Just as Moores law dictates, technology scaling still continues in recent years. Thus increasing the number of on-chip cores continues to be the facto strategy to scale performance while interconnection networks remain important in multi-core chip systems. In this paper, we present a brand-new network-on-chip (NoC) based on full-mesh architecture - Full-Mesh NoC (FMN) that can provide high connectivity with low-radix routers. We give an example of implementation for a 64-node FMN system in 28 nm process. Results show that this FMN can achieve high frequency of 700 MHz while area requirement of each router is less than 0.035 mm2. At last, we simulate a no-buffered FMN and a buffered FMN under uniform random traffic. Results show that the throughput of the no-buffered FMN can achieve the level of high-radix crossbar. The buffered FMN can achieve 2X throughput compared with the traditional buffered 2D-mesh NoC.
We present an inductorless circuit technique for a CMOS limiting amplifier (LA), which consists of an input buffer, third-order broadband gain stages, and an output buffer for driving 50-Ω transmission lines. By employing stream-mode active feedback with negative capacitance circuit (NCC), the bandwidth of the proposed circuit can be effectively enhanced while maintaining a flat frequency response within the −3 dB bandwidth. Based on TSMC 0.18-µm CMOS process, the proposed LA circuit is optimized and implemented. The measurement results shown that the −3 dB bandwidth is 7.2 GHz with a voltage gain of 41 dB, and a data rate of 9 Gb/s is successfully achieved. The fluctuation of the group delay is less than ±25 ps, and the maximum output voltage is 1 Vpp. The measured noise figure of the LA circuit is about 10 dB. Due to the absence of spiral inductors, the die only occupies a core size of 0.3 × 0.2 mm2, and consumes 79 mW from a 1.8 V supply voltage.
One multi-functional antenna with multi-band multi-mode property based on mirroring and scaling method (MSM) is proposed in this letter. The design procession could be divided into three steps: firstly, one former periodic end-fire antenna with bowtie dipoles is mirrored on the same substrate; then the mirrored part is scaled into about two thirds of the original size; finally, parameters are optimized for fine tuning target. This antenna has three different radiation modes at three different bands respectively, i.e. one +y directional mode in the lower band, one −y directional mode in the higher band and one bidirectional mode along both ±y direction in-between the two bands. Both the field distribution and parameter analysis are given to explore the working principle. One prototype is fabricated for verification. With the advantages of planar structure, wide bandwidth and radiation pattern selectivity, the proposed antenna has a potential application in wireless communication system in the future.
In this paper, an improved design of a radiation hardened memory cell (RHMC), based on the SEU (single event upset) physics mechanism and reasonable transistor size, is proposed. The memory cell can enhance the reliability for space radiation environment, which also can offer differential read operation for robust sensing. With the help of 90 nm standard digital CMOS technology, the simulation demonstrates that the proposed radiation hardened memory cell has the ability to recover an SEU in any one sensitive node and provides multiple-node upsets protection. The comparisons for previous several hardened memory cell are also executed, which shows the proposed memory cell keeps advantages of low power and high stability.
We investigate the increased amplitude of oscillating pulse edge developed in a transmission line periodically loaded with series-connected resonant-tunneling diodes (RTDs). In general, the series-connected RTDs in the middle part of the line can be stabilized in one of the multi-stable states that may disable the proper edge oscillation. We find that it becomes possible to design the line to support such a simple oscillating edge by managing the time required for the state transition of bistable states and the oscillation amplitude increases in proportion to the number of series-connected RTDs through SPICE calculations.
In this paper, a novel mainlobe interference suppression method via eigen-projection processing and covariance matrix sparse reconstruction is proposed, which is able to work when the desired signal is present in the training data. Firstly, the proposed method uses the spatial spectrum algorithm to estimate the direction of arrival (DOA) of sources and the power of sources can be estimated by compressive sensing (CS). Then, the eigen-projection matrix is calculated via the result of DOA to suppress the mainlobe interference in echo data. Finally, adaptive weight vector is obtained by SINCM reconstruction. Compared with other methods, the proposed method can achieve better performance and stronger robustness.
This paper proposed and studied a simple and novel displacement sensor that stretching two fiber Bragg gratings (FBGs) directly with a developed lever structure. The sensing principle was presented, and the corresponding theoretical model was derived and validated. Experimental results show that this design has an excellent displacement sensitivity of 23.654 pm/µm and a high resolution of 42 nm within a range of 0∼300 µm. The resonant frequency and dynamic working bandwidth of the sensor are 115 Hz and 0∼50 Hz respectively, which were obtained through finite element method (FEM) and modal testing experiment. The difference method was utilized to decouple temperature sensitivity of the FBGs, and the experimental results indicate that the designed sensor shows good temperature independence. This sensor can be utilized for micro-amplitude displacement measurement in harsh industry environment.
The mapping of IP cores to the topology is one of the most important steps for NoC (Network-on-Chip) design. Metaheuristic algorithms (MAs) are widely employed since the mapping is an NP-hard problem. Most mapping algorithms only consider small-scale NoC and ignore stability. In this letter, a stable metaheuristic algorithm called WOAGA, based on Whale Optimization Algorithm (WOA) and Genetic Algorithm (GA), is proposed for large-scale NoC mapping to achieve the low-energy consumption and stability. In the proposed algorithm, irregular crossover and mutation operations are integrated into the modified WOA. A perturbation is utilized to jump out of local optima effectively. Simulation results show that the proposed algorithm is more stable and achieve better solution with energy consumption reduced significantly.