Motivated by emerging battery-operated applications that demand intensive computation in portable environments, techniques are investigated which reduce power consumption in CMOS digital circuits while maintaining computational throughput. Techniques for low-power operation are shown which use the lowest possible supply voltage coupled with architectural, logic style, circuit, and technology optimizations. An architecturally based scaling strategy is presented which indicates that the optimum voltage is much lower than that determined by other scaling considerations. This optimum is achieved by trading increased silicon area for reduced power consumption. >

/pdf/low-power-cmos-digital-design-52v6aautu5.pdf

Low-power CMOS digital design

1-V power supply high-speed low-power digital circuit technology with 0.5-/spl mu/m multithreshold-voltage CMOS (MTCMOS) is proposed. This technology features both low-threshold voltage and high-threshold voltage MOSFET's in a single LSI. The low-threshold voltage MOSFET's enhance speed performance at a low supply voltage of 1 V or less, while the high-threshold voltage MOSFET's suppress the stand-by leakage current during the sleep period. This technology has brought about logic gate characteristics of a 1.7-ns propagation delay time and 0.3-/spl mu/W/MHz/gate power dissipation with a standard load. In addition, an MTCMOS standard cell library has been developed so that conventional CAD tools can be used to lay out low-voltage LSI's. To demonstrate MTCMOS's effectiveness, a PLL LSI based on standard cells was designed as a carrying vehicle. 18-MHz operation at 1 V was achieved using a 0.5-/spl mu/m CMOS process. >

1-V power supply high-speed digital circuit technology with multithreshold-voltage CMOS

Scaling the Si MOSFET is reconsidered. Requirements on subthreshold leakage control force conventional scaling to use high doping as the device dimension penetrates into the deep-submicrometer regime, leading to an undesirably large junction capacitance and degraded mobility. By studying the scaling of fully depleted SOI devices, the important concept of controlling horizontal leakage through vertical structures is highlighted. Several structural variations of conventional SOI structures are discussed in terms of a natural length scale to guide the design. The concept of vertical doping engineering can also be realized in bulk Si to obtain good subthreshold characteristics without large junction capacitance or heavy channel doping. >

Scaling the Si MOSFET: from bulk to SOI to bulk

Al3O3 thin film growth on Si(100) using binary reaction sequence chemistry

Dynamic random-access memory (DRAM) is the building block of modern main memory systems. DRAM cells must be periodically refreshed to prevent loss of data. These refresh operations waste energy and degrade system performance by interfering with memory accesses. The negative effects of DRAM refresh increase as DRAM device capacity increases. Existing DRAM devices refresh all cells at a rate determined by the leakiest cell in the device. However, most DRAM cells can retain data for significantly longer. Therefore, many of these refreshes are unnecessary. In this paper, we propose RAIDR (Retention-Aware Intelligent DRAM Refresh), a low-cost mechanism that can identify and skip unnecessary refreshes using knowledge of cell retention times. Our key idea is to group DRAM rows into retention time bins and apply a different refresh rate to each bin. As a result, rows containing leaky cells are refreshed as frequently as normal, while most rows are refreshed less frequently. RAIDR uses Bloom filters to efficiently implement retention time bins. RAIDR requires no modification to DRAM and minimal modification to the memory controller. In an 8-core system with 32 GB DRAM, RAIDR achieves a 74.6% refresh reduction, an average DRAM power reduction of 16.1%, and an average system performance improvement of 8.6% over existing systems, at a modest storage overhead of 1.25 KB in the memory controller. RAIDR's benefits are robust to variation in DRAM system configuration, and increase as memory capacity increases.

/pdf/raidr-retention-aware-intelligent-dram-refresh-4t380u711e.pdf

RAIDR: Retention-Aware Intelligent DRAM Refresh

A bus architecture is proposed for reducing the operating power of future ULSIs. It uses new types of bus driver circuits and bus receiver circuits to reduce the bus signal swing while maintaining a low standby current. The bus driver circuit has a source offset configuration, achieved by the use of low-V/sub T/ MOSFETs and an internal supply voltage corresponding to the reduced signal swing. The bus receiver circuit has a symmetric configuration with two-level conversion circuits, each of which consists of a transmission gate and a cross-coupled latch circuit. Fast level conversion is achieved without increasing the standby current. The combination of the new bus driver and receiver enables the bus swing to be reduced to one-third that of the conventional architecture while maintaining high-speed data transmission and a low standby current. A test circuit designed and fabricated using 0.3- mu m processes verifies the operation of the proposed architecture. Further improvements in the speed performance are possible with device optimization. >

Sub-1-V swing internal bus architecture for future low-power ULSIs

Low-temperature (77K, 4.2K) operation is proposed for bulk CMOS devices to be used in superfast VLSI applications. Symmetrical variation of the parameters of both n-channel and p-channel MOSFETs with respect to the temperature and latch-up immunity makes CMOS a very promising device technology at low temperatures. To demonstrate the performance advantage of circuit operation at low temperatures, inverter chains and 16-kb static random-access memories (RAMs) with 2-/spl mu/m gate length were measured. Average propagation delay for an inverter chain has been reduced to 175 ps (77K) and 104 ps (4.2K) from 296 ps at 300K without sacrificing power dissipation. The power-delay product is less 1 fJ, which is the smallest for silicon devices reported to date. The chip select-access time of the RAM has been reduced to 14.3 ns (77K) from 24 ns (300K).

Operation of bulk CMOS devices at very low temperatures

Direct modulation at 40 Gb/s of a 1.3-mum InGaAlAs distributed feedback ridge waveguide laser is experimentally demonstrated. By combination of the high differential gain of an InGaAlAs multiquantum well active layer, a short cavity length of 100 mum, and a low-resistance notch-free grating, it achieves high bandwidth of 29 GHz and high-extinction ratio of 5 dB at 40-Gb/s modulation. Moreover, the laser operates at a record maximum ambient temperature of 60degC under 40-Gb/s directly modulation. It also achieves 40-Gb/s modulated transmission over 2 km with a low power penalty of 0.25 dB at 25degC .

40-Gb/s Direct Modulation With High Extinction Ratio Operation of 1.3- $\mu$ m InGaAlAs Multiquantum Well Ridge Waveguide Distributed Feedback Lasers

Analytical expressions are presented for subthreshold current reduction in a decoded-driver by self-reverse biasing, which is inherently required for low-voltage, low-power, high-speed DRAM's for portable equipment. The scheme involves inserting a switching MOS transistor between the driver circuits and its power supply line. The subthreshold current of the decoded-driver is reduced to the order of 10/sup -3/ in the practical temperature range (250-350 K) with 254 mV of self-reverse biasing voltage, while the delay time is only 3% more than in conventional schemes. The transition time of 1 ms from the operating state to the low subthreshold current state is sufficient to reduce the subthreshold current. The rapid recovery time of 1 ns from the low subthreshold current state does not interrupt the start of normal operation. The subthreshold current reduction was confirmed experimentally using a test chip fabricated with 0.25- mu m technology. >

Subthreshold current reduction for decoded-driver by self-reverse biasing (DRAMs)

A kind of data-line (DL) interference noise in a scaled DRAM cell array is found and studied through analysis. The dynamic behavior of cell arrays due to sense-amplifier operation is derived analytically. Analysis shows that the amount of interference noise is more than three times larger than expected from simple data-line coupling. A novel experimental technique for precise noise determination is developed to verify the analysis. Analytical results are in good agreement with the experimental data. It is found that the interference noise plays a dominant role in determining the operating margin of the DRAM and that a novel process or a cell array architecture for minimizing the interference noise is indispensable in 16-Mb DRAM and beyond. >

Mayu Aoki

Papers

Sub-1-V swing internal bus architecture for future low-power ULSIs

Operation of bulk CMOS devices at very low temperatures

40-Gb/s Direct Modulation With High Extinction Ratio Operation of 1.3- $\mu$ m InGaAlAs Multiquantum Well Ridge Waveguide Distributed Feedback Lasers

Subthreshold current reduction for decoded-driver by self-reverse biasing (DRAMs)

The impact of data-line interference noise on DRAM scaling