In this work, a Conditional Generative Adversarial Network (CGAN) [1] is demonstrated based on the Resistive Random Access Memory (RRAM). During training, the read noise of RRAM is utilized as a random bias source to ...
详细信息
ISBN:
(数字)9781728165585
ISBN:
(纸本)9781728165592
In this work, a Conditional Generative Adversarial Network (CGAN) [1] is demonstrated based on the Resistive Random Access Memory (RRAM). During training, the read noise of RRAM is utilized as a random bias source to enrich the diversity of the generator in CGAN. Further, we evaluate the impact of both read noise (RRAM as weight storage cell) and the resolution of the AD/DA circuit on the performance of CGAN through a comprehensive simulation.
Electroencephalogram (EEG) has significant applications on medical diagnosis and Brain Computer Interface (BCI). But the main obstacle of analyzing EEG signal is various types of noises to get actual information. Extr...
Electroencephalogram (EEG) has significant applications on medical diagnosis and Brain Computer Interface (BCI). But the main obstacle of analyzing EEG signal is various types of noises to get actual information. Extracting important features is a key issue in this study. This paper uses the BCI Competition IV 2b motion imagery data, in which we provide a review of various prior art to determine the motion imaginary MI mission. Using machine learning to identify two different movements in the EEG signal, the data from nine subjects were analyzed by principal component analysis (PCA) combined with discrete wavelet (DWT) packet analysis. The extracted DWT feature is input into the support vector machine (SVM) classifier, and the experimental results shows that this method is better than traditional methods with a classification accuracy rate of 86.7%.
Two low voltage dual-modulus frequency divider based on extended true single-phase clock (E-TSPC) logic are proposed. By reducing the number of serial transistors from VDD to GND, the proposed designs can effectively ...
详细信息
One design technique that aims to reduce power consumption of MOS current mode logic dual-modulus frequency divider is presented in this paper. With combinational logic transferred the proposed scheme can obtain high ...
详细信息
A 12-bit 2.6 GS/s radio frequency digital to analog converter(RF DAC)based on 1 um GaAs heterojunction bipolar transistor(HBT)process is *** DAC integrates a 4:1 multiplexer to reduce the data rate of input ports,whic...
详细信息
A 12-bit 2.6 GS/s radio frequency digital to analog converter(RF DAC)based on 1 um GaAs heterojunction bipolar transistor(HBT)process is *** DAC integrates a 4:1 multiplexer to reduce the data rate of input ports,which greatly facilitates the *** core adopts 4+8 segmented current steering structure.R-2R ladder network is used for 8 least significant bit(LSB)to realize binary current weighting and thermometer coding is used for 4 most significant bit(MSB).Return-to-zero(RZ)technology is used to expand the effective bandwidth of DAC output to the third Nyquist *** proposed DAC has a better output power flatness and spurious-free dynamic range(SFDR).Compared to traditional DAC,measured results demonstrate that the output power of this RZ DAC is increased by 33 dB and the SFDR is enhanced by 27 dB near the second Nyquist band.
This paper presents a novel true random number generator with entropy source based on swappable matrix feedback ring oscillator (SMFRO). The proposed SMFRO structure utilizes multiple feedback mechanism to produce rel...
This paper presents a novel true random number generator with entropy source based on swappable matrix feedback ring oscillator (SMFRO). The proposed SMFRO structure utilizes multiple feedback mechanism to produce reliable chaotic oscillation. Three SMFROs implemented on FPGA with different resource configurations are tested. All of them can pass chaotic oscillation and NIST tests. Proposed design can generate random bits at 500MHz without extra post-processing.
Binary neural networks (BNNs) - neural networks with binary {-1, +1} weights and activations at run-time, have attracted much attention in recent years, which drastically reduce resource requirement and power consumpt...
Binary neural networks (BNNs) - neural networks with binary {-1, +1} weights and activations at run-time, have attracted much attention in recent years, which drastically reduce resource requirement and power consumption with tolerable accuracy loss. While, spiking neural networks (SNNs) are designed as a brain-inspired computing model, which can be deployed on large-scale distributed and event-based neuromorphic hardware efficiently. To enable a conversion of above two, we develop a variant of BNNs which employ binary-valued {0, +1} activations and ternary-valued {-1, 0, +1} weights and achieve start-of-the-art classification accuracy across two standard vision datasets, even better than the previous proposals using BNNs or SNNs. Furtherly, we propose an effective configuration method for network deployment on our designed programmable neuro-synaptic system with only one timestep and milliwatt level power consumption.
In this paper, we successfully realized a shortest 10-nm channel length transistor based on ultrathin 3.5-nm indium tin oxide channel. Using 5-nm lanthanum-doped hafnium oxide (HfLaO) high- κ dielectric, the 10- nm c...
详细信息
ISBN:
(数字)9781728188881
ISBN:
(纸本)9781728188898
In this paper, we successfully realized a shortest 10-nm channel length transistor based on ultrathin 3.5-nm indium tin oxide channel. Using 5-nm lanthanum-doped hafnium oxide (HfLaO) high- κ dielectric, the 10- nm channel ultrathin device architecture with wide bandgap exhibits excellent switching behavior with on/off ratio exceeding 10 10 and ultra-low leakage current of 40 fA/μm. Record-high on-state current of 1860 μA/μm and transconductance (g m ) of more than 1000 μS/μm have been achieved, benefited from a rather small contact resistance of 162 Ω·μm. R on of the 10-nm ITO transistor was smaller than 500 Ω·μm. Adopting the MIT Virtual Source (MVS) model, we extracted the saturation injection velocity ν x0 up to 8.8 × 10 6 cm/s and the resulting mean free path λ mfp of ITO transistor is 21.6 nm. Radio frequency transistor with 30-nm-long channel exhibited record high radio-frequency (RF) performance with small-signal current gain (f T ) of 20 GHz and maximum oscillation frequency (f max ) of 13 GHz. These two metrics contribute to √(f T × f max ) exceeding 15 GHz, confirming overwhelming superiority compared to RF transistors based on other ultrathin novel channel materials such as MoS 2 , BP and metal-oxides. Finally, we employed bootstrapped mode (BST) inverters to fabricate a 5-stage ring oscillator, and achieved the record-low propagation delay of 0.4 ns/stage among metal-oxides.
Dear editor,Ge is considered as a promising channel material to replace Si because of its high carrier mobility than Si and compatibility with conventional Si process[1–3].Strain engineering has been widely used in S...
详细信息
Dear editor,Ge is considered as a promising channel material to replace Si because of its high carrier mobility than Si and compatibility with conventional Si process[1–3].Strain engineering has been widely used in Si-based devices and can also be adopted
暂无评论