Mingoo Seok
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
Mingoo Seok has a total of 190 co-authors. Here are the top co-authors by number of shared publications:
- Bo Zhang
Electrical Engineering, Columbia University, New York, New York, USA
10 shared publications - Jonghyun Oh
Department of Electrical Engineering, Columbia University, New York, NY, USA
8 shared publications - Ram Kumar Krishnamurthy
Intel Corporation, Hillsboro, OR, USA
7 shared publications - Weifeng He
Shanghai Jiao Tong University
7 shared publications - Chuan-Tung Lin
Department of Electrical Engineering, Columbia University, New York, NY, USA
6 shared publications - Jiangyi Li
Department of Electrical Engineering, Columbia University, New York, NY, USA
6 shared publications
Latest Publications
- 1 November 2026
- Published by Institute of Electrical and Electronics Engineers (IEEE) in IEEE Transactions on Circuits and Systems I: Regular Papers
- Vol. 73 (11) , 7610-7623
KWS-CIM: A Charge-Domain eDRAM-Based Compute-in-Memory Macro for Ternary DFSMN Keyword Spotting
- 1 November 2026
- Published by Institute of Electrical and Electronics Engineers (IEEE) in IEEE Transactions on Circuits and Systems I: Regular Papers
- Vol. 73 (11) , 7610-7623
KWS-CIM: A Charge-Domain eDRAM-Based Compute-in-Memory Macro for Ternary DFSMN Keyword Spotting
- Open Access
- 26 October 2026
- Published by Association for Computing Machinery (ACM)
- p. 562-580
Wideband Real-Time Adaptive Full-Duplex Radios
- 26 October 2026
- Published by Association for Computing Machinery (ACM)
- p. 562-580
Open AccessWideband Real-Time Adaptive Full-Duplex Radios
- 2 June 2026
- Published by Institute of Electrical and Electronics Engineers (IEEE) in IEEE Journal of Solid-State Circuits
- Vol. PP (00189200) , 1-13
STAR-SRAM: 16-bit Floating-Point SRAM-Based Digital Computing-in-Memory Macro in a 28 nm
A digital computing-in-memory (DCIM) macro emerges as a promising building block in a deep neural network (DNN) accelerator. To better support DNN workloads, circuit designers aim to improve three main metrics for macros: energy efficiency, compute density, and weight density. Improvements in those metrics directly translate into reduced energy consumption, shorter inference latency, and a smaller silicon footprint at the accelerator level. This article introduces STAR-SRAM, a DCIM macro designed for 16-bit brain floating-point (BF16) arithmetic. The macro is designed using a top-down approach. We started the design process with model-based parameter optimization, in which we determined the optimal design parameters that maximize the figure of merit (FoM) of a DCIM macro. Then, we developed several circuit techniques that further improve the FoM, including pre-multiply symmetric alignment, approximate multiplier circuits, and sparsity-aware wordline, clock, and input gating techniques. Fabricated in 28-nm CMOS, STAR-SRAM achieves the state-of-the-art FoM with up to 43.06 TFLOPS/W, 1.89 TFLOPS/mm2, and 400 Kb/mm2.
- 2 June 2026
- Published by Institute of Electrical and Electronics Engineers (IEEE) in IEEE Journal of Solid-State Circuits
- Vol. PP (00189200) , 1-13
STAR-SRAM: 16-bit Floating-Point SRAM-Based Digital Computing-in-Memory Macro in a 28 nm
A digital computing-in-memory (DCIM) macro emerges as a promising building block in a deep neural network (DNN) accelerator. To better support DNN workloads, circuit designers aim to improve three main metrics for macros: energy efficiency, compute density, and weight density. Improvements in those metrics directly translate into reduced energy consumption, shorter inference latency, and a smaller silicon footprint at the accelerator level. This article introduces STAR-SRAM, a DCIM macro designed for 16-bit brain floating-point (BF16) arithmetic. The macro is designed using a top-down approach. We started the design process with model-based parameter optimization, in which we determined the optimal design parameters that maximize the figure of merit (FoM) of a DCIM macro. Then, we developed several circuit techniques that further improve the FoM, including pre-multiply symmetric alignment, approximate multiplier circuits, and sparsity-aware wordline, clock, and input gating techniques. Fabricated in 28-nm CMOS, STAR-SRAM achieves the state-of-the-art FoM with up to 43.06 TFLOPS/W, 1.89 TFLOPS/mm2, and 400 Kb/mm2.
- 1 June 2026
- Published by Institute of Electrical and Electronics Engineers (IEEE) in IEEE Journal of Solid-State Circuits
- Vol. PP (00189200) , 1-12
GUARD: A Clock and Voltage Glitch Detector With On-Demand Protection
This article presents GUARD, a fully digital, variation-tolerant detector for clock and voltage glitch attacks, featuring an integrated on-demand protection mechanism. Fabricated in 28-nm CMOS, GUARD provides robust security for digital systems by monitoring the system clock for maliciously injected faults. It employs a pair of optimized time-to-digital converters (TDCs) to detect clock pulsewidth variations as small as 100 ps and supports clock frequencies up to 1.2GHz. Upon detection, a low-latency clock-gating mechanism is immediately triggered to prevent fault propagation into a victim microprocessor unit (MCU), achieving 100% protection success in experimental attacks. With simple one-time calibration, GUARD demonstrates robust operation across a wide industrial temperature range of $- 40~{^{\circ }}$ C to $125~{^{\circ }}$ C and a supply voltage range of 0.6 to 1.1V. Furthermore, GUARD can differentiate between malicious voltage glitches and normal on-chip IR drops, thereby preventing unnecessary performance overhead. GUARD occupies $2958~\mu $ m2 and consumes 0.874 mW per 100 MHz of operating frequency.
- 1 June 2026
- Published by Institute of Electrical and Electronics Engineers (IEEE) in IEEE Journal of Solid-State Circuits
- Vol. PP (00189200) , 1-12
GUARD: A Clock and Voltage Glitch Detector With On-Demand Protection
This article presents GUARD, a fully digital, variation-tolerant detector for clock and voltage glitch attacks, featuring an integrated on-demand protection mechanism. Fabricated in 28-nm CMOS, GUARD provides robust security for digital systems by monitoring the system clock for maliciously injected faults. It employs a pair of optimized time-to-digital converters (TDCs) to detect clock pulsewidth variations as small as 100 ps and supports clock frequencies up to 1.2GHz. Upon detection, a low-latency clock-gating mechanism is immediately triggered to prevent fault propagation into a victim microprocessor unit (MCU), achieving 100% protection success in experimental attacks. With simple one-time calibration, GUARD demonstrates robust operation across a wide industrial temperature range of $- 40~{^{\circ }}$ C to $125~{^{\circ }}$ C and a supply voltage range of 0.6 to 1.1V. Furthermore, GUARD can differentiate between malicious voltage glitches and normal on-chip IR drops, thereby preventing unnecessary performance overhead. GUARD occupies $2958~\mu $ m2 and consumes 0.874 mW per 100 MHz of operating frequency.
- 27 March 2026
- Published by Institute of Electrical and Electronics Engineers (IEEE) in IEEE Journal on Emerging and Selected Topics in Circuits and Systems
- Vol. 16 (2) , 276-288
IGA-SRAM: A Compact SRAM-Based IMC Engine for Binary/Ternary Deep Neural Networks With ADC-Less and Error-Aware Training Algorithm
This work presents a compact 6T SRAM-based analog in-memory computing (A-IMC) macro for energy-efficient inference of binary and ternary neural networks (BNNs/TNNs). While conventional A-IMC architectures suffer from significant area and power overheads due to high-resolution analog-to-digital converters (ADCs), we propose an ADC-less computing scheme based on 1-bit sense amplifiers (SAs), supported by a novel training algorithm that mitigates quantization-induced performance degradation. To maintain hardware simplicity, we adopt a compact 6T bitcell structure and introduce dual wordlines (WLs) for efficient ternary input encoding. We further propose impulse Gaussian approximation (IGA), a differentiable quantization surrogate that enables stable and accurate gradient backpropagation in the presence of 1-bit partial sum quantization. Additionally, we introduce an error-aware training (EAT) method that leverages measured partial-sum (PS) error statistics from silicon to inject gradually increasing noise in the training phase, effectively compensating for hardware-induced variations. Fabricated in a 65 nm CMOS GP process, the proposed 128 × 128 A-IMC macro achieves 47.06 TOPS/mm2 and CIFAR10 accuracy of 86.89% under 1.2/0.7 V operating voltages with ResNet20.
- 27 March 2026
- Published by Institute of Electrical and Electronics Engineers (IEEE) in IEEE Journal on Emerging and Selected Topics in Circuits and Systems
- Vol. 16 (2) , 276-288
IGA-SRAM: A Compact SRAM-Based IMC Engine for Binary/Ternary Deep Neural Networks With ADC-Less and Error-Aware Training Algorithm
This work presents a compact 6T SRAM-based analog in-memory computing (A-IMC) macro for energy-efficient inference of binary and ternary neural networks (BNNs/TNNs). While conventional A-IMC architectures suffer from significant area and power overheads due to high-resolution analog-to-digital converters (ADCs), we propose an ADC-less computing scheme based on 1-bit sense amplifiers (SAs), supported by a novel training algorithm that mitigates quantization-induced performance degradation. To maintain hardware simplicity, we adopt a compact 6T bitcell structure and introduce dual wordlines (WLs) for efficient ternary input encoding. We further propose impulse Gaussian approximation (IGA), a differentiable quantization surrogate that enables stable and accurate gradient backpropagation in the presence of 1-bit partial sum quantization. Additionally, we introduce an error-aware training (EAT) method that leverages measured partial-sum (PS) error statistics from silicon to inject gradually increasing noise in the training phase, effectively compensating for hardware-induced variations. Fabricated in a 65 nm CMOS GP process, the proposed 128 × 128 A-IMC macro achieves 47.06 TOPS/mm2 and CIFAR10 accuracy of 86.89% under 1.2/0.7 V operating voltages with ResNet20.
- 24 February 2026
- Published by Institute of Electrical and Electronics Engineers (IEEE) in IEEE Transactions on Circuits and Systems I: Regular Papers
- Vol. PP (15498328) , 1-12
D6CIM: 60.4-TOPS/W All-Digital 6T-SRAM-Based Compute-in-Memory Macro Supporting 1-to-8 b Fixed-Point Arithmetic in a 28-nm CMOS
This article presents an all-digital 6T-SRAM-based compute-in-memory (D6CIM) macro that supports 1-to-8-bit fixed-point vector matrix multiplication. The D6CIM is fully digital, employing only digital standard cells and 6T SRAM bitcells. The D6CIM adopts a time-sharing architecture with an optimal degree of time-sharing to maximize the product of compute density and weight density. It also features several circuit techniques, namely non-precharge single-ended access, hybrid compressor adder-tree circuits, and bidirectional shift-accumulation. We prototype the D6CIM macro in a 28-nm CMOS technology. The maximum operating frequency is ~360 MHz in a 1.1-V supply. The energy efficiency is measured at up to 60.4 TOPS/W. It achieves a compute density of 1.46 TOPS/mm2 and a weight density of 1005 kb/mm2. Compared to prior state-of-the-art CIM macros, the D6CIM achieves 1.3-4.8X improvement in the product of energy efficiency, compute density, and weight density.
- 24 February 2026
- Published by Institute of Electrical and Electronics Engineers (IEEE) in IEEE Transactions on Circuits and Systems I: Regular Papers
- Vol. PP (15498328) , 1-12
D6CIM: 60.4-TOPS/W All-Digital 6T-SRAM-Based Compute-in-Memory Macro Supporting 1-to-8 b Fixed-Point Arithmetic in a 28-nm CMOS
This article presents an all-digital 6T-SRAM-based compute-in-memory (D6CIM) macro that supports 1-to-8-bit fixed-point vector matrix multiplication. The D6CIM is fully digital, employing only digital standard cells and 6T SRAM bitcells. The D6CIM adopts a time-sharing architecture with an optimal degree of time-sharing to maximize the product of compute density and weight density. It also features several circuit techniques, namely non-precharge single-ended access, hybrid compressor adder-tree circuits, and bidirectional shift-accumulation. We prototype the D6CIM macro in a 28-nm CMOS technology. The maximum operating frequency is ~360 MHz in a 1.1-V supply. The energy efficiency is measured at up to 60.4 TOPS/W. It achieves a compute density of 1.46 TOPS/mm2 and a weight density of 1005 kb/mm2. Compared to prior state-of-the-art CIM macros, the D6CIM achieves 1.3-4.8X improvement in the product of energy efficiency, compute density, and weight density.
- 19 February 2026
- Published by Institute of Electrical and Electronics Engineers (IEEE) in IEEE Journal of Solid-State Circuits
- Vol. PP (00189200) , 1-10
A 7.5- μ W 35-Keyword End-to-End Keyword Spotting System With Random Augmented On-Chip Training
Fully integrated keyword spotting (KWS) systems designed for low-power operation face two major challenges. First, increasing the number of supported keywords significantly raises system complexity and power consumption. Second, most existing systems are not personalized to individual users, as they are trained on data from native English speakers, leading to degraded accuracy for non-native English speakers. In this article, we present a low-power KWS system that supports 35 keywords with a low-complexity design and enables personalized on-chip training to adapt to individual accents. The system is trained using only a single recording per keyword by applying random augmentation. A prototype implemented in 28-nm CMOS achieves an average accuracy of 91.2% on the accented speech command dataset (ASCD), which includes speech data from 20 speakers across nine different countries. The proposed system consumes $5.31~\mu $ W during inference and $7.49~\mu $ W during training.
- 19 February 2026
- Published by Institute of Electrical and Electronics Engineers (IEEE) in IEEE Journal of Solid-State Circuits
- Vol. PP (00189200) , 1-10
A 7.5- μ W 35-Keyword End-to-End Keyword Spotting System With Random Augmented On-Chip Training
Fully integrated keyword spotting (KWS) systems designed for low-power operation face two major challenges. First, increasing the number of supported keywords significantly raises system complexity and power consumption. Second, most existing systems are not personalized to individual users, as they are trained on data from native English speakers, leading to degraded accuracy for non-native English speakers. In this article, we present a low-power KWS system that supports 35 keywords with a low-complexity design and enables personalized on-chip training to adapt to individual accents. The system is trained using only a single recording per keyword by applying random augmentation. A prototype implemented in 28-nm CMOS achieves an average accuracy of 91.2% on the accented speech command dataset (ASCD), which includes speech data from 20 speakers across nine different countries. The proposed system consumes $5.31~\mu $ W during inference and $7.49~\mu $ W during training.