Pub Date : 2020-12-12DOI: 10.1109/IEDM13553.2020.9371952
C. I, Yingchao Lin, Guizhen Wang
Facing the severe power consumption and energy efficiency challenges in 5G era, a novel DPD solution enabled by deep learning and big data is proposed. This is a flexible system suitable for various wireless network architectures and diverse application scenarios. The architecture, mechanism and deployment strategy along with its advantages are presented. Preliminary validation and analyses are also illustrated for the feasibility.
{"title":"A Deep Learning Enabled Universal DPD System","authors":"C. I, Yingchao Lin, Guizhen Wang","doi":"10.1109/IEDM13553.2020.9371952","DOIUrl":"https://doi.org/10.1109/IEDM13553.2020.9371952","url":null,"abstract":"Facing the severe power consumption and energy efficiency challenges in 5G era, a novel DPD solution enabled by deep learning and big data is proposed. This is a flexible system suitable for various wireless network architectures and diverse application scenarios. The architecture, mechanism and deployment strategy along with its advantages are presented. Preliminary validation and analyses are also illustrated for the feasibility.","PeriodicalId":415186,"journal":{"name":"2020 IEEE International Electron Devices Meeting (IEDM)","volume":"57 1","pages":"0"},"PeriodicalIF":0.0,"publicationDate":"2020-12-12","publicationTypes":"Journal Article","fieldsOfStudy":null,"isOpenAccess":false,"openAccessPdf":"","citationCount":null,"resultStr":null,"platform":"Semanticscholar","paperid":"116357375","PeriodicalName":null,"FirstCategoryId":null,"ListUrlMain":null,"RegionNum":0,"RegionCategory":"","ArticlePicture":[],"TitleCN":null,"AbstractTextCN":null,"PMCID":"","EPubDate":null,"PubModel":null,"JCR":null,"JCRName":null,"Score":null,"Total":0}
Pub Date : 2020-12-12DOI: 10.1109/IEDM13553.2020.9372128
Hongyan Zhou, Shibin Zhang, Zhongxu Li, Kai Huang, Pengcheng Zheng, Jinbo Wu, Chen Shen, Liping Zhang, T. You, Lianghui Zhang, Kang Liu, Huarui Sun, Hongtao Xu, Xiaomeng Zhao, X. Ou
We demonstrate groups of surface wave (SH0 mode) and Lamb wave (S0 mode) acoustic devices on lithium niobate thin films on silicon carbide (LNOSiC) heterogeneous substrate. The 4-inch LNOSiC with an excellent thermal transport property is prepared by ion-cutting process. The fabricated acoustic resonators on the LNOSiC substrate show scalable resonances from 2.0 to 4.72 GHz, in which the SH0 (S0) mode resonator shows a $k_t^2$ of 24.1% (15.5%) and a maximum Bode-Q of 976 (577) at 2.54 (3.56) GHz. Moreover, the phase velocity (Vp) of the SH0 (S0) mode is greater than 5000 (6400) m/s, which is about 1.25 (1.6) times higher than that of the conventional SAWs, so as the operating frequency. The filter with a center frequency of 2.62 GHz, an insertion loss (IL) of 1.06 dB, and a 3-dB fractional bandwidth (FBW) of 12.6% (three times larger than that of the conventional SAWs) is also achieved. The acoustic devices on heterogeneous substrate are very promising for high frequency, wideband and high power 5G front-ends.
{"title":"Surface Wave and Lamb Wave Acoustic Devices on Heterogenous Substrate for 5G Front-Ends","authors":"Hongyan Zhou, Shibin Zhang, Zhongxu Li, Kai Huang, Pengcheng Zheng, Jinbo Wu, Chen Shen, Liping Zhang, T. You, Lianghui Zhang, Kang Liu, Huarui Sun, Hongtao Xu, Xiaomeng Zhao, X. Ou","doi":"10.1109/IEDM13553.2020.9372128","DOIUrl":"https://doi.org/10.1109/IEDM13553.2020.9372128","url":null,"abstract":"We demonstrate groups of surface wave (SH0 mode) and Lamb wave (S0 mode) acoustic devices on lithium niobate thin films on silicon carbide (LNOSiC) heterogeneous substrate. The 4-inch LNOSiC with an excellent thermal transport property is prepared by ion-cutting process. The fabricated acoustic resonators on the LNOSiC substrate show scalable resonances from 2.0 to 4.72 GHz, in which the SH0 (S0) mode resonator shows a $k_t^2$ of 24.1% (15.5%) and a maximum Bode-Q of 976 (577) at 2.54 (3.56) GHz. Moreover, the phase velocity (Vp) of the SH0 (S0) mode is greater than 5000 (6400) m/s, which is about 1.25 (1.6) times higher than that of the conventional SAWs, so as the operating frequency. The filter with a center frequency of 2.62 GHz, an insertion loss (IL) of 1.06 dB, and a 3-dB fractional bandwidth (FBW) of 12.6% (three times larger than that of the conventional SAWs) is also achieved. The acoustic devices on heterogeneous substrate are very promising for high frequency, wideband and high power 5G front-ends.","PeriodicalId":415186,"journal":{"name":"2020 IEEE International Electron Devices Meeting (IEDM)","volume":"65 1","pages":"0"},"PeriodicalIF":0.0,"publicationDate":"2020-12-12","publicationTypes":"Journal Article","fieldsOfStudy":null,"isOpenAccess":false,"openAccessPdf":"","citationCount":null,"resultStr":null,"platform":"Semanticscholar","paperid":"125443901","PeriodicalName":null,"FirstCategoryId":null,"ListUrlMain":null,"RegionNum":0,"RegionCategory":"","ArticlePicture":[],"TitleCN":null,"AbstractTextCN":null,"PMCID":"","EPubDate":null,"PubModel":null,"JCR":null,"JCRName":null,"Score":null,"Total":0}
Pub Date : 2020-12-12DOI: 10.1109/IEDM13553.2020.9371940
Abhishek A. Sharma, B. Doyle, H. Yoo, I. Tung, J. Kavalieros, M. Metz, M. Reshotko, P. Majhi, Tobias L. Brown-Heft, Yu-Jin Chen, V. Le
Scaled ferroelectric transistors (Lg =76 nm) in a back- gated configuration are fabricated with a channel-last process flow. Using this approach, optimization of the ferroelectric gate oxide film can be decoupled from that of the semiconductor channel to reduce parasitic interfaces. As a result, ferroelectric transistors with 3σ memory window for fast programming time of 10 ns (including an instantaneous read-after-write) at 1.8 V and high endurance of 1012 cycles are demonstrated for the first time.
{"title":"High Speed Memory Operation in Channel-Last, Back-gated Ferroelectric Transistors","authors":"Abhishek A. Sharma, B. Doyle, H. Yoo, I. Tung, J. Kavalieros, M. Metz, M. Reshotko, P. Majhi, Tobias L. Brown-Heft, Yu-Jin Chen, V. Le","doi":"10.1109/IEDM13553.2020.9371940","DOIUrl":"https://doi.org/10.1109/IEDM13553.2020.9371940","url":null,"abstract":"Scaled ferroelectric transistors (Lg =76 nm) in a back- gated configuration are fabricated with a channel-last process flow. Using this approach, optimization of the ferroelectric gate oxide film can be decoupled from that of the semiconductor channel to reduce parasitic interfaces. As a result, ferroelectric transistors with 3σ memory window for fast programming time of 10 ns (including an instantaneous read-after-write) at 1.8 V and high endurance of 1012 cycles are demonstrated for the first time.","PeriodicalId":415186,"journal":{"name":"2020 IEEE International Electron Devices Meeting (IEDM)","volume":"177 1","pages":"0"},"PeriodicalIF":0.0,"publicationDate":"2020-12-12","publicationTypes":"Journal Article","fieldsOfStudy":null,"isOpenAccess":false,"openAccessPdf":"","citationCount":null,"resultStr":null,"platform":"Semanticscholar","paperid":"124396866","PeriodicalName":null,"FirstCategoryId":null,"ListUrlMain":null,"RegionNum":0,"RegionCategory":"","ArticlePicture":[],"TitleCN":null,"AbstractTextCN":null,"PMCID":"","EPubDate":null,"PubModel":null,"JCR":null,"JCRName":null,"Score":null,"Total":0}
Pub Date : 2020-12-12DOI: 10.1109/IEDM13553.2020.9372063
L. Capua, S. Sheibani, S. Kamaei, J. Zhang, A. Ionescu
In this work we report the fabrication, characterization and validation of a cortisol biosensor, together with a unified predictive calibrated model. We demonstrated the possibility of using a classical submicron semiconductor FET as the transducer for a cortisol biosensor, extending its gate with a graphene on platinum electrode decorated with cortisol specific aptamers. The sensor outperforms the so far any reported cortisol sensors, in terms of performance and integration capability: (i) we report sensor validation over 4 orders of concertation (1 nM - 10 μM, matching human sweat concentration range), (ii) with excellent voltage (14.7 mV/dec.) and current (80% relative change with respect baseline) sensitivity, (iii) low drift, smaller than 10 mV/h, (iv) low power consumption (sub-nW DC power), (v) record low detection limit (LOD) for cortisol of 0.2nM, and (vi) selectivity over other hormones such as testosterone. Moreover, we have developed and validated the first unified compact analog predictive calibrated model for cortisol FET sensors based on experimental data, valid from weak to strong inversion, and able to capture the output current dependence on hormone concentrations. In addition, this model is accurate in the prediction of ID, gm and transconductance efficiency, ID/gm, enabling simulation and optimization of analog design readout, together with power and signal-to-noise ratio trade-offs.
{"title":"Extended-Gate FET cortisol sensor for stress disorders based on aptamers-decorated graphene electrode: Fabrication, Experiments and Unified Analog Predictive Modeling","authors":"L. Capua, S. Sheibani, S. Kamaei, J. Zhang, A. Ionescu","doi":"10.1109/IEDM13553.2020.9372063","DOIUrl":"https://doi.org/10.1109/IEDM13553.2020.9372063","url":null,"abstract":"In this work we report the fabrication, characterization and validation of a cortisol biosensor, together with a unified predictive calibrated model. We demonstrated the possibility of using a classical submicron semiconductor FET as the transducer for a cortisol biosensor, extending its gate with a graphene on platinum electrode decorated with cortisol specific aptamers. The sensor outperforms the so far any reported cortisol sensors, in terms of performance and integration capability: (i) we report sensor validation over 4 orders of concertation (1 nM - 10 μM, matching human sweat concentration range), (ii) with excellent voltage (14.7 mV/dec.) and current (80% relative change with respect baseline) sensitivity, (iii) low drift, smaller than 10 mV/h, (iv) low power consumption (sub-nW DC power), (v) record low detection limit (LOD) for cortisol of 0.2nM, and (vi) selectivity over other hormones such as testosterone. Moreover, we have developed and validated the first unified compact analog predictive calibrated model for cortisol FET sensors based on experimental data, valid from weak to strong inversion, and able to capture the output current dependence on hormone concentrations. In addition, this model is accurate in the prediction of ID, gm and transconductance efficiency, ID/gm, enabling simulation and optimization of analog design readout, together with power and signal-to-noise ratio trade-offs.","PeriodicalId":415186,"journal":{"name":"2020 IEEE International Electron Devices Meeting (IEDM)","volume":"24 1","pages":"0"},"PeriodicalIF":0.0,"publicationDate":"2020-12-12","publicationTypes":"Journal Article","fieldsOfStudy":null,"isOpenAccess":false,"openAccessPdf":"","citationCount":null,"resultStr":null,"platform":"Semanticscholar","paperid":"127777584","PeriodicalName":null,"FirstCategoryId":null,"ListUrlMain":null,"RegionNum":0,"RegionCategory":"","ArticlePicture":[],"TitleCN":null,"AbstractTextCN":null,"PMCID":"","EPubDate":null,"PubModel":null,"JCR":null,"JCRName":null,"Score":null,"Total":0}
Pub Date : 2020-12-12DOI: 10.1109/IEDM13553.2020.9372016
R. Siddique, L. Liedtke, H. Park, S. Y. Lee, H. Raniwala, D. Y. Park, D. Lim, H. Choo
Glaucoma, one of the leading cause of irreversible blindness, is largely caused by an elevated intraocular pressure (IOP). However, current IOP monitoring techniques inherit major disadvantages such as imprecision, no real or long time monitoring, and difficult readout. Here, we report on a highly miniaturized (200 um thick) optomechanical nanophotonic sensor implant for long-term, continuous and on-demand IOP monitoring. This IOP sensor is made of a flexible 3D hybrid photonic crystals (HPC) that functions as a pressure-sensitive optical resonator (0.1 nm/mm Hg) and delivers IOP readings when interrogated with near-infrared light with an average accuracy of 0.56 mm Hg over the range of 0–40 mm Hg. A new fabrication process is developed using colloidal self-assembly leading to a single step formation of hybrid periodic and amorphous layers exploiting the inverse process of a drying "coffee-stain" effect. The HPC results in a wide-angle strong resonance of ±40° ensuring an easy and accurate remote and long readout distance. 8 sensors were mounted inside the anterior chamber in New Zealand white rabbits and provided continuous, accurate measurements of IOP with handheld detector for up to 6 months with no signs of inflammation.
青光眼是不可逆失明的主要原因之一,主要由眼压升高引起。然而,目前的IOP监测技术存在着不精确、不能实时监测或长时间监测、读数困难等主要缺点。在这里,我们报道了一种高度小型化(200um厚)的光机械纳米光子传感器植入物,用于长期、连续和按需监测IOP。该IOP传感器由柔性3D混合光子晶体(HPC)制成,可作为压敏光学谐振器(0.1 nm/mm Hg),并在近红外光下提供IOP读数,在0-40 mm Hg范围内平均精度为0.56 mm Hg。利用胶体自组装技术开发了一种新的制造工艺,利用干燥“咖啡渍”效应的逆过程,可一步形成混合周期性和非晶态层。HPC产生±40°的广角强共振,确保轻松准确的远程和长读数距离。在新西兰大白兔的前房内安装8个传感器,用手持探测器连续、准确地测量IOP长达6个月,无炎症迹象。
{"title":"Nanophotonic sensor implants with 3D hybrid periodic-amorphous photonic crystals for wide-angle monitoring of long-term in-vivo intraocular pressure","authors":"R. Siddique, L. Liedtke, H. Park, S. Y. Lee, H. Raniwala, D. Y. Park, D. Lim, H. Choo","doi":"10.1109/IEDM13553.2020.9372016","DOIUrl":"https://doi.org/10.1109/IEDM13553.2020.9372016","url":null,"abstract":"Glaucoma, one of the leading cause of irreversible blindness, is largely caused by an elevated intraocular pressure (IOP). However, current IOP monitoring techniques inherit major disadvantages such as imprecision, no real or long time monitoring, and difficult readout. Here, we report on a highly miniaturized (200 um thick) optomechanical nanophotonic sensor implant for long-term, continuous and on-demand IOP monitoring. This IOP sensor is made of a flexible 3D hybrid photonic crystals (HPC) that functions as a pressure-sensitive optical resonator (0.1 nm/mm Hg) and delivers IOP readings when interrogated with near-infrared light with an average accuracy of 0.56 mm Hg over the range of 0–40 mm Hg. A new fabrication process is developed using colloidal self-assembly leading to a single step formation of hybrid periodic and amorphous layers exploiting the inverse process of a drying \"coffee-stain\" effect. The HPC results in a wide-angle strong resonance of ±40° ensuring an easy and accurate remote and long readout distance. 8 sensors were mounted inside the anterior chamber in New Zealand white rabbits and provided continuous, accurate measurements of IOP with handheld detector for up to 6 months with no signs of inflammation.","PeriodicalId":415186,"journal":{"name":"2020 IEEE International Electron Devices Meeting (IEDM)","volume":"10 1","pages":"0"},"PeriodicalIF":0.0,"publicationDate":"2020-12-12","publicationTypes":"Journal Article","fieldsOfStudy":null,"isOpenAccess":false,"openAccessPdf":"","citationCount":null,"resultStr":null,"platform":"Semanticscholar","paperid":"129023026","PeriodicalName":null,"FirstCategoryId":null,"ListUrlMain":null,"RegionNum":0,"RegionCategory":"","ArticlePicture":[],"TitleCN":null,"AbstractTextCN":null,"PMCID":"","EPubDate":null,"PubModel":null,"JCR":null,"JCRName":null,"Score":null,"Total":0}
Pub Date : 2020-12-12DOI: 10.1109/IEDM13553.2020.9372013
Taekyong Kim, J. D. del Alamo, D. Antoniadis
We have carried out a detailed experimental study of the switching dynamics of HfZrO2 Metal-Ferroelectric-Metal (MFM) and Metal-Ferroelectric-Insulator-Metal (MFIM) structures. In order to extract the intrinsic dynamic response, our experimental methodology has paid close attention to minimizing and calibrating all circuit and sample parasitics. In MFM structures, we have found no evidence of negative capacitance (NC) effect. A new dynamic model based on the multi-domain Preisach model describes well all observed behavior including major and minor charge-voltage loops. Our study also reveals the crucial role that parasitics play in the observed device dynamics and can explain claims of NC effects in MFM structures in the literature. In our MFIM structures, we observe clear NC behavior. We not only confirm the transient quasi-static S-like ferroelectric (FE) charge-field behavior described in the literature, but for the first time, we report a dynamic response that displays a hysteretic behavior in the NC region. A model based on the Landau-Khalatnikov (L-K) equation that incorporates FE dynamics via a phenomenological frictional resistance adequately describes the observed results when that resistance is made dependent on the direction of the voltage drive vs. time. Mitigation of this hysteretic NC behavior will be crucial for the harnessing of NC in practical MOSFETs.
{"title":"Dynamics of HfZrO2 Ferroelectric Structures: Experiments and Models","authors":"Taekyong Kim, J. D. del Alamo, D. Antoniadis","doi":"10.1109/IEDM13553.2020.9372013","DOIUrl":"https://doi.org/10.1109/IEDM13553.2020.9372013","url":null,"abstract":"We have carried out a detailed experimental study of the switching dynamics of HfZrO2 Metal-Ferroelectric-Metal (MFM) and Metal-Ferroelectric-Insulator-Metal (MFIM) structures. In order to extract the intrinsic dynamic response, our experimental methodology has paid close attention to minimizing and calibrating all circuit and sample parasitics. In MFM structures, we have found no evidence of negative capacitance (NC) effect. A new dynamic model based on the multi-domain Preisach model describes well all observed behavior including major and minor charge-voltage loops. Our study also reveals the crucial role that parasitics play in the observed device dynamics and can explain claims of NC effects in MFM structures in the literature. In our MFIM structures, we observe clear NC behavior. We not only confirm the transient quasi-static S-like ferroelectric (FE) charge-field behavior described in the literature, but for the first time, we report a dynamic response that displays a hysteretic behavior in the NC region. A model based on the Landau-Khalatnikov (L-K) equation that incorporates FE dynamics via a phenomenological frictional resistance adequately describes the observed results when that resistance is made dependent on the direction of the voltage drive vs. time. Mitigation of this hysteretic NC behavior will be crucial for the harnessing of NC in practical MOSFETs.","PeriodicalId":415186,"journal":{"name":"2020 IEEE International Electron Devices Meeting (IEDM)","volume":"139 1","pages":"0"},"PeriodicalIF":0.0,"publicationDate":"2020-12-12","publicationTypes":"Journal Article","fieldsOfStudy":null,"isOpenAccess":false,"openAccessPdf":"","citationCount":null,"resultStr":null,"platform":"Semanticscholar","paperid":"132671365","PeriodicalName":null,"FirstCategoryId":null,"ListUrlMain":null,"RegionNum":0,"RegionCategory":"","ArticlePicture":[],"TitleCN":null,"AbstractTextCN":null,"PMCID":"","EPubDate":null,"PubModel":null,"JCR":null,"JCRName":null,"Score":null,"Total":0}
Pub Date : 2020-12-12DOI: 10.1109/IEDM13553.2020.9372027
I-Hsiang Wang, P. Hong, K. Peng, Horng-Chih Lin, T. George, Pei-Wen Li
Starting with our remarkable discovery of spherical germanium (Ge) quantum dot (QD) formation, we have embarked on an exciting journey of further discovery, all the while maintaining CMOS-compatible processes. We have taken advantage of the many peculiar and symbiotic interactions of Si, Ge and O interstitials to create a novel portfolio of electronic, photonic and quantum computing devices. This paper summarizes several of these completely new and counter-intuitive accomplishments. Using a coordinated combination of lithographic patterning and self-assembly, size-tunable spherical Ge QDs were controllably placed at designated spatial locations within Si-containing layers. We exploited the exquisite control available through the thermal oxidation of Si1-xGex patterned structures in proximity to Si3N4/Si layers. Our so-called "designer" Ge QDs have succeeded in opening up myriad device possibilities, including paired QDs for qubits, single-hole transistors (SHTs) for charge sensing, photodetectors and light-emitters for Si photonics, and junctionless (JL) FETs using standard Si processing.
{"title":"The Wonderful World of Designer Ge Quantum Dots","authors":"I-Hsiang Wang, P. Hong, K. Peng, Horng-Chih Lin, T. George, Pei-Wen Li","doi":"10.1109/IEDM13553.2020.9372027","DOIUrl":"https://doi.org/10.1109/IEDM13553.2020.9372027","url":null,"abstract":"Starting with our remarkable discovery of spherical germanium (Ge) quantum dot (QD) formation, we have embarked on an exciting journey of further discovery, all the while maintaining CMOS-compatible processes. We have taken advantage of the many peculiar and symbiotic interactions of Si, Ge and O interstitials to create a novel portfolio of electronic, photonic and quantum computing devices. This paper summarizes several of these completely new and counter-intuitive accomplishments. Using a coordinated combination of lithographic patterning and self-assembly, size-tunable spherical Ge QDs were controllably placed at designated spatial locations within Si-containing layers. We exploited the exquisite control available through the thermal oxidation of Si1-xGex patterned structures in proximity to Si3N4/Si layers. Our so-called \"designer\" Ge QDs have succeeded in opening up myriad device possibilities, including paired QDs for qubits, single-hole transistors (SHTs) for charge sensing, photodetectors and light-emitters for Si photonics, and junctionless (JL) FETs using standard Si processing.","PeriodicalId":415186,"journal":{"name":"2020 IEEE International Electron Devices Meeting (IEDM)","volume":"3 1","pages":"0"},"PeriodicalIF":0.0,"publicationDate":"2020-12-12","publicationTypes":"Journal Article","fieldsOfStudy":null,"isOpenAccess":false,"openAccessPdf":"","citationCount":null,"resultStr":null,"platform":"Semanticscholar","paperid":"130693000","PeriodicalName":null,"FirstCategoryId":null,"ListUrlMain":null,"RegionNum":0,"RegionCategory":"","ArticlePicture":[],"TitleCN":null,"AbstractTextCN":null,"PMCID":"","EPubDate":null,"PubModel":null,"JCR":null,"JCRName":null,"Score":null,"Total":0}
Pub Date : 2020-12-12DOI: 10.1109/IEDM13553.2020.9372055
D. Lin, Xiangyu Wu, D. Cott, D. Verreck, B. Groven, S. Sergeant, Q. Smets, S. Sutar, I. Asselberghs, I. Radu
We have engineered dual gate WS2 transistors with scaled top and back gate stacks based on a surface physisorption ALD approach for advanced logic applications. Connected dual gate MOSFET operation with a 2ML WS2 channel reaches 210μA/um drain current and 2.7μF/cm2 capacitance (>3.4×1013/cm2 sheet charge density) at 3V gate bias, with >108 on-off ratio, 120μS/um max. transconductance and 109mV/dec sub-threshold swing at 100nm Lch. This dual gate design enables us to explore EOT scaling, ambipolar I-V and C-V(capacitance-voltage) response on CVD WS2 channel.
{"title":"Dual gate synthetic WS2 MOSFETs with 120μS/μm Gm 2.7μF/cm2 capacitance and ambipolar channel","authors":"D. Lin, Xiangyu Wu, D. Cott, D. Verreck, B. Groven, S. Sergeant, Q. Smets, S. Sutar, I. Asselberghs, I. Radu","doi":"10.1109/IEDM13553.2020.9372055","DOIUrl":"https://doi.org/10.1109/IEDM13553.2020.9372055","url":null,"abstract":"We have engineered dual gate WS2 transistors with scaled top and back gate stacks based on a surface physisorption ALD approach for advanced logic applications. Connected dual gate MOSFET operation with a 2ML WS2 channel reaches 210μA/um drain current and 2.7μF/cm2 capacitance (>3.4×1013/cm2 sheet charge density) at 3V gate bias, with >108 on-off ratio, 120μS/um max. transconductance and 109mV/dec sub-threshold swing at 100nm Lch. This dual gate design enables us to explore EOT scaling, ambipolar I-V and C-V(capacitance-voltage) response on CVD WS2 channel.","PeriodicalId":415186,"journal":{"name":"2020 IEEE International Electron Devices Meeting (IEDM)","volume":"17 1","pages":"0"},"PeriodicalIF":0.0,"publicationDate":"2020-12-12","publicationTypes":"Journal Article","fieldsOfStudy":null,"isOpenAccess":false,"openAccessPdf":"","citationCount":null,"resultStr":null,"platform":"Semanticscholar","paperid":"130706409","PeriodicalName":null,"FirstCategoryId":null,"ListUrlMain":null,"RegionNum":0,"RegionCategory":"","ArticlePicture":[],"TitleCN":null,"AbstractTextCN":null,"PMCID":"","EPubDate":null,"PubModel":null,"JCR":null,"JCRName":null,"Score":null,"Total":0}
Pub Date : 2020-12-12DOI: 10.1109/IEDM13553.2020.9372124
T. Soliman, F. Müller, T. Kirchner, T. Hoffmann, H. Ganem, E. Karimov, T. Ali, M. Lederer, C. Sudarshan, T. Kämpfe, A. Guntoro, N. Wehn
This paper presents an efficient crossbar design and implementation intended for analog compute-in-memory (ACiM) acceleration of artificial neural networks based on ferroelectric FET (FeFET) technology. The novel mixed signal blocks presented in this work reduce the device-to-device variation and are optimized for low area, low power and high throughput. In addition, we illustrate the operation and programmability of the crossbar that adopts bit decomposition techniques for MAC operation. Our crossbar based ACiM accelerator achieves a record peak performance of 13714 TOPS/W.
{"title":"Ultra-Low Power Flexible Precision FeFET Based Analog In-Memory Computing","authors":"T. Soliman, F. Müller, T. Kirchner, T. Hoffmann, H. Ganem, E. Karimov, T. Ali, M. Lederer, C. Sudarshan, T. Kämpfe, A. Guntoro, N. Wehn","doi":"10.1109/IEDM13553.2020.9372124","DOIUrl":"https://doi.org/10.1109/IEDM13553.2020.9372124","url":null,"abstract":"This paper presents an efficient crossbar design and implementation intended for analog compute-in-memory (ACiM) acceleration of artificial neural networks based on ferroelectric FET (FeFET) technology. The novel mixed signal blocks presented in this work reduce the device-to-device variation and are optimized for low area, low power and high throughput. In addition, we illustrate the operation and programmability of the crossbar that adopts bit decomposition techniques for MAC operation. Our crossbar based ACiM accelerator achieves a record peak performance of 13714 TOPS/W.","PeriodicalId":415186,"journal":{"name":"2020 IEEE International Electron Devices Meeting (IEDM)","volume":"46 1","pages":"0"},"PeriodicalIF":0.0,"publicationDate":"2020-12-12","publicationTypes":"Journal Article","fieldsOfStudy":null,"isOpenAccess":false,"openAccessPdf":"","citationCount":null,"resultStr":null,"platform":"Semanticscholar","paperid":"116581330","PeriodicalName":null,"FirstCategoryId":null,"ListUrlMain":null,"RegionNum":0,"RegionCategory":"","ArticlePicture":[],"TitleCN":null,"AbstractTextCN":null,"PMCID":"","EPubDate":null,"PubModel":null,"JCR":null,"JCRName":null,"Score":null,"Total":0}
Pub Date : 2020-12-12DOI: 10.1109/IEDM13553.2020.9372040
S. H. Han, J. Lee, H. Shin, J. Lee, K. Suh, K. Nam, B. Kwon, M. Cho, J. Lee, J. Jeong, J. Park, S. C. Oh, S. O. Park, S. Hwang, S. Pyo, H. Jung, Y. Ji, J. Bak, D. S. Kim, W. S. Ham, Y. Kim, K. Lee, Y. J. Song, G. Koh, Y. Hong, G. Jeong
We present the world-first demonstration of 28-nm embedded MRAM (eMRAM) for frame buffer memory, highlighting the smallest macro size (0.08 mm2/Mb) reported to date. Compared to SRAM that is commonly used for frame buffer memory, eMRAM provides 47% area saving. For frame buffer applications, read disturbance and endurance are the most critical reliability considerations. With magnetic tunnel junction process improvements, we have verified sufficient read disturbance margins and met the endurance requirement (> 1E10 cycles) which corresponds to 10-year continuous usage. Compared to flash-type eMRAM, we have achieved 40% switching current reduction with < 50ns read/write speed.
{"title":"28-nm 0.08 mm2/Mb Embedded MRAM for Frame Buffer Memory","authors":"S. H. Han, J. Lee, H. Shin, J. Lee, K. Suh, K. Nam, B. Kwon, M. Cho, J. Lee, J. Jeong, J. Park, S. C. Oh, S. O. Park, S. Hwang, S. Pyo, H. Jung, Y. Ji, J. Bak, D. S. Kim, W. S. Ham, Y. Kim, K. Lee, Y. J. Song, G. Koh, Y. Hong, G. Jeong","doi":"10.1109/IEDM13553.2020.9372040","DOIUrl":"https://doi.org/10.1109/IEDM13553.2020.9372040","url":null,"abstract":"We present the world-first demonstration of 28-nm embedded MRAM (eMRAM) for frame buffer memory, highlighting the smallest macro size (0.08 mm2/Mb) reported to date. Compared to SRAM that is commonly used for frame buffer memory, eMRAM provides 47% area saving. For frame buffer applications, read disturbance and endurance are the most critical reliability considerations. With magnetic tunnel junction process improvements, we have verified sufficient read disturbance margins and met the endurance requirement (> 1E10 cycles) which corresponds to 10-year continuous usage. Compared to flash-type eMRAM, we have achieved 40% switching current reduction with < 50ns read/write speed.","PeriodicalId":415186,"journal":{"name":"2020 IEEE International Electron Devices Meeting (IEDM)","volume":"7 1","pages":"0"},"PeriodicalIF":0.0,"publicationDate":"2020-12-12","publicationTypes":"Journal Article","fieldsOfStudy":null,"isOpenAccess":false,"openAccessPdf":"","citationCount":null,"resultStr":null,"platform":"Semanticscholar","paperid":"131532145","PeriodicalName":null,"FirstCategoryId":null,"ListUrlMain":null,"RegionNum":0,"RegionCategory":"","ArticlePicture":[],"TitleCN":null,"AbstractTextCN":null,"PMCID":"","EPubDate":null,"PubModel":null,"JCR":null,"JCRName":null,"Score":null,"Total":0}