• ETRI Journal
• /
• v.42 no.4
• /
• pp.468-479
• /
• 2020

The proposed AI processor architecture has high throughput for accelerating the neural network and reduces the external memory bandwidth required for processing the neural network. For achieving high throughput, the proposed super thread core (STC) includes 128 × 128 nano cores operating at the clock frequency of 1.2 GHz. The function-safe architecture is proposed for a fault-tolerance system such as an electronics system for autonomous cars. The general-purpose processor (GPP) core is integrated with STC for controlling the STC and processing the AI algorithm. It has a self-recovering cache and dynamic lockstep function. The function-safe design has proved the fault performance has ASIL D of ISO26262 standard fault tolerance levels. Therefore, the entire AI processor is fabricated via the 28-nm CMOS process as a prototype chip. Its peak computing performance is 40 TFLOPS at 1.2 GHz with the supply voltage of 1.1 V. The measured energy efficiency is 1.3 TOPS/W. A GPP for control with a function-safe design can have ISO26262 ASIL-D with the single-point fault-tolerance rate of 99.64%.

• Journal of the Institute of Electronics Engineers of Korea SD
• /
• v.41 no.1
• /
• pp.71-80
• /
• 2004

A new SSN(Simultaneous Switching Noise) model is presented, which can afford to investigate SSN due to integrated circuit package. It is shown that previous SSN models are not accurate enough to be practical since they do not take decoupling capacitor into account. In this paper, a new SSN model including the decoupling capacitor is developed. It is verified that the model has excellent agreement(within $5\%$ error) with HSPICE simulation which employs TSMC 0.18um CMOS process technology.

• Proceedings of the IEEK Conference
• /
• 2005.11a
• /
• pp.607-610
• /
• 2005

In this paper, Ni Germanosilicide using Ni-Ta/Co/TiN is proposed to improve thermal stability. The sheet resistance of Ni Germanosilicide utilizing pure Ni increased dramatically after the post-silicidation annealing at $600^{\circ}C$ for 30min. However, using the proposed Ni-Ta/Co/TiN structure, low temperature silicidation and wide range of RTP process window were achieved.

• IEIE Transactions on Smart Processing and Computing
• /
• v.5 no.3
• /
• pp.164-168
• /
• 2016

A low insertion-loss, high-isolation switch based on single pole double throw (SPDT) for a 2.4GHz Bluetooth low-energy transceiver is presented in this paper. In order to increase isolation, the body floating technique is implemented. Based on characteristics whereby the ratio of the sizes of the shunt and the series transistors significantly affect the performance of the switches, the device sizes are optimized. A simple matching network is also designed to enhance the insertion loss. Thus, the SPDT switch has high isolation and low insertion loss without increasing the complexity of the circuit. The proposed SPDT is designed and simulated in a complementary metal-oxide semiconductor 65nm process. The switch has a $530{\mu}m{\times}270{\mu}m$ area and achieves 0.9dB, 1.78dB insertion loss and 40dB, 41dB isolation of transmission, reception modes, respectively.

• JSTS:Journal of Semiconductor Technology and Science
• /
• v.12 no.2
• /
• pp.186-192
• /
• 2012

Half-rail stacked drivers are used to reduce power consumption of the drivers for synchronous buck converters. In this paper, the stacked driver is optimized by matching the average charging and discharging currents used by high-side and low-side drivers. By matching the two currents, the average intermediate bias voltage can remain constant without the aid of the voltage regulator as long as the voltage ripple stays within the window defined by the hysteresis of the regulator. Thus the optimized driver in this paper can minimize the power consumption in the regulator. The current matching requirement yields the value for the intermediate bias voltage, which deviates from the half-rail voltage. Furthermore the required capacitance is also reduced in this design due to decreased charging current, which results in significantly reduced die area. The detailed analysis and design of the stacked driver is verified through simulations done using 5V MOSFET parameters of a typical 0.35-${\mu}m$ CMOS process. The difference in power loss between the conventional half-rail driver and the proposed driver is less than 1%. But the conventional half-rail driver has excess charge stored in the capacitor, which will be dissipated in the regulator unless reused by an external circuit. Due to the reduction in the required capacitance, the estimated saving in chip area is approximately 18.5% compared to the half-rail driver.

• JSTS:Journal of Semiconductor Technology and Science
• /
• v.12 no.2
• /
• pp.168-174
• /
• 2012

This paper describes a novel global on-chip interconnect scheme, in which a one UI-delayed symbol as well as the current symbol is sent for easing the sensing operation at receiver end. With this approach, the voltage swing on the channel for reliable sensing can be reduced, resulting in performance improvement in terms of power consumption, peak current, and delay spread due to PVT variations, as compared to the conventional repeater insertion schemes. Evaluation for on-chip interconnects having various lengths in a 130 nm CMOS process indicated that the proposed on-chip interconnect scheme achieved a power reduction of up to 71.3%. The peak current during data transmission and the delay spread due to PVT variations were also reduced by as much as 52.1% and 65.3%, respectively.

• Journal of the Institute of Electronics and Information Engineers
• /
• v.51 no.1
• /
• pp.90-96
• /
• 2014

This paper presents a practical linearization technique for a wide-band bang-bang digital phase locked-loop(BBDPLL) by selecting optimal loop gains. In this paper, limitation of the theoretical design method for BBDPLL is explained, and introduced how to implement practical BBDPLLs with CMOS process. In the proposed BBDPLL, the limited cycle noise is removed by reducing the proportional gain while increasing the integer array and dither gain. Comparing to the conventional BBDPLL, the proposed one shows a small area, low power, linear characteristic. Moreover, the proposed design technique can control a loop bandwidth of the BBDPLL. Performance of the proposed BBDPLL is verified using CppSim simulator.

• JSTS:Journal of Semiconductor Technology and Science
• /
• v.17 no.3
• /
• pp.438-445
• /
• 2017

This paper presents a low-power high-resolution band-pass ${\Sigma}{\Delta}$ ADC for accelerometer applications. The proposed band-pass ${\Sigma}{\Delta}$ ADC consists of a high-performance 6-th order feed-forward ${\Sigma}{\Delta}$ modulator with 1-bit quantization and a low-power, area-efficient digital filter. The ADC is fabricated in 180 nm 1P6M mixed-signal CMOS process with a die area of $5mm^2$. This high-resolution ADC got 90 dB peak signal to noise plus distortion ratio (SNDR) and 96 dB dynamic range (DR) over 4 kHz bandwidth, while the intermediate frequency (IF) is shifting from 100 KHz to 200 KHz. The power dissipation of the chip is 5.6 mW under 1.8 V (digital)/3.3 V (analog) power supply.

• Journal of IKEEE
• /
• v.16 no.3
• /
• pp.265-273
• /
• 2012

Data stability and leakage power dissipation have become a critical issue in scaled SRAM design. In this paper, an advanced 8T SRAM cell improving the read and write stability of data storage elements as well as reducing the leakage current in the idle mode is presented. During the read operation, the bit-cell keeps the noise-vulnerable data 'low' node voltage close to the ground level, and thus producing near-ideal voltage transfer characteristics essential for robust read functionality. In the write operation, a negative bias on the cell facilitates to change the contents of the bit. Unlike the conventional 6T cell, there is no conflicting read and write requirement on sizing the transistors. In the standby mode, the built-in stacked device in the 8T cell reduces the leakage current significantly. The 8T SRAM cell implemented in a 130 nm CMOS technology demonstrates almost 100 % higher read stability while bearing 20 % better write-ability at 1.2 V typical condition, and a reduction by 45 % in leakage power consumption compared to the standard 6T cell. The stability enhancement and leakage power reduction provided with the proposed bit-cell are confirmed under process, voltage and temperature variations.

• JSTS:Journal of Semiconductor Technology and Science
• /
• v.14 no.5
• /
• pp.549-556
• /
• 2014

This paper proposes a highly power-efficient single-inductor multiple-outputs (SIMO) DC-DC converter with a gate charge sharing method in which gate charges of output switches are shared to improve the power efficiency and to reduce the switching power loss. The proposed converter was fabricated by using a $0.18{\mu}m$ CMOS process technology with high voltage devices of 5 V. The input voltage range of the converter is from 2.8 V to 4.2 V, which is based on a single cell lithium-ion battery, and the output voltages are 1.0 V, 1.2 V, 1.8 V, 2.5 V, and 3.3 V. Using the proposed gate charge sharing method, the maximum power efficiency is measured to be 87.2% at the total output current of 450 mA. The measured power efficiency improved by 2.1% compared with that of the SIMO DC-DC converter without the proposed gate charge sharing method.