Note: Descriptions are shown in the official language in which they were submitted.
20~(~763
DIGITAL SIGNA~ TRANSNISSION SYSTEM
BACKGROUND OF THE INVENTION
Field of the Invention
The invention relates to a digital signal
transmission system for the simultaneous transmission
of several signals via separate wire pairs in a multi-
pair cable.
Description of the Prior Art
Digital transmission systems with maximal bit
rates of 2.048 Mbit/s for the multiple use of multi-
pair symmetrical low-frequency cables are known. AMI-
coded or HDB3-coded line ~ignal~ are predominantly
transmltted via the Nyquist-equalized channel of the
wire pair and are threshold detected. These
conventionàl procedures are ~ery ~ensitive to cro~stalk
interferences from similar transmission systems which
operate on adjacent pairs in the same cabIe. This
I,
¦ problem is partlcularly acute when bidirectional
signals, independent of each other, are transmitted in
the same cable. The proceduxes are also sensitive to
impulsive interferences caused by the character
transmission of electro-mechanical switched systems.
`~
20~0763
These interferences limit the maximum field length or
regeneration space that may be bridged and also, in the
case of a fixed field length, limit the number of
transmission systems of the same kind that can be
located in the same cable.
The main source of interference is crosstalk,
which is caused by signals of systems of the same kind
which operate on ad;oining pairs within the same
cable. Therefore, a greater field length cannot be
achieved by increasing the transmitter power, but only
by an improved system design. Further interference is
i caused by the crosstalk of rare pulses with high' energy, which originate from asymmetrical signaling
r~ procedures on pairs that are still used for traditional
analog voice transmis6ion. The spectral energy density
of such pulses is concentrated at frequencies which are
low compared to the data rate. The transmitted pulses
are stretched over many symbol lntervals T by the
di~perslve medium, resulting in a strong attenuation at
hlgh frequencies.
An optimum solution to the detection of a digital
signal with severe inter-symbol interference and
,~, ,
colored noise is achieved by using a Maximum-Likelihood-
Sequence-Estimation using the Viterbi-algorithm, as
explained by G. S. Forney, Jr., "Maximum-Likelihood-
-2-
. ~
'., - : ' t
' , . :,
2010763
Sequence-Estimation (MLSE) of Di~ital Sequences in the
Presence of Inter-Symbol Interference", IEEE Trans.
Information Theory, Vol. IT-18, pp. 363-378, May 1972.
The complete MLSE procedure would require tracing
numerous paths through a trellis and therefore cannot
be realized at the desired data rate. A modified
version using a threshold detector with decision
feedback which corresponds to a state reduction of the
MLSE procedure down to one state was implemented, as
described by M. V. Eyuboglu, S. U. Quershi, "Reduced
State Sequence Estimation With Set Partitioning and
Decision Feedback", IEEE Trans. communications, Vol.
COM-35, pp. 13-20, January 1988. For this receiver
type, as well as for the optimum detector, the highest
reliability i5 achieved by using a four-level
~quaternary) ba~e band ~ignal without redundance
(provided that no additional equipment for decoding of
hannel codes i~ tolerated).
, ~ ~
~ -3-
- 2o.~0763
SUMMARY OF THE INVENTION
The present invention contemplates a digital
transmission system for the simultaneous transmission
of several digital signals via separate wire pairs
located in a multi-pair cable.
The dominating near-end crosstalk experienced when
there are a large number of systems transmitting on
pairs within the same cable, is avoided by using a
cable which is divided into quads, with the wire pairs
; 10 within the same quad all handling signaling in the same
direction. When this configuration is used, the
resulting noise may be approximated by colored Gaussian
noise. Thus, a bundle cable with spiral quads can be
f fully wired when each spiral guad is used for the
tran3mission of signals in one direction only.
i A binary signal would be affected by the channel
;~ attenuation which grows for an lncr~asing freguency in
a much more rigorou~ manner. Therefore, a quaternary
~ignaling (1.024 M baud) is used in the present
invention. The binary, scrambled digital source
signals axe re-coded into redundance-free quaternary
transmitted signals.
,~
~,:
~ -4-
.~
2U~0763
The received signals, which are Nyquist-distorted
and suffer from low frequency attenuation, are
adaptively equalized in a symbol-interference-free
manner, and the low-frequency signal portions that have
been lost along the transmission path are recovered by
means of a quantized feedback (decision-feedback
equalization).
The signals which are equalized in an interference-
free manner are detected by means of a simple threshold
detector. From the difference between a received
signal value present at the optimum detection time and
the useful value of the received signal estimated on
~ the basis of the decision, an estimate of the
!~J interference is obtained by subtraction. On the basis
i~
of previous estimated interference values and the
¢urrent estimated interference value, a prediction of
the interference at the next detection time is made by
f~ means of a linear filter. This predicted value is
subtra¢ted from tho slgnal value at the next detection
time, in order to achieve a reduction in the effective
;~
Z~3~ interference power.
The prediction of the interference value is
possible because, on the basis of the crosstalk
frequency response and by means of the equalization,
linear statistical links within the interference
i,jlZZ ~
' ~Z
,~,Z ~ ~:
~: :
201~763
process, which is produced in its predominant part by
cross-talk of signals of the same kind in the multi-
pair cable, are formed. The interference predictor
filter can also be used adaptively for minimization of
the remaining residual interference power.
The predominant part of the receiver is realized
digitally, in accordance with the methods known from
digital signal processing. The receiver, in this way,
achieves the maximum possible interference protection
against crosstalk interferences in the case of the use
of a threshold value detector. However, only small
losses in the signal-to-noise ratio, as compared with
optimal detection (correlation receiver for all
possible symbol sequences), must be accepted.
The novel tran6mission proces6 is characterized by
the fact that, instead of the QR structure (decision-
feedback equalization) of the interference value
prediction with a double realization o~ the predictor
filter in the signal path and in the ~e~dback path,
ZO which is designated as optimal in Adaptive ~ilters, by
I
M. L. Honig and D. G. Mes6erschmitt, Kluwer Academic
Publi6hers, Bo6ton, (1984), a simple structure of the
s~ interference-value predictor with a single realization
of the predictor ~ilter is used in the present case.
This is because, in the case of the present use, the
6-
,.,
~o76~
quaternary signal is the optimal of all redundance-free
transmitted signals. The ætructure according to the
drawing is found to be superior to the QR structure for
this quaternary signal.
~ An additional characterizing feature is the
separate possibility of adjustment of the filters for
the (linear and decision-feedback) equalization and for
the interference-value prediction made possible in this
way. By this means, different setting algorithms for
these filters and different adaptation speeds for
equalization and interference-value prediction can be
selected. A setting of the interference-value
prediction that is slower as compared with the
equalization produces a distinct broadening of the
range for transmis~ion properties (cable length, wire
diameter, type of interference) within which a
7'
successful adaptation is possible for an identical
I ` coefficient number of the digital filters. In
,1
addltion, by means of the ~tructure with separate
~i 20 setting algorithms for the equalizer (zero-forcing
algorithm) and interference-value prediction (minimum-
mean-square algorithm?, a more effective interference
~::
suppression 1s aahieved in the case of stationary
operation than when a common setting procedure is
selected for all filter~.
-7-
,~ ~
,~
...... , ,,- , . ,.. ;.. ~,, ,:.; ,.,.. , .. , .. , ,. , , , ,. , . , ~ , ~ . .
- , ,, ,- - ,;, , " , , - ;,:; -
;~0~0763
- Another identifying characteristic of the novel
transmission process is that analog compromise-type
preliminary equalization is carried out to reduce the
coefficient number of the linear and of the decision-
feedback part of the adaptive, digitally realized
equalizer. This is designed in such a way that a
; whitened match filter, as in "Detectors and Optimum
Receiver Filters for Digital Signals with Intersymbol
, Interference", Part I and Part II, by J. Huber,
s 10 Frequency 41 (1987), pp. 161-167 and pp. 189-196, is
contained as a factor and that, in this way, a
transition from a continuous-~ime signal to discrete-
time scanning values without loss of information is
ensured. In this way, this compromise equalizer
corresponds to the optimal Nyquist filter for the cable
attenuation and cross-talk transmission function
q, .
expected on average, This usually results in roll-off-
~actor~ the pul~e ~haper part o~ the compromise
equalizer that are between 0.3 and 0.6 and which lead
to optimum increases in the signal-to-noise ratio.
As another characterizing feature of the
transmission system is achievement of an optimization
of the detection time on the basis of the coefficients
set in the linear adaptive equalizer, which is also
carried out in an adaptive manner. In this way, long-
-8-
,,' ', ;, ',' ,.' '' ''' '. :' ' ' : `,, ' `.', ," ' ' ;, :' ' ',
.. ;. ., ~ . : . :, . : . ,, , :, " .
X0~0763
term variations of cable properties or of the
oscillator for the timing recovery in the receiver are
equalized.
The invention can be used advantageously in local
subscriber line networks for increasing the field
length for 2.048 Mbit/s transmission. The predominant
number of subscribers is reached in this way without a
regenerative repeater. As a result of the improved
interference protection against crosstalk and impulsive
interferences, a very high degree of assignment of the
local connecting cables with 2.048 Mbit/s systems of
the same kind with a relatively large field length can
be achieved. By means of a skillful selection of the
different setting algorithms for the equalization and
the interference-value prediction, the high-pass
~iltering system becomes completely effective against
the unsteady impulsive interferences.
DESCRIPIION OF TNE DRA,WING
,
An exemplifying embodiment of the invention is
represented in the drawing which is a schematic
diagram.
:.
'3:
g _
i: :
,?:
,~ .. : . , . ,`, , ., , , .,. . : . , . , . :
2010~6~
DETAILED DESCRIPTION OF THE PREFERRED EMBODIMENT
A redundance-free four-stage (quaternary) line
signal is transmitted on a wire pair of a multi-pair
cable. The multi-pair cable is a bundle cable with
spiral quads with the pairs in each quad used for
transmission in one direction only. The signal is
etrongly dietorted and attenuated on the way to a
receiver, in accordance with the transmission function
of the wire pair, and is present at a receiver input as
an intelligence signal e(t). The intelligence signal
e(t) is superposed by crosstalk interferences from
., ,
~; transnission systems of the same kind operated on
adjacent wires in the same quad, and from unsteady
impulsive interferences of electro-mechanical switched
systQme.
Both the intelligence eignaI and all interferences
at the receiver input are ~ilterèd w1th a high-pass
~ilter. This high-pae~ filter HP reduces the low-
frequency impulsive interferences. A linear equalizer
2 and decision-feedback equalizer 3 eliminate the
impulsive interferences of the intelligence signal e(t)
according to the first Nyquist criterion. By dividing
the linear equalizer 2 into an analog compromise
equalizer KE having a fixed setting and an adjustable
~ 10-,
~ ,
: ~
:
. :v , . . -
. . ~ . ~:
20~0763
digital equalizer DE, the number of required
coefficients, cO through cL~ of the equalizer DE is
reduced.
The output signal of the analog compromise
equalizer KE is scanned by the symbol clock pulse T to
produce a sequence of discrete scanning values (ei~.
The discrete values (ei) are analyzed with the
transverse filter TF2 of the adjustable digital
equalizer ~E in accordance with the setting of the
coe~ficients aO through CL, to produce a sequence of
the discrete-time openings (di). The detection time is
established on the basis of the coefficients CO through
CL to achieve optimization of the detection time to
compensate for long-term variations of cable properties
or of the oscillator for the timing recovery. The
optimization of the detection time is related to the
moment o~ scanning. The concept iB to use the
Q~iciQnts to ad~u~t the tact. Such a moment for
: scanning may, for example, be any one of the following:
L
; 20 ¦Cj¦ = Max
o
:,
~ or
.
i ~ ~
--1 1--
,~
.... .
2o~0~63
L-l
¦ = Max
= o
or
CL = Max
A compensation signal dQRi is derived from a
detected signal vi via a transverse filter TF3 of the
decision-feedback equalizer 3 in accordance with the
coefficients bo through bM. m is compensation signal
dQRi is intended to cancel distortion caused by the
high-pass filter HP. The distorted signal, with its
scanning values ki, is formed by means of a difference
function where ki = di ~ dQRi. The distorted signal ki
contains an intelligence signal component and an
interf~rence-~lgnal component.
:The adaptive setting of the coefficients cO
:
~ through cL and bo through bM is performed as described
,
by K. P. Graf and J. Huber, two of the inventors, in
"Design and Performance of an All-Digital Adaptive
z.048 Mbit/s Data Transmission System Us1ng Noise
Prediction", Proceedings of International Symposium on
-12-
; ~
f~
: " .,, . , ", ,, , , , : -, : , , . , :- . . .
2010763
Circuits and Systems (ISCAS~, May 1989, which is
incorporated herein by reference. The method uses the
known zero-forcing algorithm as described in Principles
of Data Communication, R. W. Lucky, J~ Salæ and E. J.
Weldon, Jr., McGraw-Hill, New York, 1968.
An adaptation of an interference-value predictor 1
with coefficients Pl through PN is carried out
completely independently of the setting of the
distortion. The known minimum-mean-square algorithm,
also described in Principles of Data Communication, is
used in this case. The scanning values ri f the
predictor input signal are formed by subtraction of the
detected signals vi from the compensated and equalized
signal ki, which is delayed by the duration of one
symbol interval T, such that ri = ki_l - vi. The
signal ki i~ superposed by colored interference which
~' 7
is caused primarily by crosstalk, and there is an
absence o~ ~ignl~icant whito interference in ki. These
scanning values of the difference signals ri are
analyzed with the transverse filter TFl of the
lnterference-value predictor 1 in such a way that the
predict~ble part of the interference mi is subtracted
from the scanning values ki to form values xi which are
provided to the detector SD. Thus, only the
intelligence signal component and the interfering
-13-
.
~1:: :~
~". ,,, . ,.. , ;., .... ,., . ", . , ~.. ,.... " ., -; . ..... ....
2Q~()763
component are present at the input of the threshold-
value detector SD as a signal xi, without a correlation
within the intelligence signal component and the
interfering component. The decision process therefore
provides greater interference protection, because of
the de-correlation.
The complexity of an implementation of the present
invention may be greatly reduced by application of the
sign algorithm wherein only the signs of the signals
that are used for the adaptive filter coefficient
settings are used. It has been found that by merely
using the signs of the signals for the coefficient
correction settings rather than the values of the
signals themselve~, the hardware may be greatly
simplified while the correction produced is still
sati~factory.
With the system of the invention it is possible to
achieve a total attenuation of 44 dB. The roll-off-
~actor o~ the receiving ~ilter may be 0.5. The system
provides a possible transmission range of nearly 3 km
on copper conductors of 0.4 mm diameter and nearly 5 km
on copper conductors of 0.6 mm diameter.
An implementation of the present invention,
including experimental results, is further described in
the article, "Deslgn and Performance of an A11 Digital
:, ,
2010763
Adaptive 2.048 Mbit/s Data Transmission System Using
Noise Prediction", written by two of the inventors.
, ~
,~
~, :
L:~
~, ,
.;
.., ~
,~.~,. ,
,
.
.',5. . ' . ~ . ,
