- Email: [email protected]
The Belle Silicon Vertex Detector
ARTICLE IN PRESS
Nuclear Instruments and Methods in Physics Research A 549 (2005) 16–19 www.elsevier.com/locate/nima
The Belle Silicon Vertex Detector H. Ishino Department of Physics, Tokyo Institute of Technology, 2-12-1, Ookayama, Meguro-ku, Tokyo, Japan Available online 23 May 2005
Abstract The Belle Silicon Vertex Detector (SVD) was designed to measure B meson decay positions precisely for studies of time-dependent CP violation. Although the first version of SVD (SVD1) worked for 4 years from 1999 with excellent performance, its insufficient radiation hardness drove us to build a second generation SVD (SVD2). The SVD2 was installed in Belle in the summer of 2003 and has been working well. The strip yield is estimated to be more than 95%. The signal-to-noise ratios are obtained to be 18–36, depending on detector ladders. The intrinsic spatial resolutions are obtained to be 12:0 0:4 and 22:3 0:8 mm for f- and z-sides, where z-side measures positions along the beam direction and f side is used for the azimuthal angle measurement. In this letter, an overview and performance results are provided for both generations of the Belle SVD. r 2005 Elsevier B.V. All rights reserved. PACS: 29.40.Wk Keywords: Belle; KEKB; Silicon vertex detector
1. Introduction One of the primary physics goals of the Belle experiment is an accurate measurement of the time-dependent CP violation parameters in the neutral B meson system. Large numbers of B mesons and a precise determination of their decay times are required for that purpose. During the past 4 years the KEKB asymmetric electron–positron collider in Japan, with the world’s highest peak luminosity of 1:06 1034 cm2 s1 , has proTel.: +81 3 5734 2388; fax: +81 3 5734 2389.
E-mail address: [email protected].
vided us 152 million Uð4SÞ particles. This resonance subsequently decays into two B mesons with a Lorentz boost of bg ¼ 0:425 in the laboratory frame. A time difference of the two meson decays is obtained from a measured distance of the two decay vertex positions, a distance of 200 mm on average. Since a Silicon Vertex Detector (SVD) [1] enables us to measure the distance with an accuracy of 100 mm or better, it is essential for these measurements. The first version of the SVD (SVD1) had worked for 4 years from 1999 [2]. Its great performance produced many fruitful physics results. For instance, sin 2f1 , one of the CP
0168-9002/$ - see front matter r 2005 Elsevier B.V. All rights reserved. doi:10.1016/j.nima.2005.04.018
ARTICLE IN PRESS H. Ishino / Nuclear Instruments and Methods in Physics Research A 549 (2005) 16–19
violation parameters, is measured to be 0:73 0:06 ðstat:Þ 0:03 ðsys:Þ, the best accuracy ever achieved. Although operation of the SVD1 was successful, it had some shortcomings such as the low radiation tolerance of the front-end electronics. An upgrade version of SVD (SVD2) [3] was designed and built in order to overcome the weak points with substantial modifications and improvements. The SVD2 was installed in Belle in the summer of 2003 and has been operating successfully with expected performance. In Section 2, an overview of the SVD1 and a description of the SVD1 performance are given. Section 3 presents an overview of the SVD2 including the points of improvement, as well as performance of the SVD2 obtained from cosmic ray muon data.
2. SVD version 1 The SVD1 detector consists of three layers of independent ladders with a polar angle coverage of 23 oyo139 . Each ladder is made up of two, three and four double-sided silicon strip detectors (DSSDs) reinforced by support ribs and readout front-end electronics (VA1) [4] for the first, second and third layers, respectively. The radii of the three layers are 30, 45.5 and 60.5 mm. The DSSD fabricated by Hamamatsu Photonics (HPK S6936) has a sensitive area of 57:5 33:5 mm2 and 300 mm thickness. It comprises 640 readout lines with 50 ð84Þ mm pitch for f (z)-side, where f is the azimuthal angle, and z is the beam direction. In total 102 DSSDs are used, and there are 81,920 readout channels. On each end of both sides of a ladder, five VA1 chips readout signals from a half of the ladder. Versions of the VA1 chips manufactured in a 0:8 mm process could function up to 1 Mrad irradiation. The 640 multiplexed signal outputs from five VA1 chips are fed serially to flash analog-to-digital converters (FADC) with a 5 MHz rate through a 30 m cable. Digital signal processors (DSPs) on the FADCs carry out pedestal and common mode subtraction, sparsification and data formatting. The radiation dose was monitored using RADFET transistors attached on the beam pipe
17
directly. The total radiation dose on the first layer detector ladders was measured to be 0.9 Mrad. The signal-to-noise ratio was kept between 20 and 30 during the SVD1 operations despite of a degradation of 30% in the first layer due to the accumulated dose, implying the good separations of signal and noise for all run periods. The hitfinding efficiency was estimated to be about 94% on average. The detector occupancy with the beam collision condition was observed to be about 4%. The impact parameter (IP) resolution is a suitable parameter for a check of detector performances, since it honestly reflects the alignment accuracy and the intrinsic spatial resolution. The IP resolution was obtained using data samples of cosmic ray muons, mu-pair events, and two photon events such as gg ! r0 r0 ! pþ p pþ p . The momentum and angular dependence of the resolutions for fand z-sides were estimated to be sf ¼ 18:6 51:3= pt b sin3=2 y mm, sz ¼ 40:8 43:5=pt b sin5=2 y mm, where pt is the transverse momentum in GeV/c, and b is the velocity of the charged particle. The IP resolutions of both f- and z-sides were found to be stable to within 10% throughout all periods, demonstrating that the overall quality of vertex reconstruction with the SVD1 remained uniform.
3. SVD version 2 Although the SVD1 demonstrated satisfactory performance, it left room for improvement. An upgraded SVD, designated SVD2, was designed for the following points of improvement:
Radiation hardness: the VA1TA chip [4] is fabricated in a 0:35 mm CMOS process, permitting operation to at least 20 Mrad irradiation. Vertex resolution and tracking efficiency: the SVD2 has four layers of ladders for improved tracking efficiency. The innermost layer radius is 20 mm, which is closer to the beam collision point than that of the SVD1, improving the IP resolution by about 20% for low momentum particles less than 250 MeV=c. Enlargement of solid angle: the SVD2 has a polar angle coverage of 17 oyo150 , which is
ARTICLE IN PRESS 18
H. Ishino / Nuclear Instruments and Methods in Physics Research A 549 (2005) 16–19
the same as that of the rest of the Belle tracking volume. Protection against the pin-holes: in the SVD1 the breakdown of the AC-coupling on a DSSD happened a few times per year, inducing a large noise due to a huge current flowing into the VA1 chip. In the upgrade, the readout chips of p(n)-side have their local ground potentials floated on the detector bias at ðþÞ40 V. The voltage decoupling between the front-end and rear-end electronics is made by an opto-coupler. Trigger capability [5]: the VA1TA chip contains a faster shaper (75 or 300 ns) and a discriminator (so-called TA part) for providing a fast trigger decision. A data acquisition (DAQ) system: the readout dead-time is significantly reduced from 128 to 25:6 ms, because analog data from individual VA1TA chip, containing 128 strip signals, are digitized in parallel. In addition, a PC farm is used for readout of digitized data from the FADCs. This farm can perform sparsification with three times larger bandwidth as that of the SVD1. The processing rate is measured to be 1.3 kHz with less than 5% dead time at a 5% detector occupancy level.
The DSSD used for the first to third (fourth) layer has the area of 76:8 25:6 mm2 (73:8 33:3 mm2 ) and the strip pitch of 75 ð73Þ mm for zside and 50 ð65Þ mm for f-side. The number of readout strips of both z- and f-sides is 512, and
500
they are mated to four VA1TA chips located at the end of a ladder. From the first to fourth layer, each ladder consists of 2, 3, 5 and 6 DSSDs glued onto a mechanical support rib. In total 246 DSSDs are utilized, and the total number of channel amounts to 110,592. A wide area flexible printed circuit (FPC) with 50 mm fine pitch is employed, instead of a double metal layer (DML) for orthogonal coordinate readout, as used in the SVD1. Advantages of using a FPC include: lowering the intrinsic noise by significantly reducing the capacitance compared with the DML structure, as well as decoupling the mechanics of the front-end electronics and DSSDs, which permits reducing the radius of the innermost sensor layer in the limited physical volume available. After the construction and commissioning of the SVD2, it was installed in Belle during the summer of 2003, and a system integration test was performed with cosmic ray muon events taken without a magnetic field. Typical detector intrinsic noise is measured to be 500, 700, 1000 and 1000 electrons for the first to fourth layer ladders, respectively. Since cluster energies are measured to be about 18,000 electrons with minimum ionizing particles, the signal-to-noise ratios are obtained to be 18–36, which are comparable with those of the SVD1, although the detector area is almost doubled for the outermost layer. Fig. 1 shows the typical residual distributions of f- and z-sides of one DSSD in the second layer after the
400 350
400 300 300
250 200
200 100
150 100 50
0 -100-80 -60 -40 -20 0 20 40 60 80 100 um (a) Phi residual
0 -150 -100
( b)
-50
0
50
100
150 um
Z residual
Fig. 1. Residual distributions of (a) f- and (b) z-sides of one DSSD after the alignment. The curves show the fitted double Gaussians.
ARTICLE IN PRESS H. Ishino / Nuclear Instruments and Methods in Physics Research A 549 (2005) 16–19
alignment with the cosmic ray muon events. From the distributions, the intrinsic spatial resolutions of f- and z-sides are measured to be 12:0 0:4 and 22:3 0:8 mm, respectively. The strip hit efficiency is obtained to be 95–99%, depending on ladders as expected. The inefficiencies are due to FPC defects, wire bonding problems and malfunction of the readout chips. The discriminator threshold levels of the TA part for all p(z)-side strips have been tuned using calibration input pulses. The obtained threshold so far is about 12,000 electrons on average and overall TA hit efficiency is measured to be about 69%. Significant effort is being made to increase the efficiency to about 90%.
19
operation at higher beam collision rates and larger radiation doses.
Acknowledgements The author would like to acknowledge and thank the Belle Collaboration and the Belle SVD group for their support. The author is also grateful to the physics department of KEK for their continuous encouragement and to the mechanical engineering center of KEK for their contribution to the construction and the installation of the SVD2 system.
4. Conclusion References The excellent performance of the SVD1 produced outstanding physics results during its 4 years of operation. A second generation upgrade, the SVD2, installed in the summer of 2003 is working well as expected. Improvements implemented in the SVD2 will permit more stable
[1] S. Mori, et al. (Eds.), Nucl. Instr. and Meth. A 479 (2002) 117. [2] T. Kawasaki, Nucl. Instr. and Meth. A 494 (2002) 94. [3] Y. Ushiroda, Nucl. Instr. and Meth. A 511 (2003) 6. [4] http://www.ideas.no [5] T. Ziegler, Nucl. Instr. and Meth. A 511 (2003) 153.
Nuclear Instruments and Methods in Physics Research A 549 (2005) 16–19 www.elsevier.com/locate/nima
The Belle Silicon Vertex Detector H. Ishino Department of Physics, Tokyo Institute of Technology, 2-12-1, Ookayama, Meguro-ku, Tokyo, Japan Available online 23 May 2005
Abstract The Belle Silicon Vertex Detector (SVD) was designed to measure B meson decay positions precisely for studies of time-dependent CP violation. Although the first version of SVD (SVD1) worked for 4 years from 1999 with excellent performance, its insufficient radiation hardness drove us to build a second generation SVD (SVD2). The SVD2 was installed in Belle in the summer of 2003 and has been working well. The strip yield is estimated to be more than 95%. The signal-to-noise ratios are obtained to be 18–36, depending on detector ladders. The intrinsic spatial resolutions are obtained to be 12:0 0:4 and 22:3 0:8 mm for f- and z-sides, where z-side measures positions along the beam direction and f side is used for the azimuthal angle measurement. In this letter, an overview and performance results are provided for both generations of the Belle SVD. r 2005 Elsevier B.V. All rights reserved. PACS: 29.40.Wk Keywords: Belle; KEKB; Silicon vertex detector
1. Introduction One of the primary physics goals of the Belle experiment is an accurate measurement of the time-dependent CP violation parameters in the neutral B meson system. Large numbers of B mesons and a precise determination of their decay times are required for that purpose. During the past 4 years the KEKB asymmetric electron–positron collider in Japan, with the world’s highest peak luminosity of 1:06 1034 cm2 s1 , has proTel.: +81 3 5734 2388; fax: +81 3 5734 2389.
E-mail address: [email protected].
vided us 152 million Uð4SÞ particles. This resonance subsequently decays into two B mesons with a Lorentz boost of bg ¼ 0:425 in the laboratory frame. A time difference of the two meson decays is obtained from a measured distance of the two decay vertex positions, a distance of 200 mm on average. Since a Silicon Vertex Detector (SVD) [1] enables us to measure the distance with an accuracy of 100 mm or better, it is essential for these measurements. The first version of the SVD (SVD1) had worked for 4 years from 1999 [2]. Its great performance produced many fruitful physics results. For instance, sin 2f1 , one of the CP
0168-9002/$ - see front matter r 2005 Elsevier B.V. All rights reserved. doi:10.1016/j.nima.2005.04.018
ARTICLE IN PRESS H. Ishino / Nuclear Instruments and Methods in Physics Research A 549 (2005) 16–19
violation parameters, is measured to be 0:73 0:06 ðstat:Þ 0:03 ðsys:Þ, the best accuracy ever achieved. Although operation of the SVD1 was successful, it had some shortcomings such as the low radiation tolerance of the front-end electronics. An upgrade version of SVD (SVD2) [3] was designed and built in order to overcome the weak points with substantial modifications and improvements. The SVD2 was installed in Belle in the summer of 2003 and has been operating successfully with expected performance. In Section 2, an overview of the SVD1 and a description of the SVD1 performance are given. Section 3 presents an overview of the SVD2 including the points of improvement, as well as performance of the SVD2 obtained from cosmic ray muon data.
2. SVD version 1 The SVD1 detector consists of three layers of independent ladders with a polar angle coverage of 23 oyo139 . Each ladder is made up of two, three and four double-sided silicon strip detectors (DSSDs) reinforced by support ribs and readout front-end electronics (VA1) [4] for the first, second and third layers, respectively. The radii of the three layers are 30, 45.5 and 60.5 mm. The DSSD fabricated by Hamamatsu Photonics (HPK S6936) has a sensitive area of 57:5 33:5 mm2 and 300 mm thickness. It comprises 640 readout lines with 50 ð84Þ mm pitch for f (z)-side, where f is the azimuthal angle, and z is the beam direction. In total 102 DSSDs are used, and there are 81,920 readout channels. On each end of both sides of a ladder, five VA1 chips readout signals from a half of the ladder. Versions of the VA1 chips manufactured in a 0:8 mm process could function up to 1 Mrad irradiation. The 640 multiplexed signal outputs from five VA1 chips are fed serially to flash analog-to-digital converters (FADC) with a 5 MHz rate through a 30 m cable. Digital signal processors (DSPs) on the FADCs carry out pedestal and common mode subtraction, sparsification and data formatting. The radiation dose was monitored using RADFET transistors attached on the beam pipe
17
directly. The total radiation dose on the first layer detector ladders was measured to be 0.9 Mrad. The signal-to-noise ratio was kept between 20 and 30 during the SVD1 operations despite of a degradation of 30% in the first layer due to the accumulated dose, implying the good separations of signal and noise for all run periods. The hitfinding efficiency was estimated to be about 94% on average. The detector occupancy with the beam collision condition was observed to be about 4%. The impact parameter (IP) resolution is a suitable parameter for a check of detector performances, since it honestly reflects the alignment accuracy and the intrinsic spatial resolution. The IP resolution was obtained using data samples of cosmic ray muons, mu-pair events, and two photon events such as gg ! r0 r0 ! pþ p pþ p . The momentum and angular dependence of the resolutions for fand z-sides were estimated to be sf ¼ 18:6 51:3= pt b sin3=2 y mm, sz ¼ 40:8 43:5=pt b sin5=2 y mm, where pt is the transverse momentum in GeV/c, and b is the velocity of the charged particle. The IP resolutions of both f- and z-sides were found to be stable to within 10% throughout all periods, demonstrating that the overall quality of vertex reconstruction with the SVD1 remained uniform.
3. SVD version 2 Although the SVD1 demonstrated satisfactory performance, it left room for improvement. An upgraded SVD, designated SVD2, was designed for the following points of improvement:
Radiation hardness: the VA1TA chip [4] is fabricated in a 0:35 mm CMOS process, permitting operation to at least 20 Mrad irradiation. Vertex resolution and tracking efficiency: the SVD2 has four layers of ladders for improved tracking efficiency. The innermost layer radius is 20 mm, which is closer to the beam collision point than that of the SVD1, improving the IP resolution by about 20% for low momentum particles less than 250 MeV=c. Enlargement of solid angle: the SVD2 has a polar angle coverage of 17 oyo150 , which is
ARTICLE IN PRESS 18
H. Ishino / Nuclear Instruments and Methods in Physics Research A 549 (2005) 16–19
the same as that of the rest of the Belle tracking volume. Protection against the pin-holes: in the SVD1 the breakdown of the AC-coupling on a DSSD happened a few times per year, inducing a large noise due to a huge current flowing into the VA1 chip. In the upgrade, the readout chips of p(n)-side have their local ground potentials floated on the detector bias at ðþÞ40 V. The voltage decoupling between the front-end and rear-end electronics is made by an opto-coupler. Trigger capability [5]: the VA1TA chip contains a faster shaper (75 or 300 ns) and a discriminator (so-called TA part) for providing a fast trigger decision. A data acquisition (DAQ) system: the readout dead-time is significantly reduced from 128 to 25:6 ms, because analog data from individual VA1TA chip, containing 128 strip signals, are digitized in parallel. In addition, a PC farm is used for readout of digitized data from the FADCs. This farm can perform sparsification with three times larger bandwidth as that of the SVD1. The processing rate is measured to be 1.3 kHz with less than 5% dead time at a 5% detector occupancy level.
The DSSD used for the first to third (fourth) layer has the area of 76:8 25:6 mm2 (73:8 33:3 mm2 ) and the strip pitch of 75 ð73Þ mm for zside and 50 ð65Þ mm for f-side. The number of readout strips of both z- and f-sides is 512, and
500
they are mated to four VA1TA chips located at the end of a ladder. From the first to fourth layer, each ladder consists of 2, 3, 5 and 6 DSSDs glued onto a mechanical support rib. In total 246 DSSDs are utilized, and the total number of channel amounts to 110,592. A wide area flexible printed circuit (FPC) with 50 mm fine pitch is employed, instead of a double metal layer (DML) for orthogonal coordinate readout, as used in the SVD1. Advantages of using a FPC include: lowering the intrinsic noise by significantly reducing the capacitance compared with the DML structure, as well as decoupling the mechanics of the front-end electronics and DSSDs, which permits reducing the radius of the innermost sensor layer in the limited physical volume available. After the construction and commissioning of the SVD2, it was installed in Belle during the summer of 2003, and a system integration test was performed with cosmic ray muon events taken without a magnetic field. Typical detector intrinsic noise is measured to be 500, 700, 1000 and 1000 electrons for the first to fourth layer ladders, respectively. Since cluster energies are measured to be about 18,000 electrons with minimum ionizing particles, the signal-to-noise ratios are obtained to be 18–36, which are comparable with those of the SVD1, although the detector area is almost doubled for the outermost layer. Fig. 1 shows the typical residual distributions of f- and z-sides of one DSSD in the second layer after the
400 350
400 300 300
250 200
200 100
150 100 50
0 -100-80 -60 -40 -20 0 20 40 60 80 100 um (a) Phi residual
0 -150 -100
( b)
-50
0
50
100
150 um
Z residual
Fig. 1. Residual distributions of (a) f- and (b) z-sides of one DSSD after the alignment. The curves show the fitted double Gaussians.
ARTICLE IN PRESS H. Ishino / Nuclear Instruments and Methods in Physics Research A 549 (2005) 16–19
alignment with the cosmic ray muon events. From the distributions, the intrinsic spatial resolutions of f- and z-sides are measured to be 12:0 0:4 and 22:3 0:8 mm, respectively. The strip hit efficiency is obtained to be 95–99%, depending on ladders as expected. The inefficiencies are due to FPC defects, wire bonding problems and malfunction of the readout chips. The discriminator threshold levels of the TA part for all p(z)-side strips have been tuned using calibration input pulses. The obtained threshold so far is about 12,000 electrons on average and overall TA hit efficiency is measured to be about 69%. Significant effort is being made to increase the efficiency to about 90%.
19
operation at higher beam collision rates and larger radiation doses.
Acknowledgements The author would like to acknowledge and thank the Belle Collaboration and the Belle SVD group for their support. The author is also grateful to the physics department of KEK for their continuous encouragement and to the mechanical engineering center of KEK for their contribution to the construction and the installation of the SVD2 system.
4. Conclusion References The excellent performance of the SVD1 produced outstanding physics results during its 4 years of operation. A second generation upgrade, the SVD2, installed in the summer of 2003 is working well as expected. Improvements implemented in the SVD2 will permit more stable
[1] S. Mori, et al. (Eds.), Nucl. Instr. and Meth. A 479 (2002) 117. [2] T. Kawasaki, Nucl. Instr. and Meth. A 494 (2002) 94. [3] Y. Ushiroda, Nucl. Instr. and Meth. A 511 (2003) 6. [4] http://www.ideas.no [5] T. Ziegler, Nucl. Instr. and Meth. A 511 (2003) 153.