• Journal of the Korea Institute of Information and Communication Engineering
• /
• v.25 no.9
• /
• pp.1158-1165
• /
• 2021

This paper proposes a CNN accelerator which is optimized Pooling layer operation incorporated in Multiplication And Accumulation(MAC) to reduce the memory size. For optimizing memory and data path circuit, the quantized 8bit integer weights are used instead of 32bit floating-point weights for pre-training of MNIST data set. To reduce chip area, the proposed CNN model is reduced by a convolutional layer, a 4*4 Max Pooling, and two fully connected layers. And all the operations use specific MAC with approximation adders and multipliers. 94% of internal memory size reduction is achieved by simultaneously performing the convolution and the pooling operation in the proposed architecture. The proposed accelerator chip is designed by using TSMC65nmGP CMOS process. That has about half size of our previous paper, 0.8*0.9 = 0.72mm². The presented CNN accelerator chip achieves 94% accuracy and 77us inference time per an MNIST image.

• Applied Chemistry for Engineering
• /
• v.23 no.6
• /
• pp.515-520
• /
• 2012

In this review, the fabrication of silicon based memory capacitor and organic memory thin film transistors (TFTs) was discussed for their potential identification tag applications and biosensor applications. Metal or non-metal nanoparticles (NPs) could be capped with chemicals or biomolecules such as protein and oligo-DNA, and also be self-assembly monolayered on corresponding target biomolecules conjugated dielectric layers. The monolayered NPs were formed to be charging elements of a nano floating gate layer as forming organic memody deivces. In particular, the strong and selective binding events of the NPs through biomolecular interactions exhibited effective electrostatic phenomena in memory capacitors and TFTs formats. In addition, memory devices fabricated as organic thin film transistors (OTFTs) have been intensively introduced to facilitate organic electronics era on flexible substrates. The memory OTFTs could be applicable eventually to the development of new conceptual devices.

• JSTS:Journal of Semiconductor Technology and Science
• /
• v.8 no.1
• /
• pp.21-26
• /
• 2008

We fabricated nano-particles of ZnO, $In_2O_3$ and $SnO_2$ by using the chemical reaction between metal thin films and polyamic acid. The average size and density of these ZnO, $In_2O_3$ and $SnO_2$ nano-particles was approximately 10, 7, and 15 nm, and $2{\times}10^{11},\;6{\times}10^{11},\;2.4{\times}10^{11}cm^{-2}$, respectively. Then, we fabricated nano-floating gate memory (NFGM) devices with ZnO and $In_2O_3$ nano-particles embedded in the devices' polyimide dielectrics and silicon dioxide layers as control and tunnel oxides, respectively. We measured the current-voltage characteristics, endurance properties and retention times of the memory devices using a semiconductor parameter analyzer. In the $In_2O_3$ NFGM, the threshold voltage shift (${\Delta}V_T$) was approximately 5 V at the initial state of programming and erasing operations. However, the memory window rapidly decreased after 1000 s from 5 to 1.5 V. The ${\Delta}V_T$ of the NFGM containing ZnO was approximately 2 V at the initial state, but the memory window decreased after 1000 s from 2 to 0.4 V. These results mean that metal-oxide nano-particles have feasibility to apply NFGM devices.

• Proceedings of the Korean Institute of Electrical and Electronic Material Engineers Conference
• /
• 2002.11a
• /
• pp.35-38
• /
• 2002

In this paper, the characteristics of channel hot electron (CHE) injection for the write operation in a NOR-type SONOS flash memory with common source line were investigated. The thicknesses of he tunnel oxide, the memory nitride， and the blocking oxide layers for the gate insulator of the fabricated SONOS devices were $34{\AA}$, $73{\AA}$, and $34{\AA}$, respectively. The SONOS devices compared to floating gate devices have many advantages, which are a simpler cell structure, compatibility with conventional logic CMOS process and a superior scalability. For these reasons, the introduction of SONOS device has stimulated. In the conventional SONOS devices, Modified Folwer-Nordheim (MFN) tunneling and CHE injection for writing require high voltages, which are typically in the range of 9 V to 15 V. However CHE injection in our devices was achieved with the single power supply of 5 V. To demonstrate CHE injection, substrate current (Isub) and one-shot programming curve were investigated. The memory window of about 3.2 V and the write speed of $100{\mu}s$ were obtained. Also, the disturbance and drain turn-on leakage during CHE injection were not affected in the SONOS array. These results show that CHE injection can be achieved with a low voltage and single power supply, and applied for the high speed program of the SONOS memory devices.

• IEMEK Journal of Embedded Systems and Applications
• /
• v.12 no.4
• /
• pp.193-204
• /
• 2017

Various 3D-stacked DRAMs have been proposed to overcome the memory wall problem. Hybrid Memory Cube (HMC) is a true 3D-stacked DRAM with stacked DRAM layers on top of a logic layer. The logic die is mainly used to implement a memory controller for HMC, and it is connected through a high speed serial link called SerDes with a host that is either a processor or another HMC. In HMC, the serial link is crucial for both performance and power consumption. Therefore, it is important that the link is configured properly so that the required performance should be satisfied while the power consumption is minimized. In this paper, we propose a HMC system model included the high speed serial link to estimate performance accurately. Since the link modeling strictly follows the link flow control mechanism defined in the HMC spec, the actual HMC performance can be estimated accurately with respect to each link configuration. Various simulations are conducted in order to deduce the correlation between the HMC performance and the link configuration with regard to memory utilization. It is confirmed that there is a strong correlation between the achievable maximum performance of HMC and the link configuration in terms of both bandwidth and latency. Therefore, it is possible to find the best link configuration when the required HMC performance is known in advance, and finding the best configuration will lead to significant power saving while the performance requirement is satisfied.

• Proceedings of the Korean Institute of Electrical and Electronic Material Engineers Conference
• /
• 2003.07a
• /
• pp.319-322
• /
• 2003

To match the charge induced by the insulators $CeO_2$ with the remanent polarization of ferro electric SBT thin films, areas of Pt/SBT/Pt (MFM) and those of $Pt/CeO_2/Si$ (MIS) capacitors were ind ependently designed. The area $S_M$ of MIS capacitors to the area $S_F$ of MFM capacitors were varied from 1 to 10, 15, and 20. Top electrode Pt and SBT layers were etched with for various area ratios of $S_M\;/\;S_F$. Bottom electrode Pt and $CeO_2$ layers were respectively deposited by do and rf sputtering in-situ process. SBT thin film were prepared by the metal orgnic decomposition (MOD) technique. $Pt(100nm)/SBT(350nm)/Pt(300nm)/CeO_2(40nm)/p-Si$ (MFMIS) gate structures have been fabricated with the various $S_M\;/\;S_F$ ratios using inductively coupled plasma reactive ion etching (ICP-RIE). The leakage current density of MFMIS gate structures were improved to $6.32{\times}10^{-7}\;A/cm^2$ at the applied gate voltage of 10 V. It is shown that in the memory window increase with the area ratio $S_M\;/\;S_F$ of the MFMIS structures and a larger memory window of 3 V can be obtained for a voltage sweep of ${\pm}9\;V$ for MFMIS structures with an area ratio $S_M\;/\;S_F\;=\;6$ than that of 0.9 V of MFS at the same applied voltage. The maximum memory windows of MFMIS structures were 2.28 V, 3.35 V, and 3.7 V with the are a ratios 1, 2, and 6 at the applied gate voltage of 11 V, respectively. It is concluded that ferroelectric gate capacitors of MFMIS are good candidates for nondestructive readout-nonvolatile memories.

• The Korean Journal of Archival Studies
• /
• no.60
• /
• pp.123-159
• /
• 2019

Given that the concept of reproducing landscape is similar to that of recording historical sights, places can become special space where memories are archived through meaningful activities. Therefore, place and landscape are the important concepts for understanding the Home of Roh Moo-hyun. This research was initiated when Roh Moo-hyun Foundation's decided to return the Home of Roh Moo-hyun to the public. A research report was published as the first result of this initiative. Then an archive book was recently published based on the first research report. The research report was about philosophical and aesthetic meanings and contents, the layers of accumulated memories, the records based on the accumulated memories, and the attributes of the place, and the possibility of archiving, whereas the purpose of the archive book is to restore and to curate the original meaning of the Home of Roh Moo hyun through cultural events. There are 'three memories' of layers in the Home of Roh Moo-hyun. The first memory is about 'life and dreams' that President Roh Moo-hyun dreamed about after his retirement to the hometown. The second memory is about 'the loss of time' for 10 years of time after the decrease of the President Roh Moo-hyun. The third memory is 'the memory of citizens', which started with the public opening of the Home of Roh Moo-hyun. 'Low Roof House of President Roh Moo-hyun' is the archive book that comprises the three memories which are accumulated in the home of Roh Moo-hyun and 'record language' full of meanings.

• 한국정보디스플레이학회:학술대회논문집
• /
• 2004.08a
• /
• pp.37-40
• /
• 2004

Electrochromic displays offer the possibility of providing high brightness in reflective mode due to the fact that no polarization of the incident or reflected light is required. The use of appropriately roughened reflective layers can enable the diffuse or Lambertian reflection of light. truly imitating the optical nature of paper as a reading medium. Furthermore, the use of an electrochromic system, essentially an electrochemical device, allows for the integration of charge storage layers, endowing such devices with an on-state memory and associated low power consumption. In this paper we describe the $NanoChromics^{TM}$ system and overview the several strategies employed towards the migration of this technology to flexible display formats.

• 제어로봇시스템학회:학술대회논문집
• /
• 2005.06a
• /
• pp.1098-1100
• /
• 2005

The floorplanning problem is an essential design step in VLSI layout design and it is how to place rectangular modules as density as possible. And then, as the DSM advances, the VLSI chip becomes more congested even though more metal layers are used for routing. Usually, a VLSI chip includes several buses. As design increases in complexity, bus routing becomes a heavy task. To ease bus routing and avoid unnecessary iterations in physical design, we need to consider bus planning in early floorplanning stage. In this paper, we propose a floorplanning method for topological constraint consisting of bus constraint and memory constraint. The proposed algorithms based on Genetic Algorithm(GA) is adopted a sequence pair. For selection control, new objective functions are introduced for topological constraint. Studies on floor planning and cell placement have been reported as being applications of GA to the LSI layout problem. However, no studies have ever seen the effect of applying GA in consideration of topological constraint. Experimental results show improvement of bus and memory constraint.

• Journal of the Korean Institute of Telematics and Electronics D
• /
• v.35D no.12
• /
• pp.75-82
• /
• 1998

Many multiplexing techniques are proposed for high storage densities in a volume hologram. In this paper, we present a hybrid angularly and spatially multiplexed volume holographic memory system. Multiple holograms are recorded by using reference and object waves with different incident angles and positions that are changed by step motors. A hologram is written by exposing the crystal with recording time schedule to the interference pattern of the object beam and a reference plane wave. Finally, we show experimental results of the storage of three layers of 300 multiplexed holograms in a LiNbO$_3$ : Fe crystal.