Yu Liu
Current Organizations / Institutes
The organizations/institutes from the scholar's latest article
Previous Organizations / Institutes
The organizations / institutes from the scholar's other articles in the past ten years
Yu Liu has a total of 88 co-authors. Here are the top co-authors by number of shared publications:
- Xiulong Wu
Anhui University
8 shared publications - Zhiting Lin
Anhui University
7 shared publications - Chunyu Peng
Anhui University
6 shared publications - Xin Li
Anhui University
6 shared publications - Chenghu Dai
School of Integrated Circuits, Anhui University, Hefei, China
5 shared publications - Ao Li
The School of Information Science and Technology, University of Science and Technology of China, Hefei, China
3 shared publications
Latest Publications
- Open Access
- 16 January 2026
- Published by Association for Computing Machinery (ACM)
- p. 761-766
A “Coach-Assistant” Human-AI Collaborative Paradigm for AI-Empowered Integrated Circuit Design Education
- 16 January 2026
- Published by Association for Computing Machinery (ACM)
- p. 761-766
Open AccessA “Coach-Assistant” Human-AI Collaborative Paradigm for AI-Empowered Integrated Circuit Design Education
- 12 January 2026
- Published by Institute of Electrical and Electronics Engineers (IEEE) in IEEE Transactions on Very Large Scale Integration (VLSI) Systems
- Vol. 34 (3) , 882-890
A 28 nm 1.3 TFLOPS/mm 2 Floating-Point SRAM-Based CIM Macro With Asynchronous Normalization and Parallel Sorting Alignment for AI-Edge Chip
State-of-the-art AI edge devices require floating-point (FP) multiply-accumulate (MAC) operations with high-energy efficiency and inference accuracy. FP computing in-memory (FP-CIM) has a broader range of applications compared to integer CIM. However, FP-CIM can incur greater power, delay, and area overheads than integer CIM due to the inherent complexity of FP computational flow. In this article, we introduce a new method for asynchronous exponent normalization and parallel mantissa alignment. This approach allows us to add exponents and find the maximum sum simultaneously. We also replace the traditional subtraction and shifting for mantissa alignment with a cross-structure maximum-finding method, enabling FP-CIM to be achieved with lower delay, area, and power overheads. The macro is designed in TSMC 28 nm process, with a memory size of 6 Kb, a layout area of 0.067 mm2, and an area efficiency of 1.3 TFLOPS/mm2. Simulation results show that the macro computational frequency and energy efficiency can reach 150 MHz and 12.8 TFLOPS/W, respectively, at 900 mV, while performing FP- MAC operations.
- 12 January 2026
- Published by Institute of Electrical and Electronics Engineers (IEEE) in IEEE Transactions on Very Large Scale Integration (VLSI) Systems
- Vol. 34 (3) , 882-890
A 28 nm 1.3 TFLOPS/mm 2 Floating-Point SRAM-Based CIM Macro With Asynchronous Normalization and Parallel Sorting Alignment for AI-Edge Chip
State-of-the-art AI edge devices require floating-point (FP) multiply-accumulate (MAC) operations with high-energy efficiency and inference accuracy. FP computing in-memory (FP-CIM) has a broader range of applications compared to integer CIM. However, FP-CIM can incur greater power, delay, and area overheads than integer CIM due to the inherent complexity of FP computational flow. In this article, we introduce a new method for asynchronous exponent normalization and parallel mantissa alignment. This approach allows us to add exponents and find the maximum sum simultaneously. We also replace the traditional subtraction and shifting for mantissa alignment with a cross-structure maximum-finding method, enabling FP-CIM to be achieved with lower delay, area, and power overheads. The macro is designed in TSMC 28 nm process, with a memory size of 6 Kb, a layout area of 0.067 mm2, and an area efficiency of 1.3 TFLOPS/mm2. Simulation results show that the macro computational frequency and energy efficiency can reach 150 MHz and 12.8 TFLOPS/W, respectively, at 900 mV, while performing FP- MAC operations.
- 12 January 2026
- Published by Institute of Electrical and Electronics Engineers (IEEE) in IEEE Transactions on Device and Materials Reliability
- Vol. 26 (1) , 321-327
Capacitor-Coupled Offset-Canceled DRAM Sense Amplifier With Ultralow Offset Voltage and Hidden Cancelation Time
The sense amplifier (SA) is a critical component for DRAM to detect the small voltage difference between bit-lines and amplify it to rail-to-rail output. As DRAM technology scales down, threshold voltage mismatches between transistors become more pronounced, which increases SA offset voltage and prolongs sensing time (tSEN). This work proposes a capacitor-coupled sense amplifier (UH-COSA) with ultralow offset voltage and hidden cancellation time. The UH-COSA reduces the tSEN by hiding the offset-cancellation operation within the charge sharing operation. UH-COSA introduces a pre-sensing stage, and enables positive feedback to improve the sensing margin. UH-COSA use one capacitor to store the voltage difference between pre-charge voltage VEQ and one inverter trip-point, and one capacitor to store the voltage difference of two inverter trip-point voltages. Post-layout simulation results show the UH-COSA having high PVT stability. Even at a low supply voltage of 0.75 V and large bit-line load capacitance of 160 fF, the sensing yield can reach 99.85%. Compared with other capacitor-coupled SAs, UH-COSA shows advantages of shortest sensing time, and smallest standard deviation of offset voltage.
- 12 January 2026
- Published by Institute of Electrical and Electronics Engineers (IEEE) in IEEE Transactions on Device and Materials Reliability
- Vol. 26 (1) , 321-327
Capacitor-Coupled Offset-Canceled DRAM Sense Amplifier With Ultralow Offset Voltage and Hidden Cancelation Time
The sense amplifier (SA) is a critical component for DRAM to detect the small voltage difference between bit-lines and amplify it to rail-to-rail output. As DRAM technology scales down, threshold voltage mismatches between transistors become more pronounced, which increases SA offset voltage and prolongs sensing time (tSEN). This work proposes a capacitor-coupled sense amplifier (UH-COSA) with ultralow offset voltage and hidden cancellation time. The UH-COSA reduces the tSEN by hiding the offset-cancellation operation within the charge sharing operation. UH-COSA introduces a pre-sensing stage, and enables positive feedback to improve the sensing margin. UH-COSA use one capacitor to store the voltage difference between pre-charge voltage VEQ and one inverter trip-point, and one capacitor to store the voltage difference of two inverter trip-point voltages. Post-layout simulation results show the UH-COSA having high PVT stability. Even at a low supply voltage of 0.75 V and large bit-line load capacitance of 160 fF, the sensing yield can reach 99.85%. Compared with other capacitor-coupled SAs, UH-COSA shows advantages of shortest sensing time, and smallest standard deviation of offset voltage.
- Open Access
- 9 January 2026
- Published by Association for Computing Machinery (ACM) in ACM Journal on Emerging Technologies in Computing Systems
- Vol. 22 (2) , 1-26
Analysis and Design of Memory Testing Algorithm for Computing-in-Memory Using MBIST
Computing-in-memory (CIM), as a novel computing architecture for the future, effectively overcomes the bottlenecks in the von Neumann architecture. The CIM architecture embeds logic into the memory array to reduce the data transfer between the processor and memory. However, embedding logic into the memory array increases the test complexity. In this study, we offer a comprehensive examination of the challenges associated with CIM and introduce a novel March-like test algorithm, named March CC, tailored for CIM chips. Computational elements are added to the read/write operation sequences, combining the tests in memory mode and computing mode into one step, which significantly improves the test efficiency. In comparison to the traditional March C− test algorithm, the proposed March CC test algorithm, with a complexity of only 10 N , enhances the fault coverage from 66.7% to 79.8% for six common single-cell fault (SCF) models and nine common double-cell fault (DCF) models. Furthermore, the March CC algorithm demonstrates good compatibility and is applicable to various memory configurations, such as SRAM, RRAM, and MRAM CIM architectures.
- 9 January 2026
- Published by Association for Computing Machinery (ACM) in ACM Journal on Emerging Technologies in Computing Systems
- Vol. 22 (2) , 1-26
Open AccessAnalysis and Design of Memory Testing Algorithm for Computing-in-Memory Using MBIST
Computing-in-memory (CIM), as a novel computing architecture for the future, effectively overcomes the bottlenecks in the von Neumann architecture. The CIM architecture embeds logic into the memory array to reduce the data transfer between the processor and memory. However, embedding logic into the memory array increases the test complexity. In this study, we offer a comprehensive examination of the challenges associated with CIM and introduce a novel March-like test algorithm, named March CC, tailored for CIM chips. Computational elements are added to the read/write operation sequences, combining the tests in memory mode and computing mode into one step, which significantly improves the test efficiency. In comparison to the traditional March C− test algorithm, the proposed March CC test algorithm, with a complexity of only 10 N , enhances the fault coverage from 66.7% to 79.8% for six common single-cell fault (SCF) models and nine common double-cell fault (DCF) models. Furthermore, the March CC algorithm demonstrates good compatibility and is applicable to various memory configurations, such as SRAM, RRAM, and MRAM CIM architectures.
- 1 January 2026
- Published by Institute of Electrical and Electronics Engineers (IEEE) in IEEE Solid-State Circuits Letters
- Vol. 9, 270-273
A 28-nm Reconfigurable CAM for High-Utilization String Matching
- 1 January 2026
- Published by Institute of Electrical and Electronics Engineers (IEEE) in IEEE Solid-State Circuits Letters
- Vol. 9, 270-273
A 28-nm Reconfigurable CAM for High-Utilization String Matching
- 5 December 2025
- Published by Springer Nature in Signal, Image and Video Processing
- Vol. 19 (17) , 1412
Pre-activated simple attention network with SoftPool for optical footprint image retrieval
- 5 December 2025
- Published by Springer Nature in Signal, Image and Video Processing
- Vol. 19 (17) , 1412
Pre-activated simple attention network with SoftPool for optical footprint image retrieval
- 22 September 2025
- Published by Institute of Electrical and Electronics Engineers (IEEE) in IEEE Transactions on Device and Materials Reliability
- Vol. 25 (4) , 788-801
Error-Corrected RRAM-Based Compute-in-Memory Accelerator Using Voltage-Clamp Current Mirror
Computational reliability is a promising research direction in computing-in-memory (CIM), particularly in non-volatile memory (NVM) technologies such as resistive random-access memory (RRAM). Due to its outstanding performance and developmental potential, RRAM plays a critical role in advancing CIM research. The high computational reliability, low-cost ADC and low power consumption are the challenges for RRAM CIM. This study proposes an Error-corrected RRAM-based CIM architecture. To solve the problem of excessive calculation current, a voltage clamp current mirror (VCCM) is designed to improve the calculation accuracy and control calculation power consumption. A new quantization method is proposed to avoid the introduction of complex quantization circuits and reduce the energy consumption of quantization. To address the issue of significant errors in small RRAM HRS/LRS, a new error-correction quantization function is proposed. This function better aligns with the computational characteristics of the CIM macro, thereby improving robustness. The macro design is evaluated under TSMC 28-nm technology node. At Vref of 100 mV, the proposed macro can run at 120.5 MHz. The energy efficiency for the CIM macro reaches 78.62 TOPS/W under the condition of 4-bit input and 4-bit output. The calculation accuracy on the MNIST and CIFAR-10 dataset are 99.6% and 91.2%, even with a significant process disturbance and memristor resistance fluctuation.
- 22 September 2025
- Published by Institute of Electrical and Electronics Engineers (IEEE) in IEEE Transactions on Device and Materials Reliability
- Vol. 25 (4) , 788-801
Error-Corrected RRAM-Based Compute-in-Memory Accelerator Using Voltage-Clamp Current Mirror
Computational reliability is a promising research direction in computing-in-memory (CIM), particularly in non-volatile memory (NVM) technologies such as resistive random-access memory (RRAM). Due to its outstanding performance and developmental potential, RRAM plays a critical role in advancing CIM research. The high computational reliability, low-cost ADC and low power consumption are the challenges for RRAM CIM. This study proposes an Error-corrected RRAM-based CIM architecture. To solve the problem of excessive calculation current, a voltage clamp current mirror (VCCM) is designed to improve the calculation accuracy and control calculation power consumption. A new quantization method is proposed to avoid the introduction of complex quantization circuits and reduce the energy consumption of quantization. To address the issue of significant errors in small RRAM HRS/LRS, a new error-correction quantization function is proposed. This function better aligns with the computational characteristics of the CIM macro, thereby improving robustness. The macro design is evaluated under TSMC 28-nm technology node. At Vref of 100 mV, the proposed macro can run at 120.5 MHz. The energy efficiency for the CIM macro reaches 78.62 TOPS/W under the condition of 4-bit input and 4-bit output. The calculation accuracy on the MNIST and CIFAR-10 dataset are 99.6% and 91.2%, even with a significant process disturbance and memristor resistance fluctuation.
- 2 September 2025
- Published by Institute of Electrical and Electronics Engineers (IEEE) in IEEE Transactions on Instrumentation and Measurement
- Vol. 74 (00189456) , 1-11
High-Efficiency Calibration Methodology of Interchannel Mismatches in TIADCs Based on Digital Orthogonal Local Oscillation Signal
The paper proposes a novel foreground integrated calibration method based on the digital orthogonal local oscillation signal to address inter-channel mismatches in TIADCs. In contrast to existing approaches, this methodology eliminates the requirement for analog auxiliary circuit configuration while obviating the necessity of employing extensive matrix operations and inverse Fourier transform computations. The employment of the phase difference elimination technique, in conjunction with the approximate calculation of the inverse tangent function, has been demonstrated to yield enhanced calibration performance and optimised design resource utilisation. The RTL simulation results demonstrate that for a 12-bit four-channel TIADC model, the amplitude of the mismatch-induced spurs is suppressed below -82dB following calibration, and SNDR and SFDR reach 72.6dB and 83.7dBFS, respectively. The experimental results of the FPGA-based measurement demonstrate that, under the sinusoidal input frequencies of 498 MHz and 1448 MHz, the SNDR increased by 27.4 dB and 24.5 dB, respectively, both prior to and following calibration. Furthermore, the SFDR increased by 35.7 dB and 30.7 dB, respectively.
- 2 September 2025
- Published by Institute of Electrical and Electronics Engineers (IEEE) in IEEE Transactions on Instrumentation and Measurement
- Vol. 74 (00189456) , 1-11
High-Efficiency Calibration Methodology of Interchannel Mismatches in TIADCs Based on Digital Orthogonal Local Oscillation Signal
The paper proposes a novel foreground integrated calibration method based on the digital orthogonal local oscillation signal to address inter-channel mismatches in TIADCs. In contrast to existing approaches, this methodology eliminates the requirement for analog auxiliary circuit configuration while obviating the necessity of employing extensive matrix operations and inverse Fourier transform computations. The employment of the phase difference elimination technique, in conjunction with the approximate calculation of the inverse tangent function, has been demonstrated to yield enhanced calibration performance and optimised design resource utilisation. The RTL simulation results demonstrate that for a 12-bit four-channel TIADC model, the amplitude of the mismatch-induced spurs is suppressed below -82dB following calibration, and SNDR and SFDR reach 72.6dB and 83.7dBFS, respectively. The experimental results of the FPGA-based measurement demonstrate that, under the sinusoidal input frequencies of 498 MHz and 1448 MHz, the SNDR increased by 27.4 dB and 24.5 dB, respectively, both prior to and following calibration. Furthermore, the SFDR increased by 35.7 dB and 30.7 dB, respectively.
- 21 March 2025
- Published by Institute of Electrical and Electronics Engineers (IEEE) in IEEE Transactions on Very Large Scale Integration (VLSI) Systems
- Vol. 33 (7) , 2009-2013
A 28-nm 9T1C SRAM-Based CIM Macro With Hierarchical Capacitance Weighting and Two-Step Capacitive Comparison ADCs for CNNs
In the realm of charge-domain computing-in-memory (CIM) macros, reducing the area of capacitor ladder and analog-to-digital converter (ADC) while maintaining high throughput remains a significant challenge. This brief introduces an adjustable-weight CIM macro designed to enhance both energy efficiency and area efficiency for convolutional neural networks (CNNs). The proposed architecture uses: 1) a customized 9T1C bit cell for sensing margin improvement and bidirectional decoupled read ports; 2) a hierarchical capacitance weighting (HCW) structure that achieves a weight accumulation of 1/2/4 bits with less capacitance area and weighting time; and 3) a two-step capacitive comparison ADCs (TC-ADCs) readout scheme to improve area efficiency and throughput. The proposed 8-kb static random address memory (SRAM) CIM macro is implemented using 28-nm CMOS technology. It can achieve an energy efficiency of 224.4 TOPS/W and an area efficiency of 21.894 TOPS/mm2, and the accuracies on MNIST, CIFAR-10, and CIFAR-100 datasets are 99.67%, 89.13%, and 67.58% with a 4-b input and 4-b weight.
- 21 March 2025
- Published by Institute of Electrical and Electronics Engineers (IEEE) in IEEE Transactions on Very Large Scale Integration (VLSI) Systems
- Vol. 33 (7) , 2009-2013
A 28-nm 9T1C SRAM-Based CIM Macro With Hierarchical Capacitance Weighting and Two-Step Capacitive Comparison ADCs for CNNs
In the realm of charge-domain computing-in-memory (CIM) macros, reducing the area of capacitor ladder and analog-to-digital converter (ADC) while maintaining high throughput remains a significant challenge. This brief introduces an adjustable-weight CIM macro designed to enhance both energy efficiency and area efficiency for convolutional neural networks (CNNs). The proposed architecture uses: 1) a customized 9T1C bit cell for sensing margin improvement and bidirectional decoupled read ports; 2) a hierarchical capacitance weighting (HCW) structure that achieves a weight accumulation of 1/2/4 bits with less capacitance area and weighting time; and 3) a two-step capacitive comparison ADCs (TC-ADCs) readout scheme to improve area efficiency and throughput. The proposed 8-kb static random address memory (SRAM) CIM macro is implemented using 28-nm CMOS technology. It can achieve an energy efficiency of 224.4 TOPS/W and an area efficiency of 21.894 TOPS/mm2, and the accuracies on MNIST, CIFAR-10, and CIFAR-100 datasets are 99.67%, 89.13%, and 67.58% with a 4-b input and 4-b weight.