Using interval phase margin assignment to self-tune a PI AQM controller for TCP traffic

We propose a self-tuning PI (Proportional-Integral) controller for an AQM (Active Queue Management) router supporting TCP traffic in the Internet. Classical control theory is applied in the controller design to meet the phase margin specification in the frequency domain. By assigning a proper interval of the phase margin, we can achieve good AQM performance by making the control system adapt to dramatic load changes. Our self-tuning PI controller self-tunes only when there is a great change in the network environment that would cause the phase margin of the AQM control system to drift outside the specified interval. Based on the knowledge of the queue size, our PI controller can regulate the TCP source window size by adjusting the packet drop probability, thus clamping the steady queue size around a desirable target buffer occupancy. We demonstrate by OPNET^® simulations that with our self-tuning PI controller applied, the network exhibits a good transient behavior. A simple PID (Proportional-Integral-Derivative) controller design method is also provided.     

Convergence and loss bounds for Bayesian sequence prediction

The probability of observing x/sub t/ at time t, given past observations x/sub 1/...x/sub t-1/ can be computed if the true generating distribution /spl mu/ of the sequences x/sub 1/x/sub 2/x/sub 3/... is known. If /spl mu/ is unknown, but known to belong to a class /spl Mscr/ one can base one's prediction on the Bayes mix /spl xi/ defined as a weighted sum of distributions /spl nu/ /spl isin/ /spl Mscr/. Various convergence results of the mixture posterior /spl xi//sub t/ to the true posterior /spl mu//sub t/ are presented. In particular, a new (elementary) derivation of the convergence /spl xi//sub t///spl mu//sub t/ /spl rarr/ 1 is provided, which additionally gives the rate of convergence. A general sequence predictor is allowed to choose an action y/sub t/ based on x/sub 1/...x/sub t-1/ and receives loss /spl lscr//sub x(t)y(t)/ if x/sub t/ is the next symbol of the sequence. No assumptions are made on the structure of /spl lscr/ (apart from being bounded) and /spl Mscr/. The Bayes-optimal prediction scheme /spl Lambda//sub /spl xi// based on mixture /spl xi/ and the Bayes-optimal informed prediction scheme /spl Lambda//sub /spl mu// are defined and the total loss L/sub /spl xi// of /spl Lambda//sub /spl xi// is bounded in terms of the total loss L/sub /spl mu// of /spl Lambda//sub /spl mu//. It is shown that L/sub /spl xi// is bounded for bounded L/sub /spl mu// and L/sub /spl xi///L/sub /spl mu// /spl rarr/ 1 for L/sub /spl mu// /spl rarr/ /spl infin/. Convergence of the instantaneous losses is also proven.     

Time-frequency localisation of Shannon wavelet packets

It is well known that the 2/spl pi/ minimally supported frequency scaling function /spl phi//sup /spl alpha//(x) satisfying /spl phi//spl circ//sup /spl alpha//(/spl omega/)=/spl chi//sub (-/spl alpha/,2/spl pi/-/spl alpha/)/(/spl omega/), 0相似文献   

Theory of Direct-Coupled-Cavity Filters

A new theory is presented for the design of direct-coupled-cavity filters in transmission line or waveguide. It is shown that for a specified range of parameters the insertion-loss characteristic of these filters in the case of Chebyshev equal-ripple characteristic is given very accurately by the formula P/sub 0/ / /P/sub L/ = 1+h/sup 2/T/sub n//sup 2/[/spl omega//sub 0/ / /spl omega/ sin(/spl pi/ /spl omega/ / /spl omega//sub 0/) / sin/spl theta//sub 0/'] where h defines the ripple level, T/sub n/ is the first-kind Chebyshev polynomial of degree n, /spl omega/ / /spl omega//sub 0/ is normalized frequency, and /spl theta//sub 0/' is an angle proportional to the bandwidth of a distributed lowpass prototype filter. The element values of the direct-coupled filter are related directly to the step impedances of the prototype whose values have been tabulated. The theory gives close agreement with computed data over a range of parameters as specified by a very simple formula. The design technique is convenient for practical applications.     

Adaptive AQM controllers for IP routers with a heuristic monitor on TCP flows

We propose adaptive proportional (P) and proportional‐integral (PI) controllers for Active Queue Management (AQM) in the Internet. We apply the classical control theory in the controller design and choose a proper phase margin to achieve good performance of AQM. We have identified a simple heuristic parameter that can monitor the changes of network environment. Our adaptive controllers would self‐tune only when the dramatic change in the network parameters drift the monitoring parameter outside its specified interval. When compared to P controller, a PI controller has the advantage of regulating the TCP source window size by adjusting the packet drop probability based on the knowledge of instantaneous queue size, thus steadying the queue size around a target buffer occupancy. We have verified our controllers by OPNET simulation, and shown that with an adaptive PI controller applied, the network is asymptotically stable with good robustness. Copyright © 2005 John Wiley & Sons, Ltd.     

Analysis of Four-Frequency Nonlinear Reactance Circuits

Several advantages of multiple-frequency nonlinear reactance circuits are described in this paper. In particular, a circuit is considered in which a nonlinear reactance couples four basic frequencies: /spl omega//sub 0/, /spl omega//sub 1/,/spl omega//sub 2/, and /spl omega//sub 3/; these are so related that /spl omega//sub 2/ = /spl omega//sub 0/ + /spl omega//sub 1/ and /spl omega//sub 3/ = /spl omega//sub 0/ - /spl omega//sub 1/. Here, /spl omega//sub 0/ is taken to be the power source or pump. It is found to be desirable to allow for the possible presence of the pump harmonic, 2/spl omega//sub 0/, and individual cases are characterized by whether 2/spl omega//sub 0/, is present or not. The major results are as follows: 1) Unlimited amplification gain is theoretically possible at frequencies higher than the pump, by reflecting negative input resistance at /spl omega//sub 2/, but without relying on any effects due to pump harmonics. 2) Unlimited up- or down-conversion gains between /spl omega//sub 1/ and /spl omega//sub 2/ are theoretically possible in the additional presence of the first pump harmonic, but without reflecting negative input or output resistance. 3) Unlimited amplification gain is theoretically possible at frequencies both lower and higher than the pump fundamental, without reflecting negative input resistance.     

On The General Relation Between /spl alpha/ and Q (Correspondence)

The relation between the quality factor Q and the attenuation constant /spl alpha/ of a transmission line has been known as follows: /spl alpha/ = /spl beta/ / /2Q where /spl beta/ is the phase constant. Recently from the following relation of propagation constant at resonance /spl Gamma/(/spl omega//sub 0/) + /spl part//spl Gamma/ / /spl part//spl omega/ /spl Delta//spl omega//spl cong/ i/spl beta/(/spl omega//sub 0/), where /spl Gamma/(/spl omega//sub 0/) = /spl alpha/(/spl omega//sub 0/) + i/spl beta/(/spl omega//sub 0/). Yeh derived a general relation between Q and /spl alpha/, namely, /spl alpha/ = /spl upsi//sub p/ / /spl upsi//sub g/ /spl beta/ / /2Q where /spl upsi//sub p/, and /spl upsi//sub g/ are the phase velocity and group velocity of the wave respectively. This general relation can be derived very simply from the generally accepted definition of /spl alpha/ and Q.     

On the theory of 1/f noise of semi-insulating materials

The 1/f noise phenomena associated with devices involving semi-insulating materials, for instance GaAs MESFET's on semi-insulating GaAs, has long been a perplexing problem. In this particular case the 1/f noise corner frequency can be up to 100 MHz before the mean square noise current at the drain is dominated by the Nyquist noise associated with the channel conductance. No reasonable explanation has ever been given, although there are many different theories. 1/f noise is a common phenomena in nature and other devices involving semi-insulating materials. We propose here that this 1/f noise is a bulk phenomena associated with localized high frequency variations and long range low frequency fluctuations, the lowest frequency being limited only by the volume of the material. Specifically the proposal here is that injection of a current I into a semi-insulating material will result in a mean square noise voltage at the point of injection given by v/sub n//sup 2/~=2(kT/q)q/spl Delta/fR(/spl omega//sub c///spl omega/) Volts/sup 2/ where /spl omega//sub c/=1/t/sub t/, for the radian frequencies, /spl omega/, larger than /spl omega//sub c/ which is the reciprocal of the transit time of the carriers. For a long sample and long transit times then this 1/f noise voltage due to current injection will be larger than the Nyquist mean square noise of the sample alone as long as the DC voltage developed across the semi-insulating sample exceeds ((2kT/q)l/sup 2/(/spl omega///spl mu/))/sup 1/2/. This theory then gives the 1/f or 1//spl omega/ frequency dependence. The dc current I might be injected for instance by the substrate current in a GaAs MESFET being injected into the semi-insulating substrate, or gate current in an IGET being injected into the gate insulator.<>     

Hybrid-Frequency Cutoffs in Gyrotropic Waveguides (Short Papers)

The behavior of the Brillouin diagram /spl omega/ versus /spl beta//sub z/ is analyzed for axially magnetized gyrotropic waveguides in the vicinity of the hybrid frequencies /spl omega//sub k/ and /spl omega//sub i/. Starting from the exact dispersion relation for the modes in the waveguides under consideration it is shown that 1) the dispersion curves terminate at discrete cutoff points located along the /spl omega/ = spl omega//sub k/, /spl omega//sub i/ line and 2) the group velocity at these points is zero. These results modify the behavior of the dispersion curves existing in the literature.     

A gradient-based track-following controller optimization for hard disk drive

This paper presents a gradient-based parameter optimization method to find the optimal compensator that minimizes the standard deviation (/spl sigma//sub PES/) of the position error signal (PES) in a hard disk drive servo system. By using the plant response data and the PES gradient information based on the nominal plant model, optimal digital controllers that minimized the 3/spl sigma//sub PES/ of a plant with uncertainty were selected within a pre-found robust stable region. As a result, an optimal track-following controller that minimized the standard deviation of the measured PES (/spl sigma//sub PESm/) was able to be obtained without the prior knowledge of the disturbance and noise model. Furthermore, we proved that if the measurement noise is white, an optimal controller that minimizes the 3/spl sigma//sub PESm/ also minimizes the 3/spl sigma//sub PES/. Both simulation and implementation results suggest that such a gradient-based search process is faster than nongradient optimization methods such as random neighborhood search and genetic algorithms.     

1677 V, 5.7 m/spl Omega//spl middot/cm/sup 2/ 4H-SiC BJTs

This letter reports the development of a high-performance power 4H-SiC bipolar junction transistor (BJT) with, simultaneously, a high blocking voltage and a low specific on-resistance (R/spl I.bar//sub ON/). A single BJT cell with an active area of 0.61 mm/sup 2/ achieves an open base collector-to-emitter blocking voltage (V/sub ceo/) of 1677 V and conducts up to 3.2 A at a forward voltage drop of V/sub CE/=3.0 V, corresponding to a low R/spl I.bar//sub ON/ of 5.7 m/spl Omega//spl middot/cm/sup 2/ up to Jc=525 A/cm/sup 2/ and a record high value of V/sub B//sup 2//R/sub SP/spl I.bar/ON/ of 493 MW/cm/sup 2/.     

Dissipative Parameters in Ferrites and Insertion Losses in Waveguide Y-Circulators Below Resonance

Extensive microwave loss measurements have been performed at frequencies from 1.3 to 11 GHz on below-resonance waveguide Y circulators loaded with a wide variety of ferrite and garnet compositions. Dissipative internal and external magnetic parameters have been measured on the same compositions. Also, dielectric loss measurements have been carried out. Two classes have been distinguished, defined by the following conditions: /spl omega//sub M/ / /spl omega/ /spl les/0.8 and 0.85 /spl les//spl omega//sub M/ / /spl omega/ /spl les/1.05. It is inferred that the (insertion loss) IL of such devices is independent of /spl Delta/H and mainly depends on the internal dissipative susceptibility x/sub i/" and on the dielectric loss tan /spl delta/. The relation of the IL versus x/sub i/" and tan /spl delta/ in the case /spl omega//sub M/ / /spl omega/ /spl les/0.8 is independent of frequency and given by the semiempirical equation IL= 10 log/sub 10/ (1-2.85 x/sub i/" - 1.60 tan /spl delta/ - 0.017)/sup -1/.     

Transients on multiconductor transmission lines above dissipative earth-numerical simulation and measurement

A finite-difference time-domain (FDTD) analysis is presented to simulate, up to the high-frequency regime, transients on multiconductor transmission lines above dissipative earth. Excitation is in the high-frequency regime if the pulse has significant spectral content up to /spl omega//spl ap//spl omega//sub c/, with /spl omega//sub c/ the frequency where the conduction and displacement current densities have equal magnitude. Thus, /spl omega//sub c/=/spl sigma//sub g///spl epsi//sub 0//spl epsi//sub rg/ with /spl sigma//sub g/ and /spl epsi//sub rg/ the ground conductance and relative permittivity. The FDTD algorithm extends the formulation of Agrawal et al. (1980), by incorporating the theory of D'Amore and Sarto for the interaction of wide-band transients with the earth. Careful experimental testing of the new algorithm is presented. The measurement configuration consists of two parallel wires; first suspended above an aluminum sheet to test the code in the low-frequency regime; then above a block of absorbing material for experiments in the high-frequency regime. The end of one of the conductors is driven with a 2-ns synthesized impulse by a stepped-frequency, automatic network analyzer. Reflected and cross-talk signals are presented in the time domain. These show good agreement with the numerically predicted common mode, differential mode and mode-converted pulses.     

The inequalities of quantum information theory

Let /spl rho/ denote the density matrix of a quantum state having n parts 1, ..., n. For I/spl sube/N={1, ..., n}, let /spl rho//sub I/=Tr/sub N/spl bsol/I/(/spl rho/) denote the density matrix of the state comprising those parts i such that i/spl isin/I, and let S(/spl rho//sub I/) denote the von Neumann (1927) entropy of the state /spl rho//sub I/. The collection of /spl nu/=2/sup n/ numbers {S(/spl rho//sub I/)}/sub I/spl sube/N/ may be regarded as a point, called the allocation of entropy for /spl rho/, in the vector space R/sup /spl nu//. Let A/sub n/ denote the set of points in R/sup /spl nu// that are allocations of entropy for n-part quantum states. We show that A~/sub n/~ (the topological closure of A/sub n/) is a closed convex cone in R/sup /spl nu//. This implies that the approximate achievability of a point as an allocation of entropy is determined by the linear inequalities that it satisfies. Lieb and Ruskai (1973) have established a number of inequalities for multipartite quantum states (strong subadditivity and weak monotonicity). We give a finite set of instances of these inequalities that is complete (in the sense that any valid linear inequality for allocations of entropy can be deduced from them by taking positive linear combinations) and independent (in the sense that none of them can be deduced from the others by taking positive linear combinations). Let B/sub n/ denote the polyhedral cone in R/sup /spl nu// determined by these inequalities. We show that A~/sub n/~=B/sub n/ for n/spl les/3. The status of this equality is open for n/spl ges/4. We also consider a symmetric version of this situation, in which S(/spl rho//sub I/) depends on I only through the number i=/spl ne/I of indexes in I and can thus be denoted S(/spl rho//sub i/). In this case, we give for each n a finite complete and independent set of inequalities governing the symmetric allocations of entropy {S(/spl rho//sub i/)}/sub 0/spl les/i/spl les/n/ in R/sup n+1/.     

Input Impedance Analysis of 1:1 Balun

The 1:1 balun constructed of a bifilar delay line and an inverse 1:1 transformer is analyzed for its input characteristics. As a main result of the analyses, the following results were obtained about the balun in this paper. 1) Resonance occurs when the line length is equal to n (any positive integer) times half of a wavelength of the unbalanced transmission mode. 2) The lower cutoff frequency is determined by electromagnetic coupling coefficient "k/sub m/" when L/sub sigma/ and /spl iota/ are constant. 3) The upper cutoff frequency is also given by a combination of k/sub m/ and k/sub e/, etc. 4) Shortening of the distance between the balun and the ground makes the resonant frequency lower when the balun has a magnetic core or when the electrostatic coupling coefficient "k/sub e/" becomes smaller and k/sub m/ is kept constant. /spl omega//sub OI/ becomes lower. 5) In order to realize a wide-band balun, /spl beta//sub b/ / /spl beta//sub u/ shouId be larger and and /spl beta//sub b/ / /spl omega/ should be kept constant.     

Microwave Characterization of Partially Magnetized Ferrites

In order to assist the microwave engineer in predicting the performance of partially magnetized devices, we have characterized the microwave permeability of partially magnetized materials. The real part of the tensor permeability elements, /spl mu/, /spl kappa/, and /spl mu//sub z/, depends primarily on the parameters /spl gamma/4/spl pi/M//spl omega/ and /spl gamma/4/spl pi/M/sub s/ / /spl omega/. Empirical formulas have been developed which show the dependence. At frequencies sufficiently below /spl omega/ = /spl pi/4/spl pi/M/sub s/, the loss can be characterized by the value of /spl mu/' at 4/spl pi/M = 0./spl mu/, /spl kappa/, and /spl mu//sub z/ depend weakly on composition, whereas /spl mu/' (4/spl pi/M = 0) does depend upon the chemical composition.     

Demonstration of 1789 V, 6.68 m/spl Omega/ /spl middot/ cm/sup 2/ 4H-SiC merged-PiN-Schottky diodes

The design, fabrication and characterisation of a high performance 4H-SiC diode of 1789 V-6.6 A with a low differential specific-on resistance (R/sub SP/spl I.bar/ON/) of 6.68 m/spl Omega/ /spl middot/ cm/sup 2/, based on a 10.3 /spl mu/m 4H-SiC blocking layer doped to 6.6/spl times/10/sup 15/ cm/sup -3/, is reported. The corresponding figure-of-merit of V/sub B//sup 2//R/sub SP/spl I.bar/ON/ for this diode is 479 MW/cm/sup 2/, which substantially surpasses previous records for all other MPS diodes.     

Mixed-objective optimization of a track-following controller using linear matrix inequalities

This work describes a design method for track-following controllers in data storage drives using mixed-objective optimization. Stability and performance requirements on the track following controller are established using norm conditions, which are converted to linear matrix inequalities (LMIs). The minimization of the position error in the root-mean-square sense is shown to be an H/sub 2/ norm minimization problem, whereas the vibration rejection requirements are expressed as H/sub 2/ or H/sub /spl infin// norm constraints. The robust stability requirement is enforced by an H/sub /spl infin// norm constraint. The mixed-norm problem is converted into a constrained minimization problem in LMIs, which is solved by convex optimization algorithms. The controller is designed directly in the discrete time domain to avoid typical degradation caused by continuous-time to discrete-time conversion. The proposed controller is implemented and tested on production hard disk drives. Experiments show that the proposed controller reduces the position error by 7%-11% while maintaining comparable stability margins and vibration rejection capability compared with the conventional controllers.     

A Nonreflective Hybrid Stop-Band Filter (Correspondence)

For operation of parametric amplifiers at low frequencies and at low-impedance levels a parametric diode of very large capacitance is necessary. In general a capacitance which has an impedance of the same order as the source impedance is required, so that for a 600-ohm system at a frequency of 5 kc the static capacitance C/sub 0/ should be C/sub 0//spl tildeoverbar/ 1//spl omega//sub s/R/sub g/ /spl tildeoverbar/0.05 ~f. The capacitance variation with bias should also be large.     

Investigation of the source/drain asymmetric effects due to gate misalignment in planar double-gate MOSFETs

A planar double-gate SOI MOSFET (DG-SOI) with thin channel and thick source/drain (S/D) was successfully fabricated. Using both experimental data and simulation results, the S/D asymmetric effect induced by gate misalignment was studied. For a misaligned DG-SOI, there is gate nonoverlapped region on one side and extra gate overlapped region on the other side. The nonoverlapped region introduces extra series resistance and weakly controlled channel, while the extra overlapped region introduces additional overlap capacitance and gate leakage current. We compared two cases: bottom gate shift to source side (DG/spl I.bar/S) and bottom gate shift to drain side (DG/spl I.bar/D). At the same gate misalignment value, DG/spl I.bar/S resulted in a larger drain-induced barrier lowering effect and smaller overlap capacitance at drain side than DG/spl I.bar/D. Because of reduced drain-side capacitance, the speed of three-stage ring oscillator of DG/spl I.bar/S, with 20% gate misalignment length (L/sub mis/) over gate length (L/sub g/), or L/sub mis//L/sub g/=20%, was faster than that of two-gate aligned DG-SOI.