Hong Keun Ahn
Current Organizations / Institutes
The organizations/institutes from the scholar's latest article
Research Subjects
Hong Keun Ahn has a total of 10 co-authors. Here are the top co-authors by number of shared publications:
- Seong-Ook Jung
Yonsei University
6 shared publications - Hyun Kook Park
Thoracic and Cardiovascular Surgery, Yonsei University School of Medicine, Seoul, Korea, Republic of
2 shared publications - Sara Choi
School of Electrical and Electronic Engineering, Yonsei University, Seoul 120-749, South Korea
2 shared publications - Seung H. Kang
Qualcomm Technologies, Inc., San Diego, CA, USA
2 shared publications - Sumin Lee
School of Electrical and Electronic Engineering, Yonsei University, Seoul, South Korea
2 shared publications Byung Kyu Song
School of Electrical and Electronic Engineering, Yonsei University, Seoul 120-749, South Korea
1 shared publication
Latest Publications
- 29 December 2023
- Published by Institute of Electrical and Electronics Engineers (IEEE) in IEEE Transactions on Circuits and Systems I: Regular Papers
- Vol. 71 (3) , 1198-1207
A CNN-Based Super-Resolution Processor With Short-Term Caching for Real-Time UHD Upscaling
Super-resolution (SR) tasks, involving the restoration of low-resolution images to high-resolution images, are expected to handle larger images in the near future. This paper proposes the short-term caching (STC) layer fusion to address the increase in cache memory size for image size expansion. The proposed STC layer fusion requires only 0.2% of the memory size compared to the previous method by discarding the overlap data that must be maintained for a long time and recalculating the discarded data later. In addition, the selective-operated dual accumulator (SODA) is proposed to apply a vertically long output patch that minimizes the additional recalculation of the STC layer fusion with resolving the memory access problem. Thus, the number of processing elements and memory bandwidth are reduced by 33.6% and 35.6%, respectively. The proposed SR processor including the proposed STC layer fusion and SODA is fabricated on a 28nm CMOS process, and the die and core area occupy 12.96mm $^{2}$ and 3.45mm $^{2}$ , respectively. The SR task achieves $\times$ 2 upscaling to UHD images at 60 fps, achieving a maximum SR throughput of 497.7 Mpixels/s. Compared to previous approaches, the proposed SR processor reduces the on-chip memory size by more than 95% and achieves 4.3 times the SR throughput.
- 29 December 2023
- Published by Institute of Electrical and Electronics Engineers (IEEE) in IEEE Transactions on Circuits and Systems I: Regular Papers
- Vol. 71 (3) , 1198-1207
A CNN-Based Super-Resolution Processor With Short-Term Caching for Real-Time UHD Upscaling
Super-resolution (SR) tasks, involving the restoration of low-resolution images to high-resolution images, are expected to handle larger images in the near future. This paper proposes the short-term caching (STC) layer fusion to address the increase in cache memory size for image size expansion. The proposed STC layer fusion requires only 0.2% of the memory size compared to the previous method by discarding the overlap data that must be maintained for a long time and recalculating the discarded data later. In addition, the selective-operated dual accumulator (SODA) is proposed to apply a vertically long output patch that minimizes the additional recalculation of the STC layer fusion with resolving the memory access problem. Thus, the number of processing elements and memory bandwidth are reduced by 33.6% and 35.6%, respectively. The proposed SR processor including the proposed STC layer fusion and SODA is fabricated on a 28nm CMOS process, and the die and core area occupy 12.96mm $^{2}$ and 3.45mm $^{2}$ , respectively. The SR task achieves $\times$ 2 upscaling to UHD images at 60 fps, achieving a maximum SR throughput of 497.7 Mpixels/s. Compared to previous approaches, the proposed SR processor reduces the on-chip memory size by more than 95% and achieves 4.3 times the SR throughput.
- 7 April 2021
- Published by Institute of Electrical and Electronics Engineers (IEEE) in IEEE Transactions on Circuits and Systems I: Regular Papers
- Vol. 68 (6) , 2481-2493
Self-Referenced Single-Ended Resistance Monitoring Write Termination Scheme for STT-RAM Write Energy Reduction
Essential design requirements for a sense amplifier (SA) used in the resistance monitoring write termination (RM-WT) scheme are suggested to reduce the write energy of spin-transfer-torque random access memory (STT-RAM) while achieving a write pass yield comparable to that of a conventional write operation. In addition, a self-referenced single-ended RM-WT (SS-RM-WT) scheme is proposed. To reduce the offset voltage, a single-ended sensing circuit (SE-SC) is used in the SA. A data-aware input voltage-transfer method is also adopted in the SE-SC to maximize the input voltage difference. By adopting a capacitor between the output of the SE-SC and the input of an inverter generating a logical output used for the write termination, the conflict between maintaining and changing the output of the SE-SC is resolved. The simulation results using the industry-compatible 65-nm technology HSPICE model parameters show that the proposed SS-RM-WT scheme achieves a 44% write energy saving on average without increasing the write error rate. Area overhead is only 11.8% for a 256-kb STT-RAM array, whereas that of the previous self-referenced RM-WT schemes is up to 42.5%.
- 7 April 2021
- Published by Institute of Electrical and Electronics Engineers (IEEE) in IEEE Transactions on Circuits and Systems I: Regular Papers
- Vol. 68 (6) , 2481-2493
Self-Referenced Single-Ended Resistance Monitoring Write Termination Scheme for STT-RAM Write Energy Reduction
Essential design requirements for a sense amplifier (SA) used in the resistance monitoring write termination (RM-WT) scheme are suggested to reduce the write energy of spin-transfer-torque random access memory (STT-RAM) while achieving a write pass yield comparable to that of a conventional write operation. In addition, a self-referenced single-ended RM-WT (SS-RM-WT) scheme is proposed. To reduce the offset voltage, a single-ended sensing circuit (SE-SC) is used in the SA. A data-aware input voltage-transfer method is also adopted in the SE-SC to maximize the input voltage difference. By adopting a capacitor between the output of the SE-SC and the input of an inverter generating a logical output used for the write termination, the conflict between maintaining and changing the output of the SE-SC is resolved. The simulation results using the industry-compatible 65-nm technology HSPICE model parameters show that the proposed SS-RM-WT scheme achieves a 44% write energy saving on average without increasing the write error rate. Area overhead is only 11.8% for a 256-kb STT-RAM array, whereas that of the previous self-referenced RM-WT schemes is up to 42.5%.
- 19 October 2020
- Published by Institute of Electrical and Electronics Engineers (IEEE) in IEEE Transactions on Very Large Scale Integration (VLSI) Systems
- Vol. 28 (12) , 2650-2657
A Novel Matchline Scheduling Method for Low-Power and Reliable Search Operation in Cross-Point-Array Nonvolatile Ternary CAM
Cross-point-array nonvolatile ternary content-addressable memory (CPA nvTCAM) has recently emerged as an alternative to static random-access-memory-type TCAM, based on increased demands for high-capacity and low-power attributes. The CPA structure has various structural weaknesses such as the searchline (SL) combining with the dischargeline and the minimum line pitch of the matchline (ML). This study analyzes these weaknesses in detail for the first time and resolves the issues caused by these weaknesses using the proposed novel ML shield scheduling method with a matching probability-based flexible searching time technique (MLSS + MPFST). The proposed MLSS + MPFST resolves various issues and achieves greater than sixfold smaller cell size (8F²) than the non-CPA nvTCAM with the smallest cell size. To verify the proposed schemes, the Monte Carlo HSPICE simulations were performed using a 22-nm industry-compatible bulk FinFET model parameter in 20-nm resistive memory technology at a circuit level, and the gem5 simulations were performed at a system level. The simulation results indicated that the CPA nvTCAM with the proposed MLSS + MPFST achieved a comparable search operation time of 1 ns and a slight power consumption overhead of 10% but an acceptable system performance overhead of less than 0.6% by resolving the various issues, compared with the non-CPA nvTCAM having the best performance.
- 19 October 2020
- Published by Institute of Electrical and Electronics Engineers (IEEE) in IEEE Transactions on Very Large Scale Integration (VLSI) Systems
- Vol. 28 (12) , 2650-2657
A Novel Matchline Scheduling Method for Low-Power and Reliable Search Operation in Cross-Point-Array Nonvolatile Ternary CAM
Cross-point-array nonvolatile ternary content-addressable memory (CPA nvTCAM) has recently emerged as an alternative to static random-access-memory-type TCAM, based on increased demands for high-capacity and low-power attributes. The CPA structure has various structural weaknesses such as the searchline (SL) combining with the dischargeline and the minimum line pitch of the matchline (ML). This study analyzes these weaknesses in detail for the first time and resolves the issues caused by these weaknesses using the proposed novel ML shield scheduling method with a matching probability-based flexible searching time technique (MLSS + MPFST). The proposed MLSS + MPFST resolves various issues and achieves greater than sixfold smaller cell size (8F²) than the non-CPA nvTCAM with the smallest cell size. To verify the proposed schemes, the Monte Carlo HSPICE simulations were performed using a 22-nm industry-compatible bulk FinFET model parameter in 20-nm resistive memory technology at a circuit level, and the gem5 simulations were performed at a system level. The simulation results indicated that the CPA nvTCAM with the proposed MLSS + MPFST achieved a comparable search operation time of 1 ns and a slight power consumption overhead of 10% but an acceptable system performance overhead of less than 0.6% by resolving the various issues, compared with the non-CPA nvTCAM having the best performance.
- Open Access
- 1 January 2020
- Published by Institute of Electrical and Electronics Engineers (IEEE) in IEEE Access
- Vol. 8, 187754-187765
CNN Acceleration With Hardware-Efficient Dataflow for Super-Resolution
- 1 January 2020
- Published by Institute of Electrical and Electronics Engineers (IEEE) in IEEE Access
- Vol. 8, 187754-187765
Open AccessCNN Acceleration With Hardware-Efficient Dataflow for Super-Resolution
- 18 April 2019
- Published by Institute of Electrical and Electronics Engineers (IEEE) in IEEE Transactions on Nanotechnology
- Vol. 18 (1536125X) , 421-431
Thermoelectric Cooling Read for Resolving Read Disturb With Inrush Current Issue in OTS-PRAM
Phase change random access memory with ovonic threshold switch (OTS-PRAM) has been recently highlighted as an alternative to dynamic random access memory (DRAM) to meet the demand for memory capacity expansion in multicore processing. However, read disturb with inrush current (RDI) occurs owing to the snapback phenomenon caused by the threshold switching of the OTS selector in the OTS-PRAM cell, which should be resolved for realizing the practical use of OTS-PRAM. In this study, the RDI is analyzed in detail for the first time. Based on this analysis, a novel thermoelectric cooling read technique (TCR) that utilizes the thermoelectric effect in the storage material of the OTS-PRAM is proposed for effectively resolving the RDI issue. The proposed TCR is verified using the Sentaurus TCAD simulator, and a core structure with the proposed TCR is evaluated. HSPICE simulation results with a memory industry-compatible 0.25-μm core circuit design rule for 20-nm PRAM technology show that the proposed TCR has a power gain of 67%, system performance improvement of 13.9%, and slight area overhead of 3% as compared to the SET-back write technique, which was the most promising technique for resolving the RDI issue until now.
- 18 April 2019
- Published by Institute of Electrical and Electronics Engineers (IEEE) in IEEE Transactions on Nanotechnology
- Vol. 18 (1536125X) , 421-431
Thermoelectric Cooling Read for Resolving Read Disturb With Inrush Current Issue in OTS-PRAM
Phase change random access memory with ovonic threshold switch (OTS-PRAM) has been recently highlighted as an alternative to dynamic random access memory (DRAM) to meet the demand for memory capacity expansion in multicore processing. However, read disturb with inrush current (RDI) occurs owing to the snapback phenomenon caused by the threshold switching of the OTS selector in the OTS-PRAM cell, which should be resolved for realizing the practical use of OTS-PRAM. In this study, the RDI is analyzed in detail for the first time. Based on this analysis, a novel thermoelectric cooling read technique (TCR) that utilizes the thermoelectric effect in the storage material of the OTS-PRAM is proposed for effectively resolving the RDI issue. The proposed TCR is verified using the Sentaurus TCAD simulator, and a core structure with the proposed TCR is evaluated. HSPICE simulation results with a memory industry-compatible 0.25-μm core circuit design rule for 20-nm PRAM technology show that the proposed TCR has a power gain of 67%, system performance improvement of 13.9%, and slight area overhead of 3% as compared to the SET-back write technique, which was the most promising technique for resolving the RDI issue until now.
- 27 November 2018
- Published by Institute of Electrical and Electronics Engineers (IEEE) in IEEE Transactions on Very Large Scale Integration (VLSI) Systems
- Vol. 27 (2) , 387-397
A Decoder for Short BCH Codes With High Decoding Efficiency and Low Power for Emerging Memories
In this paper, a double-error-correcting and triple-error-detecting (DEC-TED) Bose-Chaudhuri-Hocquenghem (BCH) code decoder with high decoding efficiency and low power for error correction in emerging memories is presented. To increase the decoding efficiency, we propose an adaptive error correction technique for the DEC-TED BCH code that detects the number of errors in a codeword immediately after syndrome generation and applies a different error correction algorithm depending on the error conditions. With the adaptive error correction technique, the average decoding latency and power consumption are significantly reduced owing to the increased decoding efficiency. To further reduce the power consumption, an invalid-transition-inhibition technique is proposed to remove the invalid transitions caused by glitches of syndrome vectors in the error-finding block. Synthesis results with an industry-compatible 65-nm technology library show that the proposed decoders for the (79, 64, 6) BCH code take only 37%-48% average decoding latency and achieve more than 70% power reduction compared to the conventional fully parallel decoder under the 10⁻⁴-10⁻² raw bit-error rate.
- 27 November 2018
- Published by Institute of Electrical and Electronics Engineers (IEEE) in IEEE Transactions on Very Large Scale Integration (VLSI) Systems
- Vol. 27 (2) , 387-397
A Decoder for Short BCH Codes With High Decoding Efficiency and Low Power for Emerging Memories
In this paper, a double-error-correcting and triple-error-detecting (DEC-TED) Bose-Chaudhuri-Hocquenghem (BCH) code decoder with high decoding efficiency and low power for error correction in emerging memories is presented. To increase the decoding efficiency, we propose an adaptive error correction technique for the DEC-TED BCH code that detects the number of errors in a codeword immediately after syndrome generation and applies a different error correction algorithm depending on the error conditions. With the adaptive error correction technique, the average decoding latency and power consumption are significantly reduced owing to the increased decoding efficiency. To further reduce the power consumption, an invalid-transition-inhibition technique is proposed to remove the invalid transitions caused by glitches of syndrome vectors in the error-finding block. Synthesis results with an industry-compatible 65-nm technology library show that the proposed decoders for the (79, 64, 6) BCH code take only 37%-48% average decoding latency and achieve more than 70% power reduction compared to the conventional fully parallel decoder under the 10⁻⁴-10⁻² raw bit-error rate.