Method and apparatus for implementing the connecting phase on a digital communications path utilizing Tomlinson-Harashima preceding
The invention relates to a method according to claim 1 for implementing the connecting phase on a digital communications path utilizing Tomlinson-Harashima precoding.
The invention also relates to an apparatus for implementing the connecting phase on a digital communications path utilizing Tomlinson-Harashima precoding.
In the transmission of digital data, or a bit stream, over a communications channel 2, the bit stream is converted in a transmitter (TX) 1 into a symbol stream (S) that further is converted into an analog signal capable of passing through a communications channel. The communications channel maybe, e.g., a radio path, copper wireline or fiber-optic cable. From the received analog signal, the receiver (RX) 3 performs the recovery of the sent bit stream in a fashion as error-free as possible. The bit stream reconstruction performed in the receiver is complicated by signal distortion and noise summed with the signal on the communications channel. Due to these side-effects, a portion of the reconstructed bits are erroneous (e.g., on an average, 1 bit per 10 bits may be erroneous).
The signal distortion originating from the transmission path is conventionally compensated for by means of equalizers that are located in the receiver, the transmitter or partially in both of these. The equalizers maybe of a fixed or adaptive type. Respec- tively, the effect of noise is compensated for by means of different coding techniques such as Reed-Solomon coding, convolution coding, trellis coding, turbo coding and others.
A generally used correction method of channel distortion is the use of a linear adap- tive equalizer 5 or a fixed or adaptive feedback equalizer 4, 6, wherein the filters are implemented with the help of digital signal processing (DSP). If the feedback equalizer is situated in the receiver, it is called a decision-feedback equalizer 6
(DFE), while an equalizer located in the transmitter is called a Tomlinson-Harashima precoder 4 (TML). A system may also have both a DFE and a TML. Furthermore, the linear equalizer may be situated in the receiver, the transmitter or, alternatively, one portion of the equalizer may be situated in the transmitter while the other portion resides in the receiver.
A shortcoming of a decision-feedback equalizer is the feedback of errors (error propagation) that degrades the noise tolerance margin of the system. When the channel distortion is heavy and the bit rate per Hz of bandwidth is large, the feedback of errors may become so significant as to block the entire functionality of the system. However, the use of Tomlinson-Harashima precoders circumvents the error propagation problem. When using a Tomlmson-Harashima precoder, compensation of signal distortion, whose transfer function is not known a priori at the design stage of the apparatus, needs the use of a protocol for establishing a connection. Moreover, the connection must be bidirectional.
The benefit of a decision-feedback equalizer over a Tomlinson-Harashima precoder is a simpler connection establishment phase. When using a decision-feedback equalizer, a connection may usually be established without the need for specific training phase sequences and/or protocols, such as blind training of equalizers. In contrast, when Tomlinson-Harashima precoder is used for compensating such a signal distortion whose transfer function is unknown during the design stage of the communications apparatus, a connection establishment protocol is invariably mandatory. Moreover, a bidirectional communications channel must assumed.
A conventional connection establishment protocol comprises the following main phases: 1) The system is adapted to the channel distortion by using an adaptive linear equalizer and adaptive decision-feedback equalizer at both ends of the connection. Herein, a low number of bits per Hz of bandwidth is typical in order to prevent error feedback from preventing operation at sufficiently high efficiency. 2) As soon as the connection operates in both directions (at a low bit rate per Hz), the parameter values of the decision-feedback equalizer are sent over to the other end of the connection.
3) The parameters of the decision-feedback equalizer thus transmitted over the communications channel are next programmed into a Tomlinson-Harashima precoder adopted at both ends of the connection, whereby the use of the decision- feedback equalizer is terminated. Subsequently, the bit rate per Hz is increased to its actual operating level. Phase 2 is mandatory inasmuch as the connection cannot be expected in a general case to have identical transfer functions in both directions and, hence, the parameters of a decision-feedback equalizer cannot be used in a Tomlinson-Harashima precoder situated at the same end of the connection.
Prior-art techniques are handicapped by requiring the parameter values of a decision- feedback equalizer to be directly applicable in a Tomlinson-Harashima precoder, too. This requirement can be fulfilled only if the implementation architecture of filter 7 in the Tomlinson-Harashima precoder is equivalent to the implementation architecture of filter 8 in the decision-feedback equalizer at the opposite end of the connection, whereby both filters may be of the FIR type, for instance. In practical circumstances, however, the equipment at the opposite ends of the connection are generally delivered by different manufacturers and given the liberty by standards to allow different implementation architectures in order to avoid a predominant position of some manufacturers), as either split or branched versions of standards are encountered.
The filters of a decision-feedback equalizer and a Tomlinson-Harashima precoder may have an architecture implementing either an FIR (finite impulse response) filter or an IIR (infinite impulse response) filter. Filters of the IIR type are generally used when the channel includes a digital HR-type filter operating at the symbol transmis- sion rate such as, e.g., a bandstop filter employed in the transmitter of NDSL modem for limiting the modem output power at given frequency ranges in order to avoid interference with radio amateur activities from spurious emissions of the telecommunications connection. Herein, the poles of the IfR filter of both the decision-feedback equalizer and the Tomlinson-Harashima precoder are set equal to those of the filter on the channel.
When dealing with the present invention, an ITR-type filter of a decision-feedback
equalizer or an IIR-type filter of a Tomlinson-Harashima precoder is assumed to be a filter having fixed and known spectral poles and either some or all of the filter parameters defining the location of its spectral nulls are adaptive. The benefit of this arrangement is that a smaller number of parameters are required in the decision- feedback equalizer or the Tomlinson-Harashima precoder as compared with a solution based on an FIR filter.
In the case that one end of the connection uses filters of the FIR type in its decision- feedback equalizer and Tomlinson-Harashima precoder while the other end uses filters of the IIR type, the parameters transmitted over the communications channel are not directly suitable for use in the Tomlinson-Harashima precoder.
It is an object of the invention to provide an entirely novel type of method and apparatus capable of overcoming the problems of the above-described prior art.
The goal of the invention is achieved by virtue transmitting during the connecting phase not only the parameter values of the decision-feedback equalizer, but also information on the implementation architecture of the filter used in the decision- feedback equalizer and, in the case the filter of the Tomlinson-Harashima precoder is of a different type in regard to the filter of the decision-feedback equalizer, the parameters are converted so as to be applicable in the filter of the Tomlinson- Harashima precoder.
An alternative embodiment comprises the step of transmitting the parameters of the filter of the decision-feedback equalizer over the connection at all times corresponding to, e.g., an FIR filter architecture and, if the decision-feedback equalizer happens to include an IIR-type filter, the sent parameters are converted suitable for an FIR-type filter prior to their transmission and, respectively, if the Tomlinson- Harashima precoder happens to be of the HR type, the parameters are converted suitable for an IIR-type filter prior to switching-in the Tomlinson-Harashima precoder.
More specifically, the method according to the invention is characterized by what is stated in the characterizing part of claim 1.
Furthermore, the apparatus according to the invention is characterized by what is stated in the characterizing part of claim 6.
The transmitter according to the invention is characterized by what is stated in the characterizing part of claim 11.
The receiver according to the invention is characterized by what is stated in the characterizing part of claim 14.
The invention offers significant benefits inasmuch as it facilitates the cooperation of different equipment architectures comprising a decision-feedback equalizer and a Tomlinson-Harashima precoder.
In the following, the invention is described in more detail with reference to exemplifying embodiments by making reference to the appended drawings in which
FIG. 1 shows a block diagram of a digital communications channel equipped with a Tomlinson-Harashima precoder in one transmission direction; -
FIG. 2 shows a block diagram of a digital communications channel equipped with a decision-feedback equalizer in one transmission direction;
FIG. 3 shows a block diagram of a conventional decision-feedback equalizer with an FIR-type filter (the filter transfer function shown in the z-plane);
FIG. 4 shows a block diagram of a conventional Tomlinson-Harashima precoder with an FIR-type filter (the filter transfer function shown in the z-plane);
FIG. 5 shows a block diagram of a conventional Tomlinson-Harashima precoder with
an IIR-type filter operating in combination with an IIR-type filter of the communications channel (the filter transfer function shown in the z-plane);
FIG. 6 shows a block diagram of a conventional decision-feedback equalizer equiv- alent to the Tomlinson-Harashima precoder of FIG. 5 and having its filter implemented as an IIR-type filter (the filter transfer function shown in the z-plane);
FIG. 7 shows a block diagram of an embodiment of the generation method of parameters in a system according to the invention for an IIR-type filter from the parameters of an FIR-type filter with the help of a conventional algorithm;
FIG. 8 shows a block diagram of an embodiment of the generation and accuracy estimation method of parameters of an FIR-type filter in a system according to the invention based on the parameters of an IIR-type filter using a prior-art algorithm;
FIG. 9 shows a block diagram of an embodiment suitable for use in a system according to the invention for the accuracy testing method for the parameters of an IIR-type filter constructed on the basis of the parameter values of an FIR-type filter; and
FIG. 10 shows a block diagram of an embodiment suitable for use in a system according to the invention for the accuracy testing method for the parameters of an FIR-type filter constructed on the basis of the parameter values of an IIR-type filter.
In following text, the following abbreviations will be used:
DFE Decision-feedback equalizer
DSP Digital signal processing
FFE Feedforward equalizer
FIR Digital filter of a finite impulse response
IIR Digital filter of an infinite impulse response
LMS Least-mean-square algorithm
RX Receiver
TX Transmitter
TML Tomlinson-Harashima precoder.
Notations to be used later in the text are:
A(z_1) = aoz"1 + aiz"2 + a2z"3+... aN-1z"N Transfer function of an FIR filter in the z- plane, whereby z'1 is the intersymbol delay.
A Set of parameters ao, & , a2, ... , aN-i .
P(z_1) = p0z_1 + piz"2 + p2z"3+...pL-iZ"11 Nominator of an IIR-type filter transfer function in the z-plane (defining the spectral zeroes of the function), whereby z"1 is the intersymbol delay.
P Set of parameters p0, pi, p2, ..., p^ .
Q(z-1) = q0 + qfz 1 + q2z"3+...qMz"M Denominator of an IIR-type filter transfer function in the z-plane (defining the spectral poles), whereby z"1 is the intersymbol delay.
Q Set of parameters q0, qi, q2, ..., qM •
LNl, LN2 Pair of transmitter-receiver terminals. The pair of equipment LNl, LN2 together with the commumcations channel forms a bidirectional communications connection.
Accordingly, the invention relates to a method and apparatus capable of accomplishing the connecting phase on a bidirectional communications connection (LNl - LN2) utilizing Tomlinson-Harashima precoding even in the case the transmitter-receiver set (LNl) at one end of the connection comprising a Tomlinson-Harashima precoder and a decision-feedback equalizer include FIR-type filters (FIGS. 3 and 4), while at the other end (LN2) of the connection the corresponding filters are of the HR type (FIG. 5 and 6). The invention is limited to IIR-type filters with fixed spectral poles (nonadaptive) and information on the architecture of the filters is either known a priori at both ends of the connection or, when necessary, can be relayed between the terminal equipment.
Next, the theory related to the invention is discussed.
First we assume the filters of the Tomlinson-Harashima precoder and decision-feedback equalizer of terminal equipment LNl to be of the FIR type and the filters of the Tomlinson-Harashima precoder and decision-feedback equalizer of equipment LN2 to be of the IIR type. Let us further assume that both transmission directions of the communications channel are equipped with a digital filter operating at the symbol transmission rate and having the denominator of its transfer function written as
QF(Z_1) = q0F + qiFZ"1 + q2F "3+... qMFZ"M. (1)
During the connecting phase, terminal equipment LNl and LN2 use adaptive decision-feedback equalizers. During the training phase of the adaptive equalizers, the FIR-type filter of the decision-feedback equalizer of the receiver of equipment
LNl changes its parameters (a0, als a2,..., au-i) to values ao,u alsl, a2,l .., aΝ-1,1 (A_) and, respectively, the IIR-type filter of the decision-feedback equalizer of the receiver of equipment LN2 changes its spectral-zero-defining the parameters (p0, pi, p2,- • •» PL-I) to values p0,2, pι,2, p2,2,..., pL-ι,2 (P2). The pole-defining parameters of the IIR-type filter are q0p, qiF, q2F,- • ., qMF (QF), which set of parameters is assumed to be known a priori in both sets of terminal equipment LNl and LN2.
In the method according to the invention, continuation of the connection phase requires that parameter set Ai is converted to represent an IIR-type filter architecture in terminal equipment LNl, whereupon the set is transmitted over the communications channel to terminal equipment LN2 or alternatively, parameter set Ai is sent over the communications channel to terminal equipment LN2 and thereupon is converted to represent an IIR-type filter architecture in terminal equipment LN2. Respectively, parameter value set P2 is converted to represent an FIR-type filter architecture in terminal equipment LN2 and thereupon is sent over the communications channel to terminal equipment LNl or alternatively, the parameter value set P2 is sent over the communications channel to terminal equipment LNl and thereupon is converted to represent an HR-type filter architecture in terminal equipment LNl .
Next, exemplary methods are discussed suitable for converting parameter value sets defining FIR/IIR-type filter architectures into parameter value sets defining HR/FIR- type filter architectures. It must be noted, however, that the method according to the invention is not limited to the conversion methods discussed below.
The conversion of a parameter value set A\ to represent an HR-type filter can be performed using a method based on a z-plane equation
A1(z-1) QF(z 1) = P1(Z-1), (2)
where the coefficients (po,l5 pι,ι, p2,ι, .. •) of polynom PΪ(ZA of (Z" ) represent typical parameters computed for an HR-type filter. The degree of the polynom computed from Eq. 2 is M+N, which is higher that the degree of a polynom A^z"1) representing
an FIR-type filter and thus is controversial to the basic idea of an IIR-type filter. Since the FIR-type filter function Aι(z_1) is also partially descriptive of transfer function l/QF(z"1), parameters po5ι> pιsι, p2,ι> • • • should be small starting from a given index value I (power 1+1 of z"1), where I is smaller than number N of parameters in filter function A^z"1). If this is not the case, the FIR-type filter A^z"1) is not sufficiently long to represent the transfer function 1/QF(Z_1).
The number (1+1) of parameters p0,l5 pl5l, p2,ι, ... can be determined using the testing arrangement of FIG. 9, where I is selected such that the power level of error signal (e) becomes sufficiently low, e.g., reduced by 45 to 50 dB, as compared with the power level of the test signal (t). As to its statistical character, the test signal (t) must be similar to the input signal of the Tomlinson-Harashima precoder filter.
An alternative method of using Eq. 2 is illustrated in FIG. 7. Parameters p0,ι, pι,ι, p2,ι, ..., pι,ι are determined with the help of an adaptive control algorithm (e.g., Least Means Square, LMS). However, the number (1+1) of parameters is preferably made using the system of FIG. 9, because the error signal (e) computed in the system of FIG. 7 is not compatible with the difference between filter functions Aι(z_1) and Pι(z" 1)/QF(Z"1) in the situation corresponding to a Tomlmson-Harashima precoder.
Conversion of parameter set P2 to represent an FIR-type filter can be performed using a method wherein to the input of an IIR-type filter P2(Z"1)/QF(Z"1) is connected the following input sequence:
1, 0, 0, 0, 0, 0, 0, 0, 0, ... (3)
Then, the output sequence of the filter P2(z"1)/QF(z"1) provides the parameter set a0,2, aι,2, a2,2, ... (A2) of an FIR-type filter. The number of parameters a ,2, als2, a2,2, ... can be determined using the test arrangement of FIG. 10, where J is selected such that the power level of error signal (e) becomes sufficiently low, e.g., reduced by 45 to 50 dB, as compared with the power level of the test signal (t). As to its statistical character, the test signal (t) must be similar to the input signal of the Tomlinson-
Harashima precoder filter.
An alternative method of determining parameters ao,2, aι,2, a2,2, ..., aJs2 is illustrated in FIG. 8. The parameters are determined with the help of an adaptive control algo- rithm (e.g., Least Means Square, LMS). Also the number (J+l) of parameters can be made using the system of FIG. 8, because now the error signal (e) computed in the system of FIG. 8 is compatible with the difference between filter functions A2(z-1) ja P2(Z"1)/QF(Z"1) in the situation corresponding to a Tomlinson-Harashima precoder.
Some of the exemplary embodiments discussed above and certain ones of the claims are written describing the connecting phase of a bidirectional communications channel in one direction. Obviously, the invention is applicable in either direction on a bidirectional communications channel.