every hit

– Contribution of z ~ 550 m in 2009 compared to z ~ 1 mm in 2008

2008

pedmV

2009 pedmV

2008

zmm

2009

zm

2009

Baseline errormV

Page 9: MEG DCH Analysis

17 February 2010 DCH Analysis 9/22

Drift Performance and Calibration •Alignment of time offsets wire by wire

– Fit to leading edge of distribution of hit – track

for each end of each wire– Check procedure by comparing hit time at 2 ends– Typical precision of 1.5 ns

•Verification of time to position relationship– HV and B dependent drift using GARFIELD– Incorporate asymmetric response at edge cells – Project track from hits in planes 0,1 to common point – measure residual – Alternative measurement of resolution from residual of hit to fitted track – Measure dependence of residuals on track angles, drift time to verify drift

model– Typical single plane resolution 250 m – Some systematic effects with angle in 2008 data, being studied again in

2009 data

dR

Fitted track shape

tied to hits

Page 10: MEG DCH Analysis

17 February 2010 DCH Analysis 10/22

Z Coordinate Measurement

•Determined first from charge division on anode – calibrated by using know phase and periodicity of cathode pads vs. Z measured by anodes – Primarily used to determine

correct cycle of cathode pad– Does not enter directly into

precision of Z determination; used when pad signals missing

•Precise Z determination from charge induced on pads– Pattern of induced charge studied with image charge method – impact on

calibration– Dependence on wire-cathode distance, offsets of wire with respect to center of

pattern (in the wire plane), fluctuations of mean Z coordinate of ionization sites– Optimization of technique for measuring charge (integration time, etc.)

potentially important– Noise contribution to charge is largest known source of error in Z determination– Standard integration and charge calibration + two alternative methods studied

• Preliminary results of alternatives give essentially same performance

Show plot of fit to sine wave

Page 11: MEG DCH Analysis

17 February 2010 DCH Analysis 11/22

Cathode Pad Calibration: Z Determination •Reminder

– Z = n*(5 cm) + 5/(2)*(arctan(Ahood/Acathode)+h(c) – A = (Qu-Qd)/(Qu+Qd); Qu(d)= a+b*sin(2*z/5+u(d))

• a,b depend on pad-anode distance• Precision of dependence of A on z studied with

electrostatic calculation – correct to good approximation

•Steps in Z calibration– Correctly align time offsets in pads vs. anodes:

integrate same part of signal• Adjust time offsets on pad signals to set the mean

valueof the difference in the time of the pad and anode to zero

– Correct for relative upstream-downstream gains:• Adjust gain to get the mean asymmetry in the cathode

and pad for each were equal to zero – Correct for effect of chamber foil bowing

• Both the induced charge and the asymmetry depend on the anode-cathode distance

• Measure Qcathode/Qanode vs. z for each wire – fit to quadratic dependence on Z

• Apply phenomenological correction to each asymmetry depending on mean induced charge for that wire and Z

– Bowing correction is ~200 m in quadrature

Page 12: MEG DCH Analysis

17 February 2010 DCH Analysis 12/22

Some Details on Chamber Bowing• Distance of hood and cathode from anode wire effected by bowing due to gas

pressure, foil mass, possibly details of how foils are fixed to frames• Electrostatic calculations show effects >10% on induced charge for deflections of

order 0.5 mm• Measure the ratio of hood to cathode asymmetry amplitude (amplitude of sine wave)

by measuring RMS in each 5 cm interval in Z along the wire • Measure the ratio of the hood to cathode charge vs. Z• Make scatter plot of asymmetry amplitude ratio to charge ratio – agrees with linear

correlation predicted by electrostatic calculation

• Expect biggest effects in center of chamber, where bowing is largest, some different dependence on Z, particularly for first and last cell

wire162cell 0

wire 45, cell 0

wire 26cell 8

wire 21cell 3

Page 13: MEG DCH Analysis

17 February 2010 DCH Analysis 13/22

Pad Crosstalk from Adjacent Anodes•Effect of charge induced by hits on adjacent wires •Consider hits at same Z in two adjacent wires in same plane, indicated by

circles in figure below. – Charge induced on pads due to anode charge in same cell will have asymmetry

zero– Charge induced on pads due to anode charge in other cell will have asymmetry

different from zero; in the example shown, more charge on DH for top pads, more on UH for bottom pads

• Only relevant for in-time hits: short integration time helps• Effect tends to cancel when 2 hits averaged, cancellation

not exact, particularly when pulse heights are different• Charge induced on adjacent cell is not trivial (as much as 7-15%)

•When Z of two hits is different (for large Z), effect will be different and perhaps larger

UH

DH

UH

DH

Page 14: MEG DCH Analysis

17 February 2010 DCH Analysis 14/22

Contributions to Z Coordinate Uncertainty

Effect Parameter Parameter

value

RMS Z error [m] Notes

Pedestal fluctuation RMS Error 0.6 mV 550 Measured in data upstream-downstream gain ratio

RMS fractional gain error 2.30% 70

Guess at gain error, gain variation channel-to-channel is about 5%

charge integration start time and width

integration, start time RMS 0.9 ns 250

Estimated from data by varying relative start/stop times

Uncorrected avalanche perpendicular position Typical shift < 100 m 0

magnitude of the shift is a guess, it could be much less, use zero for nowUncorrected

avalanche (Z) position shift

Typical shift for 25 ion sites/cm 90

RMS error varies from a few to 170 mm with angle and drift distance

Difference in wire-cathode , wire-hood distance

Estimated uncorrected displacement 150 m 100 Estimated residual effect after correction

Cross talk from adjacent hit anodes 329

30% of the time, 7% cross talk from adjacent anode, very rough estimate of size of effect

Total in quadrature 704

Page 15: MEG DCH Analysis

17 February 2010 DCH Analysis 15/22

Measuring High Level Resolutions•Need PDFs for likelihood fits or acceptances for a cut and count analysis

•For the positron, these have contributions from:– Momentum response function – no fixed momentum calibration line– Positron angles (and) at the target – no fixed direction events– Positron intercept at the target – contribution to the photon angle

measurement

•Response functions not expected to be Gaussian distributions– Resolutions will depend on, for example, track length, pitch angle, etc.

•For momentum, can fit to the edge of the Michel spectrum– Sensitive to only the high energy side of the response function, the important

one – Lower energy side strongly correlated with momentum dependence of

acceptance

•For momentum and angles, can exploit tracks that have two full turns in the spectrometer, comparing momenta and angles at a common point near the axis to infer the resolution – For momentum, cannot determine separately the upper and lower edges,

must assume it is symmetrical. Complementary to fit to Michel edge– For , possible systematic differences from dependence on Z– For , technique excludes contribution from effect of uncertainty in path

length in projecting back to target: 1 mm error in path length is about 7 mrad error in

•All resolution functions should be measured after perfecting low level performance and optimizing selection criteria (not yet done) – Results are likely to improve with analysis

Page 16: MEG DCH Analysis

17 February 2010 DCH Analysis 16/22

Momentum Resolution •Fit to Michel edge

– Fit function is sum of offset Gaussians– Fit results depend on acceptance function and dataset: Michel, low

intensity, MEG sidebands – Sample fit to 2009 data before DRS correction: RMS for -1.5< E <1.5 =

0.580 MeV

•Alternative measurement from 2 turn comparison– Single Gaussian fit: RMS = 0.490 MeV – Fit to convolution of sum of 2 Gaussians:

RMS in region -1.5 < E < 1.5 = 0.447 MeV

•Third possibility to use Mott scattering of mono-energetic electron beam scattered into spectrometer to characterize momentum resolution– De-convolve energy spread in beam, energy loss

dispersion in thick scattering target

Page 17: MEG DCH Analysis

17 February 2010 DCH Analysis 17/22

Angle and Vertex Position Resolutions

•Use technique of two-turn tracks to project to common point near spectrometer axis

•Theta angle resolution– Reasonably well fit by Gaussian: RMS of = 12.7 mrad

•Z position resolution– Well fit by Gaussian: RMS of z = 3.1 mm – Roughly consistent with contribution from scattering

•Phi angle resolution – Well fit by Gaussian: RMS of = 8.1 mrad – Error is correlated with momentum error

•R position resolution – Well fit by Gaussian: RMS of R= 2.4 mm

Page 18: MEG DCH Analysis

17 February 2010 DCH Analysis 18/22

Correspondence Between Resolutions at Target and 2-Turn Comparison

• Can use MC to get correspondence between z position resolution and positron resolution– For perfect z resolution, resolution is 7 mrad– Expect ~9 mrad resolution for current Z

resolution

• Can also use MC to calculate correspondence between resolutions inferred from comparisons of 2 turns to the resolution at the target– Plot (1 -2)/√2vs. (meas-true) parametric in z – Current resolution in 1 -2 corresponds to about

10.5 mrad resolution

• Two avenues for improvement– Improve Z resolution– Understand and fix lack of agreement between

measured resolution and that predicted for current Z resolution

(

meas-

true)

Z(meas-true)

(

two

turn

s)

(meas-true)

MC vs. Z

MC 2turnvs. tgt

Page 19: MEG DCH Analysis

17 February 2010 DCH Analysis 19/22

Correlation of Momentum and Quality Measures

•Events with p>52.8 MeV/c represent poorly measured tracks; is there a correlation with track properties?

•Width of central part of momentum resolution function most important for physics background estimate

•Tails in positron momentum resolution function less important; few low momentum positrons satisfy trigger, hence few low momentum positrons can contribute to accidental background.

Page 20: MEG DCH Analysis

17 February 2010 DCH Analysis 20/22

Can We Estimate Tracking Efficiency from Data?

•Use highly pre-scaled timing counter trigger data

•~ 520 C total live protons on target 1.31 x 107 /s/mA (assume livetimesame for MEG, other triggers) Implies ~ 683 x 1010 total muon stops

N→e= 1935 muons satisfying selection criteria counted = 6.83x1012 muon stops calculated ( few percent uncertainty )

X 10-7 prescale factor known X 0.35 TIC acceptance x efficiency for Michel measured

X 0.101 fraction of Michel spectrum > 50 MeV calculated X (0.92-1.0) conditional trigger efficiency for TICmeasured*

X 0.091 Michel geometric acceptance X DCH drift chamber reconstruction & cuts unknown

DCH = 1935 x 107 / 0.35 / 0.101 / 0.96 / 0.091 / (6.83x1012) = 0.92

Need to redo TIC efficiency measurement for 2009

Page 21: MEG DCH Analysis

17 February 2010 DCH Analysis 21/22

Conclusions

•Tracking efficiency in 2009 data is much better due to improved chamber performance.

•Intrinsic resolutions are improved wrt last year’s data– Current status is really a lower limit on performance– Central part of Rresolution is close to expectations, but tails are

more than originally anticipated– Z resolution worse than planned and not fully understood from

calculated contributions, but now not a dominant contribution to angular resolution

•Angle resolutions better understood, still work to be done– Should get better agreement with MC when measured low-level

resolutions are used – Incorporate cell dependences in resolutions– Understand contribution to resolution from momentum error resulting

in error in path-length to target

Page 22: MEG DCH Analysis

17 February 2010 DCH Analysis 22/22

Conclusions

•Prospects for improvement

– Still early in optimization of even low level performance• Fitting for improved baseline subtraction (noise filtering – some indications of

possible improvements)• Drift time-distance model verification• Anode to adjacent pad crosstalk corrections• Re-optimization of integration time with fully calibrated system• Correction of edge effects (near wire ends) in Z determination

– Some possible software improvements ( preliminary results show little improvement )• Alternative alignment• Alternative integration scheme

– High level improvements • Incorporating partial turns in fitting• Improved projection to TIC using TIC signal • Incorporating track time as parameter in fitting• Understanding of 1-2 mm offset in magnet vs. spectrometer

– Hardware changes• Reduction of noise at hardware level • Additional measurements of resolution with Mott scattering