CA1205876A - Fir-type balance filter incorporated in the transmitter-receiver unit in a telecommunication system - Google Patents
Fir-type balance filter incorporated in the transmitter-receiver unit in a telecommunication systemInfo
- Publication number
- CA1205876A CA1205876A CA000414386A CA414386A CA1205876A CA 1205876 A CA1205876 A CA 1205876A CA 000414386 A CA000414386 A CA 000414386A CA 414386 A CA414386 A CA 414386A CA 1205876 A CA1205876 A CA 1205876A
- Authority
- CA
- Canada
- Prior art keywords
- filter
- group
- memory
- unit
- links
- Prior art date
- Legal status (The legal status is an assumption and is not a legal conclusion. Google has not performed a legal analysis and makes no representation as to the accuracy of the status listed.)
- Expired
Links
- 238000012937 correction Methods 0.000 claims abstract description 15
- 239000004020 conductor Substances 0.000 claims abstract description 4
- 238000005070 sampling Methods 0.000 claims description 4
- 230000003044 adaptive effect Effects 0.000 abstract description 3
- 101150087426 Gnal gene Proteins 0.000 description 12
- 241000282326 Felis catus Species 0.000 description 6
- 230000004044 response Effects 0.000 description 6
- 239000000306 component Substances 0.000 description 4
- 238000010586 diagram Methods 0.000 description 4
- 241000518994 Conta Species 0.000 description 2
- 230000000694 effects Effects 0.000 description 2
- 238000000034 method Methods 0.000 description 2
- 230000005540 biological transmission Effects 0.000 description 1
- 230000001934 delay Effects 0.000 description 1
- 230000003111 delayed effect Effects 0.000 description 1
- 238000001514 detection method Methods 0.000 description 1
- 230000002401 inhibitory effect Effects 0.000 description 1
- 230000006386 memory function Effects 0.000 description 1
- 239000002674 ointment Substances 0.000 description 1
- 239000000047 product Substances 0.000 description 1
- 230000001702 transmitter Effects 0.000 description 1
- NWONKYPBYAMBJT-UHFFFAOYSA-L zinc sulfate Chemical compound [Zn+2].[O-]S([O-])(=O)=O NWONKYPBYAMBJT-UHFFFAOYSA-L 0.000 description 1
Classifications
-
- H—ELECTRICITY
- H04—ELECTRIC COMMUNICATION TECHNIQUE
- H04B—TRANSMISSION
- H04B3/00—Line transmission systems
- H04B3/02—Details
- H04B3/20—Reducing echo effects or singing; Opening or closing transmitting path; Conditioning for transmission in one direction or the other
- H04B3/23—Reducing echo effects or singing; Opening or closing transmitting path; Conditioning for transmission in one direction or the other using a replica of transmitted signal in the time domain, e.g. echo cancellers
-
- H—ELECTRICITY
- H04—ELECTRIC COMMUNICATION TECHNIQUE
- H04B—TRANSMISSION
- H04B3/00—Line transmission systems
- H04B3/02—Details
- H04B3/20—Reducing echo effects or singing; Opening or closing transmitting path; Conditioning for transmission in one direction or the other
- H04B3/23—Reducing echo effects or singing; Opening or closing transmitting path; Conditioning for transmission in one direction or the other using a replica of transmitted signal in the time domain, e.g. echo cancellers
- H04B3/231—Echo cancellers using readout of a memory to provide the echo replica
Abstract
ABSTRACT OF THE DISCLOSURE
An FIR-type balance filter for preventing, in adaptive echo cancellation in a telecommunication system in which digital information transferred in duplex over a single conductor pair, the local data transmitter from disturbing the local data recep-tion. The filter is thereby arranged so that the total number of links in the filter is divided into at least a first group and a second group. The apparatus contains a memory addressable from the filter links in the first group. There are written into and read from said memory the correction values from and to a cor-rection unit, sending signals for correcting the filter parameters.
The memory also sends signals to a summing circuit, to which are also supplied the output signals from the second group of filter links. The output signals may be originated by table look-up in a further memory, or by vector multiplication. A second correc-tion unit corrects the filter parameters in the second group. The output signal from the summing circuit is digital-analogue converted and with its sign added to the signal coming from the remote end and transmitted to the local transmitter, thus compensating for the nonlinearity in the echo signal.
An FIR-type balance filter for preventing, in adaptive echo cancellation in a telecommunication system in which digital information transferred in duplex over a single conductor pair, the local data transmitter from disturbing the local data recep-tion. The filter is thereby arranged so that the total number of links in the filter is divided into at least a first group and a second group. The apparatus contains a memory addressable from the filter links in the first group. There are written into and read from said memory the correction values from and to a cor-rection unit, sending signals for correcting the filter parameters.
The memory also sends signals to a summing circuit, to which are also supplied the output signals from the second group of filter links. The output signals may be originated by table look-up in a further memory, or by vector multiplication. A second correc-tion unit corrects the filter parameters in the second group. The output signal from the summing circuit is digital-analogue converted and with its sign added to the signal coming from the remote end and transmitted to the local transmitter, thus compensating for the nonlinearity in the echo signal.
Description
~ s~
The present invention relates to a balance filter for a duplex data transmission in a telecommunication system, and more particularly -to a filter which is incorporated in an echo canceller for inhibiting the effect of an echo signal from the transmit side to the receive side via the hybrid coupler.
In describing the prior art reference will be made to Fig. l of the accompanying drawings, in which:-Figure l is a block diagram of a transmitter-recei~er unit ~or dup~ex data -transmissions;
Figure 2 is a block aiagram of an embodiment o the balance filter in accordance with the invention included in the unit shown in Figure l;
Figure 3 illustrates an FIR filter included in another embodiment of the balance filter according to Figure
The present invention relates to a balance filter for a duplex data transmission in a telecommunication system, and more particularly -to a filter which is incorporated in an echo canceller for inhibiting the effect of an echo signal from the transmit side to the receive side via the hybrid coupler.
In describing the prior art reference will be made to Fig. l of the accompanying drawings, in which:-Figure l is a block diagram of a transmitter-recei~er unit ~or dup~ex data -transmissions;
Figure 2 is a block aiagram of an embodiment o the balance filter in accordance with the invention included in the unit shown in Figure l;
Figure 3 illustrates an FIR filter included in another embodiment of the balance filter according to Figure
2; and Figure 4 is an impulse response diagram for the transmitter-receiver side in Figure 1.
~5 A method and apparatus are described in applicants Canadian Patent Application No. 414,564 filed November 1, 1982 ~or achieving adaptive echo cancellation. The principle of the apparatus will be appaxent from the appended Figure l.
The data sequence bn and an are transmitted on the conductor pair to and from the remote end, respectiv~ly. At the local end receive units arrive not only the signal w(t~ ~rom the remote end but also a signal h(t)~ from the local transmi-tter S. In practical cases, the amplitude of the h(t) signal may be very much grea-ter than that o~ the w(t) signal. To enable correct detection in the detector D a signal y(t) is subtracted ~rom the received signal. The former is the ~5~7~
outpu-t si~nal from the balance filter B, the input signal of which comes from the data sequence bn. The balance filter is adjusted adaptively with the aid of the correction unit KB so that v(t) and h(t~ become equal in the sampling instants kT.
After sub-traction there only remains -the w(t) signal from which date from the remote end can be ex~racted.
The balance filter should be of the FIR type (Finite Impulse Response). The output signal of such a filter is defined thus:
Yk Yk ( _ ~ Xk ) where the parameters of the filter are represented by the vector c = (cO, c1, --~ c ;. - la -.~,~ \
5~
and the ;nput signal ;s wr;tten in vector form:
~C = t~k'Xk-l' '''~ Xk-N+1 By correlat;ng ;ts ;nput signals the correct;on un;t KB calculates a correct;on of the f;lter parameters:
c (k ~ 1) = c~k) ~ ~c~k) where c(k) ;s the f;l$er parameter vector at the t;me kT and ~c(k) is a correct;on vector.
The convent;onal ;mplementat;on of an FIR f;lter ut;l;zes the vector mult;pl;cat;on:
lD Yk = c xk = ~ cnxk_n (when M = N) Here the number of parameters ;s thus N.
Another way in wh;ch to reali~e a f;lter of the FIR type ;s to utl;lize table look-up. All conce;vable output signals are stored ;n a memoryO
If each input signaL sampLe xn can assume L d;screte levels the number of poss;ble output s;gnals w;ll be~
N
M = L
For the memory sizetobemanageable, L and N must be relat;vely small.
The ;mplementat;on can be ;nterest;ng, part;cularly ;n binary data transm;ss;on (L = 2). The f;lter input s;gnal vector ~ with N compo-nents determ;nes an address j wh;ch po;nts out a def;n;te f;lter para-meter Cj ;n the memory.
Yk - Cj;~
In vector mult;pl;cat;on the correct;on vector w;ll be:
Ac(k) = (~Co~c~ CN-1 37~
i~e. all the parameters are updated simuLtaneousLy.
In the table Look-up the correction vector will ~e:
~ctk~ ~ (0,0,... ~cj,~ ...,0) i.e. only one of the parameters ;s updated at each correct;on occasion.
It will be seen f~om the above that the table f;lter r2sults ;n large memory space if N ;s large (the memory s;ze ;ncreases exponentially w;th n). S;nce in many practical cases ;t may be necessary to have a relat;vely large value of N~ the table f;lter discussed above may be unsuitable.
.
The transm;tter un;t S, coupler G and lowpass f;lter LP as well as other un;ts not shown ;n F;gure 1 ;n the transmitter-receiver un;t give rise to nonlinearit;es. The signal ~(t) coming to the hybrid cou~ler G from the remote end is comparatively weak. The signal h(t) from the trans-m;tter unit S through ~he coupler G (the leak signal) is~ sn the other hand, at least at the beginning oftransmit-rcceivem~de, relati~ely stLong before echo cancellation has managed to come into action. In practical cases~ the signal htt) may be much greater than w(t), e.g. a hundred times as great. S;nce the signal h(t) passed a nonlinear signal path, components from a given transmitted data symbol bk arise ;n the signal htt), for wh;ch a l;near balance f;lter normally cannot compensate~
Since the leak signal h(t) can be very much greater than the signal s(t) from the remote end, a minor nonl;nearity in the system may also be devastating for echo cancellation ;f thenonlinear;ty cannot be com-pensated.
In accordance ~ith the proposed invention, the totaL filter length N
in the balance filter ;s divided into two, poss;bly more~ parts ~here N=N1+N2~ the part of the filter which ;s to compensate the greatest echo signal amplitude values, and thereby the greatest nonlinear com-ponents also, has relat;vely short f;lter length and table look-up ;s 5~7~
utilized for this part of the filter. Remaining par-ts N2 may utilize table look-up or vector multiplica-tion. Since the value Nl can be compara-tively small, the total filter requires considerably less memory space, simultaneously as convergence will be quicker than if a conven-tional table look-up tech-nique were utilized for the whole filter.
The object of -the present invention is to provide a balance filter for inclusion in a transmitter-receiver unit wi-th adaptive echo cancellation, which has a filter structure such that nonlineariti.es in -the transmi-t-ter-receiver pa-th present in practically cases may be compensated.
According to -the present invention there is provided an FIR-type balance filter for inclusion in the transmit-receive unit of a telecommunication system in which digital information is transmitted in duplex over a single conductor pair and at least one correction unit is arranged for correc-tion of the filter parameters so that the balance filter out-put signal will be equal to the leak signal coming from the local send side to the receiver side at a sampling instant (k=1,2...), comprising a plurality of delay links divided into two groups with a first number Nl in one group and with a second number N2 in the other group, an addressing unit connected at least to the links of the fi.rst group to form an address from the Nl first incoming data symbols, and a coefficient memory connected to the addressing uni-t, said memory containing values of a plurality of Nl output signals, which form the fil.ter parameters, which are pointed ou-t by the addressing unit, and which are arranged for updating by said correction uni-t, and said second group further havlng means for forming an output signal constitu-ting a filter parameter.
The advantage of the fil-ter in accordance with the invention is, apar-t from compensa-ting nonlinearities in the ` ~f 7~
leak signal, that little memory space is required and a shor-tened convergence time in -the filter is obtained.
The invention will now be described in de-tail with reference to Fig.s 2 to 4 of -the accompanying drawings in-troduced above.
The new implementa-tion of the balance filter B in Figure 1 will be seen from -the block diagram of Figùre 2.
The input ~ - 4a -~ j ~
S~7~;
to the transmitter unit S on which the data symbols bk occur is con-nected to a first delay c;rcu;t T1 ;ncluded ;n the balance ~ilter ~.
In the embodiment ;llustrated in F;gure 2, the balance f;l~ter conta;ns two groups I and II of delay circuits. The first gro~p I ;ncludes the S two delay circuits T1 and T2 and the other group II the dircu;ts T3, T4 and T5. It is naturally possible to extend the number of delay c;rcu;ts as well as the number of groups. Each c;rcu;t T1-T5 delays the ;ncom;ng value bk, bk~ .. a t;me T correspond;ng to the trans-m;tted data speed.
In group I there is also ;ncluded an addressing unit AE1, the ;nputs (;n this case 3) being connected to the ;nputs of the delay circuits T1-T3. The unit AE1 is connected by ;ts outp~t to the ;nput of a memory un;t M1 storing the coefficients c; in table form, sa;d coeff;c;ents const;tut;ng the filter parameters Group I ut;l;zes tabLe look-up and the number N1 ;s here = 3. The values bk~ bk 1~ bk_2 are appl;ed to the address;ng un;t AE1, where a given address ;s formed ;n response to the value of the symbols bk, bk 1~ bk ~ The address thus formed po;nts out ;n the memory un;t M1 a g;ven coefficient ck which occurs on the out-put of the unit M1.
2û In the embodiment illustrated here, group II in the balance filter B
also utilizes table look-up as does group I. In the same ~ay as for group I a g;ven coeffic;en~ c; ;s po;nted out ;n the memory unit M2 from the addressing unit AF2 ;n response to values of the symbols bk 3 bk ~ bk 5 Here N is = 6; N1 = 3 and N2 = 3. The coeff;c;ents ck from group I and c; from group II are appl;ed to both ;nputs of a summing c;rcuit A2. The output of th;s circuit is connected to a digital-to-analogue converter DA for converting the digital sum ck+c, to the analogue signal y~t). This signal is the correction s;gnal for the an310gue s;gnal h(t)+w(t) com;ng from the coupler G to the receive side 3n and, as above, h(t) ;s the leak s;gnal and utt) the remote signal. The realtionsh;p y(t)=h(t) appl;es for correct balanc;ng. S;nce the balance filter has taken cons;derat;on only to the s;x f;rst data symbols, according to the ;llustrated embod;ment, the s;gnal y(t) w;ll only be an approx;mat;on of h(t)~ The effect of subsequent data symbols bk 6 bk 7 etce~era ;s nPgl;gable ;n practice however, and y(t) g;ves a suffic;ently good approx;mat;on of the leak s;gnal h(t)~
The output of the summ;ng c;rcuit A1 is convent;onally taccording to F;gure 1) connected to a sampl;ng rircuit SH, uh;ch samples the cnr-rected s;gnal r(t)=h5t)+w(t)-y(t)~ ~(t) at the sampl;ng ;nstants tk~ k=0,1,..., so that a sampled s;gnal rk ;5 obta;ned~ The s;gnal rk is appl;ed to the detector D on the rece;ve s;de ~o form an esti- ~
mat;on of the transm;tted data flow an from the remote end. Both flows rk and bk have to be uncorrelated for correct balanc;ng to be per-formed by the balance filter4 T~o correction un;ts KB1 and KB2 ope-rat;ng in parallel are therefore connected between the detector D and sampl;ng un;t SH. These are of the same ;mplemen~at;on and are de-scr;bed ;n the Swed;sh patent appl;cat;on ment;oned ;n the introduction.
Both un;ts KB1 and KBZ have an input connected to the ;nput of the transm;tter un;t S and thus rece;ve the data flow bk. The output of KB1 is connected to the memory un;t M1 and the output of KB2 is con-nected to the un;t MZ, to send corrert;on signals to these un;ts~
Furthermore, an OlltpUt of each of the memory un;ts M1, M2 ;s connected to an ;nput on the correction units KB1, KB2 for supplying the coeffi-cient Cj~ ck which is to be corrected. In either correction unit KB1 Ke2 a correl3tion ;s made of the values rk from the sampling unit SH.
As long as there is correlation bet~een the values rk and bk, unam-bigous correct;on s;gnals w;ll be added to the coeff;c;ents cjS ck and new values Cj+1, ck+1 will be fed tD the memory units M1, M2. When correlation ;s no longer presr-nt, th;s means that the signals rk and bk are ;ndependent and correct;on for the Leak s;gnal h(t) is made by the balance f;lter. The correlat;on ;s performed in the un;ts KB1, KB2 by the coeff;c;ent c;, ck (wh;ch ;s to be corrected) be;ng po;nted out by ~he data values bk, bk_1, bk_2 ~for KB1) and bk_3~ bk_4~ ~k-5 (for KBZ), correct;on of the po;nted-out coeff;c;ent tak;ng place ;n res-ponse to the value of rk, as descr;bed ;n the above-ment;oned patent appl;cat;on~
Group II accord;ng to F;gure 2, which carr;es out table look-up, can be replaced by a d;g;tal f;lter accord;ng to F;gure 3, for carry;ng Ollt vector mult;pl;cat;Dn. Th;s ;s su;table ;n the cases where N2 is large, thus requ;r;ng a large number of b;nary values (0 and 1) to be stored ;n S~7~
the memory M2. The filter of F;gure 3 conventionally compr;ses delay c;rcuits T3, T4~ T5, the inputs o~ which, toge~her ~;th the output of T5, are connected to a summ;ng circuit A3 via the controllable multi-pliers MP1-MP4. The output of the summing c;rcu;t A3 ;s connected to S the summ;ng c;rcu;t A2 in F;gure 2. The mult;pliers MPl~MP3 ~ultiply the delayed values:
bk 3~bk_5 by the coeff;c;ents c1,c2,c3 the express;on j=1 bk j~cj be;ng formed after add;t;on, th;s form;ng the output s;gnal from the dig;tal f;lter and corresponding to the output s;gnal from the memory unit M2 in F;gure 2. The coeffic;ents are supp~ied to the mult;pliers MP1-MP4 f~r~ a memory un;t M3 connected to the correct;on unit KB2 for updat;ng, s;m;lar w;th updat;ng of the coeffic;ents ;n the memory unit M2 according to Figure 2n In this case the correlation is carried out ;n the un;ts KB1, KB2 by summ;ng the pro-ducts of the values bk and rk over a given number of sampl;ng ;ntervals,i. e. forming the funct;on:
k= ~ k rk-j A d;agram is shown in Figure 4 for more closely il(ustrating the problem ~ith nonlinearity in the transmitter-receiver part. The diagram shows the impulse response g(t) of the impulse occurring on the transmit side and from the transm;t side to rece;ve s;de v;a the ~n;ts S, G, LP. The leak signal htt) can then be denoted:
~ oo h t~ k=-oo k 9 N
and h(k) = bk_j gj if N data symbols are considered and if the sig-nal paths S, G, LP is linear. A data symbol thus gives rise to the com-ponents bk_1 92~ bk_2 92 etcetera If the signal path is nonlinear, h(k3 may be denoted htk) = f~bk, bk_1~ bk_2~ ~ 91' 92' 93' denotation of h(k~ can be approximately a nonlinear portion and a l;near port;on according to:
30 h(k) = ~1tbk,bk_1,bk_2,91,92,93) ~ bk_3 94~bk_4 95 bk-N 9N
~5~7~
According to the inventive concept~ the balance f;lter is div;ded up according to the above ;n a first group I, ut;~iz;ng table look-up and a second group II u~i~;zing table look-up or vector mult;pl;cat;on~
Group I contains a filter part which by means of table look-up im;tates S the nonl;near ~unction f ~(bk, bk_1~ bk_2, 919293 tains an FIR-type transversal f;lter which imitates the linear funct;on bk_3 94 bk_4 95~ _ In the embodiment according to F;gure 2~ group II also compr;ses a filter utili~;ng table look-up. S;nce in table look-up the output sig-nal of the filter group I tor II) does not need to be linearly related,the nonlinear part of h(t) can be imitated by the balance f;lter.
The advantage of dividing up the balance filter in two (or more3 parts is that quisker updating of the f;lter coeff;c;ents Ck~ Cj in the memory units M1, M2 can be obtained. If the balance f;lter were only to conta;n solely one memory of storing N coeffisients each where coeff;cients may have a plural;ty of levels, e.g. L=2 levels, the num-ber of possible values which must be stored will be 2N. In the case of N=6 the memory must thus store 26=64 b;nary values. W;th the proposed balance f;lter, the memory function is div;ded up into two units M1, Z0 M2 where M1 stores N1=3 and M2 stores N2=3 coe~fic;ents. The unit M1 w;ll thus store 2 =8 and M2 2 =8 d;fferent binary values. Since these operate in parallelO the average time between updating of each un;t ~;ll only be 8/64 = 1/8 of the t;me ~or a s;ngle memory.
~5 A method and apparatus are described in applicants Canadian Patent Application No. 414,564 filed November 1, 1982 ~or achieving adaptive echo cancellation. The principle of the apparatus will be appaxent from the appended Figure l.
The data sequence bn and an are transmitted on the conductor pair to and from the remote end, respectiv~ly. At the local end receive units arrive not only the signal w(t~ ~rom the remote end but also a signal h(t)~ from the local transmi-tter S. In practical cases, the amplitude of the h(t) signal may be very much grea-ter than that o~ the w(t) signal. To enable correct detection in the detector D a signal y(t) is subtracted ~rom the received signal. The former is the ~5~7~
outpu-t si~nal from the balance filter B, the input signal of which comes from the data sequence bn. The balance filter is adjusted adaptively with the aid of the correction unit KB so that v(t) and h(t~ become equal in the sampling instants kT.
After sub-traction there only remains -the w(t) signal from which date from the remote end can be ex~racted.
The balance filter should be of the FIR type (Finite Impulse Response). The output signal of such a filter is defined thus:
Yk Yk ( _ ~ Xk ) where the parameters of the filter are represented by the vector c = (cO, c1, --~ c ;. - la -.~,~ \
5~
and the ;nput signal ;s wr;tten in vector form:
~C = t~k'Xk-l' '''~ Xk-N+1 By correlat;ng ;ts ;nput signals the correct;on un;t KB calculates a correct;on of the f;lter parameters:
c (k ~ 1) = c~k) ~ ~c~k) where c(k) ;s the f;l$er parameter vector at the t;me kT and ~c(k) is a correct;on vector.
The convent;onal ;mplementat;on of an FIR f;lter ut;l;zes the vector mult;pl;cat;on:
lD Yk = c xk = ~ cnxk_n (when M = N) Here the number of parameters ;s thus N.
Another way in wh;ch to reali~e a f;lter of the FIR type ;s to utl;lize table look-up. All conce;vable output signals are stored ;n a memoryO
If each input signaL sampLe xn can assume L d;screte levels the number of poss;ble output s;gnals w;ll be~
N
M = L
For the memory sizetobemanageable, L and N must be relat;vely small.
The ;mplementat;on can be ;nterest;ng, part;cularly ;n binary data transm;ss;on (L = 2). The f;lter input s;gnal vector ~ with N compo-nents determ;nes an address j wh;ch po;nts out a def;n;te f;lter para-meter Cj ;n the memory.
Yk - Cj;~
In vector mult;pl;cat;on the correct;on vector w;ll be:
Ac(k) = (~Co~c~ CN-1 37~
i~e. all the parameters are updated simuLtaneousLy.
In the table Look-up the correction vector will ~e:
~ctk~ ~ (0,0,... ~cj,~ ...,0) i.e. only one of the parameters ;s updated at each correct;on occasion.
It will be seen f~om the above that the table f;lter r2sults ;n large memory space if N ;s large (the memory s;ze ;ncreases exponentially w;th n). S;nce in many practical cases ;t may be necessary to have a relat;vely large value of N~ the table f;lter discussed above may be unsuitable.
.
The transm;tter un;t S, coupler G and lowpass f;lter LP as well as other un;ts not shown ;n F;gure 1 ;n the transmitter-receiver un;t give rise to nonlinearit;es. The signal ~(t) coming to the hybrid cou~ler G from the remote end is comparatively weak. The signal h(t) from the trans-m;tter unit S through ~he coupler G (the leak signal) is~ sn the other hand, at least at the beginning oftransmit-rcceivem~de, relati~ely stLong before echo cancellation has managed to come into action. In practical cases~ the signal htt) may be much greater than w(t), e.g. a hundred times as great. S;nce the signal h(t) passed a nonlinear signal path, components from a given transmitted data symbol bk arise ;n the signal htt), for wh;ch a l;near balance f;lter normally cannot compensate~
Since the leak signal h(t) can be very much greater than the signal s(t) from the remote end, a minor nonl;nearity in the system may also be devastating for echo cancellation ;f thenonlinear;ty cannot be com-pensated.
In accordance ~ith the proposed invention, the totaL filter length N
in the balance filter ;s divided into two, poss;bly more~ parts ~here N=N1+N2~ the part of the filter which ;s to compensate the greatest echo signal amplitude values, and thereby the greatest nonlinear com-ponents also, has relat;vely short f;lter length and table look-up ;s 5~7~
utilized for this part of the filter. Remaining par-ts N2 may utilize table look-up or vector multiplica-tion. Since the value Nl can be compara-tively small, the total filter requires considerably less memory space, simultaneously as convergence will be quicker than if a conven-tional table look-up tech-nique were utilized for the whole filter.
The object of -the present invention is to provide a balance filter for inclusion in a transmitter-receiver unit wi-th adaptive echo cancellation, which has a filter structure such that nonlineariti.es in -the transmi-t-ter-receiver pa-th present in practically cases may be compensated.
According to -the present invention there is provided an FIR-type balance filter for inclusion in the transmit-receive unit of a telecommunication system in which digital information is transmitted in duplex over a single conductor pair and at least one correction unit is arranged for correc-tion of the filter parameters so that the balance filter out-put signal will be equal to the leak signal coming from the local send side to the receiver side at a sampling instant (k=1,2...), comprising a plurality of delay links divided into two groups with a first number Nl in one group and with a second number N2 in the other group, an addressing unit connected at least to the links of the fi.rst group to form an address from the Nl first incoming data symbols, and a coefficient memory connected to the addressing uni-t, said memory containing values of a plurality of Nl output signals, which form the fil.ter parameters, which are pointed ou-t by the addressing unit, and which are arranged for updating by said correction uni-t, and said second group further havlng means for forming an output signal constitu-ting a filter parameter.
The advantage of the fil-ter in accordance with the invention is, apar-t from compensa-ting nonlinearities in the ` ~f 7~
leak signal, that little memory space is required and a shor-tened convergence time in -the filter is obtained.
The invention will now be described in de-tail with reference to Fig.s 2 to 4 of -the accompanying drawings in-troduced above.
The new implementa-tion of the balance filter B in Figure 1 will be seen from -the block diagram of Figùre 2.
The input ~ - 4a -~ j ~
S~7~;
to the transmitter unit S on which the data symbols bk occur is con-nected to a first delay c;rcu;t T1 ;ncluded ;n the balance ~ilter ~.
In the embodiment ;llustrated in F;gure 2, the balance f;l~ter conta;ns two groups I and II of delay circuits. The first gro~p I ;ncludes the S two delay circuits T1 and T2 and the other group II the dircu;ts T3, T4 and T5. It is naturally possible to extend the number of delay c;rcu;ts as well as the number of groups. Each c;rcu;t T1-T5 delays the ;ncom;ng value bk, bk~ .. a t;me T correspond;ng to the trans-m;tted data speed.
In group I there is also ;ncluded an addressing unit AE1, the ;nputs (;n this case 3) being connected to the ;nputs of the delay circuits T1-T3. The unit AE1 is connected by ;ts outp~t to the ;nput of a memory un;t M1 storing the coefficients c; in table form, sa;d coeff;c;ents const;tut;ng the filter parameters Group I ut;l;zes tabLe look-up and the number N1 ;s here = 3. The values bk~ bk 1~ bk_2 are appl;ed to the address;ng un;t AE1, where a given address ;s formed ;n response to the value of the symbols bk, bk 1~ bk ~ The address thus formed po;nts out ;n the memory un;t M1 a g;ven coefficient ck which occurs on the out-put of the unit M1.
2û In the embodiment illustrated here, group II in the balance filter B
also utilizes table look-up as does group I. In the same ~ay as for group I a g;ven coeffic;en~ c; ;s po;nted out ;n the memory unit M2 from the addressing unit AF2 ;n response to values of the symbols bk 3 bk ~ bk 5 Here N is = 6; N1 = 3 and N2 = 3. The coeff;c;ents ck from group I and c; from group II are appl;ed to both ;nputs of a summing c;rcuit A2. The output of th;s circuit is connected to a digital-to-analogue converter DA for converting the digital sum ck+c, to the analogue signal y~t). This signal is the correction s;gnal for the an310gue s;gnal h(t)+w(t) com;ng from the coupler G to the receive side 3n and, as above, h(t) ;s the leak s;gnal and utt) the remote signal. The realtionsh;p y(t)=h(t) appl;es for correct balanc;ng. S;nce the balance filter has taken cons;derat;on only to the s;x f;rst data symbols, according to the ;llustrated embod;ment, the s;gnal y(t) w;ll only be an approx;mat;on of h(t)~ The effect of subsequent data symbols bk 6 bk 7 etce~era ;s nPgl;gable ;n practice however, and y(t) g;ves a suffic;ently good approx;mat;on of the leak s;gnal h(t)~
The output of the summ;ng c;rcuit A1 is convent;onally taccording to F;gure 1) connected to a sampl;ng rircuit SH, uh;ch samples the cnr-rected s;gnal r(t)=h5t)+w(t)-y(t)~ ~(t) at the sampl;ng ;nstants tk~ k=0,1,..., so that a sampled s;gnal rk ;5 obta;ned~ The s;gnal rk is appl;ed to the detector D on the rece;ve s;de ~o form an esti- ~
mat;on of the transm;tted data flow an from the remote end. Both flows rk and bk have to be uncorrelated for correct balanc;ng to be per-formed by the balance filter4 T~o correction un;ts KB1 and KB2 ope-rat;ng in parallel are therefore connected between the detector D and sampl;ng un;t SH. These are of the same ;mplemen~at;on and are de-scr;bed ;n the Swed;sh patent appl;cat;on ment;oned ;n the introduction.
Both un;ts KB1 and KBZ have an input connected to the ;nput of the transm;tter un;t S and thus rece;ve the data flow bk. The output of KB1 is connected to the memory un;t M1 and the output of KB2 is con-nected to the un;t MZ, to send corrert;on signals to these un;ts~
Furthermore, an OlltpUt of each of the memory un;ts M1, M2 ;s connected to an ;nput on the correction units KB1, KB2 for supplying the coeffi-cient Cj~ ck which is to be corrected. In either correction unit KB1 Ke2 a correl3tion ;s made of the values rk from the sampling unit SH.
As long as there is correlation bet~een the values rk and bk, unam-bigous correct;on s;gnals w;ll be added to the coeff;c;ents cjS ck and new values Cj+1, ck+1 will be fed tD the memory units M1, M2. When correlation ;s no longer presr-nt, th;s means that the signals rk and bk are ;ndependent and correct;on for the Leak s;gnal h(t) is made by the balance f;lter. The correlat;on ;s performed in the un;ts KB1, KB2 by the coeff;c;ent c;, ck (wh;ch ;s to be corrected) be;ng po;nted out by ~he data values bk, bk_1, bk_2 ~for KB1) and bk_3~ bk_4~ ~k-5 (for KBZ), correct;on of the po;nted-out coeff;c;ent tak;ng place ;n res-ponse to the value of rk, as descr;bed ;n the above-ment;oned patent appl;cat;on~
Group II accord;ng to F;gure 2, which carr;es out table look-up, can be replaced by a d;g;tal f;lter accord;ng to F;gure 3, for carry;ng Ollt vector mult;pl;cat;Dn. Th;s ;s su;table ;n the cases where N2 is large, thus requ;r;ng a large number of b;nary values (0 and 1) to be stored ;n S~7~
the memory M2. The filter of F;gure 3 conventionally compr;ses delay c;rcuits T3, T4~ T5, the inputs o~ which, toge~her ~;th the output of T5, are connected to a summ;ng circuit A3 via the controllable multi-pliers MP1-MP4. The output of the summing c;rcu;t A3 ;s connected to S the summ;ng c;rcu;t A2 in F;gure 2. The mult;pliers MPl~MP3 ~ultiply the delayed values:
bk 3~bk_5 by the coeff;c;ents c1,c2,c3 the express;on j=1 bk j~cj be;ng formed after add;t;on, th;s form;ng the output s;gnal from the dig;tal f;lter and corresponding to the output s;gnal from the memory unit M2 in F;gure 2. The coeffic;ents are supp~ied to the mult;pliers MP1-MP4 f~r~ a memory un;t M3 connected to the correct;on unit KB2 for updat;ng, s;m;lar w;th updat;ng of the coeffic;ents ;n the memory unit M2 according to Figure 2n In this case the correlation is carried out ;n the un;ts KB1, KB2 by summ;ng the pro-ducts of the values bk and rk over a given number of sampl;ng ;ntervals,i. e. forming the funct;on:
k= ~ k rk-j A d;agram is shown in Figure 4 for more closely il(ustrating the problem ~ith nonlinearity in the transmitter-receiver part. The diagram shows the impulse response g(t) of the impulse occurring on the transmit side and from the transm;t side to rece;ve s;de v;a the ~n;ts S, G, LP. The leak signal htt) can then be denoted:
~ oo h t~ k=-oo k 9 N
and h(k) = bk_j gj if N data symbols are considered and if the sig-nal paths S, G, LP is linear. A data symbol thus gives rise to the com-ponents bk_1 92~ bk_2 92 etcetera If the signal path is nonlinear, h(k3 may be denoted htk) = f~bk, bk_1~ bk_2~ ~ 91' 92' 93' denotation of h(k~ can be approximately a nonlinear portion and a l;near port;on according to:
30 h(k) = ~1tbk,bk_1,bk_2,91,92,93) ~ bk_3 94~bk_4 95 bk-N 9N
~5~7~
According to the inventive concept~ the balance f;lter is div;ded up according to the above ;n a first group I, ut;~iz;ng table look-up and a second group II u~i~;zing table look-up or vector mult;pl;cat;on~
Group I contains a filter part which by means of table look-up im;tates S the nonl;near ~unction f ~(bk, bk_1~ bk_2, 919293 tains an FIR-type transversal f;lter which imitates the linear funct;on bk_3 94 bk_4 95~ _ In the embodiment according to F;gure 2~ group II also compr;ses a filter utili~;ng table look-up. S;nce in table look-up the output sig-nal of the filter group I tor II) does not need to be linearly related,the nonlinear part of h(t) can be imitated by the balance f;lter.
The advantage of dividing up the balance filter in two (or more3 parts is that quisker updating of the f;lter coeff;c;ents Ck~ Cj in the memory units M1, M2 can be obtained. If the balance f;lter were only to conta;n solely one memory of storing N coeffisients each where coeff;cients may have a plural;ty of levels, e.g. L=2 levels, the num-ber of possible values which must be stored will be 2N. In the case of N=6 the memory must thus store 26=64 b;nary values. W;th the proposed balance f;lter, the memory function is div;ded up into two units M1, Z0 M2 where M1 stores N1=3 and M2 stores N2=3 coe~fic;ents. The unit M1 w;ll thus store 2 =8 and M2 2 =8 d;fferent binary values. Since these operate in parallelO the average time between updating of each un;t ~;ll only be 8/64 = 1/8 of the t;me ~or a s;ngle memory.
Claims (3)
PROPERTY OR PRIVILEGE IS CLAIMED ARE DEFINED AS FOLLOWS:
1. An FIR-type balance filter for inclusion in the transmit-receive unit of a telecommunication system in which digital information is transmitted in duplex over a single conductor pair and at least one correction unit is arranged for correction of the filter parameters so that the balance filter output signal will be equal to the leak signal coming from the local send side to the receiver side at a sampling instant (k=1,2...), comprising a plurality of delay links divided into two groups with a first number N1 in one group and with a second number N2 in the other group, an addressing unit connected at least to the links of the first group to form an address from the N1 first incoming data symbols, and a coefficient memory connected to the addressing unit, said memory containing values of a plurality of N1 output signals which form the filter parameters, which are pointed out by the addressing unit, and which are arranged for updating by said correction unit, and said second group further having means for forming an output signal constitu-ting a filter parameter.
2. Balance filter as claimed in claim 1, wherein the second group having N2 delay links is connected to a second addressing unit to form an address from the N2-N-N11 transmitted data symbols after the N1 symbols first coming into the filter, and a second coefficient memory is connected to said second addressing unit, said second memory containing the value of a plurality of N2 output signals which form the filter parameters and which are pointed out by said second addressing unit.
3. Balance filter as claimed in claim 1, wherein the links of the second group form a part of a transversal filter which vector-multiplies the values coming into the filter group by coefficient values stored in a further co-efficient memory.
Applications Claiming Priority (2)
| Application Number | Priority Date | Filing Date | Title |
|---|---|---|---|
| SE8106444-3 | 1981-11-02 | ||
| SE8106444A SE426764B (en) | 1981-11-02 | 1981-11-02 | PROCEDURE TO ACCEPT ADAPTIVE ECO-ELIMINATION BY TRANSFER OF DIGITAL INFORMATION IN DUPLEX EQUIPMENT FOR CARRYING OUT THE PROCEDURE |
Publications (1)
| Publication Number | Publication Date |
|---|---|
| CA1205876A true CA1205876A (en) | 1986-06-10 |
Family
ID=20344930
Family Applications (2)
| Application Number | Title | Priority Date | Filing Date |
|---|---|---|---|
| CA000414386A Expired CA1205876A (en) | 1981-11-02 | 1982-10-28 | Fir-type balance filter incorporated in the transmitter-receiver unit in a telecommunication system |
| CA000414564A Expired CA1191224A (en) | 1981-11-02 | 1982-11-01 | Method of providing adaptive echo cancellation in transmission of digital information in duplex, and apparatus for performing the method |
Family Applications After (1)
| Application Number | Title | Priority Date | Filing Date |
|---|---|---|---|
| CA000414564A Expired CA1191224A (en) | 1981-11-02 | 1982-11-01 | Method of providing adaptive echo cancellation in transmission of digital information in duplex, and apparatus for performing the method |
Country Status (19)
| Country | Link |
|---|---|
| US (1) | US4539674A (en) |
| EP (1) | EP0092570B1 (en) |
| JP (1) | JPS58501850A (en) |
| KR (1) | KR860002211B1 (en) |
| AU (1) | AU553372B2 (en) |
| BR (1) | BR8207953A (en) |
| CA (2) | CA1205876A (en) |
| DE (1) | DE3265650D1 (en) |
| DK (1) | DK306483A (en) |
| ES (1) | ES8402483A1 (en) |
| FI (1) | FI74561C (en) |
| IE (1) | IE53759B1 (en) |
| IT (1) | IT1153338B (en) |
| MX (1) | MX152441A (en) |
| MY (1) | MY8600355A (en) |
| NO (1) | NO157238C (en) |
| SE (1) | SE426764B (en) |
| WO (1) | WO1983001716A1 (en) |
| YU (1) | YU244082A (en) |
Families Citing this family (12)
| Publication number | Priority date | Publication date | Assignee | Title |
|---|---|---|---|---|
| SE434789B (en) * | 1982-12-21 | 1984-08-13 | Ellemtel Utvecklings Ab | METHOD OF DETECTING DANGEROUS LINE TENSIONS OVER A DOUBLE LINE |
| FR2540314A1 (en) * | 1983-01-31 | 1984-08-03 | Trt Telecom Radio Electr | METHOD FOR INITIALIZING COEFFICIENT FILTERS IN A NEAR-AND-NEAR ECHO CANCELLATION DEVICE AND DEVICE FOR IMPLEMENTING SAID METHOD |
| JPS59151546A (en) * | 1983-02-18 | 1984-08-30 | Nec Corp | Adaptive type echo cancellor |
| DE3327467A1 (en) * | 1983-07-29 | 1985-02-14 | Siemens AG, 1000 Berlin und 8000 München | METHOD AND CIRCUIT ARRANGEMENT FOR COMPENSATING ECHO SIGNALS |
| US4707824A (en) * | 1983-12-15 | 1987-11-17 | Nec Corporation | Method and apparatus for cancelling echo |
| SE441052B (en) * | 1984-02-01 | 1985-09-02 | Ellemtel Utvecklings Ab | PROCEDURE TO TEST THE FUNCTION OF AN ADAPTIVE ECO-ELIMINATION DEVICE |
| IT1178913B (en) * | 1984-03-26 | 1987-09-16 | Cselt Centro Studi Lab Telecom | ZIONAL TWO-WIRE TRANSMISSION SYSTEM WITH D ECHO CANCELLATION |
| US4672665A (en) * | 1984-07-27 | 1987-06-09 | Matsushita Electric Industrial Co. Ltd. | Echo canceller |
| US4989221A (en) * | 1987-03-30 | 1991-01-29 | Codex Corporation | Sample rate converter |
| US5062102A (en) * | 1988-12-01 | 1991-10-29 | Nec Corporation | Echo canceller with means for determining filter coefficients from autocorrelation and cross-correlation coefficients |
| US5852661A (en) * | 1995-02-17 | 1998-12-22 | Advanced Micro Devices, Inc. | Adaptive echo cancellation used with echo suppression to reduce short and long duration echoes |
| US6965578B1 (en) * | 1998-09-30 | 2005-11-15 | Conexant Systems, Inc. | Echo canceling method and apparatus for digital data communication system |
Family Cites Families (7)
| Publication number | Priority date | Publication date | Assignee | Title |
|---|---|---|---|---|
| FR2277470A1 (en) * | 1974-07-03 | 1976-01-30 | Cit Alcatel | CENTERING DEVICE FOR A TRANSVERSAL FILTER OF AN ECHO CANCELER |
| NL170688C (en) * | 1976-06-28 | 1982-12-01 | Philips Nv | DEVICE FOR SIMULTANEOUS TWO WAY DATA TRANSMISSION ON TWO WIRE CONNECTIONS. |
| US4074086A (en) * | 1976-09-07 | 1978-02-14 | Bell Telephone Laboratories, Incorporated | Joint adaptive echo canceller and equalizer for two-wire full-duplex data transmission |
| DE2801375C3 (en) * | 1977-01-17 | 1982-01-28 | Télécommunications Radioélectriques et Téléphoniques T.R.T., 75013 Paris | Digital echo canceller for a modem for data transmission with the aid of modulation of a carrier |
| DE2729109A1 (en) * | 1977-06-28 | 1979-01-11 | Siemens Ag | Echo compensation circuit for async. carrier system - gives rapid compensation for echoes with continuous or intermittent speech using Hilbert transformer |
| US4268727A (en) * | 1979-03-14 | 1981-05-19 | International Telephone And Telegraph Corporation | Adaptive digital echo cancellation circuit |
| CA1173128A (en) * | 1981-10-13 | 1984-08-21 | Pok F. Lee | Echo cancellation using transversal filters |
-
1981
- 1981-11-02 SE SE8106444A patent/SE426764B/en not_active IP Right Cessation
-
1982
- 1982-10-28 CA CA000414386A patent/CA1205876A/en not_active Expired
- 1982-10-29 ES ES516965A patent/ES8402483A1/en not_active Expired
- 1982-10-29 IE IE2605/82A patent/IE53759B1/en unknown
- 1982-10-29 MX MX195004A patent/MX152441A/en unknown
- 1982-11-01 YU YU02440/82A patent/YU244082A/en unknown
- 1982-11-01 CA CA000414564A patent/CA1191224A/en not_active Expired
- 1982-11-02 IT IT24025/82A patent/IT1153338B/en active
- 1982-11-02 WO PCT/SE1982/000363 patent/WO1983001716A1/en active IP Right Grant
- 1982-11-02 KR KR8204929A patent/KR860002211B1/en active
- 1982-11-02 DE DE8282903352T patent/DE3265650D1/en not_active Expired
- 1982-11-02 BR BR8207953A patent/BR8207953A/en not_active IP Right Cessation
- 1982-11-02 AU AU91222/82A patent/AU553372B2/en not_active Ceased
- 1982-11-02 US US06/514,825 patent/US4539674A/en not_active Expired - Lifetime
- 1982-11-02 JP JP82503355A patent/JPS58501850A/en active Pending
- 1982-11-02 EP EP82903352A patent/EP0092570B1/en not_active Expired
-
1983
- 1983-06-17 FI FI832228A patent/FI74561C/en not_active IP Right Cessation
- 1983-07-01 DK DK306483A patent/DK306483A/en not_active Application Discontinuation
- 1983-07-01 NO NO83832420A patent/NO157238C/en unknown
-
1986
- 1986-12-30 MY MY355/86A patent/MY8600355A/en unknown
Also Published As
| Publication number | Publication date |
|---|---|
| EP0092570B1 (en) | 1985-08-21 |
| DK306483D0 (en) | 1983-07-01 |
| FI74561B (en) | 1987-10-30 |
| FI832228L (en) | 1983-06-17 |
| DK306483A (en) | 1983-07-01 |
| IE822605L (en) | 1983-05-02 |
| IE53759B1 (en) | 1989-02-01 |
| KR840002785A (en) | 1984-07-16 |
| AU553372B2 (en) | 1986-07-10 |
| EP0092570A1 (en) | 1983-11-02 |
| ES516965A0 (en) | 1984-01-16 |
| MY8600355A (en) | 1986-12-31 |
| KR860002211B1 (en) | 1986-12-31 |
| NO157238C (en) | 1988-02-10 |
| DE3265650D1 (en) | 1985-09-26 |
| WO1983001716A1 (en) | 1983-05-11 |
| ES8402483A1 (en) | 1984-01-16 |
| US4539674A (en) | 1985-09-03 |
| SE426764B (en) | 1983-02-07 |
| AU9122282A (en) | 1983-05-18 |
| NO832420L (en) | 1983-07-01 |
| CA1191224A (en) | 1985-07-30 |
| BR8207953A (en) | 1983-10-04 |
| JPS58501850A (en) | 1983-10-27 |
| IT1153338B (en) | 1987-01-14 |
| FI832228A0 (en) | 1983-06-17 |
| IT8224025A0 (en) | 1982-11-02 |
| MX152441A (en) | 1985-07-15 |
| FI74561C (en) | 1988-02-08 |
| YU244082A (en) | 1985-04-30 |
| NO157238B (en) | 1987-11-02 |
Similar Documents
| Publication | Publication Date | Title |
|---|---|---|
| CA1205876A (en) | Fir-type balance filter incorporated in the transmitter-receiver unit in a telecommunication system | |
| EP0240055B1 (en) | Arrangement for compensating for non-linear distortion in an input signal to be digitized and an echo cancelling system comprising such an arrangement | |
| US4362909A (en) | Echo canceler with high-pass filter | |
| US4007341A (en) | Echo cancelling device | |
| US4817081A (en) | Adaptive filter for producing an echo cancellation signal in a transceiver system for duplex digital communication through one single pair of conductors | |
| US4688245A (en) | Method and circuit arrangement for compensating cross-talk and/or echo signals | |
| EP0021317B1 (en) | Adaptive filter with tap coefficient leakage | |
| DE2740141A1 (en) | COMMON ADAPTIVE ECHO CANCELLATION AND EQUALIZATION DEVICE FOR A TWO WIRE FULLY DUPLEX DATA TRANSFER ARRANGEMENT | |
| JPS594336A (en) | Terminal unit in composite transmission system | |
| EP0092569B1 (en) | A fir-type balance filter incorporated in the transmitter-receiver unit in a telecommunication system | |
| WO1993004539A1 (en) | Adaptive hybrid | |
| KR20010033507A (en) | Adaptive non-linear echo compensator | |
| GB2075313A (en) | Echo cancellers | |
| CA2407453A1 (en) | Far-end cross-talk cancellation for data transmission on multiple propagation modes | |
| JP2928801B2 (en) | Echo cancellation method and subscriber line audio processing circuit | |
| CA1265594A (en) | Echo cancellers for bidirectional digital transmission systems | |
| SE447777B (en) | PROCEDURE FOR SETTING A DIGITAL EQUIPMENT FILTER AT THE SAME TIME OF ADAPTIVE ECO-ELIMINATION AND ADAPTIVE ELIMINATION OF INTERFACES INCURRED BY INTERSYMBOL INTERFERENCE, AND PROCEDURE FOR IMPLEMENTATION OF PROCEDURE | |
| US6400760B1 (en) | Apparatus and method for an error signal compression technique in a fast adaptive equalizer circuit | |
| US5751769A (en) | Programmable digital linear and nonlinear transversal equalizer | |
| US4837780A (en) | Transmit line buildout circuits | |
| KR20020059739A (en) | Device and method for echo compensation in a two-wire full duplex channel transmission method | |
| JP3419130B2 (en) | Echo canceller device | |
| US5027370A (en) | Circuit arrangement for the equalization of digital signals received in analog form | |
| CA1266896A (en) | Transmit line buildout circuits | |
| Kanemasa et al. | An adaptive filter convergence method for echo cancellation and decision feedback equalization |
Legal Events
| Date | Code | Title | Description |
|---|---|---|---|
| MKEX | Expiry |