Documenti di Didattica
Documenti di Professioni
Documenti di Cultura
Based on statistics of cosmic runs accumulated since November 1998 to January 1999 with BaBar Drift Chamber, several systematics directly related to the dE=dx resolution are considered.
Abstract
1 2
This research was partly supported by Ministry of Science of Russian Federation and SLAC. Temporary in SLAC, from Budker INP, Novosibirsk, 630090, Russia. 3 SLAC.
Contents
1 Waveform
1.1 Test of waveform and charge quality . . . . . . . . . . . . . . . . . . . . . 1.1.1 Calibration le and coe cients . . . . . . . . . . . . . . . . . . . . 1.1.2 Charge calculation from waveform . . . . . . . . . . . . . . . . . . 1.1.3 Comparison of pedestal . . . . . . . . . . . . . . . . . . . . . . . . 1.1.4 Comparison of charge . . . . . . . . . . . . . . . . . . . . . . . . . 1.1.5 Check of address correctness . . . . . . . . . . . . . . . . . . . . . 1.1.6 Waveform quality control . . . . . . . . . . . . . . . . . . . . . . . 1.1.7 Summary for this section . . . . . . . . . . . . . . . . . . . . . . . 1.2 Correlations of waveform parameters and reconstruction of clipped charge 1.2.1 De nition of waveform parameters . . . . . . . . . . . . . . . . . . 1.2.2 Correlations of A32 with charge Q and Noverflow . . . . . . . . . . 1.2.3 Correlations of charge Q with Qclip and Noverflow . . . . . . . . . . 1.2.4 Clipped charge reconstruction . . . . . . . . . . . . . . . . . . . . . 1.2.5 Summary for this section . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . .
4
4 4 4 5 5 5 6 6 6 7 7 7 7 8
2.1 Parameters for track selection . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 2.2 Spectra of momentum . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 2.3 Angular distributions of cosmic statistics . . . . . . . . . . . . . . . . . . . . . . . . .
8
8 8 8
3.1 Wire-to-wire gain correction . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 9 3.2 Dip angular correction for mean amplitude . . . . . . . . . . . . . . . . . . . . . . . 9 3.3 Distance of closest approach correction for the mean amplitude . . . . . . . . . . . . 10 4.1 4.2 4.3 4.4 4.5 4.6 Evaluation of dE/dx & Truncated mean algorithm . . . . . . . . . . . . . . . . Optimization of the truncated mean algorithm . . . . . . . . . . . . . . . . . . Parameterization of ionization strength vs ................... Dependence of dE=dx resolution on applied corrections and cut of parameters . Dependence of dE=dx resolution on sample length . . . . . . . . . . . . . . . . Dependence of dE=dx resolution on number of amplitude samples . . . . . . . . . . . . . . . . . . . . . . . . . .
4 dE/dx analysis
10
10 11 11 12 12 12
5.1 Reconstruction of clipped charge . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 13 5.2 Amplitude dependence on entrance angle . . . . . . . . . . . . . . . . . . . . . . . . 13 5.3 Correction of the truncated mean biasing e ect . . . . . . . . . . . . . . . . . . . . . 14 2
13
5.4 Mean amplitude versus time correction . . . . . . . . . . . . 5.5 Dip-angular amplitude correction for spectrum . . . . . . . 5.5.1 Procedure . . . . . . . . . . . . . . . . . . . . . . . . 5.5.2 Comparison of new and old dip-angular corrections . 5.5.3 Summary for this section . . . . . . . . . . . . . . .
. . . . . . . . . . . . . . . .
. . . . . . . . . . . . . . . .
. . . . . . . . . . . . . . . .
. . . . . . . . . . . . . . . .
. . . . . . . . . . . . . . . .
. . . . . . . . . . . . . . . .
. . . . . . . . . . . . . . . .
. . . . . . . . . . . . . . . .
. . . . . . . . . . . . . . . .
. . . . . . . . . . . . . . . .
. . . . . . . . . . . . . . . .
. . . . . . . . . . . . . . . .
. . . . . . . . . . . . . . . .
. . . . . . . . . . . . . . . .
14 14 15 15 16
6.1 Single cell spectrum simulation . . . . . . . . . . . . 6.1.1 Simulation algorithm . . . . . . . . . . . . . . 6.1.2 Contribution of the gas gain uctuations . . 6.1.3 Fit of simulated spectra . . . . . . . . . . . . 6.1.4 Dependence of spectral parameters on < m > 6.2 Experimental dependence of single cell spectrum on 6.2.1 Procedure of spectra accumulation . . . . . . 6.2.2 Fit of experimental spectra . . . . . . . . . . 6.2.3 Dependence of t parameters on ..... 6.3 Comparison of experimental and simulated spectra . 6.4 Summary for this section . . . . . . . . . . . . . . .
. . . . . . . . . .
. . . . . . . . . .
. . . . . . . . . . .
. . . . . . . . . . .
16
16 16 17 17 17 18 18 18 18 18 19
7 Summary 8 Acknowledgments
19 20
Introduction
From October 1998 to January 1999 cosmic runs with BaBar Drift Chamber separately and integrated with other detector's systems have been conducted before the beginning of running at PEP-II on May 7, 1999. By November '98, the Drift Chamber reached stable operation and later statistics are used for this analysis. At the beginning of this work in the middle of February 1999, the working version of standard BaBar code for dE/dx was almost complete. Some of the usual amplitude corrections were tested and implemented in on-line dE/dx calibration. Expected dE=dx resolution 6:5 7% were obtained at > 10 20, but for the most useful range for BaBar (2 < < 10) there was an unexpected deterioration of resolution of up to 8 9%. The question arose of possibly improving the resolution in this range. A few problems directly related to dE=dx resolution are considered in this note. All e orts were concentrated in two directions: to improve as much as possible the dE=dx resolution and to try understand to what extent results di er from expected on the basis of ionization statistics. The contents of this note are divided into the following almost independent parts: waveform analysis; amplitude systematic corrections; dE=dx resolution evaluation and optimization of the truncated mean algorithm; check on some of the classic statistical hypothesis in uencing single cell spectra and leading to the dependence of dE=dx on sample length and number of samples.
1 Waveform
1.1 Test of waveform and charge quality
The main idea of this test is to check that the ROM \feature extraction" algorithm does not contain rough errors in calculation of charge from the waveform and that electronics calibration coe cients are used in the proper way. Probabilities of \bad" waveforms have been estimated for di erent values of high voltage. This test is done using n-tuples with waveforms for runs 2246, 2248, 2250, 2252 (HV = 2080, 2020, 1960, 1900 V) with corresponding calibration le 160.
where A0 is ADC count for current sample, Alin is a resulting linear amplitude. 2. Linear interpolation of missed amplitude samples due to the time codes in the waveform. 3. Integration of the charge: starting from leading edge (with threshold 1 ADC count above pedestal level) to the 32-nd sample. 4. Correction of the charge Q according to the calibration coe cients:
Q=2 (
nint
where nint is a number of integrated samples, leads to the units of roughly fC=2.
Private communication.
( 15%) at larger total charge. Hence, combining these two parameters, charge resolution could be improved in all range. The simplest linear combination of charges Qcorr = (Qcorr:1 + Qcorr:2 )=2 has been tested, without optimization of relative weights. Results of this procedure are shown in Fig. 9a, as a correlation between Qcorr and total charge Q. Resolution of corrected charge on charge is shown in Fig. 9b. Fig. 10 contains the same resolution plot with two others, corresponding to the charge reconstruction using Qcorr:1 and Qcorr:2 separately. Signi cant improvement in case of linear combination is clearly observed.
95% of hits have an entrance angle inside the ranges 20o w.r.t. indicated directions. As shown in Fig. 16, the \desert" area outside of peaked ranges is populated mainly by hits of tracks with low . Non-uniform distribution of cosmic statistics could strongly a ect investigations of amplitude dependence on . An amplitude dependence on entrance angle should exist (at least due to the drift cell geometry). Hence, real experimental events should be used for this calibration. In the present analysis the dependence on entrance angle should not have signi cant in uence due to the narrow range.
to-layer gain calibration. Individual layer calibration for doca and dip angle amplitude dependences is applied instead of calibration for the layer groups. Polynomial ts are used everywhere instead of Chebyshev polynomials.
Q cos is used instead of dq=dx for each of drift cells. Wire to wire gain calibration based on 5 104 events of one run is used instead of layer-
for historical reasons. Dependence of mean amplitude < Q cos > versus jsin j for separate layers and all layers with low and high gas gain are shown in Figs 24, 25, 26, 27 before and after corrections as well. A third order polynomial is used for parameterization of the dependence for each layer. Polynomial coe cients are stored after calibration and used in following to restore the function for correction purposes. As a curiosity should be mentioned that the dependence of mean amplitude on jsin j1=2 (see Fig. 32) can be parameterized by an almost straight line. A higher degree could be used to re ne the t.
4 dE/dx analysis
4.1 Evaluation of dE/dx & Truncated mean algorithm
After application of amplitude correction procedures the value Q cos can be considered as a dE=dx for each sample, expressed in relative units. Mean value of all amplitude samples along the track are considered in the simplest case as a dE=dx estimator. Truncated mean method (see the survey in 6]) is usually applied to get more stable estimation of the dE=dx value for each track. In this case a certain fraction of high amplitude samples fhigh are discarded to escape a long amplitude tail of Landau distribution. Sometimes, a fraction of low amplitude samples flow is also discarded to suppress a contribution of noise uctuations. The values flow : fhigh expressed in percents will be used in the following to specify the truncated mean algorithm parameters. Ionization strength strongly depends on momentum of the particle or . Typical scatter plot of dE=dx versus is shown in Fig.33. Systematic corrections wire-to-wire gain, spectrum dip angular (see Section 5.5), doca as well as 10%:25% truncated mean were applied. Two approaches were exploited to get the mean ionization strength and its spread as functions of . All the statistics were shared for equidistant in logarithmic scale of histogram bins. In rst case mean dE=dx and RMS for each bin are evaluated. The Gaussian t and its parameters (most probable value Im:p: and width of distribution ) were exploited in the second case for each bin, as shown in Fig.34. Comparison of this two approaches in terms of normalized mean ionization strength and dE=dx resolution versus are presented in Figs 35, Fig. 36 respectively. For Fig 35 the normalization was done for the same value Imin in both cases. There are no signi cant di erence in the mean ionization strength. Fig. 36 shows, that Gaussian approximation provides more stable de nition of the value, than simple RMS. 10
I = p1 3 (p2 + 2ln( ) 2p
2p3
(2)
with xed additional parameter p3 = 1, and modi ed Bethe-Bloch formulae 12] providing phenomenological account for ionization density e ect in the region of Fermi plateau:
I = p1 5 (p2 2p
2p5
(3)
with xed parameter p5 = 1. Both approximations are shown in Fig. 39. Absence of cosmic statistics in the region < 2 does not allow one to x de nitively a shape of the function in the 1= 2 range. Obviously, that the function (3) with additional parameters provides a better description of experimental data over the full range.
This work contains very serious error | it does not take in to account dependence of mean ionization strength. To shadow their problems and get better dE/dx resolution, authors used very speci c trick: several cosmic events were combined together and truncated mean procedure was applied to all of their amplitude samples. Hence, the nal resolution results are very doubtful.
5
11
of the truncated mean, is clearly seen as a step-wise dependence. Fig 45b shows the dependence of p2 resolution on number of hits. Resulting degree p2 = 0:55 of tted function p1 =Nhit is even larger than 0.5 due to the strong in uence of the truncated mean at small number of amplitude samples. The same results for Nhit 80 without and with optimal truncated mean 10%:25% are shown in Fig. 46. In this case biasing of the dE=dx mean value has not so regular behavior because of changed truncated mean algorithm. Approximation of resolution without truncated mean algorithm leads to p2 = 0:47 (see t parameter in Fig. 46b) at 2 =Ndof = 394=86. The truncated mean algorithm provides about two-fold improvement of resolution. The dependence of resolution on number of samples in this can not be interpolated successfully form Nhit = 1 to 80. Results of three variants of interpolation are shown in Table 1. Full range from Nhit = 1 to 80 provides poor 2 =N dof = 1711=80. The degree of the function at the beginning of the scale (2 < Nhit < 10) is equal to 0.57, while the rest part (10 < Nhit < 80) shows degree of 0.45 with very good 2 =Ndof . The last t could be used for resolution approximation in the wide range of Nhit > 10. Table 1: Results of the t of dE=dx resolution on number of amplitude samples by the function p2 p1 =Nhit . Di erent ranges of Nhit are presented to show the di erence in resulting parameter p2 .
2 =N Fitted range of Nhit p1 p2 dof 1 80 0.4660 0.0057 0.499 0.004 1711/80 2 10 0.5264 0.0075 0.575 0.008 23.0/9 10 80 0.3931 0.0034 0.453 0.002 42.7/71
5.5.1 Procedure
An adequate transformation of the spectra can be done using normalized integrated spectral distributions. To get these statistics of few runs ( 2 105 events) are treated separately for layers with low and high gain. Very ne granulated (500 bins) spectra of Q cos have been accumulated for 50 equidistant slices in the range of 0 < jsin j < 1, integrated and normalized. Results are shown in Fig. 48. The last bin of each histogram contains the over ow for this histogram (> 2500 fC=2). This method allows one to compare the location of equivalent parts (at least on the point of view probability) of di erent spectra and to perform the transformation. Indeed, another projection of this distribution in units of Q cos versus jsin j, for constant levels of probability with step of 0.05 is shown in Figs 50, 51, and provides the desired histograms for amplitude transformation versus dip angle. A third degree polynomial is su cient to t each of these histograms. To show the relative size of the correction, the same tted functions are plotted in Figs 52, 53 with normalization at = 30o . A wide spread of correction functions is observed for di erent values of Q cos . Dependence of polynomial coe cients on probability level index are shown in Fig. 54, and, in principle, allows their parameterization in order to go from discrete to continuum correction functions. For this study a simpli ed method has been applied for the correction purpose. The nearest function from the set presented in Fig. 50 and Fig. 51 is used for each amplitude hit with coordinates: (jsin j, Q cos ). The transformation brings all the amplitudes to the same angle (for example = 30o or 0o ), where all the functions are well de ned. The result of application of this procedure to the normalized integrated probability distributions is shown in Fig. 49. For all angular slices the normalized integrated probabilities and hence the spectra are coincide with good precision.
be estimated, if we plot the same ionization curves as a function of K and momentum (see Figs 57) and compare their di erence in units of resolution for , shown in Fig. 58, 59. New correction algorithm shows advantage for all range of momentum.
n=
m X i=1
ki :
Obtained spectra of normalized number of electrons n= < m > are shown in Fig. 60d as a dashed histogram and in Fig. 61 for few values of < m >. E ects of ionization drift, electron attachment, di usion etc. are ignored. Drift cell geometry e ects are partly corrected in experimental spectra and also omitted at simulation. At the next stage, gas gain uctuations were accounted for using a form of Yule-Furry (exponential) distribution (Fig. 60c), instead of Polya distribution, which has parameters not well de ned for our gas mixture and strongly depend on gas gain factor. Individual gas gain factor Gi randomly chosen from exponential distribution with mean value < Gain >= 5 104 (see Fig. 60c) was applied to each of n electrons and a nal charge Q is de ned as a sum:
The cluster size distribution is taken from experimental measurements for He, but it has roughly similar dependence for iso C4 H10 .
6
16
Q=
n X i=1
Gi :
Spectra of normalized charge Q= < m > = < Gain > are shown in Fig. 60d as a hatched histogram and in Fig. 62 for few values of < m >.
f (Q) = p1 exp
where is a linear function of charge:
1 log2 (Q) + p2 ; 4 2 p2 4
(4)
and parameters p1 , p2 , p3 associated with Gaussian normalization factor, Qm:p: and respectively. Additional parameter p4 accounts for left-right tail asymmetry. To get the commonly used resolution parameter for single cell spectra, FWHM=Qm:p:, one uses 2:36 =Qm:p: .
17
18
7 Summary
Several systematics related to dE=dx resolution of the BaBar drift chamber are considered using cosmic data. The correctness of charge calculation from the waveform in the ROM code is checked as well as the correct application of calibration coe cients. Correlations of waveform parameters are considered. New parameters for ag of \bad waveform" are proposed. New approach to the reconstruction of clipped waveform charge is suggested and was shown that it provides 10 20% accuracy at three-fold extension of the charge scale. Standard systematic amplitude corrections are repeated for dE/dx evaluation. Dependences of dE=dx resolution versus on di erent parameters and applied methods are considered. Optimal parameters of the truncated mean method are found correspond to 10%:25% of discarded hits with low and high amplitude respectively. Dependence of dE=dx resolution on number of hits per track is considered and obtained interpo0 lation Nhits:45 could be used in wide range of 10 < Nhits < 80. There is no obvious interpolation to Nhits = 1 was found, while, the function with additional parameters could be used for this aim. Several new amplitude corrections are suggested. Spectra-dip-angular correction is most promising of them. It provides signi cant improvement (from 1% to 0.6%) in dE=dx resolution as well as some improvement in K= separation, comparing with standard dip-angular correction for mean amplitude. Statistical nature of the observed single cell spectra is tested. Good coincidence of experimental and simulated spectra have been achieved at number of primary electrons from 15 to 30. 19
8 Acknowledgments
We wish to thank Dr. William Dunwoodie for helpful discussions and many of our colleagues discussing related problems at numerous O ine Calibration Friday Meetings.
References
1] M.Hauschild et al., Nucl.Instr. and Meth. A314 (1992) 74{85. 2] A.Boyarski et al., Nucl.Instr. and Meth. A283 (1989) 617{621. 3] M.Dubrovin, D.Coupal, Amplitude Analysis of Cosmic Runs with BaBar Drift Chamber Prototype-II & dE/dx, BaBar Note 426, April 20, 1998. 4] V.Blinov, TNDC-97-397, October 28, 1997. 5] Report of Fergus Wilson presented on February'99 BaBar Collaboration Meeting. 6] W.Blum, L.Rolandi, Particle Detection with Drift Chambers, Springer Verlag Berlin Heidelberg 1993. 7] M.Berthet et al., BaBar Note 479, January 20, 1999. 8] V.Blinov, BaBar Note 480, January, 1999. 9] M.Milek, P.M.Patel, BaBar Note 485, March 19, 1999. 10] K.K.Gan et al., Nucl.Instr. and Meth. A374 (1996) 27{33. 11] Bethe, Bloch, 1930. 12] M.Hauschild, CERN PPE/91-130. 13] W.W.M.Allison, Ann. Rev. Nucl. Sc. 30, 253 (1980). 14] A.H.Walenta, Nucl.Instr. and Meth. 161 (1979) 45{58. 15] Results of Federico Colecchia presented on February'99 BaBar Collaboration Meeting. 16] J.Vavra et al., Nucl.Instr. and Meth. A203 (1982) 109{118. 17] D.Boutigny, private communication. 18] D.Coupal, TNDC 95-16, 1996. 19] H.Fischle et al., Nucl. Instr. Meth. A301 (1991) 202.
20
Calibration 160
C(7,:)
Channels
Entries Mean RMS
C(8,:)
Channels
7424 4.817 0.3194
200 100 0
4.5
Pedestal Channels
7424 1.429 0.1570
5.5
0 0.03
0.035
R.M.S. of Pedestal
Entries Mean RMS 7424 1.365 0.2357
0.04
0.045
C(11,:)
Channels
300 200 100 0 0.75 1 1.25 1.5 1.75
Entries Mean RMS
C(22,:)
400 200 0
0.5
1.5
C(23,:)
Channels Channels
800 600 400 200 0 0
7424 0.4806 0.6659E-01
C(27,:)
800 600 400 200
Entries Mean RMS 7424 0.4957 0.6858E-01
0.2
0.4
0.2
0.4
21
Run 2248
Hits
20000
Entries Mean RMS Constant Mean Sigma
Hits
15000
(a)
10000
(b)
5000
1000 0
-4
-2
ADC counts
Entries Mean RMS Constant Mean Sigma 91057 7.595 14.95 0.1263E+05/ 74 0.2090E+05 5.006 5.978
2000
4000
6000
Qwf, fC/2
91057 1289. 1037.
8000
Hits
Hits
(c)
(d)
-50
50
100
2000
4000
Qntup, fC/2
6000
8000
Figure 2: a{spectra of pedestal di erence, b; d{two type of charges, and c{charge di erence.
22
Run 2248
Ev 11 L24 W 46 dQ 872.3
100 50 0
Entries Mean RMS 32 19.92 7.869
Ev 11 L23 W 46 dQ1188.3
100 50
Entries Mean RMS 32 21.32 8.206
10
20
30
10
20
30
Ev 12 L19 W 64 dQ 73.2
100 50 0 0 10 20 30
Entries Mean RMS 32 12.75 6.308
Ev 14 L31 W 39 dQ 68.2
10 5 0 0 10 20 30
Entries Mean RMS 32 19.87 7.877
Ev 14 L32 W 38 dQ 54.4
100 50 0 0 10 20 30
Entries Mean RMS 32 12.22 7.170
Ev 15 L33 W 26 dQ 52.4
20 10 0 0 10 20 30
Entries Mean RMS 32 14.58 7.911
Ev 17 L13 W 24 dQ 62.4
Entries Mean RMS 32 19.10 7.316
20
10
20
30
20 10 0
Ev 17 L13 W 25 dQ 63.7
Entries Mean RMS 32 19.77 7.544
10 5 0
Ev 17 L14 W 25 dQ 61.8
Entries Mean RMS 32 19.48 7.886
10
20
30
10
20
30
Run 2248
Waveforms
Entries Mean RMS
Waveforms
N integrated samples
91057 26.09 4.423
10 5 10 4 10 3 10 2 10
N overflowed(64) samples
Entries Mean RMS 91057 0.1879 1.446
10
20
Nint.samples Waveforms
91057 3.248 3.986
30
10
20
Nover.samples
91057 2.356 1.315
30
N time samples
30000 20000 10000
Entries Mean RMS
Nbi-lin.samples
Entries Mean RMS 91057 7.633 3.707
30
10
20
Ntime samples
30
24
Povrefow , in %
12 10 8 6 4 2 0
90 80 70 60 50 40 30 20 10 0
High Voltage, V
1900
1950
2000
High Voltage, V
2050
2100
High Voltage, V
2050
2100
80
20 18 16 14 12 10 8 6 4 2 0
1900
1950
2000
High Voltage, V
2050
2100
25
ADC counts
40
Noverflow 31 Level 31
20
Pmeasured
TDC-samples A32
Reference pedestal
10 15 20 25 30
Sample No.
Figure 6: Defenitions some of the waveform parameters.
Run 2248
A32 , ADC count
25 20 15 10 5 0 0 2000
(a)
30
40 35 30 25 20 15 10
(b)
ENTRIES 0. 1.00 0.
656. 0.890E+05 0.
5 0 2
ENTRIES 0. 0. 0.
620. 0.894E+05 0.
90165 133. 0. 0.
4000
6000
Q, fC/2
8000
Noverflow 63
Figure 7: Scatter plots of A32 versus charge Q (a) and Noverflow 63 (b). 26
Run 2248
Q, fC/2
(a)
Q, fC/2
6000
(c)
Clipped Q, fC/2
2000
3000
10
Noverflow 31
15
Q / <Q>
Q / <Q>
(b)
(d)
Clipped Q, fC/2
3000
10
Noverflow 31
Figure 8: a { correlations between charge Q and clipped charge Qclip. Points with errors are the mean pro le of this distribution. Fitted curve is a six-order polynom. b { relative width of charge distribution versus clipped charge. c { correlations of charge Q and number of over owed samples Noverflow 31 . Points with errors are the mean pro le of this distribution. Fitted curve is a second-order polynom. d { relative width of charge distribution versus Noverflow 31 . 27
Run 2248
(a)
Q corr. / <Qcorr.> Qcorr. , fC/2
6000 5000 4000 3000 2000 1000 0 0 2000 4000 6000 0.7 0.6 0.5 0.4 0.3 0.2 0.1 0 0 2000 4000 6000
(b)
Q, fC/2
Q, fC/2
Figure 9: a { corrected charge Qcorr: as a function of total charge Q. Points with errors are the mean pro le of this distribution tted by the straight line. b { relative width of the corrected charge distribution versus total charge.
Q corr. / <Qcorr.>
0.25 0.2 0.15 0.1 0.05 0
- (Q1+Q2)/2 - Q1(Nover.31) - Q2(Qclip.31)
Q, fC/2
Figure 10: Relative width of the corrected charge distribution versus total charge for three cases of charge reconstruction. 28
Run 268
Events
(a)
Events
50000
(b)
2500
-0.4
-0.2
0.2
0.4
Ntrk Events
5000
(c)
5000
(d)
0 40
60
80
100
120
0 40
60
80
100
120
Ndigi Events
nAllDigi Events
5000
(e)
(f)
1000
0 10
20
30
40
50
nsampl.f(1) Events
(g)
2000
5000
(h)
0 -100
100
200
-40
-20
20
40
z0 , cm
d0 , cm
29
Runs 268,269
(a)
Entries Mean RMS 47119 368.1 143.0
Probability
Events
(b)
2000
0 -1 10
10
10
10
0 -1 10
10
10
10
PT , GeV
PT , GeV Probability
1
Events
(c)
(d)
2000
0 -1 10
10
10
10
0 -1 10
10
10
10
P , GeV
P , GeV Probability
1
Events
(e)
(f)
2000
0 1 10 10
10
10
0 1 10 10
10
10
Figure 12: Spectra of transverse momentum PT , momentum P and and their normalized integrals (b; d; f ) respectively.
30
Events
Hits
1500
1000
0.2
0.4
0.6
0.8
|sin|
-90
, degree
90
180
|sin|
, degree
180 90 0 -90
-180 1 10 10
10
10
Figure 16: Scatter plot of entrance angle versus for 1000 events. Each event is presented by vertical strike of dots, corresponding to the track hits. 31
Run 268
Wire
250
200
150
100
50
10
15
20
25
30
35
Layer
40
32
Runs 268-272
Spectrum for L 1 W 25
400 200 0 0 500
Entries Mean RMS Constant Mean Sigma 5751 644.5 560.8 22.77 / 15 377.5 487.6 312.0
Spectrum for L 2 W 25
400 200 0 0 500
Entries Mean RMS 32.32 Constant Mean Sigma 5666 570.7 505.7 / 15 436.0 443.4 271.7
Spectrum for L 3 W 25
400 200 0 0 500
Entries Mean RMS Constant Mean Sigma 5801 622.3 478.0 37.66 / 15 405.7 495.2 285.3
Spectrum for L 4 W 25
Entries Mean RMS
200 0
500
Spectrum for L 5 W 25
Entries Mean RMS
Spectrum for L 6 W 25
400 200 0 0 500
Entries Mean RMS 53.07 Constant Mean Sigma 4810 563.6 523.3 / 15 360.7 397.8 293.1
200
500
Spectrum for L 7 W 25
400 200 0 0 500
Entries Mean RMS Constant Mean Sigma 4936 614.5 513.6 54.70 / 15 350.0 472.6 283.8
Spectrum for L 8 W 25
Entries Mean RMS
200
500
Spectrum for L 9 W 25
200 100 0 0 500
Entries Mean RMS Constant Mean Sigma 4442 848.8 659.7 17.18 / 15 246.1 660.6 311.0
Spectrum for L 10 W 25
Entries Mean RMS 33.91
200 0
500
Figure 18: Spectra of raw Q cos for several wires. Corrections were not applied.
33
Spectrum for L 2 W 25
1000 500 0 0 500
Entries Mean RMS Constant Mean Sigma 11001 676.9 522.2 48.13 / 10 1060. 540.4 182.3
Spectrum for L 3 W 25
1000 500 0 0 500
Entries Mean RMS Constant Mean Sigma 11185 673.6 527.1 103.6 / 10 1076. 536.9 179.0
Spectrum for L 4 W 25
1000 500 0 0 500
Entries Mean RMS Constant Mean Sigma 10708 662.1 517.9 161.8 / 10 1081. 527.4 172.0
1000 500 0
Spectrum for L 5 W 25
Entries Mean RMS Constant Mean Sigma 9889 661.7 474.8 68.20 / 10 927.1 529.1 186.3
Spectrum for L 6 W 25
Entries Mean RMS 52.16
500 0
500
500
Spectrum for L 7 W 25
1000 500 0
Entries Mean RMS Constant Mean Sigma 9549 651.6 521.3 128.5 / 10 947.1 512.0 175.0
Spectrum for L 8 W 25
1000 500 0
Entries Mean RMS 94.60 Constant Mean Sigma 9472 660.5 493.1 / 10 912.7 520.1 181.3
500
500
Spectrum for L 9 W 25
Entries Mean RMS
Spectrum for L 10 W 25
Entries Mean RMS
500 0
500 0
500
500
Figure 19: Spectra of Q cos for several wires. Wire-to-wire gain, doca and dip angle corrections were applied.
34
Runs 268-272
Sum of single wire spectra Hits
8000 6000 4000
Entries Mean RMS P1 P2 P3 P4 14113946 705.7 608.5 0.3632E+06/ 196 419.2 188.3 0.1480E+07 2.839
Hits
x 10
2 10 Sum
Run 268
of single wire spectra
Entries Mean RMS P1 P2 P3 P4 2993198 632.9 491.7 0.8770E+05/ 196 511.8 108.4 0.4395E+06 1.975
3000
2000
1000 2000 0 0 1000 2000 3000 4000 0 0 1000 2000 3000 4000
Figure 20: Spectrum of raw Q cos summarized for all wires. Corrections were not applied. Fitted curve is Moyal analytical form for Landau distribution approximation.
Run 268
Mean charge
600 400 200 0 0 20 40 60 80 100 120
Q cos
Figure 21: Spectrum of Q cos summarized for all wires. Wire to wire gain, doca and dip angular corrections were applied. Fitted curve is Moyal analytical form for Landau distribution approximation.
Run 268
<Q cos>
800 700 600 500
Q cos
Layer 10
Layer 10
71.67 A0 / 127 643.8
20
40
60
80
A0
100
962.8
120
Wire
/ 39 635.1
40
600
10
15
20
25
30
35
40
Figure 22: Intermediate results of wire-towire gain calibration: top gure | < Q cos > vs wire number for one of the layers, bottom gure | < Q cos > vs layer number. Statistical errors are shown.
Layer
Figure 23: The same, that in Fig. 22, but wire-to-wire gain, doca and dip angle corrections were applied. 35
Layer
Layer 3 Q cos
900 800 700 0 0.2
P1 P2 P3 P4 P1 P2 P3 P4 330.3 / 42 676.1 698.3 -1034. 583.2
Run 268
Q cos
1200 1000 800 600
P1 P2 P3 P4 95.73 / 43 661.6 -57.89 126.7 -91.71
Layer 3
0.4
0.6
0.8
Layer 4 Q cos
900 800 700 0 0.2 0.4 0.6 0.8
295.7 / 42 668.2 774.6 -1114. 605.4
|sin|
400
0
P1 P2 P3 P4
0.2
80.66
0.4
/ 43 675.6 -66.07 108.5 -58.22
0.6
0.8
Q cos
Layer 5
|sin|
|sin|
400
0.2
0.4
0.6
0.8
|sin|
Figure 24: Dependence of Q cos on jsin j for two layers. Wire-to-wire gain correction was applied.
Layers with lower gain Q cos
900 800 700 0 0.2 0.4 0.6 0.8
Figure 25: Dependence of Q cos on jsin j for two layers. Wire-to-wire gain, doca and dip angle corrections were applied.
1200 1000 800 600 1 400 0 0.2 0.4 0.6 0.8 1
Run 268
|sin|
Q cos
Q cos
|sin|
|sin|
400
0.2
0.4
0.6
0.8
|sin|
Figure 26: Dependence of Q cos on jsin j. for all layers. Wire-to-wire gain correction was applied. 36
Figure 27: Dependence of Q cos on jsin j for all layers. Wire-to-wire gain, doca and dip angle corrections were applied.
Layer 3 Q cos
1773. A0 A1 A2 A3 A4 A5 A6 A7
Run 268
/ 14 1007. -13.84 -1380. 247.7 1405. -528.7 -757.3 290.2
Layer 3
137.9 A0 A1 A2 A3 A4 A5 A6 A7 / 14 645.1 4.391 -169.7 -56.42 606.1 146.9 -505.8 -99.52
Q cos
Layer 5 Q cos
1000 750 500 250 0 -1 0
A0 A1 A2 A3 A4 A5 A6 A7
Doca / dl
4680. / 14 959.0 267.3 -879.6 -819.2 473.1 891.4 -317.7 -324.5
Layer 5 Q cos
1000 750 500 250 0 -1 0
A0 A1 A2 A3 A4 A5 A6 A7
Doca / dl
747.6 / 14 645.0 -10.92 -207.1 98.86 702.1 -222.3 -567.9 138.4
Doca / dl
Doca / dl
Figure 28: Dependence of Q cos on normalized doca for two layers. Wire-to-wire gain correction was applied.
Layers with lower gain Q cos
1000 750 500 250 0 -1 0 1
Figure 29: .Dependence of Q cos on normalized doca for two layers. Wire-to-wire gain, doca and dip angle corrections were applied.
Layers with lower gain Q cos
1000 750 500 250 0 -1 0 1
Run 268
Doca / dl
Doca / dl
Figure 30: Dependence of Q cos on normalized doca for all layers. A t for the 40-th layer is plotted in both gures and it is di erent from all other curves. Wire-to-wire gain correction was applied.
Figure 31: Dependence of Q cos on normalized doca for all layers. Wire-to-wire gain, doca and dip angle corrections were applied. 37
Run 268
Q cos
1200 1000 800 600 400
P1 P2 P3 P4 83.25 / 42 632.6 374.5 -135.9 20.30
Layer 3
dE/dx
0.6 0.8
600 500
0
P1 P2 P3 P4
0.2
0.4
Q cos
Layer 4
118.3 / 42 637.1 212.7 420.9 -397.6
|sin|
1/2
400 300 10 10
10
0.2
0.4
0.6
0.8
|sin|
1/2
Figure 32: Dependence of Q cos on jsin j1=2 for two layers. Wire-to-wire gain correction was applied. Straight line could be used for approximation.
Figure 33: Scatter plot of dE=dx versus for 10 K events of run 268. Systematic corrections: wire-to-wire gain; dip angle; doca as well as 10%:25% truncated mean were applied to dE=dx.
38
Band 4: 1.99526<<2.51189
40.30
1.4
P1 P2 P3
I / Imin
Band 8: 5.01187<<6.30957
114.7 P1 P2 P3 / 20 690.6 392.1 29.77
1 10 10
2
Figure 35: Relative dE=dx versus . For open circles I and were de ned as Mean and RMS values; black dots | dE=dx distribution for each bin was tted by Gaussian and its parameters were used for I and .
10
0.1
250 0 100
/I
35.70
0.09
- RMS / Mean
0.08 0.07
Events
0.06 0.05 10 10
2
800
1000
dE/dx
Figure 34: dE=dx spectra for several bins of . Gaussian t is used for de nition of the most probable and width ( ) of distribution. 39
Figure 36: dE=dx resolution for the cases presented in Fig. 35.
10
a) <dE/dx>
0.6 0.4 0.2 0
b) dE/dx / <dE/dx>
N 20 lo
w dis c.
20 0 0
N high disc
arded
N 20 lo
w dis c.
20 0 0
N high disca
rded
c) Detailed map
dE/dx / <dE/dx> Nlow discarded
10 8 6 4 0.08 2 0 10 0.06 0.14 0.12 0.1
Nhigh discarded
15
10
20
Ndiscarded
30
Figure 37: Truncated mean optimization at 2 < P < 4 GeV and Nsamples = 40. The location of the optimum does not show signi cant dependence on momentum. 40
a) <dE/dx>
800 700 600 500 400
Run 268
b) dE/dx / <dE/dx>
N 20 lo
w dis c.
20 0 0
Nh
scard igh di
ed
N 20 lo
w dis c.
20 0 0
N high disca
rded
c) Detailed map
dE/dx / <dE/dx> Nlow discarded
20 15 10 5 0.05 0 15 20 25 0.04 0.1 0.09 0.08 0.07 0.06
Nhigh discarded
10
20
Ndiscarded
30
Figure 38: Truncated mean optimization at 2 < P < 4 GeV and Nsamples = 80. The location of the optimum con rms, that the aspect ratio for truncated mean could be used as 10% : 25% of discarded hits with lowest and highest amplitudes respectively. 41
I / I min
(a)
P1 P2 P3
2
10
10
I / I min
(b)
P1 P2 P3 P4 P5 10
2
2p3
10
a)
p1
2 p3 (
p2 + 2ln(
2p5
b)
p1
2 p5 (
p2
ln(p3 + (
)p4 ))
Figure 39: Fit of the normalized ionization strength: a | approximation of Bethe-Bloch formulae 11] with additional parameter p3 ; b | modi ed approximation 12] taking into account \density e ect". 42
Run 268
RMS / <dE/dx>
0.1
(a)
0.08 0.06 0.04 10 10
2 3
- All hits of 2 tracks - 1 - st track
RMS / <dE/dx>
10
(b)
10
Figure 40: Comparison of dE=dx resolution versus were applied in each case.
10
10
43
Run 268
RMS / <dE/dx>
0.1 0.09 0.08 0.07 0.06 0.05 10 10
- 1 - st track - |Doca/dl| < 0.9 - Nhit(1)=40 2
(a)
RMS / <dE/dx>
0.1
10
(b)
0.08 0.06 0.04 10
10
Figure 41: Comparison of dE=dx resolution versus were applied in each case.
44
/<dE/dx>
- 35<Nhit<46, hits of 1-st track - 71<Nhit<91, edge 18 layers - 71<Nhit<91, inner 22 layers
/<dE/dx>
0.1
10
10
0.05
Figure 42: Comparison of dE=dx resolution versus for di erent cases. 10%:25% truncated mean were applied in each case.
10
10
Figure 43: Comparison of dE=dx resolution versus for di erent cases. Event selection and algorithm of truncated mean procedure are strongly in uenced for dE=dx resolution.
Run 268
dE/dx /<dE/dx>
1000 800 600 400 200 0 0 0.5 1
(a)
0.1
(b)
0.02 0 0 0.5 1
|sin| |sin| Figure 44: a | scatter plot of dE=dx versus jsin j obtained at 2 < P < 4 GeV and 38 Nhit 42; b | dE=dx resolution versus jsin j, solid curve | tted expected dependence on l (amplitude p sample length) l 0:32 13]; dashed curve | statistical limit 1= l.
45
Run 268
/ <dE/dx> <dE/dx>
600
27.66 P1 P2 / 29 0.5216 0.5518
550
(a)
0.15
(b)
450
Nhits
40
10
20
30
Nhits
40
Figure 45: Dependences of a | < dE=dx > and b | dE=dx resolution on number of samples per track. Truncated mean algorithm: 3 hits with low amplitude and 25% with high amplitude are discarded. 1-st track selection conditions 2 < P < 4 GeV and initial Nhits = 40 have been applied. Each track contributes one time per each point with randomly selected set of descending in number amplitude samples. Integer number in uence at 25% truncated mean level is clearly appeared as a p2 step-wise dependence in a. Standard parameterization p1 =Nhit was used to t points in b. Rather 0:5 fast slope than Nhit is observed at small Nhit .
46
<dE/dx>
700
600
(a)
- without truncated mean - 10:25% truncated mean
500
400
10
20
30
40
50
60
70
80
Nhits
90
/ <dE/dx>
0.1805E-02 0.6379E-03
(b)
- without truncated mean - 10:25% truncated mean
10
20
30
40
50
60
70
80
Nhits
90
Figure 46: The same that in Fig. 45, but all hits of the event are treated as for one track without limitations on initial value Nhits 80. Track cut-o parameters: 2 < P1 < 4 GeV , jP1 P2 j < 0:2 GeV . Open circles and table of the t parameters correspond to evaluation of dE=dx without excluding of amplitude samples. Filled circles | for optimal truncated mean algorithm (10% : 25%). Detailed results of the t for truncated mean are listed in Table 1.
47
(a)
700
650
600
100
200
300
Point/1000 events
400
500
Q / <Q>
1.4
A0 0.2770E+06/ 512 0.9638
(b)
1.2
0.8
100
200
300
Point/1000 events
400
500
Figure 47: Long term behavior of summarized spectrum mean value | a and relative r.m.s. width of spectrum | b for runs 268 281 (13:38 | 23:30 January 9, 1999). Wire to wire gain, doca and dip angular corrections were de ned for run 268 and applied for all next runs.
48
Probability
250
500
750
1000
1250
1500
1750
2000
2250
Q cos
2500
Probability
250
500
750
1000
1250
1500
1750
2000
2250
Q cos
2500
Probability
250
500
750
1000
1250
1500
1750
2000
2250
Q cos
2500
Probability
250
500
750
1000
1250
1500
1750
2000
2250
Q cos
2500
Q cos for 48 (of 50) slices on jsin j. Top curves correspond to jsin j 0, bottom | jsin j 1. All statistics shared between low (a) and high (b) gain layers.
Figure 49: The same, that in Fig. 48, but after application of considered correction.
49
Q cos
Q cos
2000
0.2
0.4
0.6
0.8
|sin|
0.2
0.4
0.6
0.8
|sin|
Figure 50: Correction functions for low gain layers, which should be applied to di erent Q cos on jsin j.
Figure 51: The same that in Fig. 50 for layers with high gain.
Q cos / Q cos30
1.2
1.2
0.8
|sin|
1
50
0.2
0.4
0.6
0.8
|sin|
Figure 52: Relative corrections to be applied for di erent Qcos on jsin j. Statistics processed for low gain layers only.
Figure 53: The same that in Fig. 52, but statistics processed for high gain layers.
P1
19 12.93 5.146
P2
19 14.45 4.488
15
20
10
15
20
P3
P4
10
15
20
10
15
20
Figure 54: Fit parameters versus probability level number. Solid line histogram | for low gain layers, dashed | for high gain layers.
51
I / I min
/I
1.4
0.09
- Old corr.
1 10 10
2
10
0.05 10 10
10
Figure 55: Normalized distributions of ionization lose strength versus for three types of dip angle corrections.
Figure 56: dE=dx resolution versus three types of dip angle corrections.
for
I / I min , IK / I min
- -K
1.4
1.2
1 10
-1
10
P , GeV
10
|IK - I| /
15 10 5 0 1 10
- Old corr. - New corr. =30o o - New corr. =0
|IK - I| /
20
1.5 1 0.5 0 1
- Old corr. - New corr. =30 - New corr. =0
o o
P , GeV
10
P , GeV
10
Figure 58: Expected K = separation as a function of momentum for three types of dip angle corrections.
Figure 59: The same that in Fig. 58, zoomed at P > 1 GeV .
53
Simulation
Events
Entries Mean RMS 100000 20.00 4.466
Events
10000
10 6 10 5
7500
(a)
5000
10 4 10 3
(b)
2500
10 2 10 10 20 30 40
50
20
40
60
80
100
Events
10
Events
8000 6000
10 4 10 3 10 2 0 1000
(c)
4000 2000 0 0
(d)
2000
3000
Q / <m> / <Gain>
Figure 60: Intermediate distributions at simulation of single cell spectra: a | simulated Poisson distribution P<m> (m) for mean number of primary electrons per track < m >= 20; b | simulated cluster size distribution 19] for case of He; c | simulated Yule-Furry (exponential) distribution of gas gain uctuations with mean value 5 104 ; d | simulated spectrum of number of electrons per cluster: dashed histogram shows the spectra without gas gain uctuations (Gain = 1), hatched histogram | all indicated statistical contributions are taken in to account.
54
0.5
Mean RMS
1.736 1.699
0.05
Mean RMS
2.127 2.001
<m> = 1
0 0.2
Mean RMS 1.490 1.415
<m> = 1
0 0.05
Mean RMS 1.722 1.640
<m> = 2
0 0.1
Mean RMS 1.517 1.176
<m> = 2
0 0.05
Mean RMS 1.611 1.286
<m> = 5
0 0.05
Mean RMS 1.623 1.051
<m> = 5
0 0.05
Mean RMS 1.672 1.124
<m> = 10
0 0.05 0 0.1
Mean RMS 1.727 1.003
<m> = 10
0 0.05 0 0.1
Mean RMS 1.752 1.038
<m> = 20
Mean RMS 1.839 0.9531
<m> = 20
Mean RMS 1.846 0.9612
<m> = 50
0 0.1
Mean RMS 1.935 0.9869
<m> = 50
0 0.1
Mean RMS 1.942 0.9851
<m> = 100
Probability
Mean RMS 1.898 0.3214
<m> = 100
Probability
0 0.1 0
Mean RMS 1.904 0.3295
0 0.2
<m> = 1000
0 0 1 2 3 4 5 6 7 8
<m> = 1000
0 1 2 3 4 5 6
n / <m>
10
Q / <m> / <Gain>
10
Figure 61: Evolution of simulated spectra for several values of < m >. Gas gain uctuations are not accounted.
Figure 62: The same that in Fig. 61, but gas gain uctuations are taken in to account at < Gain >= 5 104 .
55
1.5
(a) - Qm.p. -
0.5
<m>
10
/ Qm.p.
(b)
<m>
10
Figure 63: a | dependence of the most probable charge and the width of spectral distribution on < m >; b | relative width of the spectra as a function of < m >.
56
(a)
90000 80000 70000 60000 50000 40000 30000 20000 10000 0 1000 20 30 2250 2000 1750 1500 1250 1000 750 500 250 0 1000
(c)
Q
co r.
10 0
Index
Qcor.
1500
Q
co r.
10 0
20
30
Index
Qcor.
1500
(b)
(d)
1000
1000
500
500
10
15
20
25
10
15
20
25
Index
Index
Figure 64: Single cell spectra of corrected charge versus index of band (proportional to ln ). a; b | sum of spectra of all drift cells; c; d | sum of spectra of 1-st layer drift cells.
57
Band 4: 1.99526<<2.51189
0.1
452.4 P1 P2 P3 P4 / 15 0.1010 410.2 190.4 0.2944
0 0.1
Band 8: 5.01187<<6.30957
1550. P1 P2 P3 P4 / 15 0.1112 393.9 175.1 0.3201
0 0.1
0 0.1
0 0.1
0 0.1
0 0.1
Probability
0.1
500
1000
1500
2000
2500
Qcor.
Figure 65: Fit of single cell spectra obtained for di erent bands of .
58
Ratios
1.6
1.4
(a)
1.2
0.8
10
10
/ Qm.p.
0.5
0.45
(b)
0.4
0.35
0.3 1 10 10
10
10
Figure 66: Dependence of spectral t parameters on for: a | most probable corrected charge and width of spectral distribution, both normalized on its minimal values; b | relative width of spectral distribution. Spectra of all drift cells were summarized together, but similar dependence with increased statistical errors were also observed for spectra summarized for each layer.
59