Nec Corporation

Nippon Electric Co., Ltd.

Nippon Telegraph and Telephone Public Corporation

Shin'ichi Koike

Eiichi Kobayashi

Masaru Yamaguchi

Masanobu Arai

Yuji Inoue

Haruhiko Tsuchiya

Takenori Ogata

Tatsuhiro Ono

An adaptive filter which is realized as an FIR filter employing stochastic gradient algorithm comprises a step size controller for controlling step size &agr;,so as to minimize the mean squared estimation error &egr;,at each time n. The step size controller regards a step size &agr;,which satisfies &dgr;&egr;,/&dgr;&agr;,&equals;0 at each time n as a theoretical optimum step size, and sets the step size &agr;,at each time n so as to be approximately equal to the theoretical optimum step size &agr;,. For example, when the adaptive filter employs stochastic gradient sign algorithm as the tap weight control algorithm, the step size controller provided with a leakage accumulator calculates a vector q,one by one according to the update equation: q,&equals;(1&minus;&rgr;)q,&plus;&rgr;sgn(e,&plus;&ugr;,)(&verbar;e,&plus;&ugr;,&verbar;),a,(&rgr;: leakage coefficient, e,: an error signal sequence component, &ugr;,: an additive noise sequence component, a,: an input signal sequence vector), and sets the step size &agr;,at each time n as: &agr;,&equals;&bgr;&par;q,&par;,(&bgr;&equals;1.07870520/trace(Ra,), Ra&equals;E&lsqb;a,a,&rsqb;). By the step size controller, both faster convergence of the adaptive filter and smaller residual mean squared estimation error after convergence can be attained.

Adaptive filter, step size control method thereof, and record medium therefor

An rms value of a sum signal series x=en+vn calculated by adding a noise series to an error signal series of a known signal series and an unknown signal series, is calculated by an rms calculator. By means of an amplitude controller, x is output as is when .linevert split.x.linevert split. is smaller than a threshold value A proportional to the rms value. When .linevert split.x.linevert split. is greater than the threshold value A, a product signal calculated by multiplying the polarity of x with the threshold value is output. Utilizing the output of the amplitude controller, respective tap weights of filter is corrected to generate be tap weights Ck.sup.(n). By this, switching control between a probability gradient LMS algorithm (LA) and a probability gradient sign algorithm (SA) can be performed smoothly. Thus, an adaptively controlled filter having an advantage of small residual error and fast convergence of LA and an advantage of satisfactory strength against disturbance, such as burst noise or so forth of SA, can be obtained.

Adaptively controlled filter

An input digital signal is applied to a subtractor where the intersymbol interference component of the signal is removed by a cancelling signal and the ISI-free signal is fed to a decision circuit to produce a binary output, from which a decision error is then detected. A greater part of the cancelling signal is generated by a nonrecursive filter connected to the decision circuit and the remainder is generated by a second-order recursive filter whose tapped delay line is connected through an adder to the output of the tapped delay line of the nonrecursive filter. The tap signals of the recursive filter are respectively multiplied with first and second feedback coefficients and applied to the adder as feedback signals. The first feedback coefficient is equal to a sum of an attenuation constant of constant value (r.sub.f) and an attenuation coefficient of variable quantity (r) and the second feedback coefficient is equal to a product of the two attenuation coefficients. The variable quantity attenuation coefficient is adaptively updated by a correlation between the decision error and a signal derived from the output of the recursive filter.

Decision feedback equalizer with second-order recursive filter

An echo canceller comprises a finite impulse response filter (6) connected to the incoming port of a hybrid (1), and a tail canceller (7) which includes an infinite impulse response filter (24; 40) having a multiplier (14; 33) for recursively multiplying an output digital symbol from the tapped-delay line (8) of the FIR filter with an attenuation coefficient R. A digital symbol from the IIR filter or one from the tapped-delay line (8) is multiplied with a tap-weight coefficient C.sub.1 and summed with the outputs of tap-weight multipliers of the FIR filter to produce a replica of an echo symbol, which is subtracted from a digital symbol y.sub.n from the outgoing port of the hybrid (1), producing a residual echo symbol e.sub.n. The coefficient C.sub.1 is derived from correlation between residual echo e.sub.n and a symbol from the tapped-delay line. In one embodiment, symbol from the tapped-delay line is further delayed (17) for a unit interval on the one hand, and multiplied (18) with coefficient R on the other, the difference between them being detected (19) and multiplied (20) with a coefficient K, which is either sgn(C.sub.1) or 1/C.sub.1, to generate a multiplied difference. Correlation is detected (22) between the difference and residual echo e.sub.n to derive the coefficient R. In a second embodiment, symbol from the tapped-delay line is further delayed (36) for a unit interval and multiplied (37) with K to produce a delayed-multiplied symbol. Correlation is detected (35) between it and residual echo e.sub.n to derive the coefficient R.

Echo canceller having FIR and IIR filters for cancelling long tail echoes

The whole or part of a waveform is cancelled by a circuit comprising a nonrecursive filter and a recursive filter. The nonrecursive filter has a plurality of first multipliers having first tap weights for modifying symbols successively shifted along a shift register. The recursive filter is connected in a series circuit to the shift register and has a second tap weight for recursively modifying the output of the shift register. The recursively modified symbol is modified by a second multiplier having a third tap weight and combined in an adder with the symbols modified by the first tap weights to produce a replica of an undesired waveform. The replica is destructively combined in a subtractor with an incoming symbol having an undesired waveform. The output of subtractor is utilized to derive the tap weights to adaptively control the inputs to the adder.

Circuit for cancelling whole or part of a waveform using nonrecursive

A quadriphase phase modulation system having a serial to parallel converter for converting first symbols of two-bit serial data into second symbols of two-bit parallel data, a circuit for generating waveform information in response to the second symbols, a memory for storing basic waveforms and a connection waveform and for reading out basic and connecting waveforms in response to the received waveform information. The basic waveform information is selected on the basis of the present waveform information and the connecting waveform is selected on the basis of the present waveform information and the immediately preceding waveform information. A digital to analog converter converts the basic and connecting waveforms output by the memory means into an analog signal.

Quadriphase phase modulation system

In a communication system comprising a two-wire bus, bidirectional communication is carried out in a time division multiple access fashion through the two-wire bus between a plurality of terminal equipments (26) coupled to the two-wire bus and a central terminator (24) connected to one end of the two-wire bus which has the other end terminated by an active terminator (25). The central terminator transmits, to the two-wire bus, a down burst comprising a down synchronization signal and a down information signal. Each terminal equipment establishes synchronization with reference to the down synchronization signal and thereafter receives the down information signal. The active terminator sends, towards the central terminator, only an up synchronization signal produced with reference to the down synchronization signal extracted from the down burst. Each terminal equipment transmits an up information signal towards the central terminator in the form of an up burst with reference to the up synchronization signal. The up information signal may be sent back to the two-wire bus by the central terminator to carry out communication among the terminal equipments.

Bidirectional communication system of a two-wire bus comprising an

In a transmission system of a class IV partial response code succession produced in accordance with a predetermined encoding rule, a transmitter intentionally violates the encoding rule to carry a subsidiary signal, so that an intentional violation appears in a receiver as two consecutive violation signals located in every other time slot, differing from an unintentional violation occurring during transmission. In the transmitter, an intermediate succession from a precoder is monitored by a detection circuit to detect an appearance of each predetermined pattern and to give the intentional violation to the code succession. In the receiver, a violation detector detects both of the intentional and the unintentional violations in accordance with a detecting rule to produce a violation signal sequence. The two consecutive violation signals are distinguished from the violation signal sequence by another detector. As the predetermined patterns, two and twelve patterns are usable in a binary and a quaternary system, respectively.

Growing community of inventors

Tokyo, Japan

Shin'ichi Koike