Jiawei Xu
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
Research Subjects
Research Interests
Jiawei Xu has a total of 155 co-authors. Here are the top co-authors by number of shared publications:
- Zhiliang Hong
Fudan University
48 shared publications - Tianxiang Qu
State Key Laboratory of Integrated Chips and Systems, Fudan University, Shanghai, China
17 shared publications - Peng Cao
Fudan University
13 shared publications - Liheng Liu
State Key Laboratory of Integrated Chips and Systems, Fudan University, Shanghai, China
8 shared publications - Yijie Li
State Key Laboratory of Integrated Chips and Systems, Fudan University, Shanghai, China
8 shared publications - Nick Van Helleputte
IMEC
7 shared publications
Latest Publications
- 1 October 2026
- Published by Institute of Electrical and Electronics Engineers (IEEE) in IEEE Journal of Solid-State Circuits
- Vol. 61 (10) , 5243-5255
A High-Efficiency Multilevel Class-G Supply Modulator Using a Configurable Capacitor Array for Wideband Power Amplifier
- 1 October 2026
- Published by Institute of Electrical and Electronics Engineers (IEEE) in IEEE Journal of Solid-State Circuits
- Vol. 61 (10) , 5243-5255
A High-Efficiency Multilevel Class-G Supply Modulator Using a Configurable Capacitor Array for Wideband Power Amplifier
- 2 July 2026
- Published by Springer Nature
- p. 163-184
Current-Balancing Instrumentation Amplifier: Principle, Circuit Techniques, and Applications
- 2 July 2026
- Published by Springer Nature
- p. 163-184
Current-Balancing Instrumentation Amplifier: Principle, Circuit Techniques, and Applications
- 1 July 2026
- Published by Institute of Electrical and Electronics Engineers (IEEE) in IEEE Transactions on Power Electronics
- Vol. 41 (7) , 11147-11159
An Ultra-Fast AOT Buck Converter Using Capacitor Current Sensing Technique
This paper presents an ultra-fast adaptive on-time (AOT) buck converter that utilizes a capacitor current sensing technique for superior transient performance. The design features an on-chip self-calibrated RC network that automatically matches the output capacitor impedance, enabling accurate real-time load current detection. This sensed current directly drives an optimal timing generator that computes and applies ideal switching patterns within a single cycle to counteract load transients. Consequently, the implemented architecture achieves time-optimal control with minimal voltage deviation. Fabricated in a 180 nm BCD process, the converter occupies 3.67 mm$^{2}$ and achieves 93.9% peak efficiency. Experimental results demonstrate consistent time-optimal response within one switching cycle across varying operating conditions, including different output voltages, output capacitors, and inductor values.
- 1 July 2026
- Published by Institute of Electrical and Electronics Engineers (IEEE) in IEEE Transactions on Power Electronics
- Vol. 41 (7) , 11147-11159
An Ultra-Fast AOT Buck Converter Using Capacitor Current Sensing Technique
This paper presents an ultra-fast adaptive on-time (AOT) buck converter that utilizes a capacitor current sensing technique for superior transient performance. The design features an on-chip self-calibrated RC network that automatically matches the output capacitor impedance, enabling accurate real-time load current detection. This sensed current directly drives an optimal timing generator that computes and applies ideal switching patterns within a single cycle to counteract load transients. Consequently, the implemented architecture achieves time-optimal control with minimal voltage deviation. Fabricated in a 180 nm BCD process, the converter occupies 3.67 mm$^{2}$ and achieves 93.9% peak efficiency. Experimental results demonstrate consistent time-optimal response within one switching cycle across varying operating conditions, including different output voltages, output capacitors, and inductor values.
- 27 May 2026
- Published by Institute of Electrical and Electronics Engineers (IEEE) in IEEE Journal of Solid-State Circuits
- Vol. PP (00189200) , 1-12
A 36 V Current-Balancing Instrumentation Amplifier With ±24 V Input Range, 5.6 MHz BW, and 140 dB CMRR at All Gain Settings
This article presents a 36V high-voltage current-balancing instrumentation amplifier (HV-CBIA), providing a new solution for industrial signal-acquisition front ends, including PLC input modules and motor drivers, where large input range, wide bandwidth, and high common-mode rejection ratio (CMRR) are simultaneously required. The proposed HV-CBIA addresses these requirements through three innovations: 1) a class-AB transconductance (TC) stage with adaptive biasing is proposed to extend the differential-mode (DM) input range while minimizing quiescent current; 2) a source-input buffer is employed to enhance the driving capability of transimpedance (TI) stage, while a negative capacitance (NC) technique is further incorporated to nullify parasitic capacitance, thereby improving the bandwidth–efficiency trade-off; and 3) an HV chopper employing area-efficient HV switches is adopted to maintain high CMRR and low offset under significant common-mode (CM) and DM swings. Fabricated in a 0.18- $\mu $ m BCD process, the proposed HV-CBIA achieves a ±24V DM input signal range, 5.6MHz bandwidth, and 140dB CMRR across all gain settings (0.4-50). It delivers a 280MHz gain–bandwidth product (GBW) at a gain of 50 while consuming only 0.7mA from an 8–36V supply, corresponding to a power efficiency of 400MHz/mA.
- 27 May 2026
- Published by Institute of Electrical and Electronics Engineers (IEEE) in IEEE Journal of Solid-State Circuits
- Vol. PP (00189200) , 1-12
A 36 V Current-Balancing Instrumentation Amplifier With ±24 V Input Range, 5.6 MHz BW, and 140 dB CMRR at All Gain Settings
This article presents a 36V high-voltage current-balancing instrumentation amplifier (HV-CBIA), providing a new solution for industrial signal-acquisition front ends, including PLC input modules and motor drivers, where large input range, wide bandwidth, and high common-mode rejection ratio (CMRR) are simultaneously required. The proposed HV-CBIA addresses these requirements through three innovations: 1) a class-AB transconductance (TC) stage with adaptive biasing is proposed to extend the differential-mode (DM) input range while minimizing quiescent current; 2) a source-input buffer is employed to enhance the driving capability of transimpedance (TI) stage, while a negative capacitance (NC) technique is further incorporated to nullify parasitic capacitance, thereby improving the bandwidth–efficiency trade-off; and 3) an HV chopper employing area-efficient HV switches is adopted to maintain high CMRR and low offset under significant common-mode (CM) and DM swings. Fabricated in a 0.18- $\mu $ m BCD process, the proposed HV-CBIA achieves a ±24V DM input signal range, 5.6MHz bandwidth, and 140dB CMRR across all gain settings (0.4-50). It delivers a 280MHz gain–bandwidth product (GBW) at a gain of 50 while consuming only 0.7mA from an 8–36V supply, corresponding to a power efficiency of 400MHz/mA.
- 26 May 2026
- Published by Institute of Electrical and Electronics Engineers (IEEE) in IEEE Solid-State Circuits Letters
- Vol. 9, 178-181
A Zero-Static-Power ADC With Integrated Dynamic Reference and Driver-Free Operation Achieving 58.6 dB SNDR From –40 °C to 85 °C
This letter presents a 10-bit ENOB charge-sharing SAR ADC with a fully integrated dynamic bandgap reference (BGR), enabling first-order noise shaping and ultra-low-power operation. The charge-sharing ADC and dynamic BGR form an ideal pair: both operate without static current, allowing compact integration and high precision. The SAR uses only three key capacitors: CREF, CSMP, and CINT, eliminating the conventional large DAC array and relaxing the BGR drive requirement. By retaining the residual voltage on CINT between cycles, quantization noise is shaped, and precision is further improved. The dynamic BGR combines PTAT and CTAT outputs from two PNP-based capacitive DACs, pre-charged to the supply voltage and discharged for an adjustable duration. Implemented in a 65-nm LP CMOS process, the proposed ADC achieves > 58.6 dB SNDR from -40 °C to 85 °C while consuming 5.18μW. The fully integrated dynamic BGR, as part of the ADC, achieves < 10.9 ppm/°C from -40 °C to 125 °C, enabling the full ADC’s state-of-the-art energy efficiency and temperature stability.
- 26 May 2026
- Published by Institute of Electrical and Electronics Engineers (IEEE) in IEEE Solid-State Circuits Letters
- Vol. 9, 178-181
A Zero-Static-Power ADC With Integrated Dynamic Reference and Driver-Free Operation Achieving 58.6 dB SNDR From –40 °C to 85 °C
This letter presents a 10-bit ENOB charge-sharing SAR ADC with a fully integrated dynamic bandgap reference (BGR), enabling first-order noise shaping and ultra-low-power operation. The charge-sharing ADC and dynamic BGR form an ideal pair: both operate without static current, allowing compact integration and high precision. The SAR uses only three key capacitors: CREF, CSMP, and CINT, eliminating the conventional large DAC array and relaxing the BGR drive requirement. By retaining the residual voltage on CINT between cycles, quantization noise is shaped, and precision is further improved. The dynamic BGR combines PTAT and CTAT outputs from two PNP-based capacitive DACs, pre-charged to the supply voltage and discharged for an adjustable duration. Implemented in a 65-nm LP CMOS process, the proposed ADC achieves > 58.6 dB SNDR from -40 °C to 85 °C while consuming 5.18μW. The fully integrated dynamic BGR, as part of the ADC, achieves < 10.9 ppm/°C from -40 °C to 125 °C, enabling the full ADC’s state-of-the-art energy efficiency and temperature stability.
- 1 May 2026
- Published by Institute of Electrical and Electronics Engineers (IEEE) in IEEE Journal of Solid-State Circuits
- Vol. PP (00189200) , 1-13
A 760 mV PP -Input-Range, 103.6dB-SNDR Direct-Digitization Sensor Readout With Pseudo-Differential Integrator and Impedance-Equalized RDAC
A high-precision, direct-digitization sensor readout (DD-RO) based on a continuous-time delta-sigma modulator (CT- $\Delta \Sigma $ M) is presented. The proposed DD-RO incorporates several key innovations to enhance performance. First, a pseudo-differential current-balancing integrator (PD-CBI) significantly extends the linear input range, whereas its intrinsically limited common-mode input range and CMRR are both improved by a common-mode cancellation loop (CMCL). Second, the PD-CBI enables quasi-quiet chopping for 1/ $f$ noise suppression and provides high input impedance. A multibit resistive DAC (RDAC) further increases the input impedance by reducing the transconductor’s net input voltage. Finally, an impedance-equalization technique employing auxiliary RDACs alleviates the signal-dependent nonlinearity arising from reference-path parasitics and DAC current variation. Fabricated in a 0.18- $\mu $ m CMOS process, the prototype achieves an input-referred noise (IRN) of 66nV/ $\surd $ Hz, a 760mVPP linear input range at 1.2V supply, and a calibration-free input impedance of 7.83 G $\Omega $ and a CMRR of 82dB, both measured at 50Hz. The DD-RO achieves a peak SNDR of 103.6dB and a dynamic range (DR) of 105dB, corresponding to state-of-the-art figures of merit of FoM ${}_{\mathrm {SNDR}}= 173.5$ dB and FoM ${}_{\mathrm {DR}} = 174.9$ dB, respectively.
- 1 May 2026
- Published by Institute of Electrical and Electronics Engineers (IEEE) in IEEE Journal of Solid-State Circuits
- Vol. PP (00189200) , 1-13
A 760 mV PP -Input-Range, 103.6dB-SNDR Direct-Digitization Sensor Readout With Pseudo-Differential Integrator and Impedance-Equalized RDAC
A high-precision, direct-digitization sensor readout (DD-RO) based on a continuous-time delta-sigma modulator (CT- $\Delta \Sigma $ M) is presented. The proposed DD-RO incorporates several key innovations to enhance performance. First, a pseudo-differential current-balancing integrator (PD-CBI) significantly extends the linear input range, whereas its intrinsically limited common-mode input range and CMRR are both improved by a common-mode cancellation loop (CMCL). Second, the PD-CBI enables quasi-quiet chopping for 1/ $f$ noise suppression and provides high input impedance. A multibit resistive DAC (RDAC) further increases the input impedance by reducing the transconductor’s net input voltage. Finally, an impedance-equalization technique employing auxiliary RDACs alleviates the signal-dependent nonlinearity arising from reference-path parasitics and DAC current variation. Fabricated in a 0.18- $\mu $ m CMOS process, the prototype achieves an input-referred noise (IRN) of 66nV/ $\surd $ Hz, a 760mVPP linear input range at 1.2V supply, and a calibration-free input impedance of 7.83 G $\Omega $ and a CMRR of 82dB, both measured at 50Hz. The DD-RO achieves a peak SNDR of 103.6dB and a dynamic range (DR) of 105dB, corresponding to state-of-the-art figures of merit of FoM ${}_{\mathrm {SNDR}}= 173.5$ dB and FoM ${}_{\mathrm {DR}} = 174.9$ dB, respectively.
- 2 April 2026
- Published by Institute of Electrical and Electronics Engineers (IEEE) in IEEE Journal of Solid-State Circuits
- Vol. 61 (7) , 3296-3306
A 60-V High-Side / In-Line Instrumentation Amplifier With 54-V/ μ s CM Transient Tolerance and 115-dB Dynamic Range
A high-voltage instrumentation amplifier (HV-IA) is proposed for high-side/in-line current sensing. To achieve low noise and high common-mode (CM) transient tolerance, the IA uses an HV transient-tailored (HV-TT) PMOS input pair with a bidirectional transient shifter (BTS) in the input chopper, achieving 54-V/ $\mu $ s CM transient tolerance (17–37-V CM step) and a wide static input CM voltage range of 0.7–56.3 V. The CM-dependent gain error, caused by the input CM voltage variations, is minimized to 0.4% through a segmented gain adjustment (SGA) technique. Moreover, an automatic gain control (AGC) loop extends the current sensing range, achieving a 115-dB cross-scale dynamic range (DR). Implemented in a 0.18- $\mu $ m BCD process, the HV-IA exhibits 19-nV/ $\surd $ Hz input-referred noise, 140-dB CM rejection ratio (CMRR), and 10- $\mu $ V input offset, while consuming $305~\mu $ A from a 60/4.2-V dual supply.
- 2 April 2026
- Published by Institute of Electrical and Electronics Engineers (IEEE) in IEEE Journal of Solid-State Circuits
- Vol. 61 (7) , 3296-3306
A 60-V High-Side / In-Line Instrumentation Amplifier With 54-V/ μ s CM Transient Tolerance and 115-dB Dynamic Range
A high-voltage instrumentation amplifier (HV-IA) is proposed for high-side/in-line current sensing. To achieve low noise and high common-mode (CM) transient tolerance, the IA uses an HV transient-tailored (HV-TT) PMOS input pair with a bidirectional transient shifter (BTS) in the input chopper, achieving 54-V/ $\mu $ s CM transient tolerance (17–37-V CM step) and a wide static input CM voltage range of 0.7–56.3 V. The CM-dependent gain error, caused by the input CM voltage variations, is minimized to 0.4% through a segmented gain adjustment (SGA) technique. Moreover, an automatic gain control (AGC) loop extends the current sensing range, achieving a 115-dB cross-scale dynamic range (DR). Implemented in a 0.18- $\mu $ m BCD process, the HV-IA exhibits 19-nV/ $\surd $ Hz input-referred noise, 140-dB CM rejection ratio (CMRR), and 10- $\mu $ V input offset, while consuming $305~\mu $ A from a 60/4.2-V dual supply.
- 18 February 2026
- Published by Institute of Electrical and Electronics Engineers (IEEE) in IEEE Journal of Solid-State Circuits
- Vol. PP (00189200) , 1-12
An Energy-Efficient and High-Accuracy Dual-ModeECG AI Processor via Deep-Fold and Zero-Free Fine-Grained Quantization
Smart wearable electrocardiography (ECG) devices require classification processors that combine high accuracy, low power dissipation, and real-time operation under stringent energy constraints. Conventional neural network approaches improve accuracy but are limited by high computational and memory costs, restricting on-chip deployment. This work presents an energy-efficient ECG AI processor enabled by algorithm-hardware co-design. Key contributions include: 1) a unified lightweight convolutional neural network (CNN) with a shared convolutional backbone and dual task-specific heads for binary and five-class classification; 2) a hardware-oriented deep-fold technique using weight symmetry to reduce parameters and multiplications by 39.1% and 40.1%, respectively; and 3) a zero-free fine-grained (ZFFG) quantization scheme that preserves accuracy at low bit-widths while enabling multiplier-less inference. Fabricated in 65-nm CMOS, the processor achieves 98.70% (binary) and 98.72% (five-class) classification accuracy with an ultra-compact 2.37 KB on-chip SRAM and only $0.09~{\mu }$ J per inference, representing a state-of-the-art, energy-efficient solution for real-time wearable ECG monitoring.
- 18 February 2026
- Published by Institute of Electrical and Electronics Engineers (IEEE) in IEEE Journal of Solid-State Circuits
- Vol. PP (00189200) , 1-12
An Energy-Efficient and High-Accuracy Dual-ModeECG AI Processor via Deep-Fold and Zero-Free Fine-Grained Quantization
Smart wearable electrocardiography (ECG) devices require classification processors that combine high accuracy, low power dissipation, and real-time operation under stringent energy constraints. Conventional neural network approaches improve accuracy but are limited by high computational and memory costs, restricting on-chip deployment. This work presents an energy-efficient ECG AI processor enabled by algorithm-hardware co-design. Key contributions include: 1) a unified lightweight convolutional neural network (CNN) with a shared convolutional backbone and dual task-specific heads for binary and five-class classification; 2) a hardware-oriented deep-fold technique using weight symmetry to reduce parameters and multiplications by 39.1% and 40.1%, respectively; and 3) a zero-free fine-grained (ZFFG) quantization scheme that preserves accuracy at low bit-widths while enabling multiplier-less inference. Fabricated in 65-nm CMOS, the processor achieves 98.70% (binary) and 98.72% (five-class) classification accuracy with an ultra-compact 2.37 KB on-chip SRAM and only $0.09~{\mu }$ J per inference, representing a state-of-the-art, energy-efficient solution for real-time wearable ECG monitoring.