A 130-nm 6-GHz 256 × 32 bit leakage-tolerant register file

Describes a 256-word × 32-bit 4-read, 4-write ported register file for 6-GHz operation in 1.2-V 130-nm technology. The local bitline uses a pseudostatic technique for aggressive bitline active leakage reduction/tolerance to enable 16 bitcells/bitline, low-V_t usage, and 50% keeper downsizing. Gate-source underdrive of -V _cc on read-select transistors is established without additional supply/bias voltages or gate-oxide overstress. 8% faster read performance and 36% higher dc noise robustness is achieved compared to dual-V_t bitline scheme optimized for high performance. Device-level measurements in the 130-nm technology show 703× bitline active leakage reduction, enabling continued V_t scaling and robust bitline scalability beyond 130-nm generation. Sustained performance and robustness benefit of the pseudostatic technique against conventional dynamic bitline with keeper-upsizing is also presented     

A 6-GHz 16-kB L1 cache in a 100-nm dual-V/sub T/ technology using a bitline leakage reduction (BLR) technique

This work describes an aggressive SRAM cell configuration using dual-V/sub T/ and minimum channel length to achieve high performance. A bitline leakage reduction technique is incorporated into an L1 cache design using the new cell in a 100-nm dual-V/sub T/ technology to eliminate impacts of bitline leakage on performance and noise margin with minimal area overhead. Bitline delay is 23% better than the best conventional design, thus enabling 6-GHz operation at with 15% higher energy.     

Forward body bias for microprocessors in 130-nm technology generation and beyond

Device and test chip measurements show that forward body bias (FBB) can be used effectively to improve performance and reduce complexity of a 130-nm dual-V/sub T/ technology, reduce leakage power during burn-in and standby, improve circuit delay and robustness, and reduce active power. FBB allows performance advantages of low-temperature operation to be realized fully without requiring transistor redesign, and also improves V/sub T/ variations, mismatch, and saturation transconductance and output resistance product (g/sub m//spl times/r/sub o/).     

A 4-GHz 130-nm address generation unit with 32-bit sparse-tree adder core

This paper describes a 32-bit address generation unit designed for 4-GHz operation in 1.2-V 130-nm technology. The AGU utilizes a 152-ps sparse-tree adder core to achieve 20% delay reduction, 80% lower interconnect complexity, and a low (1%) active energy leakage component. The dual-V/sub T/ semidynamic implementation of the adder core provides the performance of a dynamic CMOS design with an average energy profile similar to static CMOS, enabling 71% savings in average energy with a good sub-130-nm scaling trend.     

5-GHz 32-bit integer execution core in 130-nm dual-V/sub T/ CMOS

A 32-bit integer execution core containing a Han-Carlson arithmetic-logic unit (ALU), an 8-entry /spl times/ 2 ALU instruction scheduler loop and a 32-entry /spl times/ 32-bit register file is described. In a 130 nm six-metal, dual-V/sub T/ CMOS technology, the 2.3 mm/sup 2/ prototype contains 160 K transistors. Measurements demonstrate capability for 5-GHz single-cycle integer execution at 25/spl deg/C. The single-ended, leakage-tolerant dynamic scheme used in the ALU and scheduler enables up to 9-wide ORs with 23% critical path speed improvement and 40% active leakage power reduction when compared to a conventional Kogge-Stone implementation. On-chip body-bias circuits provide additional performance improvement or leakage tolerance. Stack node preconditioning improves ALU performance by 10%. At 5 GHz, ALU power is 95 mW at 0.95 V and the register file consumes 172 mW at 1.37 V. The ALU performance is scalable to 6.5 GHz at 1.1 V and to 10 GHz at 1.7 V, 25/spl deg/C.     

A bitline leakage compensation scheme for low-voltage SRAMs

A bitline leakage current of an SRAM, induced by leakage current of the transmission transistors in the cells that are associated with the bitline, increases as the threshold voltage (V_TH) of the transistors is reduced for high performance at low power-supply voltage (V_DD). The increased bitline leakage causes slow or incorrect read/write operation of an SRAM because the leakage current acts as noise current for a sense amplifier. In this paper, the problem has been solved from a circuitry point of view, and the scheme which detects the bitline leakage current in a precharge cycle and compensates for it during a read/write cycle is proposed. Employing this scheme, the SRAM with 360-μA bitline leakage current can perform a read/write operation at the same speed as one that has no bitline leakage current. This enables a 0.1-V reduction in V_TH, and keeps the V_TH and delay scalability of a high-performance SRAM in technology progress. An experimental 8-Kb SRAM with 256 rows is fabricated in a 0.25-μm CMOS technology, which demonstrates the effectiveness of the scheme     

A 1.5-GHz 130-nm Itanium/sup /spl reg// 2 Processor with 6-MB on-die L3 cache

This 130-nm Itanium 2 processor implements the explicitly parallel instruction computing (EPIC) architecture and features an on-die 6-MB 24-way set-associative level-3 cache. The 374-mm/sup 2/ die contains 410 M transistors and is implemented in a dual-V/sub t/ process with six Cu interconnect layers and FSG dielectric. The processor runs at 1.5 GHz at 1.3 V and dissipates a maximum of 130 W. This paper reviews circuit design and package details, power delivery, the reliability, availability, and serviceability (RAS) features, design for test (DFT), and design for manufacturability (DFM) features, as well as an overview of the design and verification methodology. The fuse-based clock deskew circuit achieves 24-ps skew across the entire die, while the scan-based skew control further reduces it to 7 ps. The 128-bit front-side bus has a bandwidth of 6.4 GB/s and supports up to four processors on a single bus.     

Measurements and analysis of SER-tolerant latch in a 90-nm dual-V/sub T/ CMOS process

We designed a soft error rate (SER) tolerant latch utilizing local redundancy. We implemented a test chip containing both the standard and SER-tolerant latches in a 90-nm dual-V/sub T/ CMOS process. Accelerated measurements with a neutron beam at Los Alamos National Laboratory demonstrated 10/spl times/ better reliability of the SER-tolerant latch over the standard latch at no speed degradation. The worst case energy and area penalties were 39% and 44%, respectively. Both the energy and area penalties are negligible for standard-latch transistor sizes at least double the minimum width. We analyzed the effects of the recovery time, threshold voltage assignment, and leakage on the SER robustness. The proposed latch can improve reliability of critical sequential logic elements in microprocessors and other circuits.     

A 130-nm triple-V/sub t/ 9-MB third-level on-die cache for the 1.7-GHz Itanium/spl reg/ 2 processor

The 18-way set-associative, single-ported 9 MB cache for the Itanium 2 processor uses 210 identical 48-kB sub-arrays with a 2.21-/spl mu/m/sup 2/ cell in a 130-nm 6-metal technology. The processor runs at 1.7 GHz at 1.35 V and dissipates 130 W. The 432-mm/sup 2/ die contains 592 M transistors, the largest transistor count reported for a microprocessor. This paper reviews circuit design and implementation details for the L3 cache data and tag arrays. The staged mode ECC scheme avoids a latency increase in the L3 tag. A high V/sub t/ implant improves the read stability and reduces the sub-threshold leakage.     

A 90-nm logic technology featuring strained-silicon

A leading-edge 90-nm technology with 1.2-nm physical gate oxide, 45-nm gate length, strained silicon, NiSi, seven layers of Cu interconnects, and low-/spl kappa/ CDO for high-performance dense logic is presented. Strained silicon is used to increase saturated n-type and p-type metal-oxide-semiconductor field-effect transistors (MOSFETs) drive currents by 10% and 25%, respectively. Using selective epitaxial Si/sub 1-x/Ge/sub x/ in the source and drain regions, longitudinal uniaxial compressive stress is introduced into the p-type MOSEFT to increase hole mobility by >50%. A tensile silicon nitride-capping layer is used to introduce tensile strain into the n-type MOSFET and enhance electron mobility by 20%. Unlike all past strained-Si work, the hole mobility enhancement in this paper is present at large vertical electric fields in nanoscale transistors making this strain technique useful for advanced logic technologies. Furthermore, using piezoresistance coefficients it is shown that significantly less strain (/spl sim/5 /spl times/) is needed for a given PMOS mobility enhancement when applied via longitudinal uniaxial compression versus in-plane biaxial tension using the conventional Si/sub 1-x/Ge/sub x/ substrate approach.     

Analysis of buck converters for on-chip integration with a dual supply voltage microprocessor

An analysis of an on-chip buck converter is presented in this paper. A high switching frequency is the key design parameter that simultaneously permits monolithic integration and high efficiency. A model of the parasitic impedances of a buck converter is developed. With this model, a design space is determined that allows integration of active and passive devices on the same die for a target technology. An efficiency of 88.4% at a switching frequency of 477 MHz is demonstrated for a voltage conversion from 1.2-0.9 volts while supplying 9.5 A average current. The area occupied by the buck converter is 12.6 mm/sup 2/ assuming an 80-nm CMOS technology. An estimate of the efficiency is shown to be within 2.4% of simulation at the target design point. Full integration of a high-efficiency buck converter on the same die with a dual-V/sub DD/ microprocessor is demonstrated to be feasible.     

Millimeter-wave CMOS design

This paper describes the design and modeling of CMOS transistors, integrated passives, and circuit blocks at millimeter-wave (mm-wave) frequencies. The effects of parasitics on the high-frequency performance of 130-nm CMOS transistors are investigated, and a peak f/sub max/ of 135 GHz has been achieved with optimal device layout. The inductive quality factor (Q/sub L/) is proposed as a more representative metric for transmission lines, and for a standard CMOS back-end process, coplanar waveguide (CPW) lines are determined to possess a higher Q/sub L/ than microstrip lines. Techniques for accurate modeling of active and passive components at mm-wave frequencies are presented. The proposed methodology was used to design two wideband mm-wave CMOS amplifiers operating at 40 GHz and 60 GHz. The 40-GHz amplifier achieves a peak |S/sub 21/| = 19 dB, output P/sub 1dB/ = -0.9 dBm, IIP3 = -7.4 dBm, and consumes 24 mA from a 1.5-V supply. The 60-GHz amplifier achieves a peak |S/sub 21/| = 12 dB, output P/sub 1dB/ = +2.0 dBm, NF = 8.8 dB, and consumes 36 mA from a 1.5-V supply. The amplifiers were fabricated in a standard 130-nm 6-metal layer bulk-CMOS process, demonstrating that complex mm-wave circuits are possible in today's mainstream CMOS technologies.     

Low-frequency noise behavior of SiO/sub 2/--HfO/sub 2/ dual-layer gate dielectric nMOSFETs with different interfacial oxide thickness

The low-frequency noise has been studied in nMOSFETs with an HfO/sub 2/--SiO/sub 2/ gate stack, for different thickness of the SiO/sub 2/ interfacial layer (IL). It is observed that the 1/f-like noise in linear operation, is about 50 times higher in the HfO/sub 2/ devices with a 0.8-nm chemical oxide IL, compared with the 4.5-nm thermal oxide reference n-channel transistors. This is shown to relate to the correspondingly higher trap density in the dielectric material. In addition, it is demonstrated that the noise rapidly reduces with increasing thickness of the IL. From the results for a 2.1-nm SiO/sub 2/ IL, it is derived that at a certain gate voltage range, electron tunneling to a defect band in the HfO/sub 2/ layer may contribute to a pronounced increase in the flicker noise.     

A 4-kB 500-MHz 4-T CMOS SRAM using low-V/sub THN/ bitline drivers and high-V/sub THP/ latches

The design and physical implementation of a prototypical 500-MHz CMOS 4-T SRAM is presented in this work. The latch of the proposed SRAM cell is realized by a pair of cross coupled high-V/sub THP/ pMOS transistors, while the bitline drivers are realized by a pair of low-V/sub THN/ nMOS transistors. The wordline voltage compensation circuit and bitline boosting circuit, then, are neither needed to enhance the data retention of memory cells. Built-in self-refreshing paths make the data retention possible without the appearance of any external refreshing mechanism. The advantages of dual threshold voltage transistors can be used to reduce the access time, and maintain data retention at the same time. Besides, a new design of cascaded noise-immune address transition detector is also included to filter out the unwanted chip select glitches when the SRAM is asynchronously operated.     

A 0.5-V 25-MHz 1-mW 256-kb MTCMOS/SOI SRAM for solar-power-operated portable personal digital equipment - sure write operation by using step-down negatively overdriven bitline scheme

Multithreshold-voltage CMOS (MTCMOS) technology has a great advantage in that it provides high-speed operation with low supply voltages of less than 1 V. A logic gate with low-V/sub th/ MOSFETs has a high operating speed, while a low-leakage power switch with a high-V/sub th/ MOSFET eliminates the off-leakage current during sleep time. By using MTCMOS circuits and silicon-on-insulator (SOI) devices, the authors have developed a 256-kb SRAM for solar-power-operated digital equipment. A double-threshold-voltage MOSFET (DTMOS) is adopted for the power switch to further reduce the off leakage. As regards the SRAM core design, we consider a hybrid configuration consisting of high-V/sub th/ and low-V/sub th/ MOSFETs (that is, multi-V/sub th/ CMOS). A new memory cell with a separate read-data path provides a larger readout current without degrading the static noise margin. A negatively overdriven bitline scheme guarantees sure write operation at ultralow supply voltages close to 0.5 V. In addition, a charge-transfer amplifier integrated with a selector and data latches for intrabus circuitry are installed to enhance the operating speed and/or reduce power dissipation. A 32K-word /spl times/ 8-bit SRAM chip, fabricated with the 0.35-/spl mu/m multi-V/sub th/ CMOS/SOI process, has successfully operated at 25 MHz under typical conditions with 0.5-V (SRAM core) and 1-V (I/O buffers) power supplies. The power dissipation during sleep time is less than 0.4 /spl mu/W and that for 25-MHz operation is 1 mW, excluding that of the I/O buffers.     

SRAM design on 65-nm CMOS technology with dynamic sleep transistor for leakage reduction

A 70-Mb SRAM is designed and fabricated on a 65-nm CMOS technology. It features a 0.57-/spl mu/m/sup 2/ 6T SRAM cell with large noise margin down to 0.7 V for low-voltage operation. The fully synchronized subarray contains an integrated leakage reduction scheme with dynamically controlled sleep transistor. SRAM virtual ground in standby is controlled by programmable bias transistors to achieve good voltage control with fine granularity under process skew. It also has a built-in programmable defect "screen" circuit for high volume manufacturing. The measurements showed that the SRAM leakage can be reduced by 3-5/spl times/ while maintaining the integrity of stored data.     

A 1.6 GB/s DDR2 128 Mb Chain FeRAM With Scalable Octal Bitline and Sensing Schemes

An 87.7 mm² 1.6 GB/s 128 Mb chain FeRAM with 130 nm 4-metal CMOS process is demonstrated. In addition to small bitline capacitance inherent to chain FeRAM architecture, three new FeRAM scaling techniques - octal bitline architecture, small parasitic capacitance sensing scheme, and dual metal plateline scheme - reduce bitline capacitance from 100 fF to 60 fF. As a result, a cell signal of ±220 mV is achieved even with the small cell size of 0.252 ?m². An 800 Mb/s/pin read/write bandwidth at 400 MHz clock is realized by installing SDRAM compatible DDR2 interface, and performance is verified by simulation. The internal power-line bounce noise due to 400 MHz clock operation is suppressed to less than 50 mV by an event-driven current driver, which supplies several hundreds of mA of current within 2 ns response. The precise timing and voltage controls are achieved by using the data stored in a compact FeRAM-fuse, which consists of extra FeRAM memory cells placed in edge of normal array instead of conventional laser fuse links. This configuration minimizes area penalty to 0.2% without cell signal degradation.     

n/sup +/-InAs-InAlAs recess gate technology for InAs-channel millimeter-wave HFETs

We report a submicrometer, self-aligned recess gate technology for millimeter-wave InAs-channel heterostructure field effect transistors. The recess gate structure is obtained in an n/sup +/-InAs-InAlAs double cap layer structure with a citric-acid-based etchant. From molecular-beam epitaxy-grown material functional devices with 1000-, 500-, and 200-nm gate length were fabricated. From all three device geometries we obtain drive currents of at least 500 mA/mm, gate leakage currents below 2 mA/mm, and RF-transconductance of 1 S/mm. For the 200-nm gate length device f/sub /spl tau// and f/sub max/ are 162 and 137 GHz, respectively. For the 500-nm gate length device f/sub /spl tau// and f/sub max/ are 89 and 140 GHz, respectively. We observe scaling limitations at 200-nm gate length, in particular a negative threshold voltage shift from -550 to -810 mV, increased kink-effect, and a high gate-to-drain capacitance of 0.5 pF/mm. The present limitations to device scaling are discussed.     

A quick intelligent page-programming architecture and a shieldedbitline sensing method for 3 V-only NAND flash memory

This paper describes a quick intelligent page-programming architecture with a newly introduced intelligent verify circuit for 3 V-only NAND flash memories. The new verify circuit, which is composed of only two transistors, results in a simple intelligent program algorithm for 3 V-only operation and a reduction of the program time to 56%. This paper also describes a shielded bitline sensing method to reduce a bitline-bitline capacitive coupling noise from 700 mV to 35 mV. The large 700 mV noise without the shielded bitline architecture is mainly caused by the NAND-type cell array structure. A 3 V-only experimental NAND flash memory, developed in a 0.7-μm NAND flash memory process technology, demonstrates that the programmed threshold voltages are controlled between 0.4 V and 1.8 V by the new verify circuit. The shielded bitline sensing method realizes a 2.5-μs random access time with a 2.7-V power supply. The page-programming is completed after the 40-μs program and 2.8-μs verify read cycle is iterated 4 times. The block-erasing time is 10 ms     

A Voltage Scalable 0.26 V, 64 kb 8T SRAM With V$_{min}$ Lowering Techniques and Deep Sleep Mode

A voltage scalable 0.26 V, 64 kb 8T SRAM with 512 cells per bitline is implemented in a 130 nm CMOS process. Utilization of the reverse short channel effect in a SRAM cell design improves cell write margin and read performance without the aid of peripheral circuits. A marginal bitline leakage compensation (MBLC) scheme compensates for the bitline leakage current which becomes comparable to a read current at subthreshold supply voltages. The MBLC allows us to lower ${rm V}_{min}$ to 0.26 V and also eliminates the need for precharged read bitlines. A floating read bitline and write bitline scheme reduces the leakage power consumption. A deep sleep mode minimizes the standby leakage power consumption without compromising the hold mode cell stability. Finally, an automatic wordline pulse width control circuit tracks PVT variations and shuts off the bitline leakage current upon completion of a read operation.