Hae-Seung Lee
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 Interests
Hae-Seung Lee has a total of 35 co-authors. Here are the top co-authors by number of shared publications:
- Anantha P. Chandrakasan
Department of Electrical Engineering and Computer Science, Massachusetts Institute of Technology, Cambridge, USA
2 shared publications - Charles G. Sodini
Massachusetts Institute of Technology, Cambridge, MA, USA
2 shared publications A. Betts
Cypress Semicond., Cambridge, MA
1 shared publicationCarlos E. Munoz
North Chelmsford
1 shared publication- Christina Wang
The Lundquist Institute at Harbor-UCLA Medical Center
1 shared publication - Denis C. Daly
Maxim Integrated
1 shared publication
Latest Publications
- 12 September 2025
- Published by Institute of Electrical and Electronics Engineers (IEEE) in Proceedings of the IEEE
- Vol. 113 (6) , 586-604
Protecting the Mixed-Signal Domain: Secure ADCs for Internet of Things Devices
Analog-to-digital converters (ADCs) are a standard building block of most Internet of Things (IoT) devices, used to convert between analog signals of the physical world and digital values for efficient storage and computation. While decades of research have explored methods to optimize the traditional power, performance, and area parameters of these circuits, a new requirement has emerged in the last five years for ADCs—security. Due to the deployment of these devices at the edge outside of a trusted computing base, there is a potential for various eavesdropping and tampering attacks. This can have a severe impact on the privacy and integrity of sensor data, which cannot be corrected for through the secure design of other blocks that follow the front end. In this article, we explore the recent work in ADC security and analyze what has been accomplished as well as what remains to be done for the successful deployment of secure ADCs in commercial systems.
- 12 September 2025
- Published by Institute of Electrical and Electronics Engineers (IEEE) in Proceedings of the IEEE
- Vol. 113 (6) , 586-604
Protecting the Mixed-Signal Domain: Secure ADCs for Internet of Things Devices
Analog-to-digital converters (ADCs) are a standard building block of most Internet of Things (IoT) devices, used to convert between analog signals of the physical world and digital values for efficient storage and computation. While decades of research have explored methods to optimize the traditional power, performance, and area parameters of these circuits, a new requirement has emerged in the last five years for ADCs—security. Due to the deployment of these devices at the edge outside of a trusted computing base, there is a potential for various eavesdropping and tampering attacks. This can have a severe impact on the privacy and integrity of sensor data, which cannot be corrected for through the secure design of other blocks that follow the front end. In this article, we explore the recent work in ADC security and analyze what has been accomplished as well as what remains to be done for the successful deployment of secure ADCs in commercial systems.
- 22 March 2024
- Published by Institute of Electrical and Electronics Engineers (IEEE) in IEEE Transactions on Nanotechnology
- Vol. 23 (1536125X) , 281-285
Biosensor Chip for Point-of-Care Diagnostics: Carbon Nanotube Sensing Platform for Bacterial Detection and Identification
Emerging technologies, such as carbon nanotubes (CNTs), show exciting benefits for next-generation electronic systems and have recently been incorporated within commercial fabrication facilities and foundries. Here we demonstrate a real-world application of carbon nanotube field-effect transistors (CNFETs) leveraging these foundry capabilities: performing disease diagnostics by detecting and identifying infectious pathogens in model biological samples. This Biosensor chip leverages the benefits of 200 mm wafer-scale and VLSI-compatible foundry fabrication of CNFETs, which can be seamlessly integrated with the mature sensing modality of antibody-based targeted antigen binding, to realize a robust and manufacturable biosensor technology. As a demonstration, we realize Biosensor chips that are capable of detecting and identifying three different infectious pathogens simultaneously through multiplexing across hundreds of functionalized CNFET-antibody biosensors within each chip.
- 22 March 2024
- Published by Institute of Electrical and Electronics Engineers (IEEE) in IEEE Transactions on Nanotechnology
- Vol. 23 (1536125X) , 281-285
Biosensor Chip for Point-of-Care Diagnostics: Carbon Nanotube Sensing Platform for Bacterial Detection and Identification
Emerging technologies, such as carbon nanotubes (CNTs), show exciting benefits for next-generation electronic systems and have recently been incorporated within commercial fabrication facilities and foundries. Here we demonstrate a real-world application of carbon nanotube field-effect transistors (CNFETs) leveraging these foundry capabilities: performing disease diagnostics by detecting and identifying infectious pathogens in model biological samples. This Biosensor chip leverages the benefits of 200 mm wafer-scale and VLSI-compatible foundry fabrication of CNFETs, which can be seamlessly integrated with the mature sensing modality of antibody-based targeted antigen binding, to realize a robust and manufacturable biosensor technology. As a demonstration, we realize Biosensor chips that are capable of detecting and identifying three different infectious pathogens simultaneously through multiplexing across hundreds of functionalized CNFET-antibody biosensors within each chip.
- 11 December 2023
- Published by Institute of Electrical and Electronics Engineers (IEEE) in IEEE Journal of Solid-State Circuits
- Vol. 59 (4) , 1158-1170
A 6.4-GS/s 1-GHz BW Continuous-Time Pipelined ADC With Time-Interleaved Sub-ADC-DAC Achieving 61.7-dB SNDR in 16-nm FinFET
We introduce a continuous-time (CT) pipeline analog-to-digital converter (ADC) featuring a time-interleaved sub-ADC-digital-to-analog converter (DAC) path in its first stage. The proposed sub-ADC-DAC path enhances the ADC’s bandwidth by improving the signal cancellation at the summing node of the first stage. In addition, we have developed an inductorless delay line for the first stage, improving the amplitude and phase matching, thus minimizing the input signal leakage into the backend ADC. Furthermore, the theoretical jitter limitations in the CT pipeline architecture have been explored, and the proposed theory is compared against the measured results. The prototype ADC was fabricated in a 16-nm FinFET process. The ADC achieves a peak signal-to-noise ratio (SNR) of 61.7 dB at low frequencies and 60.8 dB at high frequencies across a 1-GHz bandwidth. The active area of the ADC is 0.77 mm $^2$ , and it consumes 240 mW. The Schreier figure of merit (FOM $_S$ ) is 157.9 dB, which is amongst the best for CT ADCs with bandwidth greater than 500 MHz.
- 11 December 2023
- Published by Institute of Electrical and Electronics Engineers (IEEE) in IEEE Journal of Solid-State Circuits
- Vol. 59 (4) , 1158-1170
A 6.4-GS/s 1-GHz BW Continuous-Time Pipelined ADC With Time-Interleaved Sub-ADC-DAC Achieving 61.7-dB SNDR in 16-nm FinFET
We introduce a continuous-time (CT) pipeline analog-to-digital converter (ADC) featuring a time-interleaved sub-ADC-digital-to-analog converter (DAC) path in its first stage. The proposed sub-ADC-DAC path enhances the ADC’s bandwidth by improving the signal cancellation at the summing node of the first stage. In addition, we have developed an inductorless delay line for the first stage, improving the amplitude and phase matching, thus minimizing the input signal leakage into the backend ADC. Furthermore, the theoretical jitter limitations in the CT pipeline architecture have been explored, and the proposed theory is compared against the measured results. The prototype ADC was fabricated in a 16-nm FinFET process. The ADC achieves a peak signal-to-noise ratio (SNR) of 61.7 dB at low frequencies and 60.8 dB at high frequencies across a 1-GHz bandwidth. The active area of the ADC is 0.77 mm $^2$ , and it consumes 240 mW. The Schreier figure of merit (FOM $_S$ ) is 157.9 dB, which is amongst the best for CT ADCs with bandwidth greater than 500 MHz.
- 1 September 2023
- Published by The Endocrine Society in The Journal of Clinical Endocrinology & Metabolism
- Vol. 109 (2) , e757-e764
The Effect of Testosterone Replacement Therapy on Nonalcoholic Fatty Liver Disease in Older Hypogonadal Men
Context: Male hypogonadism is associated with visceral obesity and the metabolic syndrome: factors important for the development of nonalcoholic fatty liver disease (NAFLD). The Testosterone Trials (The T Trials) showed testosterone (T) treatment compared with placebo in older hypogonadal men was associated with decreases in cholesterol and insulin levels suggesting that T treatment may improve NAFLD. Objective: Compare effects of T vs placebo treatment on NAFLD scores and liver scans in elderly hypogonadal men. Methods: Secondary data analyses from 479 older hypogonadal men with total T < 275 ng/dL from The T Trials were performed. Three clinical liver fat scores—lipid accumulation product index, hepatic steatosis index, nonalcoholic fatty liver disease-metabolic syndrome score—and liver computed tomography (CT) Hounsfield units and liver to spleen ratio were evaluated at baseline and 12 months after treatment. Results: There were no statistically significant differences of change in lipid accumulation product index (P = .98), hepatic steatosis index (P = .67), and nonalcoholic fatty liver disease-metabolic syndrome (P = .52) in 246 men treated with T compared with 233 treated with placebo for 12 months. Liver CT showed no statistically significant difference of change in Hounsfield units (P = .24; n = 71 for T, n = 69 for placebo) and liver to spleen ratio (P = .74; n = 55 for T, n = 62 for placebo) between the 2 groups. Conclusions: Our study did not show improvement of NAFLD in older hypogonadal men after 12 months of T vs placebo treatment, as assessed by 3 clinical scores and liver CT for hepatic steatosis. Future studies with longer treatment duration and additional NAFLD diagnostic modalities as primary outcome are warranted.
- 1 September 2023
- Published by The Endocrine Society in The Journal of Clinical Endocrinology & Metabolism
- Vol. 109 (2) , e757-e764
The Effect of Testosterone Replacement Therapy on Nonalcoholic Fatty Liver Disease in Older Hypogonadal Men
Context: Male hypogonadism is associated with visceral obesity and the metabolic syndrome: factors important for the development of nonalcoholic fatty liver disease (NAFLD). The Testosterone Trials (The T Trials) showed testosterone (T) treatment compared with placebo in older hypogonadal men was associated with decreases in cholesterol and insulin levels suggesting that T treatment may improve NAFLD. Objective: Compare effects of T vs placebo treatment on NAFLD scores and liver scans in elderly hypogonadal men. Methods: Secondary data analyses from 479 older hypogonadal men with total T < 275 ng/dL from The T Trials were performed. Three clinical liver fat scores—lipid accumulation product index, hepatic steatosis index, nonalcoholic fatty liver disease-metabolic syndrome score—and liver computed tomography (CT) Hounsfield units and liver to spleen ratio were evaluated at baseline and 12 months after treatment. Results: There were no statistically significant differences of change in lipid accumulation product index (P = .98), hepatic steatosis index (P = .67), and nonalcoholic fatty liver disease-metabolic syndrome (P = .52) in 246 men treated with T compared with 233 treated with placebo for 12 months. Liver CT showed no statistically significant difference of change in Hounsfield units (P = .24; n = 71 for T, n = 69 for placebo) and liver to spleen ratio (P = .74; n = 55 for T, n = 62 for placebo) between the 2 groups. Conclusions: Our study did not show improvement of NAFLD in older hypogonadal men after 12 months of T vs placebo treatment, as assessed by 3 clinical scores and liver CT for hepatic steatosis. Future studies with longer treatment duration and additional NAFLD diagnostic modalities as primary outcome are warranted.
- Open Access
Learning to Design Circuits
Analog IC design relies on human experts to search for parameters that satisfy circuit specifications with their experience and intuitions, which is highly labor intensive, time consuming and suboptimal. Machine learning is a promising tool to automate this process. However, supervised learning is difficult for this task due to the low availability of training data: 1) Circuit simulation is slow, thus generating large-scale dataset is time-consuming; 2) Most circuit designs are propitiatory IPs within individual IC companies, making it expensive to collect large-scale datasets. We propose Learning to Design Circuits (L2DC) to leverage reinforcement learning that learns to efficiently generate new circuits data and to optimize circuits. We fix the schematic, and optimize the parameters of the transistors automatically by training an RL agent with no prior knowledge about optimizing circuits. After iteratively getting observations, generating a new set of transistor parameters, getting a reward, and adjusting the model, L2DC is able to optimize circuits. We evaluate L2DC on two transimpedance amplifiers. Trained for a day, our RL agent can achieve comparable or better performance than human experts trained for a quarter. It first learns to meet hard-constraints (eg. gain, bandwidth), and then learns to optimize good-to-have targets (eg. area, power). Compared with grid search-aided human design, L2DC can achieve $\mathbf{250}\boldsymbol{\times}$ higher sample efficiency with comparable performance. Under the same runtime constraint, the performance of L2DC is also better than Bayesian Optimization.Open AccessLearning to Design Circuits
Analog IC design relies on human experts to search for parameters that satisfy circuit specifications with their experience and intuitions, which is highly labor intensive, time consuming and suboptimal. Machine learning is a promising tool to automate this process. However, supervised learning is difficult for this task due to the low availability of training data: 1) Circuit simulation is slow, thus generating large-scale dataset is time-consuming; 2) Most circuit designs are propitiatory IPs within individual IC companies, making it expensive to collect large-scale datasets. We propose Learning to Design Circuits (L2DC) to leverage reinforcement learning that learns to efficiently generate new circuits data and to optimize circuits. We fix the schematic, and optimize the parameters of the transistors automatically by training an RL agent with no prior knowledge about optimizing circuits. After iteratively getting observations, generating a new set of transistor parameters, getting a reward, and adjusting the model, L2DC is able to optimize circuits. We evaluate L2DC on two transimpedance amplifiers. Trained for a day, our RL agent can achieve comparable or better performance than human experts trained for a quarter. It first learns to meet hard-constraints (eg. gain, bandwidth), and then learns to optimize good-to-have targets (eg. area, power). Compared with grid search-aided human design, L2DC can achieve $\mathbf{250}\boldsymbol{\times}$ higher sample efficiency with comparable performance. Under the same runtime constraint, the performance of L2DC is also better than Bayesian Optimization. - 18 May 2020
- Published by Institute of Electrical and Electronics Engineers (IEEE) in IEEE Journal of Biomedical and Health Informatics
- Vol. 25 (1) , 163-174
Non-Invasive Evaluation of a Carotid Arterial Pressure Waveform Using Motion-Tolerant Ultrasound Measurements During the Valsalva Maneuver
This work details the non-invasive evaluation of a carotid arterial blood pressure (ABP) waveform during the Valsalva maneuver. Unfocused and wide acoustic beams are insonated on the carotid artery to achieve motion-tolerant measurements with a simple two-element ultrasound scanner. Arterial flow and distension waveforms are reliably estimated from spectral Doppler and M-mode ultrasound images whose qualities are consistently maintained in different phases of the maneuver despite possible displacements of the artery. A local pulse wave velocity is estimated using a flow-area method, and it is then combined with the distension waveform to produce the ABP waveform. Human subject validation on seven healthy subjects shows that the bias in pulse pressure estimates across subjects is 0.47±13.1 mmHg. The average root mean square deviations of the ultrasonically measured waveform across subjects is 10.1±2.43 mmHg, excluding the strain phase of the Valsalva maneuver, and 17.7±6.30 mmHg in all phases. The mean correlation coefficient between the ultrasonically measured and reference waveform is calculated to be 0.92±0.04 across subjects. Detailed morphological features and their changes across different phases are observed as reported. This uninterrupted central ABP waveform monitoring under hemodynamics changes supports the idea of a novel stress test to evaluate the health and dynamics of the cardiovascular system at a spot check in clinical settings.
- 18 May 2020
- Published by Institute of Electrical and Electronics Engineers (IEEE) in IEEE Journal of Biomedical and Health Informatics
- Vol. 25 (1) , 163-174
Non-Invasive Evaluation of a Carotid Arterial Pressure Waveform Using Motion-Tolerant Ultrasound Measurements During the Valsalva Maneuver
This work details the non-invasive evaluation of a carotid arterial blood pressure (ABP) waveform during the Valsalva maneuver. Unfocused and wide acoustic beams are insonated on the carotid artery to achieve motion-tolerant measurements with a simple two-element ultrasound scanner. Arterial flow and distension waveforms are reliably estimated from spectral Doppler and M-mode ultrasound images whose qualities are consistently maintained in different phases of the maneuver despite possible displacements of the artery. A local pulse wave velocity is estimated using a flow-area method, and it is then combined with the distension waveform to produce the ABP waveform. Human subject validation on seven healthy subjects shows that the bias in pulse pressure estimates across subjects is 0.47±13.1 mmHg. The average root mean square deviations of the ultrasonically measured waveform across subjects is 10.1±2.43 mmHg, excluding the strain phase of the Valsalva maneuver, and 17.7±6.30 mmHg in all phases. The mean correlation coefficient between the ultrasonically measured and reference waveform is calculated to be 0.92±0.04 across subjects. Detailed morphological features and their changes across different phases are observed as reported. This uninterrupted central ABP waveform monitoring under hemodynamics changes supports the idea of a novel stress test to evaluate the health and dynamics of the cardiovascular system at a spot check in clinical settings.
- 29 May 2014
- Published by Institute of Electrical and Electronics Engineers (IEEE) in IEEE Journal of Solid-State Circuits
- Vol. 49 (6) , 1366-1382
A 12 bit 200 MS/s Zero-Crossing-Based Pipelined ADC With Early Sub-ADC Decision and Output Residue Background Calibration
A 12 bit 200 MS/s analog-to-digital converter (ADC) applies techniques of zero-crossing-based circuits as a replacement for high-gain high-speed op-amps. High accuracy in the residue amplifier is achieved by using a coarse phase in ZCBC followed by a level-shifting capacitor for fine phase. Sub-ADC flash comparators are strobed immediately after the coarse phase to achieve a high sampling rate. The systematic offset voltage between the coarse and fine phase manifests itself as systematic offset in the sub-ADC comparators. This offset is caused by the coarse phase undershoot and the fine phase overshoot. In this work, the offset is cancelled with background calibration by residue range correction circuits in the following stage's sub-ADC. In addition, the sub-ADC's random comparator offset is calibrated with a discrete-time charge-pump based background calibration technique. The reference buffer, bias circuitry, and digital error correction circuits are all integrated on a single chip. The ADC occupies an area of 0.282 mm 2 in 55 nm CMOS technology and dissipates 30.7 mW. It achieves 64.6 dB SNDR and 82.9 dBc SFDR at 200 MS/s for a FOM of 111 fJ/conversion-step. The SNDR degrades gracefully above the designed sampling frequency to 62.9 dB at 250 MS/s, and remains above 50 dB at 300 MS/s.
- 29 May 2014
- Published by Institute of Electrical and Electronics Engineers (IEEE) in IEEE Journal of Solid-State Circuits
- Vol. 49 (6) , 1366-1382
A 12 bit 200 MS/s Zero-Crossing-Based Pipelined ADC With Early Sub-ADC Decision and Output Residue Background Calibration
A 12 bit 200 MS/s analog-to-digital converter (ADC) applies techniques of zero-crossing-based circuits as a replacement for high-gain high-speed op-amps. High accuracy in the residue amplifier is achieved by using a coarse phase in ZCBC followed by a level-shifting capacitor for fine phase. Sub-ADC flash comparators are strobed immediately after the coarse phase to achieve a high sampling rate. The systematic offset voltage between the coarse and fine phase manifests itself as systematic offset in the sub-ADC comparators. This offset is caused by the coarse phase undershoot and the fine phase overshoot. In this work, the offset is cancelled with background calibration by residue range correction circuits in the following stage's sub-ADC. In addition, the sub-ADC's random comparator offset is calibrated with a discrete-time charge-pump based background calibration technique. The reference buffer, bias circuitry, and digital error correction circuits are all integrated on a single chip. The ADC occupies an area of 0.282 mm 2 in 55 nm CMOS technology and dissipates 30.7 mW. It achieves 64.6 dB SNDR and 82.9 dBc SFDR at 200 MS/s for a FOM of 111 fJ/conversion-step. The SNDR degrades gracefully above the designed sampling frequency to 62.9 dB at 250 MS/s, and remains above 50 dB at 300 MS/s.
- 1 June 2007
- Published by Institute of Electrical and Electronics Engineers (IEEE)
- p. 6-9
Limits of Power Consumption in Analog Circuits
In analog circuits, power consumption is fundamentally related to noise and the signal bandwidth. In this paper, the theoretical minimum and practical power consumption of various implementations of analog circuits is analyzed. Continuous-time and discrete-time implementations of analog circuits are contrasted. Recent circuit approaches such as comparator and zero-crossing-based circuits are also considered. Finally, the effects of scaling on the power consumption are discussed.
- 1 June 2007
- Published by Institute of Electrical and Electronics Engineers (IEEE)
- p. 6-9
Limits of Power Consumption in Analog Circuits
In analog circuits, power consumption is fundamentally related to noise and the signal bandwidth. In this paper, the theoretical minimum and practical power consumption of various implementations of analog circuits is analyzed. Continuous-time and discrete-time implementations of analog circuits are contrasted. Recent circuit approaches such as comparator and zero-crossing-based circuits are also considered. Finally, the effects of scaling on the power consumption are discussed.
- 1 January 2006
- Published by Institute of Electrical and Electronics Engineers (IEEE)
- No. 01936530,p. 2004-2013
A 3MPixel Low-Noise Flexible Architecture CMOS Image Sensor
An image sensor with a 2.54mum pixel fabricated in a 0.18mum CMOS technology is presented. The 3T pixel with drain-side row-select reduces reset noise by cascoded feedback reset and increases responsivity with common-source readout. The reset noise is 13e- and the responsivity at 550nm is 1.12V/lux-s in source-follower readout mode and 5.6V/lux-s in common-source readout mode
- 1 January 2006
- Published by Institute of Electrical and Electronics Engineers (IEEE)
- No. 01936530,p. 2004-2013
A 3MPixel Low-Noise Flexible Architecture CMOS Image Sensor
An image sensor with a 2.54mum pixel fabricated in a 0.18mum CMOS technology is presented. The 3T pixel with drain-side row-select reduces reset noise by cascoded feedback reset and increases responsivity with common-source readout. The reset noise is 13e- and the responsivity at 550nm is 1.12V/lux-s in source-follower readout mode and 5.6V/lux-s in common-source readout mode