This paper presents an optimum tuning procedure for high-order low-pass (LP) elliptic filters. Since elliptic filters are often used to satisfy very tight specifications, they often need to be tuned accurately. In this paper, we describe the tuning of one biquad, the ‘tuning biquad’, in a cascade of biquads. It is shown by Matlab simulations that the best choice for the tuning biquad consists of the pole pair with the highest pole Q (‘maximum-Q poles’) combined with the zero pair with the lowest frequency (‘minimum-frequency zeros’). We also show how standard tuning procedures, such as those for the Tow-Thomas biquad, lead to excellent results. As an example, the tuning procedure is performed on a normalized sevent-horder elliptic LP filter.

A delta operator based ladder synthesis procedure is presented for the realization of three different types of transfer functions. The procedure overcomes the numerical problem of the conventional z-domain LDI approach when the filter poles are clustered near z=1 and does not require any s-domain polynomial transformation to achieve spectral factorization [4]. Also, the entire operation involves only rational polynomials.

It is a known fact that blind algorithms have convergence times of an order of magnitude longer than their nonblind counterparts. However, as shown in this letter, the knowledge of a subset of signals can greatly accelerate the convergence of blind source separation. The convergence behavior of the proposed algorithm is compared with the blind-only case.

A new switched-current integrator which accomplishes forward and backward difference integration and achieves double-sampling capabilities is described. The signal-flow graphs (SFGs) of several double-sampling integrators are derived and mapped into transistor circuits. Furthermore, as an example, a biquad based on an exact design and operating in differential and double-sampling mode is described and simulated. It is shown that the pole sensitivity of this biquad is reduced by a factor of 2 and the coefficient spread by a factor of 3 when compared with a bilinearly transformed reference filter.

A computationally simple nonlinearity in the form of a threshold device for the blind separation of sub-Gaussian signals is derived. Convergence is shown to be robust, fast, and comparable to that of more complex polynomial nonlinearities. Together with the known signum nonlinearity for super-Gaussian distributions, which basically is a threshold device with the threshold set to zero, the general threshold nonlinearity (with an appropriate threshold) can separate any non-Gaussian signals.

The spurious-free dynamic range (SFDR) of a MOSFET-C filter can be increased greatly by generating its tuning voltage with a charge pump. In this paper, we apply this technique to build a Sallen-and-Key lowpass filter with a pole frequency of 24 MHz and a pole Q of 3. It has an SFDR better than 50 dB and consumes 16 mW from a 3.3 V supply. Implemented with a double-poly triple-metal 0.6-/spl mu/m CMOS process, it covers an area of only 0.11 mm/sup 2/. In addition to a description of the filter and the charge pump, we also discuss linear and non-linear clock feed-through from the charge pump's own ring oscillator, and derive a formula for the optimum voltage swing at the MOSFET-C network nodes.

This paper presents a new approach to the adaptation of a wavelet filterbank based on perceptual and rate-distortion criteria. The system makes use of a wavelet-packet transform where each subband can have an individual time-segmentation. Boundary effects can be avoided by using overlapping blocks of samples and therefore switching bases is possible at every tree-level without affecting other subbands. A modified psychoacoustic model using perceptual entropy can control the switching of the wavelet filterbank and the individual time-segmentation of every subband allows one to take advantage of temporal masking. Additionally a rate-distortion measure can control the filterbank for lossless audio coding applications or in cases where large coding gains can be achieved without using perceptual criteria. The weight of the perceptual measure as well as the weight of the rate-distortion measure can be selected individually, enabling to trade lossless coding versus perceptual coding.

Many publications and articles describe the electrical properties of muscular tissue. In these articles the terms "low pass filter" and "purely resistive medium" often occur in parallel. From the viewpoint of electrical engineering (namely from a network theoretic viewpoint) one would not expect that a purely resistive medium behaves as a low pass filter. Instead the existence of some capacitive and/or inductive elements would be expected. We present a network theoretic description of the low pass filter effect of muscular tissue. We use a descretized and simplified tissue model, e.g. a network of resistors. For this model the low pass characteristic of muscular tissue will be verified. The fundamental reason for the low pass characteristic of muscular tissue is the fact that electrical activity travels with some constant velocity along a given path (e.g. the muscle fibres) while the measurement of this activity is done at some fixed point. The independant variables "time" and "space" can be mapped onto each other. The low pass filter effect occurs in space domain.

A novel deconvolution algorithm is proposed for ultrasonic (US) echoes from layered materials. The basic idea is to model the reflectivity functions as well-defined sparse sequences and to perform a maximum-likelihood (ML) sequence estimation over the set of these sequences. A tree-based sequence estimation exhibits an exponentially growing complexity, but can in this case be converted into a trellis-based estimation with a linearly growing complexity. Deconvolution results of measured US-signals are given to confirm the theory. They demonstrate that the a priori knowledge introduced by the reduction of the solution space improves the deconvolution considerably.

A by-inspection analysis method for multiphase switched-current (SI) circuits using signal-flow graph (SFG) techniques is presented. The derivation of the SFG is on the transistor level and is primarily useful for the hand analysis and design of small and medium-size SI circuits (e.g., SI filters, decimators, interpolators). Tables of commonly used SI circuits, in which the corresponding SFGs are given, make the derivation easy and fast. Not only the overall discrete-time transfer functions, but also the ones between individual switching phases, are obtainable.

Discrete-time nonlinear systems represented in the state-space form are considered. Input (drive signal) and output (response signal) are assumed to be measurable. The problem of approximating the external behavior of such a system-in the form of an Input-Output (I-O) model-is addressed. It is already known that a system with fading-memory can be uniformly I-O approximated by a nonlinear MA filter fed with the input signal. In this paper, we prove a more general result: Almost any (in a precise sense) continuous nonlinear system is uniformly I-O approximable by a nonlinear ARMA filter fed with the input and output signals. In other words, an I-O model can be designed that tracks the external behavior of the system. We suggest this result constitutes a mathematical justification for the practice of "black box" nonlinear system identification.

Based on the properties of stable all pass functions (defined for delta-operator formulated discrete-time systems) a stability testing procedure is developed in this paper.

In adaptive noise cancelling, linear digital filters have been used to minimize the mean squared difference between filter outputs and the desired signal. However, for non-Gaussian probability density functions of the involved signals, nonlinear filters can further reduce the mean squared difference, thereby improving the signal-to-noise ratio at the system output. This is illustrated with a two-microphone beamformer for cancelling directional interference. In the case of a single uniformly distributed interference, we establish the optimum nonlinear performance limit. To approximate optimum performance, we realize two nonlinear filter architectures, the Volterra filter and the multilayer perceptron. The Volterra filter is also examined for speech interference. The beamformer is adapted to minimize the mean squared difference, but performance is measured with the intelligibility weighted gain. This criterion requires the signal-to-noise ratio at the beamformer output. For the nonlinear processor, this can only be determined when no target components exist in the reference channel of the noise canceller so that the target is transmitted without distortion. Under these ideal conditions and at equal filter lengths, the quadratic Volterra filter improves the intelligibility-weighted gain by maximally 2 dB relative to the linear filter.

Given a data set, the problem of designing a dynamic model either for an observed signal or for an input/output system appears in numerous situations where no physical model is available. The performance of this fitting crucially depends on the model class used. In this paper, the very first issue-that of model order selection-is addressed. Interestingly, the best performance realizable can be estimated directly from the data, without having to choose and test different model classes. An estimator of this performance bound for each order is proposed, allowing the determination of the optimal order, irrespectively of any specific model class.>

The paper proposes: (1) linearization procedure for nonlinear state equations obtained from two-dimensional (2-D) analog nonlinear circuits and; (2) characterization of the equilibrium point stability based on various two-variable (2-V) Hurwitz properties. It is concluded that at the equilibrium point (EP), 2-D nonlinear circuit is asymptotically stable if the characteristic equation obtained from linearized system exhibits 2-V very strict Hurwitz property. The necessary condition for the asymptotic stability imposes the 2-V scattering Hurwitz property on the characteristic polynomial of the linearized system.>

A method of noise reduction based on discrimination function minimization (DFM) is presented. This method requires the definition of a function that discriminates between clean and noisy speech. A particular discrimination function is derived for a certain type of noise, such that the minimization of this function reduces the noise in the noisy speech signal. The DFM noise reduction method does not assume stationarity of the noise over an entire speech period. It is shown how an abrupt change in piecewise-stationary noise can be tracked. The method has been successfully applied to the reduction of automobile noise.>

The authors report a method for the reduction of the near-end echo in the subscriber loop environment for primary-rate ISDN full-duplex data transmission. The technique uses an RC network as the matching input line impedance at the transmitter to reduce the near-end echo. Data rates ranging from the ISDN basic rate with (2B+D) channels to primary rate with (23B+D) channels are covered. The results indicate that a single basic configuration for the balancing line impedance may be adapted to keep the transhybrid loss in the range of about 22 dB in the worst case over the entire frequency range of interest. The balancing procedure for the hybrid takes place in a training sequence during which the component values of the RC matching circuit are adjusted from their initial values by an iterative adaptation procedure. Two techniques for optimizing the component values have been verified. For all cases examined, including worst-case line configurations, the components of the RC matching circuit have converged to final values, for which the minimum transhybrid loss had a maximum above 20 dB.>

A class of practical fast algorithms is introduced for the discrete cosine transform (DCT). For an 8-point DCT only 11 multiplications and 29 additions are required. A systematic approach is presented for generating the different members in this class, all having the same minimum arithmetic complexity. The structure of many of the published algorithms can be found in members of this class. An extension of the algorithm to longer transformations is presented. The resulting 16-point DCT requires only 31 multiplications and 81 additions, which is, to the authors' knowledge, less than required by previously published algorithms.>

A strategy for tactile hearing aids that is useful in detection of acoustic alarm signals is proposed. Through TIPS (for tactile identification of preclassified signals), a significant decrease in learning time and an improvement of recognition performance is achieved compared with currently used devices. Experimental results for different classification methods computed with a database of 200 alarm signals, and the performance of the wearable experimental device SIGNALON are presented. A unifying interpretation of the performance of the applied methods and some preliminary results for further improvements are given.>

A new circuit for the simulation of floating inductors, capacitors, and FDNRs is proposed. The circuit is easy to design and to tune. It permits the realization of impedances with high element values and is therefore well suited for the design of low-frequency active filters.