As NAND flash memory manufacturers scale down to smaller process technology nodes and store more bits per cell, reliability and endurance of flash memory reduce. Wear-leveling and error correction coding can improve both reliability and endurance, but finding effective algorithms requires a strong understanding of flash memory error patterns. To enable such understanding, we have designed and implemented a framework for fast and accurate characterization of flash memory throughout its lifetime. This paper examines the complex flash errors that occur at 30-40nm flash technologies. We demonstrate distinct error patterns, such as cycle-dependency, location-dependency and value-dependency, for various types of flash operations. We analyze the discovered error patterns and explain why they exist from a circuit and device standpoint. Our hope is that the understanding developed from this characterization serves as a building block for new error tolerance algorithms for flash memory.

/pdf/error-patterns-in-mlc-nand-flash-memory-measurement-598fgyt3mx.pdf

Error patterns in MLC NAND flash memory: measurement, characterization, and analysis

Neuromorphic systems are gaining increasing importance in an era where CMOS digital computing techniques are meeting hard physical limits. These silicon systems mimic extremely energy efficient neural computing structures, potentially both for solving engineering applications as well as understanding neural computation. Towards this end, the authors provide a glimpse at what the technology evolution roadmap looks like for these systems so that Neuromorphic engineers may gain the same benefit of anticipation and foresight that IC designers gained from Moore's law many years ago. Scaling of energy efficiency, performance, and size will be discussed as well as how the implementation and application space of Neuromorphic systems are expected to evolve over time.

/pdf/finding-a-roadmap-to-achieve-large-neuromorphic-hardware-4vszjr2j41.pdf

Finding a roadmap to achieve large neuromorphic hardware systems.

Nonvolatile memories with fast write operation at low voltage are required as storage devices to exceed flash memory performance [1–3]. We develop an 8Mb multi-layered cross-point ReRAM macro with 443MB/s write throughput (64b parallel write per 17.2ns cycle), which is almost twice as fast as existing methods, using the fast-switching performance of TaO x ReRAM [4] and the following three techniques to reduce the sneak current in bipolar type cross-point cell array structure in an 0.18μm process. First, memory cell and array technologies reduce the sneak current with a newly developed bidirectional diode as a memory cell select element for the first time. Second, we use a hierarchical bitline (BL) structure for multi-layered cross-point memory with fast and stable current control. Third, we implement a multi-bit write architecture that realizes fast write operation and suppresses sneak current. This work is applicable to both high-density stand-alone and embedded memory with more stacked memory arrays and/or scaling memory cells.

An 8 Mb Multi-Layered Cross-Point ReRAM Macro With 443 MB/s Write Throughput

As technology continues to scale down, NAND Flash memory has been increasingly relying on error-correction codes (ECCs) to ensure the overall data storage integrity. Although advanced ECCs such as low-density parity-check (LDPC) codes can provide significantly stronger error-correction capability over BCH codes being used in current practice, their decoding requires soft-decision log-likelihood ratio (LLR) information. This results in two critical issues. First, accurate calculation of LLR demands fine-grained memory-cell sensing, which nevertheless tends to incur implementation overhead and access latency penalty. Hence, it is critical to minimize the fine-grained memory sensing precision. Second, accurate calculation of LLR also demands the availability of a memory-cell threshold-voltage distribution model. As the major source for memory-cell threshold-voltage distribution distortion, cell-to-cell interference must be carefully incorporated into the model. However, these two critical issues have not been ever addressed in the open literature. This paper attempts to address these open issues. We derive mathematical formulations to approximately model the threshold-voltage distribution of memory cells in the presence of cell-to-cell interference, based on which the calculation of LLRs is mathematically formulated. This paper also proposes a nonuniform memory sensing strategy to reduce the memory sensing precision and, thus, sensing latency while still maintaining good error-correction performance. In addition, we investigate these design issues under the scenario when we can also sense interfering cells and hence explicitly estimate cell-to-cell interference strength. We carry out extensive computer simulations to demonstrate the effectiveness and involved tradeoffs, assuming the use of LDPC codes in 2-bits/cell NAND Flash memory.

On the Use of Soft-Decision Error-Correction Codes in nand Flash Memory

High-capacity NAND flash memory can achieve high density storage by using multi-level cells (MLC) to store more than one bit per cell. Although this larger storage capacity is certainly beneficial, the increased density also increases the raw bit error rate (BER), making powerful error correction coding necessary. Traditional flash memories employ simple algebraic codes, such as BCH codes, that can correct a fixed, specified number of errors. This paper investigates the application of low-density parity-check (LDPC) codes which are well known for their ability to approach capacity in the AWGN channel. We obtain soft information for the LDPC decoder by performing multiple cell reads with distinct word-line voltages. The values of the word-line voltages (also called reference voltages) are optimized by maximizing the mutual information between the input and output of the multiple-read channel. Our results show that using this soft information in the LDPC decoder provides a significant benefit and enables us to outperform BCH codes over a range of block error rates.

/pdf/soft-information-for-ldpc-decoding-in-flash-mutual-10c6o4wlhq.pdf

Soft Information for LDPC Decoding in Flash: Mutual-Information Optimized Quantization

Techniques and corresponding circuitry are presented for the detection of wordline leakage in a memory array. In an exemplary embodiment, a capacitive voltage divider is used to translate the high voltage drop to low voltage drop that can be compared with a reference voltage to determine the voltage drop due to leakage. An on-chip self calibration method can help assure the accuracy of this technique for detecting leakage limit. In other embodiments, the current drawn by a reference array, where a high voltage is applied to the array with all wordlines non-selected, is compared to the current drawn by an array where the high voltage is applied and one or more selected wordlines. In these current based embodiments, the reference array can be a different array, or the same array as that one selected for testing.

Detection of word-line leakage in memory arrays: current based approach

Techniques and corresponding circuitry are presented for the detection of wordline leakage in a memory array. In an exemplary embodiment, a capacitive voltage divider is used to translate the high voltage drop to low voltage drop that can be compared with a reference voltage to determine the voltage drop due to leakage. An on-chip self calibration method can help assure the accuracy of this technique for detecting leakage limit.

Detection of word-line leakage in memory arrays

A charge pump system for supplying an output voltage to a load is described. It includes a regulation circuit connected to receive the output voltage and derive an enable signal from it and multiple charge pump circuits connected in parallel to supply the output voltage. Each of the charge pump circuits is also connected to receive a clock signal and the enable signal. The system also includes one or more delay circuit elements, where a corresponding one or more, but less than all, of the charge pump circuits are connectable to receive the enable signal delayed by the corresponding delay circuit element.

Techniques of Ripple Reduction for Charge Pumps

A 16 Gb 8-level NAND flash chip on 56 nm CMOS technology has been fabricated and is being reported for the first time. This is the first 3-bit per cell (X3) chip published with all-bitline (ABL) architecture, which doubles the write performance compared with conventional shielded bitline architecture. A new advanced cache program algorithm provides another 15% improvement in write performance. This paper also discusses a technique for resolving the sensing error resulting from cell source line noise, which usually varies with the data pattern. The new architecture and advanced algorithm enable an 8 MB/s write performance that is comparable to previously published 2-bit per cell (4-level) NAND performance. Considering the significant cost reduction compared to 4-level NAND flash based on the same technology, this chip is a strong candidate for many mainstream applications.

A 16 Gb 3-Bit Per Cell (X3) NAND Flash Memory on 56 nm Technology With 8 MB/s Write Rate

Techniques of providing a low output voltage, high current capability charge pump are given. The charge pump has multiple capacitors along with switching circuitry. In an initialization phase, the first plate of each of the capacitors is connected to receive a regulator voltage and the second plate of each capacitor is connected to ground. In a transfer phase, the capacitors are connected in series, where, for each capacitor after the first, the second plate is connected to the first plate of the preceding capacitor in the series. The output voltage of the pump is from the first plate of the last capacitor in the series. Regulation circuitry generates the regulator voltage from a reference voltage to have a value responsive to the output voltage level of the pump.

Jonathan Huynh

Papers

Detection of word-line leakage in memory arrays: current based approach

Detection of word-line leakage in memory arrays

Techniques of Ripple Reduction for Charge Pumps

A 16 Gb 3-Bit Per Cell (X3) NAND Flash Memory on 56 nm Technology With 8 MB/s Write Rate

Low voltage charge pump with regulation