Miao Sun

Ms. Miao Sun


Research Focus & Metrics

Academic Career Length

9Years (2018-Present)

(Calculated once Scilit publication data is synced: current year − year of first publication + 1)

Representative Publications

View All Publications
19 December 2022 in IEEE Journal of Solid-State Circuits
Verified
Journal Article
Closed access
SZShenglong Zhuo
Shenglong Zhuo
TXTao Xia
Tao Xia
LZLei Zhao
Lei Zhao
MSMiao Sun
Miao Sun
YWYifan Wu
Yifan Wu
PCPatrick Yin Chiang
Patrick Yin Chiang
An entire solid-state direct time-of-flight (dToF) light detection and ranging (LiDAR) system that incorporates innovations for both the transmitter (TX) and the receiver (RX) is presented in this work. For the illumination TX, we demonstrate solid-state channel addressability, which significantly reduces the transmit power and improves the ranging distance by dividing the field of view (FoV) into separately illuminated sub-regions. In the RX, we introduce single-photon avalanche diodes (SPADs) pixel binning, which enables reconfigurability of the sensor’s spatial resolution. Finally, we introduce a machine learning (ML) technique that enables this pixel-binned depth sensor to upscale its spatial resolution after training/inference fusion with the intensity image. The laser diode driver (LDD) chip is implemented in the 180-nm bipolar-CMOS-DMOS (BCD) process and is capable of pumping more than 8-A peak current into a multi-junction vertical-cavity surface-emitting laser (VCSEL) array, producing up to 20.3-W optical pulses under 12.5-V supply voltage. The sensor chip is also implemented in the 180-nm BCD process with a $128 \times 128$ SPAD array and reconfigurable pixel binning. Hardware and software co-optimization under low signal-to-noise ratio (SNR) conditions with ML-based spatial resolution upscaling is demonstrated.

21 October 2022 in IEEE Transactions on Circuits and Systems II: Express Briefs
Verified
Journal Article
Closed access
SZShenglong Zhuo
Shenglong Zhuo
YWYuwei Wang
Yuwei Wang
TXTao Xia
Tao Xia
YWYifan Wu
Yifan Wu
WZWei Zheng
Wei Zheng
PCPatrick Yin Chang
Patrick Yin Chang
The applications requiring three-dimensional information have been growing rapidly, including facial recognition, gesture recognition, auto-focus for 2D camera, self-driving autonomous vehicles, etc. Indirect Time-of-Flight (iToF) technique has been adopted widely for 3D depth imaging due to its high spatial resolution and excellent depth accuracy performance. In this paper, a 200 MHz pulsed optical Illuminator for iToF applications is reported. A 940 nm GaAs-based triple-junction VCSEL array is adopted and an integrated pulsed VCSEL driver ASIC in 180 nm BCD process is realized. Optical pulses with peak power up to 14 W can be transmitted by the laser illuminator with a maximum driving current of 6.5 A by the driver ASIC. The modulation frequency can be up to 200 MHz. Automatic optical power control (AOPC) is realized for stable optical power output and laser eye safety protection. An on-chip DLL-based time interpolator is proposed to eliminate the mechanical moving components used during system calibration. The measured absolute accuracy of the time interpolator is +/-50 ps with a resolution of 800 ps and tuning range of 49.6 ns. This work effectively extends the measurement distance of the iToF system. The calibration efficiency is also improved with the proposed method.

26 September 2025 in ACM Transactions on Embedded Computing Systems
Verified
Journal Article
Open access
JKJiyong Kim
Jiyong Kim
JLJaeho Lee
Jaeho Lee
JLJiahao Lin
Jiahao Lin
AKAlish Kanani
Alish Kanani
MSMiao Sun
Miao Sun
JPJaehyun Park
Jaehyun Park
State Space Model (SSM)-based machine learning architectures have recently gained significant attention for processing sequential data. Mamba, a recent sequence-to-sequence SSM, offers competitive accuracy with superior computational efficiency compared to state-of-the-art transformer models. While this advantage makes Mamba particularly promising for resource-constrained edge devices, no hardware acceleration frameworks are currently optimized for deploying it in such environments. This article presents eMamba, a comprehensive end-to-end hardware acceleration framework explicitly designed for deploying Mamba models on edge platforms. eMamba maximizes computational efficiency by replacing complex normalization layers with lightweight hardware-aware alternatives and approximating expensive operations, such as SiLU activation and exponentiation, considering the target applications. Then, it performs an approximation-aware neural architecture search (NAS) to tune the learnable parameters used during approximation. Evaluations with Fashion-MNIST, CIFAR-10, and MARS, an open-source human pose estimation dataset, show eMamba achieves comparable accuracy to state-of-the-art techniques using 1.63–19.9× fewer parameters. In addition, it generalizes well to large-scale natural language tasks, demonstrating stable perplexity across varying sequence lengths on the WikiText2 dataset. We also quantize and implement the entire eMamba pipeline on an AMD ZCU102 FPGA and ASIC using GlobalFoundries (GF) 22 nm technology. Experimental results show 4.95–5.62× lower latency and 2.22–9.95× higher throughput, with 4.77× smaller area, 9.84× lower power, and 48.6× lower energy consumption than baseline solutions while maintaining competitive accuracy.

27 June 2022 in IEEE Transactions on Circuits and Systems II: Express Briefs
Verified
Journal Article
Closed access
TXTao Xia
Tao Xia
XCXuefeng Chen
Xuefeng Chen
YWYifan Wu
Yifan Wu
YWYuwei Wang
Yuwei Wang
YLYuan Li
Yuan Li
PCPatrick Yin Chiang
Patrick Yin Chiang
An integrated laser diode driver for multi-junction vertical-cavity surface-emitting laser (VCSEL) arrays is presented in this brief. The driver chip, fabricated in 180nm bipolar-CMOS-DMOS (BCD) process, working under 12V supply, can pump more than 8A peak current into a triple-junction VCSEL array, producing up to 20.3W optical pulses. Measured optical pulses have tunable widths from 276ps to 2352ps with 120ps/160ps rise/fall time. When emitting 10MHz-frequency 960ps-width optical pulses with peak power ranging from 3.6W to 17.5W, the driver-VCSEL system has over 20% electrical-to-optical power conversion efficiency. The driver can record its output current waveform with a built-in pulse monitor. The peak data from the recorded waveform is used for automatic peak current control.

21 March 2023 in IEEE Transactions on Circuits and Systems II: Express Briefs
Verified
Journal Article
Closed access
MSMiao Sun
Miao Sun
YCYingjie Cao
Yingjie Cao
JQJian Qian
Jian Qian
JLJie Li
Jie Li
SZSifan Zhou
Sifan Zhou
SZShenglong Zhuo
Shenglong Zhuo
Light Detection and Ranging (LiDAR) is becoming a critical requirement for future computer vision applications, such as AR/VR (iPhone-LiDAR) and ADAS (Automotive-LiDAR). A depth point-cloud input has different characteristics than a conventional RGB image input, such that the CNN depthinference implementation is unique when compared with a standard super-resolution CNN(SR-CNN). In this brief, we present a heterogeneous AI-accelerator SoC, which is specific to depth image completion computation. Three key innovations are introduced to improve SoC’s performance. First, to accommodate the unique input data structure of a depth input, a fully-filled dataflow management engine is proposed to pre-process the RGB+Depth input, significantly improving processing element utilization (PEU). Second, to improve the efficiency of the instruction configurations of the CNN accelerator, a hardware-tiling co-processor is proposed that performs the tiling strategy of the CNN accelerator, assigning each sub-job to the PE array directly, therefore reducing the time for task assignments. Third, due to the large number of vector operations required for the postprocess in the neural network, a RISC-V core is incorporated to execute vector computations better. The SoC is implemented in 40nm CMOS process, achieving 2TOPs/W energy efficiency with 34fps throughput under VGA-resolution output for realtime LiDAR systems.

3 August 2023 in Sensors
Verified
Journal Article
Open access
PDF file
JWJier Wang
Jier Wang
JLJie Li
Jie Li
YWYifan Wu
Yifan Wu
HYHengwei Yu
Hengwei Yu
LCLebei Cui
Lebei Cui
PCPatrick Yin Chiang
Patrick Yin Chiang
Light detection and ranging (LiDAR) technology, a cutting-edge advancement in mobile applications, presents a myriad of compelling use cases, including enhancing low-light photography, capturing and sharing 3D images of fascinating objects, and elevating the overall augmented reality (AR) experience. However, its widespread adoption has been hindered by the prohibitive costs and substantial power consumption associated with its implementation in mobile devices. To surmount these obstacles, this paper proposes a low-power, low-cost, single-photon avalanche detector (SPAD)-based system-on-chip (SoC) which packages the microlens arrays (MLAs) and a lightweight RGB-guided sparse depth imaging completion neural network for 3D LiDAR imaging. The proposed SoC integrates an 8 × 8 SPAD macropixel array with time-to-digital converters (TDCs) and a charge pump, fabricated using a 180 nm bipolar-CMOS-DMOS (BCD) process. Initially, the primary function of this SoC was limited to serving as a ranging sensor. A random MLA-based homogenizing diffuser efficiently transforms Gaussian beams into flat-topped beams with a 45° field of view (FOV), enabling flash projection at the transmitter. To further enhance resolution and broaden application possibilities, a lightweight neural network employing RGB-guided sparse depth complementation is proposed, enabling a substantial expansion of image resolution from 8 × 8 to quarter video graphics array level (QVGA; 256 × 256). Experimental results demonstrate the effectiveness and stability of the hardware encompassing the SoC and optical system, as well as the lightweight features and accuracy of the algorithmic neural network. The state-of-the-art SoC-neural network solution offers a promising and inspiring foundation for developing consumer-level 3D imaging applications on mobile devices.