Data Storage Using a Non-integer Number of Bits per Cell

Similar documents
Nonlinear Multi-Error Correction Codes for Reliable MLC NAND Flash Memories Zhen Wang, Mark Karpovsky, Fellow, IEEE, and Ajay Joshi, Member, IEEE

Basics of Error Correcting Codes

Analog to digital and digital to analog converters

LDPC Decoding: VLSI Architectures and Implementations

A Study of Polar Codes for MLC NAND Flash Memories

Low-complexity Low-Precision LDPC Decoding for SSD Controllers

The counterpart to a DAC is the ADC, which is generally a more complicated circuit. One of the most popular ADC circuit is the successive

C06a: Digital Modulation

Extending NAND Endurance with Advanced Controller Technology

IES Digital Mock Test

Mixed Synchronous/Asynchronous State Memory for Low Power FSM Design

SUCCESSIVE approximation register (SAR) analog-todigital

An Interconnect-Centric Approach to Cyclic Shifter Design

Advanced Digital Signal Processing Part 2: Digital Processing of Continuous-Time Signals

Outline. Communications Engineering 1

Mitigating Inter-Cell Coupling Effects in MLC NAND Flash via Constrained Coding

Digital Television Lecture 5

A High Speed Encoder for a 5GS/s 5 Bit Flash ADC

FORWARD ERROR CORRECTION PROPOSAL FOR EPOC PHY LAYER

Cmos Full Adder and Multiplexer Based Encoder for Low Resolution Flash Adc

Computer Architecture and Organization:

Digital Logic Circuits

1. The decimal number 62 is represented in hexadecimal (base 16) and binary (base 2) respectively as

TSTE17 System Design, CDIO. General project hints. Behavioral Model. General project hints, cont. Lecture 5. Required documents Modulation, cont.

An Analysis of Multipliers in a New Binary System

A Multiplexer-Based Digital Passive Linear Counter (PLINCO)

ENGN8637, Semster-1, 2018 Project Description Project 1: Bit Interleaved Modulation

AREA AND DELAY EFFICIENT DESIGN FOR PARALLEL PREFIX FINITE FIELD MULTIPLIER

Embedded System Hardware

Data Acquisition: A/D & D/A Conversion

MS Project :Trading Accuracy for Power with an Under-designed Multiplier Architecture Parag Kulkarni Adviser : Prof. Puneet Gupta Electrical Eng.

Error Detection & Vertical LDPC ECC for Reliable 3D-TLC NAND Flash

Combining Modern Codes and Set- Partitioning for Multilevel Storage Systems

CT-516 Advanced Digital Communications

CMOS High Speed A/D Converter Architectures

DIGITAL LOGIC CIRCUITS

IEEE C /02R1. IEEE Mobile Broadband Wireless Access <

Volume 2, Issue 9, September 2014 International Journal of Advance Research in Computer Science and Management Studies

Performance Evaluation of Bit Division Multiplexing combined with Non-Uniform QAM

A Low-Noise Self-Calibrating Dynamic Comparator for High-Speed ADCs

2. ADC Architectures and CMOS Circuits

Analogue to Digital Conversion

Towards 100G over Copper

Unit 3. Logic Design

Swedish College of Engineering and Technology Rahim Yar Khan

Tuesday, March 1st, 9:15 11:00. Snorre Aunet Nanoelectronics group Department of Informatics University of Oslo.

Low Complexity Cross Parity Codes for Multiple and Random Bit Error Correction

An Energy-Division Multiple Access Scheme

CSC/MATA67 Tutorial, Week 12

Combinational logic. ! Regular logic: multiplexers, decoders, LUTs and FPGAs. ! Switches, basic logic and truth tables, logic functions

C802.16a-02/76. IEEE Broadband Wireless Access Working Group <

Joint Relaying and Network Coding in Wireless Networks

K-Best Decoders for 5G+ Wireless Communication

High-Speed Analog to Digital Converters. ELCT 1003:High Speed ADCs

LDPC FEC PROPOSAL FOR EPOC. Richard S. Prodan Broadcom Corporation

Low Power Error Correcting Codes Using Majority Logic Decoding

Digital Integrated CircuitDesign

05/11/2006. Lecture What does a computer do? Logic Manipulation. Data manipulation

Local Asynchronous Communication. By S.Senthilmurugan Asst.Professor/ICE SRM University. Chennai.

IMAGE PROCESSING: POINT PROCESSES

EE 435/535: Error Correcting Codes Project 1, Fall 2009: Extended Hamming Code. 1 Introduction. 2 Extended Hamming Code: Encoding. 1.

Construction of Efficient Amplitude Phase Shift Keying Constellations

CONSTRUCTION AND PERFORMANCE STUDIES OF A PSEUDO-ORTHOGONAL CODE FOR FIBER OPTIC CDMA LAN

Using TCM Techniques to Decrease BER Without Bandwidth Compromise. Using TCM Techniques to Decrease BER Without Bandwidth Compromise. nutaq.

Introduction to Communications Part Two: Physical Layer Ch3: Data & Signals

B.E SEMESTER: 4 INFORMATION TECHNOLOGY

Lecture 13 February 23

HY448 Sample Problems

Thus there are three basic modulation techniques: 1) AMPLITUDE SHIFT KEYING 2) FREQUENCY SHIFT KEYING 3) PHASE SHIFT KEYING

Information Theory and Communication Optimal Codes

A Study on Comparator and Offset Calibration Techniques in High Speed Nyquist ADCs. Chi Hang Chan, Ivor

DIGITALLY controlled and area-efficient calibration circuits

8.2 Union, Intersection, and Complement of Events; Odds

High Speed Flash Analog to Digital Converters

Chapter 5 - Elementary Probability Theory

comparasion to BPSK, to distinguish those symbols, therefore, the error performance is degraded. Fig 2 QPSK signal constellation

Arda Gumusalan CS788Term Project 2

Hamming net based Low Complexity Successive Cancellation Polar Decoder

Analog-to-Digital Converter (ADC) And Digital-to-Analog Converter (DAC)

Convolutional Coding Using Booth Algorithm For Application in Wireless Communication

Cooperative Networked Radar: The Two-Step Detector

Lecture 02: Logic Families. R.J. Harris & D.G. Bailey

LDPC Codes for Rank Modulation in Flash Memories

Module 3 Greedy Strategy

A NEW APPROACH TO DESIGN LOW POWER CMOS FLASH A/D CONVERTER

Performance of Combined Error Correction and Error Detection for very Short Block Length Codes

EE521 Analog and Digital Communications

A New, Low-Cost, Sampled-Data, 10-Bit CMOS A/D Converter

Optimal Power Allocation for Type II H ARQ via Geometric Programming

SEN366 Computer Networks

Lecture 3: Wireless Physical Layer: Modulation Techniques. Mythili Vutukuru CS 653 Spring 2014 Jan 13, Monday

A/D Conversion and Filtering for Ultra Low Power Radios. Dejan Radjen Yasser Sherazi. Advanced Digital IC Design. Contents. Why is this important?

Reading: Schwarz and Oldham (light on non-ideal) and comparator viewgraphs. Lecture 14: October 17, 2001

Index terms: Analog to digital converter, Flash ADC, Pseudo NMOS logic, Pseudo Dynamic CMOS logic multi threshold voltage CMOS inverters.

Information Theory at the Extremes

NAND Structure Aware Controller Framework

Modified Partial Product Generator for Redundant Binary Multiplier with High Modularity and Carry-Free Addition

Outage Probability of a Multi-User Cooperation Protocol in an Asynchronous CDMA Cellular Uplink

The Critical Role of Firmware and Flash Translation Layers in Solid State Drive Design

A REVIEW ON 4 BIT FLASH ANALOG TO DIGITAL CONVERTOR

Transcription:

Data Storage Using a Non-integer Number of Bits per Cell Naftali Sommer June 21st, 2017

The Conventional Scheme Information is stored in a memory cell by setting its threshold voltage 1 bit/cell - Many cells are programmed and read simultaneously Conventional Flash memory devices store an integer number of bits per cell 2 bits/cell - Number of voltage states is a power of 2 3 bits/cell Easy to program, read and protect by Error Correcting Codes (ECC) using simple schemes USB USB USB USB USB

The Conventional Scheme - 3 Bits/Cell Example Multi-level coding - 3 independent logical data pages are written to the same cells - Denoted LSB,, USB Due to the Gray mapping, each page can be independently read with a relatively small number of sense operations - LSB - single sense (RV34), - 2 senses (RV32, RV36), USB - 4 senses - Average read latency of (1+2+4)/3 = 2.33 senses per page - Beneficial for random read scenarios A binary Error Correcting Code (ECC) can be independently applied to each page

Motivation for a Non Integer Number of Bits/cell Voltage window Distribution width The distribution width and the available voltage window define the equivalent Signal to Noise Ratio (SNR) This SNR limits the net number of bits/cell that can be reliably stored Example: assume that the available SNR enables net storage of 2.5 bits/cell - 2 possible schemes: - Advantages of the 7 states scheme - Simpler implementation of the ECC - lower power consumption and die size - Higher code rate, less errors to correct - Faster to program Number of voltage states Number of bits/cell ECC code rate Net number of bits/cell 8 log2(8) = 3 0.83 2.5 7 log2(7) = 2.8 0.89 2.5

Problem Statement We would like to design a practical mapping scheme for a non power-of-2 number of states (e.g. 9, 7, 6, 5, 3) that will enable: - Simple ECC scheme - Short read latency for random read Potential solution - use non-binary symbols, each symbol mapped to a cell - Practical non-binary ECC schemes exist (e.g. LDPC, BCH) - However - random read operations will have high latency - For example, 6 sense operations are needed to distinguish between 7 states - Much higher than the 2.33 sense operations shown in the previous slides for 8 voltage states Main challenge - how to map several independent pages to the same cells, such that different pages are read with disjoint sense operations

Proposed Scheme for the 7 States Case Proposed scheme - map 3 independent pages as done for 8 states, but don t use the highest level - Problem - what to do if {LSB,, USB} = {0, 1, 1}? - Solution - write LSB & pages as before, but map USB bits only to cells for which {LSB, } = {11, 10, 00} - Only 3/4 of the cells can be used for storing the USB page - USB page size will be 3/4 of the LSB and page sizes - For example: LSB and are 16 KByte, USB is 12 KByte - Conditionally invert the LSB and data before programming in order to make sure that at least 3/4 of the cells will enable a USB bit programming Equivalent to storing 2.75 bits/cell - Close to the theoretical limit of log2(7) = 2.8 bits/cell

Proposed Scheme for the 7 States Case - Cont d A binary ECC is applied independently to each page, similar to the conventional scheme Read latency - First 2 pages are read with low-latency in the same way as in the conventional scheme - Single sense for LSB, 2 senses for - For reading the 3/4-sized USB page, we first need to read and ECC decode the first 2 pages - Need to know in which cells the 3/4 page is stored, and must know it error-free - A total of 6 sense operations are needed to read the USB page - Average read latency: (1*1 + 1*2 + 0.75*6)/2.75 = 2.73 sense operations - Worse than the 2.33 of the conventional scheme, but still much better than the 6 of the non binary symbols scheme

Extensions The scheme can be extended to any number of voltage states - Examples: - 2.5 bits/cell using 6 voltage states by programming 2 full pages + a half-sized 3rd page to half of the cells - 1.5 bits/cell using 3 voltage states by programming a full page + a half-sized 2nd page to half of the cells - 3.25 bits/cell using 10 voltage states by programming 3 full pages + a 1/4-sized 4th page to 1/4 of the cells The scheme can be extended to other types of Gray mapping