Variable Bandwidth Smoothing Based on Cluster Analysis

Piecewise variable bandwidth local polynomial smoothing was studied. A local polynomial smoothing method was given based on kernel functions and the optimal variable bandwidth choice was discussed using cluster analysis. Computer simulations show that the current method is better than Donoho's wavelet shrinkage method.

Local Linear Estimation by TSLS with Variable Bandwidth for Semi-parametric Simultaneous Equation Models in Econometrics

Econometric simultaneous equation models play an important role in making economic policies, analyzing economic structure and economic forecasting. This paper presents local linear estimators by TSLS with variable bandwidth for every structural equation in semi-parametric simultaneous equation models in econometrics. The properties under large sample size were studied by using the asymptotic theory when all variables were random. The results show that the estimators of the parameters have consistency and asymptotic normality, and their convergence rates are equal to n^-1/2. And the estimator of the nonparametric function has the consistency and asymptotic normality in interior points and its rate of convergence is equal to the optimal convergence rate of the nonparametric function estimation.

A compact and reconfigurable low noise amplifier employing combinational active inductors and composite resistors feedback techniques

A compact and reconfigurable low noise amplifier(LNA)is proposed by combining an input transistor,composite transistors with Darlington configuration as the amplification and output transistor,T-type structure composite resistors instead of a simplex structure resistor,a shunt inductor feedback realized by a tunable active inductor(AI),a shunt inductor peaking technique realized by another tunable AI.The division and collaboration among different resistances in the T-type structure composite resistor realize simultaneously input impedance matching,output impedance matching and good noise performance;the shunt feedback and peaking technique using two tunable AIs not only extend frequency bandwidth and improve gain flatness,but also make the gain and frequency band can be tuned simultaneously by the external bias of tunable AIs;the Darlington configuration of composite transistors provides high gain;furthermore,the adoption of the small size AIs instead of large size passive spiral inductor,and the use of composite resistors make the LNA have a small size.The LNA is fabricated and verified by GaAs/InGaP hetero-junction bipolar transistor(HBT)process.The results show that at the frequency of 7 GHz,the gain S_(21)is maximum and up to 19 dB;the S_(21)can be tuned from 17 dB to 19 dB by tuning external bias of tunable AIs,that is,the tunable amount of S_(21)is 2 dB,and similarly at 8 GHz;the tunable range of 3 dB bandwidth is 1 GHz.In addition,the gain S_(21)flatness is better than 0.4 dB under frequency from 3.1 GHz to 10.6 GHz;the size of the LNA only has 760μm×1260μm(including PADs).Therefore,the proposed strategies in the paper provide a new solution to the design of small size and reconfigurable ultra-wideband(UWB)LNA and can be used further to adjust the variations of gain and bandwidth of radio frequency integrated circuits(RFICs)due to package,parasitic and the variation of fabrication process and temperature.

An On-line Finite Element Temperature Field Model for Plate Ultra Fast Cooling Process

Taking the element specific-heat interpolation function into account, a one-dimensional （l-D） finite ele- ment temperature field model for the on-line control of the ultra fast cooling process was developed based on the heat transfer theory. This 1-D model was successfully implemented in one 4 300 mm plate production line. To improve the calculation accuracy of this model, the temperature-dependent material properties inside an element were considered during the modeling process. Furthermore, in order to satisfy the real-time requirements of the on-line model, the variable bandwidth storage method and the Cholesky decomposition method were used in the programming to storage the data and carry out the numerical solution. The on-line application of the proposed model indicated that the devia- tion between the calculated cooling stop temperature and the measured one was less than ± 15 ℃.