A 126.6-mm2 AND-type 512-Mb flash memory with 1.8-Vpower supply

A 512-Mb flash memory, which is applicable to removable flash media of portable equipment such as audio players, has been developed. The chip is fabricated with a 0.18-μm CMOS process on a 126.6-mm² die, and uses a multilevel technique (2 bit/1 cell). The memory cell is AND-type, which is suitable for multilevel operation. This paper reports new techniques adopted in the 512-Mb flash memory. First, techniques for low voltage operation are described. The charge pump, control of pumps, and the reference voltage generator are improved to generate internal voltage stably for multilevel flash memory. Next, a method for reducing total memory cost in the removable flash media is described. A new operation mode named read-modify-write is introduced on the chip. This feature makes the memory system simple, because the controller does not have to track sector-erase information     

A 120-mm2 64-Mb NAND flash memory achieving 180 ns/Byteeffective program speed

Emerging application areas of mass storage flash memories require low cost, high density flash memories with enhanced device performance. This paper describes a 64 Mb NAND flash memory having improved read and program performances. A 40 MB/s read throughput is achieved by improving the page sensing time and employing the full-chip burst read capability. A 2-μs random access time is obtained by using a precharged capacitive decoupling sensing scheme with a staggered row decoder scheme. The full-chip burst read capability is realized by introducing a new array architecture. A narrow incremental step pulse programming scheme achieves a 5 MB/s program throughput corresponding to 180 ns/Byte effective program speed. The chip has been fabricated using a 0.4-μm single-metal CMOS process resulting in a die size of 120 mm² and an effective cell size of 1.1 μm²     

A 29-mm2, 1.8-V-only, 16-Mb DINOR flash memory withgate-protected-poly-diode (GPPD) charge pump

A 29-mm², 16-Mb divided bitline NOR (DINOR) flash memory is fabricated using 0.25-μm triple-well three-layer-metal CMOS technology. Read access time is 72 ns at 1.8 V. A poly diode charge-pump technique improves pump efficiency and eliminates the body effect problem     

40-mm2 3-V-only 50-MHz 64-Mb 2-b/cell CHE NOR flashmemory

This paper presents a 3-V-only 64-Mb 4-level-cell (2-b/cell) NOR-type channel-hot-electron (CHE) programmed flash memory fabricated in 0.18-μm shallow-trench isolation CMOS technology. The device (die size 40 mm²) is organized in 64 1-Mb sectors. Hierarchical column and row decoding ensures complete isolation between different sectors during any operation, thereby increasing device reliability while still providing layout area optimization. Staircase gate-voltage programming is used to achieve narrow threshold-voltage distributions. The same program throughput as for bilevel CHE-programmed memories is obtained, thanks to parallel programming. A mixed balanced/unbalanced sensing approach allows efficient use of the available threshold window. Asynchronous (130-ns access time) and burst-mode (up to 50-MHz data rate) reading is possible. Both column and row redundancy is provided to ensure extended failure coverage. Error correction code techniques, correcting 1 failed over 32 data cells, are also integrated     

A 117-mm2 3.3-V only 128-Mb multilevel NAND flash memoryfor mass storage applications

For a quantum step in further cost reduction, the multilevel cell concept has been combined with the NAND flash memory. Key requirements of mass storage, low cost, and high serial access throughput have been achieved by sacrificing fast random access performance. This paper describes a 128-Mb multilevel NAND flash memory storing 2 b per cell. Multilevel storage is achieved through tight cell threshold voltage distribution of 0.4 V and is made practical by significantly reducing program disturbance by using a local self-boosting scheme. An intelligent page buffer enables cell-by-cell and state-by-state program and inhibit operations. A read throughput of 14.0 MB/s and a program throughput of 0.5 MB/s are achieved. The device has been fabricated with 0.4-μm CMOS technology, resulting in a 117 mm² die size and a 1.1 μm² effective cell size     

A 220-mm2, four- and eight-bank, 256-Mb SDRAM withsingle-sided stitched WL architecture

A 220-mm², 256-Mb SDRAM has been fabricated in fully planarized 0.22-μm CMOS technology with buried strap trench cell. The single-sided stitched word-line (WL) architecture employs asymmetric block activation and shared row decoders to realize 86.7% cell/chip-length efficiency (57.3% cell/chip efficiency). A staggered WL driver arrangement makes it possible to build the drivers on a 0.484-μm WL pitch in limited space. An intraunit address increment pipeline scheme having two logical 8-Mb arrays within one physical 16-Mb unit results in a burst frequency up to 200 MHz for single data rate, while allowing four- and eight-bank organizations. A data rate of 270 Mbits/s was confirmed with a 135-MHz frequency doubling test mode. Single-ended addresses and a single ended read-write-drive bus reduce the ICC₄ current to ~90 mA for 100-MHz seamless burst operation. A detailed shmoo analysis demonstrates address-access time of 13.5 ns and clock-access time of 5 ns. This design also uses a selectable row domain and divided column redundancy scheme that repairs up to ~1400 faults/chip with only 8% chip overhead     

Performance of 1-mm2 Silicon Photomultiplier

A silicon photomultiplier (SPM) is a large area detector consisting of a parallel array of photon counting microcells. Each microcell consists of a Geiger Mode photodiode with an integrated quenching element. Each microcell is then connected to a common output. The microcells have a uniform gain of up to 10 and provide an identical charge output signal for each photon detected. Under illumination the summed output of the detector is proportional to the number of Geiger pulses and hence proportional to the incident photon flux. This combination gives extremely high performance comparable to that of a conventional photomultiplier tube (PMT). We report on the characterization of two different 1 mm² SPM detector designs with 620 and 920 microcells at room temperature (20 deg) and down to . We assess detection efficiency, breakdown voltage, gain, dark rate, crosstalk, timing jitter and dynamic range. The SPM detector operates over the visible region of the spectrum, characterized here from 400 to 800 nm. The peak photon detection efficiency of 15% occurs at 500 nm with a cooled () dark rate of 600 at a bias voltage of 31 V. In a test for positron emission tomography (PET), an energy resolution of 25% was recorded for the detection of 511 keV gamma radiation using 1 mm1 mm15 mm LYSO scintillator crystal. The SPM has many applications such as medical imaging, microscopy, high-energy physics, and homeland security.     

A 0.79-mm2 29-mW Real-Time Face Detection Core

A 0.79-mm² 29-mW real-time face detection core is fabricated in a 0.13-mum CMOS technology. It consists of 75-kgate logic, 58-kbit SRAM, and an ARM AMBA bus interface. Comprehensive optimization in both algorithm and hardware design improves performance and reduces area and power dissipation. Two kinds of templates with facial features are proposed to achieve high speed and yet accurate face detection. A Steady State Genetic Algorithm is employed for high-speed hardware implementation of template matching. To reduce area and power dissipation, frame memory is optimized at minimum and the detection engine is shared for two kinds of template matching. The core can detect eight faces in each frame of moving pictures at 30 frames/second. Face detection accuracy is 92%     

A universal 0.03-mm2 one-pin crystal oscillator in CMOS

A one-pin crystal oscillator with an integrated load capacitance of 15 pF has been realized in a standard 0.35-μm CMOS technology. Due to the structure of the oscillator and the use of MOS gate capacitance for the load capacitors, the chip area can be very small. The total active area including load capacitors is less than 0.03 mm². The design can be operated with supply voltages in the range of 1.4-3.6 V and allows crystal frequencies in the range of 3-30 MHz. The current consumption of the oscillator core is 180 μA at 10 MHz with 3.3-V power supply. It produces a rail-to-rail output swing, regulated by an amplitude control loop, and has the same flexibility and ease of frequency tuning as a common Pierce oscillator. As no special IC process options are required, the design is very suitable for clock generation in digital very-large-scale-integration chips     

A 14-bit 1.8-V 20-mW 1-mm2 CMOS DAC

A low-voltage low-power small-area and high-resolution digital-to-analog converter (DAC) for mixed-signal applications is Introduced. A binary weighted current steering DAC is a power-efficient architecture, because almost all the current taken from the supply is used for the output signal. The current steering architecture is also highly suitable for high-speed operation. Typically, the architecture suffers from poor linearity characteristics, but the problem can be prevented with a novel calibration method, where the currents generated for the most significant bits are fine tuned. As a result, a very compact and low-power solution can be implemented by using a low-voltage digital technology     

A 98 mm2 die size 3.3-V 64-Mb flash memory with FN-NORtype four-level cell

In order to realize high-capacity and low-cost flash memory, we have developed a 64-Mb flash memory with multilevel cell operation scheme. The 64-Mb flash memory has been achieved in a 98 mm² die size by using four-level per cell operation scheme, NOR type cell array, and 0.4-μm CMOS technology. Using an FN type program/erase cell allows a single 3.3 V supply voltage. In order to establish fast programming operation using Fowler-Nordheim (FN)-NOR type memory cell, we have developed a highly parallel multilevel programming technology. The drain voltage controlled multilevel programming (DCMP) scheme, the parallel multilevel verify (PMV) circuit, and the compact multilevel sense-amplifier (CMS) have been implemented to achieve 128 b parallel programming and 6.3 μs/Byte programming speed     

A robust 8F2 ferroelectric RAM cell with depletiondevice (DeFeRAM)

This paper describes an area-penalty-free, leakage-compensated, and noise-immune 8F² cell design suitable for high-density, low-power ferroelectric RAM (FeRAM) generations. The new concept features a 1T1C ferroelectric memory cell containing an additional depletion device (DeFeRAM) controlled by the passing word line in a folded bit-line architecture. The depletion device permits the use of a common cell plate at intermediate voltage level. A highly reliable three-level word-line driver circuit design is discussed     

A 130-mm2, 256-Mbit NAND flash with shallow trenchisolation technology

A 256-Mbit flash memory has been developed using a NAND cell structure with a shallow trench isolation (STI) process. A tight bit-line pitch of 0.55 μm is achieved with 0.25-μm STI. The memory cell is shrunk to 0.29 μm², which realizes a 130-mm² , 256-Mbit flash memory. Peripheral transistors are scaled with memory cells in order to reduce fabrication process steps. A voltage down converter, which generates 2.5-V constant internal power source, is applied to protect the scaled transistors. An improved bit-line clamp sensing scheme achieves 3.8-μs first access time in spite of long and tight pitch bit-line. A 1-kbyte page mode with 35-ns serial data out realizes 25-Mbyte/s read throughput. An incremental step pulse with a bit by bit verify scheme programs 1-k cells in 1-V Vt distribution within 200 μs. That realizes 4.4-Mbyte/s programming throughput     

A 56-nm CMOS 99-mm2 8-Gb Multi-Level NAND Flash Memory With 10-MB/s Program Throughput

A single 3.3-V only, 8-Gb NAND flash memory with the smallest chip to date, 98.8 mm², has been successfully developed. This is the world's first integrated semiconductor chip fabricated with 56-nm CMOS technologies. The effective cell size including the select transistors is 0.0075 mum² per bit, which is the smallest ever reported. To decrease the chip size, a very efficient floor plan with one-sided row decoder, one-sided page buffer, and one-sided pad is introduced. As a result, an excellent 70% cell area efficiency is realized. The program throughput is drastically improved to twice as large as previously reported and comparable to binary memories. The best ever 10-MB/s programming is realized by increasing the page size from 4kB to 8kB. In addition, noise cancellation circuits and the dual VDD-line scheme realize both a small die size and a fast programming. An external page copy achieves a fast 93-ms block copy, efficiently using a 1-MB block size     

A 0.004-mm2 Portable Multiphase Clock Generator Tile for 1.2-GHz RISC Microprocessor

Portable multiphase clock generators capable of adjusting its clock phase according to input clock frequencies have been developed both in a 0.18-mum and in a 0.13-mum CMOS technologies. They consist of a full-digital CMOS circuit design that leads to a simple, robust, and portable IP. In addition, their open-loop architecture lead to no jitter accumulation and one-cycle lock characteristic that enables clock-on-demand circuit structures. The implemented low power clock generator tile in a 0.13-mum CMOS technology occupies only 0.004 mm ² and operates at variable input frequencies ranging from 625 MHz to 1.2 GHz within a plusmn 2% phase error having one-cycle lock time.     

A 390-mm2, 16-bank, 1-Gb DDR SDRAM with hybrid bitlinearchitecture

A 390-mm², 16-bank, 1-Gb, double-data-rate (DDR) synchronous dynamic random access memory (SDRAM) has been fabricated in fully planarized 0.175-μm, 8F² trench cell technology. The 1-Gb SDRAM employs a hybrid bitline architecture with 512 cells/local-bitline (LBL). Four LBL pairs are connected through multiplexers to each sense amplifier (SA). Two of the LBL pairs are coupled to the SA by wiring over two other LBL pairs using hierarchical bitlines. This results in a reduction of the number of the SA's to 1/4, reducing the chip size by 6%. A hierarchical column-select-line scheme is incorporated with a hierarchical dataline (MDQ) architecture. This makes 16-bank organization possible while sharing hierarchical column decoders and second sense amplifiers. A hierarchical 8-b prefetch scheme employs four MDQ's for each read-write drive (RWD) and two RWD's for each DQ. This reduces the frequencies of the MDQ's and the RWD's to 1/8 and 1/2, respectively. A 1-V swing signaling on the RWD is used to reduce the burst current by 18 mA. The 1-V swing signaling is successfully converted to 2.1 V with self-timed first-in, first-out circuitry. The hardware data demonstrate 400-Mb/s/pin operation with a 16-mm TSOP-II package. Seamless burst operation at various frequencies has also been confirmed. These features result in a 1.6-Gb/s data rate for ×32 200-MHz DDR operation with a cell/chip area efficiency of 67.5%     

A 99-mm2 0.7-W single-chip MPEG-2 422P@ML video, audio,and system encoder with a 64-Mb embedded DRAM for portable 422P@HLencoder system

A scalable single-chip 422P@ML MPEG-2 video, audio, and system encoder LSI for portable 422P@HL system is described. The encoder LSI is implemented using 0.13-μm embedded DRAM technology. It integrates 3-M logic gates and 64-Mb DRAM in an area of 99-mm². The power consumption is suppressed to 0.7 W by adopting a low-power DRAM core. It performs real-time 422P@ML video encoding, audio encoding, and system encoding with no external DRAM. Furthermore, the encoder LSI realizes a 422P@HL video encoder with multichip configuration, due to its scalable architecture. This results in a PC-card size 422P@HL encoder for portable HDTV codec system     

A 16-Mb CMOS SRAM with a 2.3-μm2 single-bit-linememory cell

A novel architecture that enables fast write/read in poly-PMOS load or high-resistance polyload single-bit-line cells is developed. The architecture for write uses alternate twin word activation (ATWA) with bit-line pulsing. A dummy cell is used to obtain a reference voltage for reading. An excellent balance between a normal cell signal line and a dummy cell signal line is attained using balanced common data-line architecture. A newly developed self-bias-control (SBC) sense amplifier provides excellent stability and fast sensing performance for input voltages close to V_CC at a low power supply of 2.5 V. The single-bit-line architecture is incorporated in a 16-Mb SRAM, which was fabricated using 0.25-μm CMOS technology. The proposed single-bit-line architecture reduces the cell area to 2.3-μm² , which is two-thirds of a conventional two-bit-line cell with the same processes. The 16-Mb SRAM, a test chip for a 64-Mb SRAM, shows a 15-ns address access time and a 20-ns cycle time     

An embedded 240-mW 10-b 50-MS/s CMOS ADC in 1-mm2

A distributed-gain preamplifier uses averaging to improve resolution by 4 b in differential nonlinearity (DNL) and 2 b in integral nonlinearity (INL) in a flash analog-to-digital converter (ADC). Fabricated in a 0.5-μm, triple-metal, single-poly CMOS process, the circuit measures 1.4 mm×1.4 mm including a bandgap and a sample-and-hold (SH), while the ADC itself occupies 1-mm². At a conversion rate of 50-MS/s the ADC dissipates 170 mW, the SH dissipates 70 mW, and the untrimmed ADC-plus-SH exhibits 54 dB S/(N+D) with a 12-MHz 90% full-scale input     

High-density chain ferroelectric random access memory (chain FRAM)

A new chain ferroelectric random access memory-a chain FRAM-has been proposed. A memory cell consists of parallel connection of one transistor and one ferroelectric capacitor, and one memory cell block consists of plural memory cells connected in series and a block selecting transistor. This configuration realizes the smallest 4 F² size memory cell using the planar transistor so far reported, and random access. The chip size of the proposed chain FRAM can be reduced to 63% of that of the conventional FRAM when 16 cells are connected in series. The fast nondriven half-V_dd cell-plate scheme, as well as the driven cell-plate scheme, are applicable to the chain FRAM without polarization switching during the standby cycle thanks to short-circuiting ferroelectric capacitors. It results in fast access time of 45 ns and cycle time of 70 ns without refresh operation