Preamble

A Silicon Microstrip Detector for Power-Limited and Large Sensitive Area Applications
Dexing Miao,¹,² Zijun Xu,¹ Zhiyu Xiang,¹ Pingcheng Liu,³ Giovanni Ambrosi,⁴ Mattia Barbanera,⁴ Mengke Cai,¹ Xudong Cai,⁵ Hsin-Yi Chou,⁶ Matteo Duranti,⁴ Valerio Formato,⁷ Maria Ionica,⁴ Yaozu Jiang,⁴ Liangchenglong Jin,¹ Vladimir Koutsenko,⁵ Qinze Li,¹,² Cong Liu,³ Xingjian Lv,¹,² Alberto Oliva,⁸ Wenxi Peng,¹ Rui Qiao,¹ Gianluigi Silvestre,⁴ Zibing Wu,⁹ Xuhao Yuan,¹ Hongyu Zhang,¹⁰ Xiyuan Zhang,¹ and Jianchun Wang¹,†

¹Institute of High Energy Physics, Beijing 100049, China
²University of Chinese Academy of Sciences, Beijing 100049, China
³Shandong Institute of Advanced Technology, Jinan 250100, China
⁴INFN Sezione di Perugia, 06100 Perugia, Italy
⁵Massachusetts Institute of Technology (MIT), Cambridge, Massachusetts 02139, USA
⁶Institute of Physics, Academia Sinica, Nankang, Taipei, 11529, Taiwan
⁷INFN Sezione di Roma Tor Vergata, 00133 Roma, Italy
⁸INFN Sezione di Bologna, 40126 Bologna, Italy
⁹Shandong University (SDU), Jinan, Shandong 250100, China
¹⁰Key Laboratory of Nuclear Physics and Ion-beam Application (MOE) and Institute of Modern Physics, Fudan University, 220 Handan Road, Shanghai, 200433, China

A silicon microstrip detector (SSD) has been developed to achieve state-of-the-art spatial resolution while being well-suited for large-area, power-limited applications. The design incorporates three floating strips with their bias resistors inserted between two aluminum readout strips. Beam test measurements with a single sensor confirmed that this configuration achieves a total detection efficiency of 99.8% and a spatial resolution of 7.6 µm for minimum ionizing particles (MIPs). A double-η algorithm was developed to optimize hit position reconstruction for this SSD. The design can be readily adapted for large-area silicon detectors.

Keywords: Silicon Microstrip Detector, Bias resistor, Test Beam, Detection Efficiency, Spatial Resolution

Introduction

Silicon microstrip detectors (SSD) \cite{1,2} are extensively used for tracking purposes across numerous experiments in particle physics \cite{3–5}, nuclear physics \cite{6}, as well as in many space-borne experiments \cite{7–15} including the Alpha Magnetic Spectrometer (AMS-02) \cite{16,17}. Due to stringent constraints on power consumption in space-borne experiments, the number of available readout electronic channels for SSDs is limited, which necessitates careful optimization of detector designs.

The AMS-02 experiment, operating aboard the International Space Station (ISS), aims to precisely measure the composition and flux of cosmic rays throughout the ISS's lifetime until 2030. To enhance the detection acceptance by a factor of three, a new large-area silicon tracker layer (Layer-0) will be installed on top of AMS-02 in 2026. Layer-0 is composed of two back-to-back detector layers, each with an active area of approximately 3.5 m². To achieve such a large active area with limited power consumption, we designed very long silicon detector modules, as shown in Fig. 1 [FIGURE:1]. The majority of power consumption in a detector system comes from the readout electronics. To address this, we designed the modules such that 8, 10, or 12 SSDs (Fig. 1 shows a 12-SSD configuration) are connected in series via wire bonding to share a front-end readout board. Compared to applications where each sensor is connected to a dedicated readout board, this design reduces the number of electronic channels by a factor of O(10). With this approach, each module consumes 1.5 W of power and contains 1024 readout channels, resulting in an average power consumption of 1.46 mW per channel. The full AMS-L0 system consists of 72 such modules, covering a total active area of approximately 7 m², which corresponds to a power density of about 1.54 mW/cm². However, the design of serially connected detector modules imposes unique technical requirements on the SSDs. To meet these stringent demands, we developed a custom-designed SSD specifically tailored for this purpose, which has been successfully adopted in the production of the AMS Layer-0 detectors.

In this paper, we present the detailed design of this SSD, its characterization through test beam measurements, and performance evaluations focused on detection efficiency and spatial resolution.

II. Designs of the Silicon Microstrip Detector

The SSD is a p⁺-in-n sensor manufactured by Hamamatsu Photonics K.K, with a total area of 113 × 80 mm² and thickness of 320 µm. Its schematic is presented in Fig. 2 [FIGURE:2].

The design incorporates two specific features: the use of three floating strips to enhance spatial resolution without increasing the number of electronic channels, and a configuration of large bias resistors and aluminum readout strips to ensure a large sensitive area. These designs result in different regions of the SSD, as indicated in Fig. 2:

• Region-A: with three floating p⁺ strips, without bias resistors
• Region-B: with three floating p⁺ strips and their bias resistors located between two aluminum readout strips
• Region-C: without floating p⁺ strips, with bias resistors for p⁺ readout strips

A p⁺ strip AC-coupled to an aluminum readout strip is referred to as a readout strip, while the strips located between two readout strips are called floating strips. In the SSD, which consists of 4095 p⁺ strips with a pitch of 27.25 µm, one out of every four p⁺ strips is a readout strip. There are three floating strips between adjacent readout strips. These floating strips facilitate charge sharing to readout strips, thereby enhancing spatial resolution \cite{18}. The impact of floating strips is quantitatively studied by comparing the performance between Region-C and Region-A.

In a ladder, aluminum readout strips of the N (N = 8, 10, 12) SSDs are connected in series along the strip direction, forming a parallel connection between the SSDs. Consequently, the coupling capacitors and bias resistors are connected in parallel. This arrangement leads to the coupling capacitance (bias resistance) increasing (decreasing) by a factor of N. Since the coupling capacitance cannot easily be adjusted during SSD fabrication by HPK, the SSDs must incorporate a bias resistance that is N² ∼ 100 times larger than the typical value in order to achieve proper impedance matching.

For a conventional SSD, the bias resistance is O(1) MΩ, implemented using a polysilicon resistor that occupies approximately 100 µm in length at the end of the aluminum readout strip \cite{19–22}. In the SSD design, the O(100) MΩ bias resistor would occupy a length of O(10) mm. Placing it directly behind the aluminum readout strips would lead to a bonding wire of about 14 mm in length, which poses a potential risk of short-circuit. Therefore, a specialized bias resistor configuration was designed, as illustrated in Fig. 2. The bias resistors for the three floating strips are inserted between two aluminum readout strips, allowing the bonding pads to be placed at the left end of the aluminum strips. On the right end, the bias resistors of the readout strips occupy that space solely, requiring only 2.9 mm in length. Thus, the length of bonding wires between two SSDs is reduced to about 5 mm.

The bias resistors between two aluminum readout strips could potentially influence their AC coupling or distort the internal electric field. Therefore, we carried out studies to compare the performance between regions with (Region-B) and without (Region-A) the bias resistors.

III. Beam Test Setup

The beam test described in this work was conducted in May 2024 at the Super Proton Synchrotron (SPS) at CERN. We performed the beam test behind the dumper of beam line H6. The beam mostly consisted of muons with momentum in the tens of GeV, generated by the proton beam striking the dumper, thus behaving as minimum ionizing particles (MIPs).

We designed the single SSD frame board as shown in Fig. 3 FIGURE:3. The gray rectangle in the middle is the SSD, which is wire-bonded to 16 front-end readout chips (IDE1140) \cite{23} on the right, resulting in a total of 1024 channels. Each channel has a gain of 2.6 µA/fC and a peaking time for positive signals of 8.5 µs. The output signals from IDE1140 are amplified by the amplifier (AD8031) in two stages with a gain of 6.97 and then routed to a 14-bit analog-to-digital converter (ADC), LTC2313ITS8-14, with a maximum input of 4096 mV. The data are then encoded by a field-programmable gate array (FPGA) and transmitted to the back-end.

The beam monitor was constructed by several layers of single SSD frame boards, as depicted in the right panel of Fig. 3(b). The frame board is designed with 90-degree rotational symmetry, allowing flexible orientation combinations for the layers. For layers with strips along the vertical (horizontal) direction, the hit position in the X (Y) coordinate can be measured, denoted as X-layer (Y-layer). In this beam test, we arranged 5 X-layers and 5 Y-layers, as illustrated in Fig. 4 [FIGURE:4], with the indicated detector under test (DUT) and the telescope. Scintillator detectors (140 × 80 mm²) \cite{24} were deployed to generate trigger signals, forming a trigger region that fully covered the DUT along the strip direction. This arrangement enabled performance evaluation of the DUT across Regions A, B, and C.

IV. Data Analysis

A. Raw Data Processing

For the i-th channel, the raw value obtained in the j-th individual readout, denoted as ADCⁱⱼ, can be considered as the sum of the following four components:

ADCⁱⱼ = Pedi + CNⁱⱼ + Noiseⁱⱼ + Sⁱⱼ

where Pedi is the pedestal of channel i, defined as the average value in the absence of particle incidence. Bias voltage fluctuation would cause an overall fluctuation of all channels, named common mode noise (CN) \cite{25}. The single event Noiseⁱⱼ is a random noise resulting from combining SSD and electronic noise, whose absolute value cannot be obtained, though the noise level of channel i (Noiseᵢ) can be calculated as the standard deviation of ADC values after CN is subtracted. Finally, Sⁱⱼ is the signal introduced by the incident particle.

The pedestal and noise level are determined through a one-minute calibration run conducted every hour during the beam test. As shown in Fig. 5 FIGURE:5, the noise level of one channel is about 3 Least Significant Bits (LSB), which corresponds to an Equivalent Noise Charge (ENC) of ∼240 e⁻. Channels with high noise (Noiseᵢ > 10 LSB) were defined as hot channels, while those with low noise (Noiseᵢ < 2 LSB) were defined as dead channels.

For each event, CN is calculated by:

CNʲ = (1/Nₐ) Σ (ADCⁱⱼ - Pedi)

where Nₐ is the number of valid channels on the a-th IDE1140. Hot channels and fired channels (ADCⁱⱼ - Pedi > 10 LSB) are not used in the CN calculation. The common noise level is sensitive to environmental conditions. As shown in Fig. 5(b), it is approximately 0.5 LSB per chip during the data-taking run.

Subsequently, for the data-taking run, the signal of channel i of event j is calculated as:

Sⁱⱼ = ADCⁱⱼ - Pedi - CNʲ

A cluster is identified by a channel where Sⁱⱼ > thseed · Noiseᵢ. The cluster expands on both sides until the value of channel x satisfies Sˣⱼ < thside · Noiseₓ. In this study, thseed = 3.5 and thside = 2.0 are chosen to reduce noise influence while maintaining detection efficiency.

B. Double-η Position Finding Algorithm

The particle impact position is typically reconstructed from the top few channels with the highest values in a cluster. Numerous studies have shown that the η algorithm \cite{18,26–28} is a proper position-finding algorithm for this purpose. This algorithm reconstructs the particle impact position by using the two adjacent channels with the highest values in a cluster. Denote Sᴸ (Sᴿ) as the value of the left (right) channel, and xᴸ (xᴿ) as the position of the readout strip corresponding to the left (right) channel. Define the variable η:

η = Sᴸ / (Sᴸ + Sᴿ)

The particle impact position xη follows a function of η:

xη = xᴸ + P · f(η)

where P is the readout strip pitch and f(η) is the cumulative distribution function (CDF) of η:

f(η) = ∫₀^η φ(η') dη'

where φ(η) is the probability density function (PDF) of η, which can be estimated from test beam data.

An issue arises when this algorithm is applied to the SSD, which exhibits significant charge sharing. For incidents near one readout strip, the two adjacent readout strips often share similar signals. Under the influence of random noise, the second-highest strip may be incorrectly identified. In these instances, choosing the two highest strips for position reconstruction can lead to a flip across the central strip, causing a larger position error. To address this issue, we propose an iterative position-finding method, namely the "double-η algorithm":

• Pick the channel with the highest value in a cluster as the first channel. Among its two neighboring channels, the one with the larger value is designated as the second channel, and the other as the third channel.
• Use the first and second (third) channels to calculate two η values according to Eq. 5, denoted as η₁₂ (η₁₃). Estimate two impact positions using Eq. 6, denoted as x₁₂ (x₁₃).
• Use x₁₂ from all layers to reconstruct the initial trajectory.
• For each layer, compare the residual of x₁₂ and x₁₃ to the initial trajectory, and select the smaller one as x⋆. Use x⋆ from all layers for final track reconstruction.

The spatial resolution comparison between this algorithm and the traditional η algorithm is discussed in Sec. V C 1.

C. Track Reconstruction

Due to multiple particle incidences or electronic noise fluctuations, a detector layer might have two or more clusters in a single event. In this beam test, approximately 10% of events contained two or more clusters. For these events, we evaluated all possible trajectory combinations and selected the one with the lowest χ² value.

Track fitting is performed using the General Broken Lines (GBL) algorithm \cite{29}, widely employed in experiments such as AMS-02 \cite{30} and CMS \cite{31}. GBL is capable of accurately accounting for multiple scattering effects. Additionally, GBL provides a comprehensive covariance matrix for all track parameters upon refitting, making it especially advantageous for calibration and alignment tasks when integrated with Millepede II \cite{32}, which is a global parameter optimization tool designed to efficiently handle up to approximately one hundred thousand global parameters.

Detector alignment is performed using 1% of the collected data. Subsequently, track fitting is carried out for all events that used telescope layers. Following this procedure, more than 10 million tracks with precise hit information were reconstructed for analyzing the performance of the DUT.

V. Performance of the SSD

We studied the impact of the SSD designs by comparing performance across different regions, with a primary focus on charge collection and sharing performance, detection efficiency, and spatial resolution.

A. Charge Sharing and Charge Collection

Charge sharing refers to the phenomenon where a single particle incident induces signals on multiple readout channels. The sharing is primarily due to capacitive coupling between neighboring strips and carrier diffusion. The charge sharing effect can be quantified by the number of channels within a cluster, referred to as the cluster size. The right panel of Fig. 6 [FIGURE:6] shows the variation of cluster size along the strip direction. The average cluster size is 2.01 for Region-A and 1.32 for Region-C, indicating the enhancement of charge sharing by the floating strips.

Additional information can be gained by investigating the cluster size as a function of the relative position between two readout strips. For each track, the hit position inner strip is defined as the position at the DUT mapped onto the range from [-54.5, 54.5], where -54.5 is the center of the nth strip and 54.5 is the center of the (n+1)th strip. The fraction of different cluster sizes as a function of hit position inner strip is shown in Fig. 7 [FIGURE:7] for Region-A and Region-C. When floating strips are present, if the hit positions are slightly offset from the readout strips, there is a rapid increase in the fraction of clusters with size ≥ 2. Conversely, in areas without floating strips, clusters of size ≥ 2 become dominant only when the hit occurs near the midpoint between two adjacent readout strips. This further underlines the role of floating strips in enhancing charge sharing.

The impact of the bias resistors for charge sharing located between two aluminum readout strips can be evaluated by comparing Region-B and Region-A. The magnified plot in Fig. 6 illustrates that the cluster size in Region-B is about 1% larger than that in Region-A. This indicates that the presence of bias resistors leads to a few percent increase in charge sharing between aluminum readout strips, possibly due to the additional equivalent capacitance introduced by the polysilicon.

While enhancing charge sharing, the presence of floating strips leads to a reduction in charge collection efficiency (CCE) due to capacitive coupling to the backplane, signal loss through bias resistors, etc. \cite{19}. As the absolute CCE cannot be determined directly, the most probable value (MPV) of a cluster signal is used to represent the relative CCE across different regions. By binning the particle incident position along strips in steps of 500 µm, the MPV was determined by fitting each bin with a Landau function convoluted with a double Gaussian function, as shown in Fig. 9 [FIGURE:9]. Figure 10 [FIGURE:10] illustrates how the MPVs change along the strip direction. From left to right, the MPV gradually rises from 45 LSB to 46 LSB, likely due to closer proximity to the readout chip. Upon reaching Region-C, the MPV suddenly increases to 65 LSB, reflecting a notable improvement in CCE once the three floating strips are no longer present. While the three floating strips reduce the MPV, the SSD maintains high detection efficiency due to the low noise level (3 LSB), as detailed in Sec. V B.

The variable η also serves as an effective indicator of the degree of charge sharing. An η value close to 0 or 1 indicates that most of the signal is collected by a single readout channel, while a value near 0.5 implies substantial sharing between two adjacent readout channels. Figure 8 [FIGURE:8] (a)(c) shows a comparison of the η distributions between regions with and without three floating strips. It can be seen that in the region without floating strips, two prominent peaks appear near η = 0 and η = 1. In contrast, the presence of three floating strips introduces three additional peaks in the η distribution. As shown in Fig. 8(b), the f(η) values corresponding to these three peaks align precisely with the relative positions of the three floating strips. These peaks indicate that a specific fraction of charges is shared with the neighboring readout strip with the help of floating strips.

B. Detection Efficiency

In this study, detection efficiency refers to the ability of the DUT to detect MIPs. It is one of the essential performance metrics for silicon detectors, evaluated through the following procedure:

• The telescope predicts a hit position xpred of a track on the DUT plane
• The DUT then searches for a cluster within a window of xpred ± 109 µm
• If a cluster is found, the track is considered to have been successfully detected by the DUT

Based on this, the efficiency along the strip is analyzed to determine the total sensitive area, and the relationship between detection efficiency and the hit position between two readout strips (hit position inner strip) is then investigated. Owing to the large telescope coverage, tracks spanning the full length of the SSD along the strip direction can be reconstructed. Figure 11 [FIGURE:11] shows the detection efficiency along the strip direction. The detection efficiency remains close to 1 across the entire SSD (total efficiency = 99.8%), with the detector edges clearly delineated. This indicates that: (1) the bias resistors located between two aluminum readout strips did not affect the efficiency, and (2) the introduction of three floating strips does not lead to a reduction in detection efficiency.

The detection efficiency versus inner strip hit position for different regions is shown in Fig. 12 [FIGURE:12]. It can be seen that all regions exhibit the highest detection efficiency near the two readout strips and the lowest in the middle, which is consistent with the charge collection behavior of the SSD. For Region-A and Region-B, the efficiency decreases gradually from the readout strip to the first floating strip but drops more rapidly after crossing the first floating strip. For Region-C, with no floating strips, the efficiency remains relatively stable over a wider range, with only a small drop in the most central range. Generally, the total detection efficiency of all three regions is 99.8%. In addition, we have studied the fake hit rate of the detector and found that, under the current cluster selection criteria, the per-channel fake hit rate is 0.022%. Therefore, the potential overestimation of detection efficiency due to fake hits can be considered negligible.

C. Spatial Resolution

Spatial resolution is another key performance metric of silicon tracking detectors, determined by both the detector hardware and the reconstruction algorithm. In the following, we first present the effect of the double-η algorithm and then evaluate the improvements introduced by the three floating strips.

The measured unbiased resolution (σmea) is defined as the standard deviation of the difference between the position reconstructed by the DUT and the telescope prediction. It comprises two components:

σ²mea = σ²dut + σ²tele

where σdut is the intrinsic spatial resolution of the DUT and σtele is the pointing resolution of the telescope. Assuming each layer has the same intrinsic spatial resolution, the pointing resolution of the telescope can be expressed as a factor k times the intrinsic resolution of the DUT:

σ²tele = kσ²dut

where the coefficient k depends on the relative positioning of the telescope planes \cite{33}:

k = Σ(zi - zdut)² / [N Σ(zi - zdut)² - (Σ(zi - zdut))²]

where N is the number of telescope layers, and zi, zdut denote the position along the beam direction of the i-th telescope layer and DUT, respectively. In our setup, the DUT is located at 60 mm, and the telescope layers are positioned at 6, 66, 96, and 120 mm. Based on this configuration, the coefficient is calculated to be k = 0.27. Thus:

σmeas = √(σ²dut + σ²tele) = √(1 + k) σdut

Therefore, the DUT resolution can be expressed as:

σdut = (1/√(1 + k)) σmeas = 0.89 σmeas

1. Effect of the Double-η Algorithm

The measured residual distributions for Region-A obtained using the η algorithm and the double-η algorithm are shown in Fig. 13 FIGURE:13(c), respectively. It can be seen that the first distribution exhibits a broader shape with pronounced tails on both sides. Meanwhile, in the two-dimensional distribution of measured residual versus inner strip hit position shown in Fig. 13(b), a clear Z-shape is observed, with the two horns contributing to the tails in Fig. 13(a). These events correspond to the "swapping phenomenon" discussed in Sec. IV B.

After applying the correction using the double-η algorithm, the Z-shaped structure disappears, as shown in Fig. 13(d). As a result, σmea was optimized from 10.6 µm to 8.1 µm, and the intrinsic σdut improved from 9.4 µm to 7.2 µm. All spatial resolution results discussed below are obtained using the double-η algorithm.

2. Improvement from the Three Floating Strips

The measured residual distributions with and without the three floating strips are shown in Fig. 14 FIGURE:14(c). The three floating strips significantly improve σmea from 17.2 µm to 8.1 µm (σdut from 15.2 µm to 7.2 µm). This enhancement is primarily due to the larger charge sharing effect introduced by the floating strips. This can also be demonstrated by comparing the η distributions in the two regions, as shown in Fig. 8(a)(c). The three floating strips lead to a flatter η distribution, indicating increased sensitivity of the η value to the inner strip hit position. As a result, the function f(η) becomes closer to the linear function (f(η) = η), as shown in Fig. 8(b)(d), which leads to better spatial resolution \cite{18}. In addition, Fig. 14(b) shows the same plot of the residual in Region-B. The σmea (σdut) of this region is 8.1 µm (7.2 µm), indicating that placing bias resistors between two aluminum readout strips does not affect the spatial resolution.

According to the SSD design, along the strip direction, the total length of Region-A and Region-B is 75.8 mm, while Region-C has a length of 2.9 mm. As the spatial resolution is uniform in the direction perpendicular to the strips, the overall effective spatial resolution of the SSD can be calculated using an area-weighted average as follows:

σeff = √[(7.2² × 75.8 + 15.2² × 2.9) / (75.8 + 2.9)] = 7.6 µm

VI. Summary

We have designed an SSD for use in serially connected configurations, making it well-suited for large-area, power-limited applications. A detailed performance characterization of a single sensor was carried out using beam tests. Introducing three floating strips and inserting their bias resistors between two aluminum readout strips are the distinctive features of this SSD. Our study shows that the three floating strips significantly enhance the charge sharing effect, which improves spatial resolution from 15.2 µm to 7.2 µm without compromising detection efficiency. Furthermore, the presence of bias resistors between two aluminum strips does not degrade the spatial resolution or detection efficiency in the corresponding region. As a result, the SSD achieves an overall efficiency of 99.8% and a spatial resolution of 7.6 µm for MIPs. During this study, a double-η algorithm for hit position reconstruction was developed to suit this SSD. The SSD from this design has been successfully applied in the AMS Layer-0 tracker upgrade, and the design principles can be utilized for future silicon detectors in space-borne cosmic ray experiments.

Acknowledgments

This study was supported by the National Key Programme for S&T Research and Development (Grant NO.: 2022YFA1604800), and the National Natural Science Foundation of China (Grant NO.: 12342503). We express our gratitude to our colleagues in the CERN accelerator departments for the excellent performance of the SPS.

References

[1] Sally Seidel. Silicon strip and pixel detectors for particle physics experiments. Phys. Rept., 828:1–34, 2019.
[2] Jia-Ju Wei, Jian-Hua Guo, and Yi-Ming Hu. Characterization of silicon microstrip sensors for space astronomy. Nucl. Sci. Tech., 31(10):97, 2020.
[3] A. Augusto Alves, Jr. et al. The LHCb Detector at the LHC. JINST, 3:S08005, 2008.
[4] S. Chatrchyan et al. The CMS Experiment at the CERN LHC. JINST, 3:S08004, 2008.
[5] G. Aad et al. The ATLAS Experiment at the CERN Large Hadron Collider. JINST, 3:S08003, 2008.
[6] Rui He et al. Advances in nuclear detection and readout techniques. Nucl. Sci. Tech., 34(12):205, 2023.
[7] S. Straulino et al. The PAMELA silicon tracker. Nucl. Instrum. Meth. A, 530:168–172, 2004.
[8] L. Bonechi et al. Status of the PAMELA silicon tracker. Nucl. Instrum. Meth. A, 570:281–285, 2007.
[9] P. W. Cattaneo et al. First results about on-ground calibration of the Silicon Tracker for the AGILE satellite. Nucl. Instrum. Meth. A, 630:251–257, 2011.
[10] A. Bulgarelli et al. The AGILE silicon tracker: Pre-launch and in-flight configuration. Nucl. Instrum. Meth. A, 614:213–226, 2010.
[11] C. Sgro et al. Construction, test and calibration of the GLAST silicon tracker. Nucl. Instrum. Meth. A, 583:9–13, 2007.
[12] Claudia Cecchi. GLAST: The Gamma ray Large Area Space Telescope. J. Phys. Conf. Ser., 120:062017, 2008.
[13] Philipp Azzarello et al. The DAMPE silicon–tungsten tracker. Nucl. Instrum. Meth. A, 831:378–384, 2016.
[14] Yi-Fan Dong, Fei Zhang, Rui Qiao, Wen-Xi Peng, Rui-Rui Fan, Ke Gong, Di Wu, and Huan-Yu Wang. DAMPE silicon tracker on-board data compression algorithm. Chin. Phys. C, 39(11):116202, 2015.
[15] Juliana T. Vievering, Lindsay Glesener, P. S. Athiray, Juan Camilo Buitrago-Casas, Sophie Musset, Daniel Ryan, Shin-nosuke Ishikawa, Jessie Duncan, Steven Christe, and Säm Krucker. FOXSI-2 Solar Microflares II: Hard X-ray Imaging Spectroscopy and Flare Energetics. Astrophys. J., 913(1):15, 2021.
[16] K. Lubelsmeyer et al. Upgrade of the Alpha Magnetic Spectrometer (AMS-02) for long term operation on the International Space Station (ISS). Nucl. Instrum. Meth. A, 654:639–648, 2011.
[17] Matteo Duranti. The AMS-02 Silicon Tracker after 500 days in space. PoS, Vertex2012:052, 2013.
[18] R. Turchetta. Spatial resolution of silicon microstrip detectors. Nucl. Instrum. Meth. A, 335:44–58, 1993.
[19] G. Casse, P. Dervan, D. Forshaw, A. Greenall, T. Huse, I. Tsurin, and M. Wormald. Degradation of charge sharing after neutron irradiation in strip silicon detectors with different geometries. Nucl. Instrum. Meth. A, 730:54–57, 2013.
[20] R. Aaij et al. Performance of the LHCb Vertex Locator. JINST, 9:P09007, 2014.
[21] A. J. Furgeri, W. de Boer, and F. Hartmann. Results of irradiation quality assurance of CMS silicon microstrip detectors. Nucl. Instrum. Meth. A, 573:264–267, 2007.
[22] V. Latoňová et al. Characterization of the polysilicon resistor in silicon strip sensors for ATLAS inner tracker as a function of temperature, pre- and post-irradiation. Nucl. Instrum. Meth. A, 1050:168119, 2023.
[23] https://ideas.no/products/ide1140/.
[24] Hongyu Zhang, Xiyang Wang, et al. Design and test for the CEPC muon subdetector based on extruded scintillator and SiPM. JINST, 19(06):P06020, 2024.
[25] Yu-Xin Cui et al. Common-mode noise analysis of silicon microstrip detectors. Nucl. Instrum. Meth. A, 1059:168962, 2024.
[26] Andrea Abba et al. Testbeam studies of pre-prototype silicon strip sensors for the LHCb UT upgrade project. Nucl. Instrum. Meth. A, 806:244–257, 2016.
[27] L. Hubbeling et al. Measurement of spatial resolution of a double sided AC coupled microstrip detector. Nucl. Instrum. Meth. A, 310:197–202, 1991.
[28] H. W. Kraner, R. Beuttenmuller, T. Ludlam, A. L. Hanson, K. W. Jones, V. Radeka, and E. H. M. Heijne. CHARGE COLLECTION IN SILICON STRIP DETECTORS. (TALK). IEEE Trans. Nucl. Sci., 30:405–414, 1983.
[29] Claus Kleinwort. General Broken Lines as advanced track fitting method. Nucl. Instrum. Meth. A, 673:107–110, 2012.
[30] Qi Yan and Vitaly Choutko. Alignment of the Alpha Magnetic Spectrometer (AMS) in space. Eur. Phys. J. C, 83:245, 2023.
[31] Younes Otarid. Pixel-Strip Modules for the CMS Tracker Phase-2 Upgrade: From DAQ test system development to module assembly and qualification. PhD thesis, Hamburg U., 2023.
[32] Volker Blobel, Claus Kleinwort, and Frank Meier. Fast alignment of a complex tracking detector using advanced track models. Comput. Phys. Commun., 182:1760–1763, 2011.
[33] L. K. Li, M. Y. Dong, Z. Gao, et al. Effect of multiple coulomb scattering on the beam tests of silicon pixel detectors. Nuclear Science and Techniques, 35(83), 2024.