General Datacomm, Inc.

Vbrick Systems, Inc.

Paul D Cole

Yuri Goldstein

Yuri Okunev

William Hanna

William E Abdelsayed

Method for clock recovery in MPEG systems

A shaping filter for a PAM mapper is provided which manipulates incoming PAM code level signals to produce outgoing PAM code level signals whose frequency spectrum is shaped to meet desired criteria. In manipulating input code level values in(t), the shaping filter preferably utilizes previous output signal values without using previous incoming signal values. In a preferred embodiment, the output code level values out(t) are constrained to an output alphabet which is a predefined subset of code levels within the .mu.-law constellation, and are obtained by manipulating the input signals according to an integer-valued shifting function such that out(t) are code level values within the output alphabet. The preferred shifting function S 'shifts' in(t) by a number of places in the output alphabet in the direction opposite the sign of ##EQU1## to provide out(t), where out(t) are .mu.-law code output level values at times t, and a.sub.i are weighting factors chosen to shape the frequency spectrum of out(t). The shaping filter may be used with multidimensional PAM constellations.

Shaping filter for high data rate signalling

A spectral and power shaping mapper for high data rate signaling modem functions in a first mode having spectral and power shaping and using convolutional encoding, a second mode having power shaping, a third mode having power shaping using convolutional encoding, and a fourth mode having straight-forward mapping. Based on data rates and other criteria, incoming bits are grouped together, with a first subgroup of bits used in the first mode to generate two sign vectors (and in the other modes a single vector), a second subgroup of bits used in conjunction with an appended bit in the first, second and third modes to generate at least two magnitude manipulation vectors, and a third subgroup of bits used in a modulus converter to generate a plurality of quotients. The quotients and magnitude manipulation vector(s) are used to select one, two or four sequences of constellation points from a multidimensional constellation depending on mode. The sequences of selected points are combined with the sign vector(s) to generate, one, two or four sequences of octets for a shaping selector. In the first and third modes, the shaping selector employs a convolutional encoder state machine which uses criteria such as lowest cost paths and/or lowest average power in order to select which sequence of octets to output. In the second mode, selection is based on power only. In the fourth mode, the generated octets are output as there is no choice of octets.

Spectral and power shaping mapper for high data rate signalling

A spectral and power shaping mapper for high data rate signaling modem is provided. Based on data rates and other criteria, groups of incoming bits are grouped together, with a first subgroup of bits being used to generate two sign vectors, a second subgroup of bits being used in conjunction with an appended bit to generate two magnitude manipulation vectors, and a third subgroup of bits being used in a modulus converter to generate a plurality of quotients (r.sub.k). The quotients and magnitude manipulation vectors are used to select two or more sequences of constellation points from a multidimensional constellation. The sequences of selected points are combined with the sign vectors to generate four sequences of octets for a shaping selector. The shaping selector, preferably in the form of a convolutional encoder state machine uses criteria such as lowest cost paths and/or lowest average power in order to select which sequence of octets to output.

A sixty-four state convolutional code known as the 'Cole code' is disclosed and take systematic and non-systematic forms. The state variables of the Cole code are r(n-1), r(n-2), s(n-1), s(n-2), p(n-1), and x(n-1). In the four-to-five non-systematic form, four inputs p(n), q(n), r(n) and s(n) are used to generate five outputs y.sub.4 (n), y.sub.3 (n), y.sub.2 (n), y.sub.1 (n), and y.sub.0 (n) according to: y.sub.4 (n)=q(n); y.sub.3 (n)=p(n)+r(n)r(n-1)+r(n-2)+r(n-1)s(n)+r(n-1)s(n-2); y.sub.2 (n)=r(n-1)+s(n)+s(n-2); y.sub.1 (n)=r(n)+r(n-1)r(n-2)+p(n-1)+p(n-2)+q(n-1); and y.sub.0 (n)=s(n-1), where p(n-2)+q(n-1)=x(n-1). New state variables are obtained directly from the inputs and from the previous state variables, except for x(n) which is equals q(n)+p(n-1). In a three-to-four non-systematic implementation of the Cole code, q(n) is not input, y.sub.4 is not output, and q(n-1) is set equal to zero. In the four-to-five systematic form of the Cole code, the inputs y.sub.4, y.sub.3, y.sub.2, y.sub.1 are taken as outputs and are also used on conjunction with the convolutional encoder to generate a fifth output y.sub.0. With state variables x(n-1), p(n-1), r(n-1), r(n-2), s(n-1), s(n-2) relabled as w.sub.1, w.sub.2, w.sub.3, w.sub.4, w.sub.5, and w.sub.6, output y.sub.0 is taken as state variable w.sub.5. The state variables are updated according to: new w.sub.6 =w.sub.5 ; new w.sub.5 =y.sub.2 +w.sub.3 +w.sub.6 ; new w.sub.4 =w.sub.3 ; new w.sub.3 =y.sub.1 +w.sub.3 w.sub.4 +w.sub.2 +w.sub.1 ; new w.sub.2 =y.sub.3 +(new w.sub.3)w.sub.3 +w.sub.4 +w.sub.3 (new w.sub.5)+w.sub.3 w.sub.6 ; and new w.sub.1 =y.sub.4 +w.sub.2. In the three-to-four systematic Cole code, y.sub.4 is taken as zero.

Convolutional encoders for modems which implement the 'Cole code'

A modem is provided with a Trellis encoder which utilizes a double-branch feedback, with a first branch of the feedback including a convolutional encoder, and a second branch of the feedback including a parity generator which computes the parity of consecutive offsets. The Trellis encoder includes a four-dimensional (4D) mapper which generates pairs of 2D points u(k). Each point is offset by a value c(k) to obtain offset points y(k) which are then predistorted by a value p(k) to obtain an output x(k). The output x(k) is filtered to update the predistortion value p(k) which in turn is used to update the offset value c(k). The offset value c(k) is chosen from a 2D constellation lattice which is closest to the predistortion value p(k). The parity C0 of consecutive offset values c(k) is used in conjunction with a redundant bit output Y0 of a convolutional encoder to generate a rotational feedback bit input U0 to the 4D mapper, where the rotational bit input is the parity of two 2D points chosen by the mapper. The mapper uses bit U0 in generating the second of the pair of points u(k). The second point is likewise offset to obtain another offset value y(k), and then predistorted to obtain another output. Two offset values y(k) are used by the convolutional encoder to generate the next redundant bit output with is eventually used with the parity of the next two offset values to generate the next bit U0 for the second of the next pair of points.

Modem utilizing parity and convolutional encoder feedback

Apparatus and methods in both modem transmitters and receivers are provided for accommodating secondary channel data without disturbing the trellis encoding, mapping, and Viterbi decoding of primary channel data, and without disturbing distribution preserving Tomlinson precoding and decoding of the primary channel data. In order to accommodate the secondary channel data, the secondary channel data are used to make quadrant decisions in a mapper, and concurrently therewith, dummy data or tertiary channel data corresponding in time with the secondary channel data are provided to the trellis encoder so that the continuous working of the trellis encoder can be maintained. In order to guarantee proper working of the trellis encoder, and to guarantee reliability of the secondary channel data, the secondary channel constellation points within a quadrant are clustered together and far from the quadrant boundaries, the selected points in each quadrant are the rotations of selected points in the other quadrants, and the points include a single point from each two-dimensional partition element of the 2 D grid in each quadrant. With primary, secondary, and tertiary data encoded in this manner, the encoded data are precoded according to Tomlinson precoding or according to any of several disclosed distribution preserving Tomlinson techniques. Decoding is dictated by the manner in which the primary and secondary data were encoded and precoded.

High speed modem systems incorporating distribution preserving Tomlinson

A transmitter is provided with a distribution preserving Tomlinson coder which predistorts shaped data signals such that the power of each data signal exiting the coder is substantially similar to the power of the data signal entering the coder and such that upon transmission of the predistorted data signal over a channel, the effect of ISI of the channel is substantially removed. The transmitter is primarily intended for coded modulation systems utilizing a 'coset' code, and the predistortion is preferably accomplished according to a linear function ##EQU1## where r.sub.k is a data signal entering the coder, a and.sub.i b.sub.i are the coefficients of polynomials relating to the channel impulse response, x.sub.k is the predistorted data signal exiting the coder, and s.sub.k is chosen to cause the signal power of x.sub.k on average to approximately equal the signal power of r.sub.k on average. Different methods for so choosing s.sub.k are disclosed. Receivers which cooperate with the provided transmitters are also provided.

Methods and apparatus employing distribution preserving Tomlinson

A transmitter is provided with a distribution preserving Tomlinson coder which predistorts shaped data signals such that the power of each data signal exiting the coder is substantially similar to the power of the data signal entering the coder and such that upon transmission of the predistorted data signal over a channel, the effect of ISI of the channel is substantially removed. The transmitter is primarily intended for coded modulation systems utilizing a 'coset' code, and the predistortion is preferably accomplished according to a linear function ##EQU1## where r.sub.k is a data signal entering the coder, a.sub.l and b.sub.l are the coefficients of polynomials relating to the channel impulse response, x.sub.k is the predistorted data signal exiting the coder, and s.sub.k is a multiple of a given value (N) which is chosen to cause x.sub.k and r.sub.k to occupy identical defined regions in space, where the total length of each defined region is the given value N. One and two dimensional regions are disclosed which are chosen in several ways. Receivers which cooperate with the provided transmitters are also provided. The receivers receive a sequence of signals, and process the sequence to provide a sequence of signals whose k'th term is an estimate of r.sub.k -s.sub.k. From the estimate of r.sub.k -s.sub.k, the receiver generates an estimate of x.sub.k by multiplying the received estimate by the channel ISI. From the estimate of x.sub.k and the estimate of r.sub.k -s.sub.k, the receiver determines the particular value of s.sub.k, and hence the value of an estimate of r.sub.k.

Apparatus and methods employing distribution preserving tomlinson

An encoder is provided for generating rotationally invariant trellis codes when used with a signal constellation having m subconstellations indexed by j, each subconstellation having a plurality of points, where each subconstellation is a distinct rotation by j(360/m) degrees of another subconstellation such that no subconstellation has a point in common with another subconstellation. The encoder has a state block and a state update block. The state means receives q inputs i, and generates m outputs j, where m=pq and where p and q are integers greater than one and have no common factors. Outputs m are generated according to inputs i and states s of the state block according to j=i mod q and jd=s mod p, where the state block can assume n states, and n=md where d is an integer greater than one and has no common factors with p. The state update block updates the state block to a new state t chosen from the allowable n states based on the output j and previous state s according to t=jd+F(s-jd) mod n, where F is the permutation of 0, . . . , n-1, and is a function given by F(am+b)=bd+a for a=0, . . . , d-1, and for b=0, . . . , m-1.

Channel encoder for generating rotationally invariant trellis codes

A method is disclosed for determining with only sixty-four distance comparisions the distances between a received eight-dimensional vector comprised of four two-dimensional coordinates and each of sixteen non-overlapping subconstellations. In order to limit the comparisons, each subconstellation is represented by a collection of sixteen representational four-coordinate vectors, with each collection having four groups of vectors of the form

Growing community of inventors

Fairfield, CT, United States of America

Paul D Cole