서브메뉴
검색
Peripheral Circuit Design Techniques for Emerging Memory
Peripheral Circuit Design Techniques for Emerging Memory
상세정보
- 자료유형
- 학위논문 서양
- 최종처리일시
- 20250211153119
- ISBN
- 9798346786429
- DDC
- 620
- 서명/저자
- Peripheral Circuit Design Techniques for Emerging Memory
- 발행사항
- [Sl] : Stanford University, 2024
- 발행사항
- Ann Arbor : ProQuest Dissertations & Theses, 2024
- 형태사항
- 208 p
- 주기사항
- Source: Dissertations Abstracts International, Volume: 86-06, Section: A.
- 주기사항
- Advisor: Wong, H. S. Philip.
- 학위논문주기
- Thesis (Ph.D.)--Stanford University, 2024.
- 초록/해제
- 요약As the capabilities of consumer electronics grow, so too does the need for low-energy, low-cost, and process-compatible embedded non-volatile memory (NVM). Resistive RAM (RRAM) is a technology that can meet this NVM need using materials and processes already available in a wide range of CMOS processes. However, (1) the read/write circuits required to access the information stored in a cell are often larger than the RRAM cells themselves in a macro, (2) there are competing pressures of decreasing read latency and increasing average read energy in an RRAM macro, and (3) the information stored in an RRAM cell must be retained even as the cell conductance experiences time-based variation. In this dissertation, we explore design methods for high-density, low energy RRAM storage macros. First, we present modeling approaches for RRAM macro readout circuit latency, energy, and thermal noise - spanning both voltage-based and current-based conductance sense amplifiers. From these models, we investigate the tradeoff between average read energy and input-referred thermal noise and find that there is not an inherent advantage in the voltage-based or current-based sensing scheme across all macro use cases. We then present a method to design small driver and pass gate circuits while providing the current and voltage necessary for RRAM cell write. The readout circuit models and write circuit area models are combined to generate prospective energy-area tradeoff curves for macro designs given RRAM cell parameters. This method allows designers to estimate read energy and macro density limits given an RRAM technology. Next, we demonstrate the Efficient Multiple-Bits-per-Cell Embedded RRAM (EMBER) macro, which uses the aforementioned read and write circuit design models/techniques to achieve a low read energy (1.0 pJ/bit single-bit-per-cell) with small unit cell area (5.6e-3 F2 single-bit-per-cell, F = 40 nm) compared to prior art. We also use experimental data obtained from EMBER to determine conductance level allocations most suitable for fast, low-BER write as well as conductance level allocations most suitable for low-energy, low-BER read over an extended period of time. Finally, we demonstrate that our read/write circuit design approach is not just applicable to RRAM, but to other emerging resistive NVMs as well.
- 일반주제명
- Random access memory
- 일반주제명
- Space exploration
- 일반주제명
- Energy
- 일반주제명
- Macros
- 일반주제명
- Transistors
- 일반주제명
- Design techniques
- 일반주제명
- Aerospace engineering
- 일반주제명
- Computer science
- 일반주제명
- Design
- 기타저자
- Stanford University.
- 기본자료저록
- Dissertations Abstracts International. 86-06A.
- 전자적 위치 및 접속
- 로그인 후 원문을 볼 수 있습니다.
MARC
008250123s2024 us c eng d■001000017165064
■00520250211153119
■006m o d
■007cr#unu||||||||
■020 ▼a9798346786429
■035 ▼a(MiAaPQ)AAI31757686
■035 ▼a(MiAaPQ)Stanforddj427mq8691
■040 ▼aMiAaPQ▼cMiAaPQ
■0820 ▼a620
■1001 ▼aUpton, Luke Robert.
■24510▼aPeripheral Circuit Design Techniques for Emerging Memory
■260 ▼a[Sl]▼bStanford University▼c2024
■260 1▼aAnn Arbor▼bProQuest Dissertations & Theses▼c2024
■300 ▼a208 p
■500 ▼aSource: Dissertations Abstracts International, Volume: 86-06, Section: A.
■500 ▼aAdvisor: Wong, H. S. Philip.
■5021 ▼aThesis (Ph.D.)--Stanford University, 2024.
■520 ▼aAs the capabilities of consumer electronics grow, so too does the need for low-energy, low-cost, and process-compatible embedded non-volatile memory (NVM). Resistive RAM (RRAM) is a technology that can meet this NVM need using materials and processes already available in a wide range of CMOS processes. However, (1) the read/write circuits required to access the information stored in a cell are often larger than the RRAM cells themselves in a macro, (2) there are competing pressures of decreasing read latency and increasing average read energy in an RRAM macro, and (3) the information stored in an RRAM cell must be retained even as the cell conductance experiences time-based variation. In this dissertation, we explore design methods for high-density, low energy RRAM storage macros. First, we present modeling approaches for RRAM macro readout circuit latency, energy, and thermal noise - spanning both voltage-based and current-based conductance sense amplifiers. From these models, we investigate the tradeoff between average read energy and input-referred thermal noise and find that there is not an inherent advantage in the voltage-based or current-based sensing scheme across all macro use cases. We then present a method to design small driver and pass gate circuits while providing the current and voltage necessary for RRAM cell write. The readout circuit models and write circuit area models are combined to generate prospective energy-area tradeoff curves for macro designs given RRAM cell parameters. This method allows designers to estimate read energy and macro density limits given an RRAM technology. Next, we demonstrate the Efficient Multiple-Bits-per-Cell Embedded RRAM (EMBER) macro, which uses the aforementioned read and write circuit design models/techniques to achieve a low read energy (1.0 pJ/bit single-bit-per-cell) with small unit cell area (5.6e-3 F2 single-bit-per-cell, F = 40 nm) compared to prior art. We also use experimental data obtained from EMBER to determine conductance level allocations most suitable for fast, low-BER write as well as conductance level allocations most suitable for low-energy, low-BER read over an extended period of time. Finally, we demonstrate that our read/write circuit design approach is not just applicable to RRAM, but to other emerging resistive NVMs as well.
■590 ▼aSchool code: 0212.
■650 4▼aRandom access memory
■650 4▼aSpace exploration
■650 4▼aEnergy
■650 4▼aMacros
■650 4▼aTransistors
■650 4▼aDesign techniques
■650 4▼aAerospace engineering
■650 4▼aComputer science
■650 4▼aDesign
■690 ▼a0791
■690 ▼a0538
■690 ▼a0984
■690 ▼a0389
■71020▼aStanford University.
■7730 ▼tDissertations Abstracts International▼g86-06A.
■790 ▼a0212
■791 ▼aPh.D.
■792 ▼a2024
■793 ▼aEnglish
■85640▼uhttp://www.riss.kr/pdu/ddodLink.do?id=T17165064▼nKERIS▼z이 자료의 원문은 한국교육학술정보원에서 제공합니다.


