Technique for improving the blind convergence of an adaptive equalizer using a transition algorithm
Info
 Publication number
 CA2217093A1 CA2217093A1 CA 2217093 CA2217093A CA2217093A1 CA 2217093 A1 CA2217093 A1 CA 2217093A1 CA 2217093 CA2217093 CA 2217093 CA 2217093 A CA2217093 A CA 2217093A CA 2217093 A1 CA2217093 A1 CA 2217093A1
 Authority
 CA
 Grant status
 Application
 Patent type
 Prior art keywords
 cma
 algorithm
 yn
 mma
 fig
 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.)
 Abandoned
Links
Classifications

 H—ELECTRICITY
 H04—ELECTRIC COMMUNICATION TECHNIQUE
 H04L—TRANSMISSION OF DIGITAL INFORMATION, e.g. TELEGRAPHIC COMMUNICATION
 H04L25/00—Baseband systems
 H04L25/02—Details ; Arrangements for supplying electrical power along data transmission lines
 H04L25/03—Shaping networks in transmitter or receiver, e.g. adaptive shaping networks ; Receiver end arrangements for processing baseband signals
 H04L25/03006—Arrangements for removing intersymbol interference
 H04L25/03012—Arrangements for removing intersymbol interference operating in the time domain
 H04L25/03019—Arrangements for removing intersymbol interference operating in the time domain adaptive, i.e. capable of adjustment during data reception
 H04L25/03038—Arrangements for removing intersymbol interference operating in the time domain adaptive, i.e. capable of adjustment during data reception with a nonrecursive structure
 H04L25/0305—Arrangements for removing intersymbol interference operating in the time domain adaptive, i.e. capable of adjustment during data reception with a nonrecursive structure using blind adaptation

 H—ELECTRICITY
 H04—ELECTRIC COMMUNICATION TECHNIQUE
 H04L—TRANSMISSION OF DIGITAL INFORMATION, e.g. TELEGRAPHIC COMMUNICATION
 H04L25/00—Baseband systems
 H04L25/38—Synchronous or startstop systems, e.g. for Baudot code
 H04L25/40—Transmitting circuits; Receiving circuits
 H04L25/49—Transmitting circuits; Receiving circuits using code conversion at the transmitter; using predistortion; using insertion of idle bits for obtaining a desired frequency spectrum; using three or more amplitude levels ; Baseband coding techniques specific to data transmission systems
 H04L25/4917—Transmitting circuits; Receiving circuits using code conversion at the transmitter; using predistortion; using insertion of idle bits for obtaining a desired frequency spectrum; using three or more amplitude levels ; Baseband coding techniques specific to data transmission systems using multilevel codes
 H04L25/4919—Transmitting circuits; Receiving circuits using code conversion at the transmitter; using predistortion; using insertion of idle bits for obtaining a desired frequency spectrum; using three or more amplitude levels ; Baseband coding techniques specific to data transmission systems using multilevel codes using balanced multilevel codes
 H04L25/4921—Transmitting circuits; Receiving circuits using code conversion at the transmitter; using predistortion; using insertion of idle bits for obtaining a desired frequency spectrum; using three or more amplitude levels ; Baseband coding techniques specific to data transmission systems using multilevel codes using balanced multilevel codes using quadrature encoding, e.g. carrierless amplitudephase coding

 H—ELECTRICITY
 H04—ELECTRIC COMMUNICATION TECHNIQUE
 H04N—PICTORIAL COMMUNICATION, e.g. TELEVISION
 H04N5/00—Details of television systems
 H04N5/14—Picture signal circuitry for video frequency region
 H04N5/21—Circuitry for suppressing or minimising disturbance, e.g. moiré, halo, even if the automatic gain control is involved
 H04N5/211—Ghost signal cancellation
Abstract
(CMA) and the "multimodulus algorithm" (MMA) during blind startup. This approach provides the basis for a "transition algorithm." One example of a transition algorithm is the CMAMMA transition algorithm in which an adaptive filter simply switches from CMA to MMA. Other examples are variations of the CMAMMA transition algorithm and are illustrated by the "Constant Rotation CMAMMA" transition algorithm and the "Dynamic Rotation CMAMMA" transition algorithm.
Description
 Technique for Improvin~ the Blind Conver~ence of an Adaptive Equalizer Usin~ a Transition Al~orithm CrossReference to Related Applications Related subject matter is disclosed in the copending, commonly assigned, U.S.
Patent applications of Werner et al., entitled "Blind Equalization," serial No. 08/646404, filed on May 7, 1996; and Werner et al., entitled "Technique for Improving the Blind Convergence of a TwoFilter Adaptive Equalizer," serial No. XXX, filed on September 19, 1996.
k~.round Of the Invention The present invention relates to communications equipment, and, more particularly, to blind equalization in a receiver.
In blind equalization, the adaptive filters of a receiver are converged without the use of a training signal. As known in the art, there are two techniques for blind equalization: one is referred to herein as the "reduced constellation algorithm" (RCA) (e.g., see Y. Sato, "A Method of SelfRecovering Equalization for Multilevel AmplitudeModulation Systems," IEEE Trans. Commun., pp. 679  682, June 1975; and U.S. Patent No. 4,227,152, issued October 7, 1980 to Godard); and the other technique is the socalled "constant modulus algorithm" (CMA) (e.g., see D. N. Godard, "SelfRecovering Equalization and Carrier Tracking in TwoDimensional Data Communications Systems,"
IEEE Trans. Commun., vol. 28, no. 11, pp. 18671875, Nov. 1980; and N. K. Jablon, "Joint Blind Equalization, Carrier Recovery, and Timing Recovery for HighOrder QAM
Signal Constellations", IEEE Trans. Signal Processing, vol. 40, no. 6, pp. 13831398, 1992.) Further, the copending, commonly assigned, U.S. Patent application of Werner et al., entitled "Blind Equalization," serial No. 08/646404, filed on May 7, 1996, presents an new blind equalization techniquethe multimodulus algorithm (MMA)as an alternative to the abovementioned RCA and CMA approaches.
However, for all blind equalization approaches the most filn~ment~l performance issue is the ability to achieve reliable initial convergenceelse the adaptive filter may converge to a wrong solution such as the wellknown "diagonal solution."
Generally spe~king, the RCA algorithm has less reliable convergence than either the CMA or MMA algorithms. As between the CMA and MMA algorithms, these algorithms have both benefits and drawbacks. For example, the CMA algorithm provides more reliable convergence  thus avoiding incorrect diagonal solutions  but the CMA
algorithm requires an expensive rotator. In comparison, the MMA algolilhl.l does not require an expensive rotator but is more susceptible than the CMA algorithm to incorrect convergence.
Summary of the Invention We have realized a blind equalization technique that uses both the CMA and MMA algorithms to achieve complementary results. This approach provides the basis for a "transition algorithm" that yields more reliable blind convergence without creating diagonal solutions, and avoids the expense of a rotator.
In an embodiment of the invention, one example of a transition algorithm is referred to herein as the CMAMMA transition algorithm. In this approach, an adaptive 10 filter simply switches from CMA to MMA.
In other embodiments of the invention, modified CMAMMA transition algorithms are presented which provide for improved constellation rotation in comparison to the basic CMAMMA approach. Illustrative examples of these modified CMAMMA
approaches are the "Constant Rotation CMAMMA" transition algorithm and the 15 "Dynamic Rotation CMAMMA" transition algorithm.
BriefDer~ ;~,lionofthe Dr~
FIG. 1 is an illustrative block diagram of a portion of a communications system embodying the principles of the invention;
FIG. 2 is an illustrative block diagram of a phasesplitting equalizer;
FIG. 3 is an illustrative block diagram of a portion of an adaptive filter for use in an equalizer;
FIG. 4 is an illustrative block diagrarn of a crosscoupled equalizer;
FIG. 5 is an illustrative block diagram of a fourfilter equalizer;
FIG. 6 is an illustrative signal point plot of an output signal of an equalizer before 25 convergence;
FIG. 7 is an illustrative signal point plot of an output signal of an equalizer for a system using the MMA blind equalization method;
FIG. 8 is an illustrative signal point plot illustrating the reduced signal point constellation of the RCA blind equalization method;
FIG. 9 is an illustrative signal point plot illustrating the circular contour of the CMA blind equalization method;
FIG. 10 is an illustrative signal point plot illustrating the piecewise linear contours of the MMA blind equalization method;
FIGs. 11 and 12 are illustrative block diagrams of a portion of a receiver 35 embodying the principles of the invention;
FIGs. 13, 14, and 15, are illustrative signal point plots illustrating the piecewise linear contours of the MMA blind equalization method for a nonsquare constellation;
FIGs. 16 and 17 are illustrative signal point plots of an output signal of an equalizer for a communications system using a twostep MMA blind equalization method;
FIG. 18 shows a table providing a general comparison between the RCA, CMA, and MMA, blind equalization methods, without CHCF;
FIG. 19 shows a table of illustrative data values for use in the RCA, CMA, and MMA, blind equalization methods;
FIG. 20 is an illustrative graph of an incorrect diagonal solution for a 64CAP
signal point constellation;
FIG. 21 shows an illustrative block diagram of a twofilter structure incol~o~dLing a CMAMMA transition algorithm;
FIG. 22 shows an illustrative flow diagram for use in the structure of FIG. 21;
FIG. 23 shows an illustrative block diagram of a twofilter structure incorporating a CRCMA transition algorithm;
FIG. 24 shows an illustrative flow diagram for use in the structure of FIG. 23;
FIG. 25 shows an illustrative block diagram of a twofilter structure incorporating a DRCMA transition algorithm;
FIG. 26 shows an illustrative flow diagram for use in the structure of FIG. 25; and FIGs. 27 and 28 show illustrative signal point constellations recovered using the CRCMA and DRCMA approaches, respectively.
Detailed Description An illustrative highlevel block diagram of a portion of a communications system25 embodying the principles of the invention is shown in FIG. 1. For illustrative purposes only, it is assumed that receiver 10 receives a CAP (carrierless, amplitude modulation, phase modulation) signal, which can be represented by:
r(t) = ~, [anp(t  nT)  bn p(t  nT)] + ~(t) (1) where an and bn are discretevalued multilevel symbols, p(t) and p (t) are impulse 30 responses which form a Hilbert pair, T is the symbol period, and ~,(t) is additive noise introduced in the channel.
It is assumed that the CAP signal in equation (1) has been distorted while prop~ting through communications channel 9 and experiences intersymbol interference (ISI). This ISI consists of intrachannel ISI (an or bn symbols interfering with each other) and interchannel ISI (an and bn symbols interfering with each other). The purpose of receiver 10 is to remove the ISI and minimi7~ the effect of the additive noise ~(t) to provide signal r'~). The inventive concept will illustratively be described in the context of a combined CMA and MMA blind equalization algorithm for use within receiver 10.
5 However, before describing the inventive concept, some background information on adaptive filters and the abovemention RCA, CMA, and MMA algorithms is presented.
Also, as used herein, an adaptive filter is, e.g., a fractionally spaced linear equalizer, which is hereafter simply referred to as an FSLE equalizer or, simply, an equalizer.
Equalizer Structures An illustrative phasesplitting FSLE equalizer 100 is shown in FIG. 2. It is assumed that FSLE equalizer 100 operates on an input signal comprising two dimensions:
an inphase component and a quadrature component. FSLE equalizer 100 comprises two parallel digital adaptive filters implemented as finite impulse response (FIR) filters 110 and 120. Equalizer 100 is called a "phasesplitting FSLE" because the two FIR filters 110 and 120 converge to inphase and quadrature filters. Some illustrative details of the equalizer structure are shown in FIG. 3. The two FIR filters 110 and 120 share the same tapped delay line 115, which stores sequences of successive AnalogtoDigital Converter (A/D) 125 samples rk. The sampling rate l/T' of A/D 125 is typically three to four times higher than the symbol rate l/T and is chosen in such a way that it satisfies the sampling 20 theorem for real signals. It is assumed that T/T' = i, where i is an integer.The outputs of the two adaptive FIR filters 110 and 120 as shown in FIG. 3 are computed at the symbol rate l/T. The equalizer taps and input samples can be represented by a corresponding Ndimensional vector. As such, the following relationships are now defined:
rnT = [rk, rk l, , rk N ] = vector of A/D samples in delay line; (2) CnT = [Co~ Cl, C2, ~ ~ ~, CN] = vector of inphase tap coefficients; and (3) dnT = [do~ d~, d2, , dN] = vector of quadrature phase tap coeff1cients; (4) where the superscript T denotes vector transpose, the subscript n refers to the symbol period nT, and k = in.
Let Yn and ynbe the computed outputs of the inphase and quadrature filters, respectively, and:
Yn = Cn rn; and (5) Yn = dn rn. (6) An X/Y display of the outputs Yn and Yn or, equivalently, of the complex output  Yn= Yn + rn, is called a signal constellation. FIGs. 6 and 17 show an 64CAP
constellation before and after illustrative convergence using the MMA algorithm. (The term "64CAP," refers to the number of predefined symbols in the signal space or signal constellation each symbol representing 6 bits since 26 = 64. Additional information on a CAP communications system can be found in J. J. Werner, "Tutorial on CarrierlessAM/PM  Part I  Fundamentals and Digital CAP Transmitter," Contribution to ANSIX3T9.5 TP/PMD Working Group, Minneapolis, June 23, 1992.) After convergence, thesignal constellation consists of a display of the complex symbols An = an + jbn corrupted by some small noise and ISI.
In the normal mode of operation, decision devices (or slicers) 130 and 135 shownin FIG. 2 compare the sampled outputs Yn and Yn Of equalizer 100 with valid symbol values an and bn and makes a decision on which symbols have been transmitted. These sliced symbols will be denoted ân and bn . The receiver then computes the following inphase and quadrature errors en and en en = Yn ân ~ (7a) en = Ynbn ~ (7b) and the tap coefficients of the two adaptive filters are updated using the f~rnili~r leastmeansquare (LMS) algorithm, i.e., Cn+l= Cn ~ a enrn~ (8a) dn+l = dn~ a en rn~ (8b) where a is the step size used in the tap adjustment algorithm.
Turning now to FIG. 4, a crosscoupled FSLE, 200, is shown. For this equalizer structure, the A/D samples are first fed to two fixed inphase and quadrature FIR filters, 210 and 205, respectively. In this case, the sampling rate l/T' of A/D 125 is typically equal to four times the symbol rate l/T. The outputs of the two fixed FIR filters are computed at a rate 1/T" that is consistent with the sampling theorem for analytic signals as known in the art. The output signals are then fed to equalizer 200 having a socalled crosscoupled structure. Typically, l~T" is twice the symbol rate 1/T.
The crosscoupled equalizer 200 uses two adaptive FIR filters 215a and 215b, each with tap vectors cn and dn. For simplicity, the same tap vector notations cn and dn (which have been used for the previous described equalizer 100 of FIG. 2) are used again.
However, it should be clear to those skilled in the art that the tap vectors are different for the two types of equalizers. These two filters are each used twice to compute the outputs  Yn and Yn Of the equalizer. Let rn and rn be the output vectors of the inphase and quadrature filters that are used to compute the outputs of the crosscoupled equalizer.
The following definitions can be made:
Cn = cn +jdn~ (9a) Rn = rn + jrn, and (9b) Yn Yn +JYn ~ (9c) The complex output Yn of the equalizer can be written in the following compact way:
Yn = Cn Rn~ (10) 10 where the asterisk * denotes complex conjugate. Making the following definitions for the sliced complex symbol An and the complex error En:
An = ân + ibn ~ (1 la) En = YnAn (1 lb) The LMS algorithm for updating the complex tap vector Cn can be written as:
Cn+/ = Cna En Rn ( 12) Turning now to FIG. 5, a fourfilter FSLE is shown. Fourfilter equalizer 300 has the same general structure as crosscoupled FSLE 200 shown in FIG. 4, except that the adaptive portion consists of four different filters rather than two filters which are used twice. For this reason it is called a fourfilter FSLE. The two output signals of equalizer 20 300 are computed as follows:
Yn = CInrn + d2 nrn, and (13a) Y n C2n r n ~ dlnrn (13b) Using the definitions for the inphase and quadrature errors en and en in equations (7a) and (7b), the following tap updating algorithms for the four filters result:
cl n+l = cl n~ a en rn~ (14a) dl n+l = dl,n + a en rn~ (14b) C2,n+1= C2,n a enrn~ and (15a) d2n+, = d2,n  aenrn (15b) Having generally described the structure of some priorart equalizers as shown in FIGs. 2  5, a general overview of the concept of blind equalization will now be described using the equalizer structure of FIG. 2.
Concept of Blind Equalization In the normal (steadystate) mode of operation, the decision devices in FIG. 2, i.e., slicers 130 and 135, compare the equalizer complex output samples, Yn~ (where Yn = Yn +
~Yn ), with all the possible transmitted complex symbols, An (where An = an + jbn)~ and selects the symbol An which is the closest to Yn. The receiver then computes an error, En~ where:
En=Yn An~ (16) which is used to update the tap coefficients of equalizer 100. This type of tap adaptation is called "decision directed", because it uses the decisions of slicers 130 and 135. The most common tap updating algorithm is the LMS algorithm, which is a stochastic gradient algorithm that minimi7~s the mean square error (MSE), which is defined as:
MSE _ E[l Enl2] = E [IYn  Anl2] = E[en] + E[en2]~ (17) where E[ ] denotes expectation and en and en are inphase and quadrature errors,respectively.
At the beginning of startup, the output signal of equalizer 100, Yn~ is corrupted by a lot of intersymbol interference, as illustrated in FIG. 6. The latter represents 20 experimental data obtained for a 64CAP receiver using a phasesplitting FSLE as represented by FIG. 2.
When a training sequence is used during startup (i.e., a predefined sequence of An symbols), the receiver can compute meaningful errors En by using the equalizer output signal Yn and the known sequence of transmitted symbols An. In this case, tap adaptation 25 is said to be done with "ideal reference" to distinguish it from decision directed tap adaptation.
However, when no training sequence is available, equalizer 100 has to be converged blindly. In this case, a decisiondirected tap updating algorithm cannot be used to converge the equalizer, because the slicer makes too many wrong decisions, as 30 should be apparent from FIG. 6.
As such, the philosophy of blind equalization is to use a tap adaptation algorithm that minimi~es a cost function that is better suited to provide initial convergence of equalizer lO0 than the MSE represented by equation (17). The cost functions used in the RCA, CMA, and MMA algorithms are described below.
Convergence of an equalizer during blind startup usually consists of two main  steps. First, a blind equalization algorithm is used to open the "eye diagram." (Hereafter, this will be referred to as "it opens the eye.") Once the eye is open enough, the receiver switches to a decision directed tap adaptation algorithm.
5 Reduced Constellation Al~orithm (RCA) This section provides a general overview of the RCA algorithm. This general overview is then followed with a description of the RCA algorithm in the context of each of the illustrative equalizer structures, described above.
With the RCA algorithrn, the error used in the tap updating algorithm is derived10 with respect to a signal constellation that has a smaller number of points than the received constellation. As illustration, it is again assumed that the signal constellation comprises 64 symbols. In the RCA algorithm, the reduced constellation typically consists of four signal points only, as shown in FIG. 8. It should be noted that the RCA algorithm requires the use of a decision device, e.g., a slicer, to select the closest signal point from 15 the reduced constellation. The error between the received sarnple Yn and the closest signal point Ar n of the reduced constellation is the complex number:
Ern = ern + jern = Yn ~ Arn ~ where (18) Arn = ârn + jbrn = R [sgn(yn)+ jsgn(yn)]~ and (l9) where sgn ( ) is the signum function and the expression on the right corresponds to the 20 case where the reduced constellation consists of four points. The reduced constellation algorithm minimi7~s the following cost function:
CF=E[IErnl2]=E[er2~n+ er.n]=E[lYn~ Atnl ]~ (20) where E [ ] denotes expectation and where er n refers to the slicer error.
Now, consider the phasesplitting equalizer structure shown in FIG. 2. Using 25 equations (5), (6), and (20), the following equations result:
ern =Yn~ âr,n = CnrrnRsgn(Yn) ~ (21a) er n = Ynbr n = dnrrnR sgn(yn ) (21 b) The gradients of the cost function represented by equation (20) with respect to the tap vectors cn and dn are equal to:
Vc(CF) = 2E[ernrn], and (22a) Vd(CF) = 2E[ e r nrn] (22b) These gradients are equal to zero when the channel is perfectly eql~li7~1 i.e.when the received samples Yn are equal to the symbol values An. This condition leads to the following value of R:
R= E[~ ] (23) For example, consider the gradient with respect to the tap vector cn. From the left of equations (21a) and (21b) there is the condition: E[(Yn R sgn(yn))rn ] = O. With perfect equalization Yn = an. Also, if it is assumed that dirrerent symbols are uncorrelated, then: E[anrn] = knE[an2], where kn is a fixed vector whose entries are a function of the channel. The above condition can then be written as: E[an2]  R E[sgn(an)an] = O.
Noting that sgn (an)a~ = I an I and solving for R, equation (23) results.
The nonaveraged gradients in equations (22a) and (22b) can be used in a stochastic gradient algorithm to adapt the tap coefficients of the equalizer, so that the following tap updating algorithms result:
cn+l = cn  a~yn  R sgn(Yn) ]rn~ and (24a) dn+l = dn ~ a[ Yn ~ R sgn ( Yn )]rn (24b) Turning now to the crosscoupled FSLE structure illustrated by FIG. 4, the complex output Yn of this equalizer is computed from equation (10). Using this expression in equation (20), the gradient of the cost function with respect to the complex tap vector Cn is:
Vc = E[( YnAr n ) Rn] (25) Assuming a perfectly equalized channel the following ~x~res~ion for R results:
R = E[ I Anl2 ] E[ I Anl2 ] (26) E[lanl]+E[lbnl] 2E[Ianl ]
where the e~lession on the right is the same as the one in equation (23) for theusual case where E[lanl]=E[Ibnl]~ The tap updating algorithm for the complex tap vector 25 Cn is given by Cn+l = Cn  a(yn  Arn) Rn (27) Turning now to the fourfilter FSLE structure illustrated by FIG. 5, the outputs Yn and Yn of this fourfilter equalizer structure are computed from equations (13a) and (13b). The gradients of the cost function in equation (20) with respect to the four tap vectors are similar to the ones given in equations (22a) and (22b) and will not be repeated here. The tap updating algorithms are given by:
CI,n+l= Cl~n~a~yn~ R sgn(yn)] rn, (28a) d/ n+l = dl n + a[ Yn ~ R sgn ( Yn )]rn (28b) C2 n+l = C2 n ~ a[ Yn ~ R sgn( Yn )] rn~ and (28c) d2 n+l = d2n~ a[ Yn ~ R sgn (Yn)] rn~ (28d) where the constant R is the same as in equation (23).
The main advantage of RCA is its low cost of implementation because it is typically the least complex of blind equalization algorithms. The tap updating algorithms represented by equations (24a), (24b), (27) and (28) are the same as the standard LMS
algorithms represented by equations (8a) and (8b) except that the slicer uses a different number of points.
The main disadvantages of RCA are its unpredictability and lack of robustness.
The algorithm is known to often converge to socalled "wrong solutions." These 15 solutions are quite acceptable from a channel equalization perspective, but do not allow the receiver to recover the right data. It should be pointed out that the equalizer structure in FIG. 2 is much more likely to converge to wrong solutions than the structure in FIG. 4.
This is due to the fact that the former has many more degrees of freedom than the latter.
A wrong solution that is often observed with the equalizer structure in FIG. 2 is 20 the socalled diagonal solution. In this case, the inphase and quadrature filters both converge to the same filter, so that they both generate the same output samples. As a result, the signal constellation at the output of the equalizer consists of points clustered along a diagonal as illustrated in FIG. 20 for a 64CAP signal point constellation. It has been found that frequency of occurrence of diagonal solutions is mostly communications 25 channel dependent. Specifically, it is created when certain fractional propagation delay offsets are introduced in the channel. (As a point of contrast, FIG. 16 shows anillustrative correct solution for a 64CAP signal point constellation using the MMA blind equalization algorithm.) Other wrong solutions can occur when the inphase and quadrature filters 30 introduce propagation delays which differ by an integral number of symbol periods. As an example, at a given sampling instant, an may appear at the output of the inphase filter while bn l appears at the output of the quadrature filter. This kind of wrong solution can generate points in the signal constellation at the output of the equalizer that do not correspond to transmitted symbols. For example, a 32point signal constellation may be converted into a 36point constellation and the 128point constellation in FIGs. 13, 14, and 15 may be converted into a 144point constellation.
Constant Modulus Al~orithm (CMA) This section provides a general overview of the CMA algorithm. This general 5 overview is then followed with a description of the CMA algorithm in the context of each of the illustrative equalizer structures, described above.
The CMA algol;lhlll minimi7es the dispersion of the equalized samples Yn with respect to a circle with radius R. This is graphically illustrated in FIG. 9. The CMA
algorithm minimi7~s the following cost function:
CF = E[¢YnlLRL) ] ~ (29) where L is a positive integer. The case L=2 is the most commonly used in practice. The cost function in equation (29) is a true twodimensional cost function which minimi7es the dispersion of the equalizer complex output signal Yn with respect to a circular twodimensional contour.
Now, consider the phasesplitting equalizer structure shown in FIG. 2. The gradients of the cost function with respect to the tap vectors cn and dn are given by:
Vc (CF) = 2L x E[(IYnlL~RL)IYnlL~2Ynrn]~ and (30a) Vd(CF) = 2L X E[(IYnlL ~ RL)IYnlL2yn rn] (30b) Assuming a perfectly equalized channel the following value for RL results:
R_ E[lAnl an] = EllAnl ] (31) E[IAni an] E[IAnl ]
where the expression on the right holds for the usual case where the statistics of the symbols an and bn are the same. For L=2, the following stochastic gradient tap updating algorithms results:
cn+/ = cn  a (yn2 + yn2 _ R2 )ynrn, and (32a) dn+l = dn  a (yn2 + ~ _ R2 ) Yn rn. (32b) Turning now to the crosscoupled FSLE structure illustrated by FIG. 4, the gradient of the cost function represented by equation (29) with respect to the complex tap vector Cn is equal to:
Vc (CF) = 2L x El(lynlL  RL)IynlL2 Yn Rn]. (33) For L=2, the tap updating algorithm for the complex tap vector becomes:
Cn+l = Cn  a(lYnl2  R2) Yn Rn~ (34) where R is given by the expression on the right in equation (31).
Turning now to the fourfilter FSLE structure illustrated by FIG. 5, the gradients of the cost function represented by equation (29) with respect to the four tap vectors are 5 similar to the ones given by equations (30a) and (30b). For L=2, the tap updating algorithms become:
cl n+/ = cl n  a (yn2 + yn2 _ R2 )Ynrn~ (35a) d~ n+l = d~ n + a (yn2 + yn2 _ R2 ) Yn rn~ (35b) C2n+/ = C2n  a (yn2 + yn2 _ R2) Yn rn, and (35c) d2n+~ = d2,n a(Yn + YnR ) Ynrn (35d) The constant R is the same as in equation (31).
The main advantages of CMA are its robustness and predictability. Unlike RCA, it rarely converges to wrong solutions. For some applications, other than those considered here, it also has the advantage of being able to partially equalize the channel in the presence of carrier phase variations. The main disadvantage of CMA is its cost of implementation. The CMA tap updating algorithm is more complex than that of the RCA
algorithm and the MMA algorithm and, in addition, the CMA algorithm requires a socalled "rotator" at the output of the equalizer. As a result, once a certain degree of convergence is achieved, the output signal of the equalizer must be counterrotated before switching to a decisiondirected tap adaptation algorithm. The need to use a rotator after the equalizer increases the cost of implementation of CMA for the some type of applications. It should be pointed out, however, that there are other applications, such as voiceband and cable modems, where the rotator function is required anyway for other purposes, such as tracking frequency offset introduced in the channel. In these latter cases, the need to do a rotation does not increase the cost of implementation, and CMA
becomes a very attractive approach.
Multimodulus Al~orithm (MMA) The MMA algorithm minimi7es the dispersion of the equalizer output samples Yn and Yn around piecewise linear inphase and quadrature contours. For the special case of square signal constellations of the type used for 16, 64, and 256CAP systems, the contours become straight lines. This is graphically illustrated in FIG. 10 for a 64point constellation. The multimodulus algorithm minimi7es the following cost function:
CF = E [(yL _ RL(y ))2 +(yL _ RL(y))z] (36) where L is a positive integer and R(Yn) and R(Yn)take discrete positive values, which depend on the equalizer outputs Yn Multimodulus Al~orithm (MMA.)  Square Constellations For square constellations, R(Yn) = R(Yn)= R = constant, so that the cost function of equation (36) becomes:
CF = CFI + CFQ = E[(ynLRL)2 +(yL _ RL)2~ (37) Unlike the cost function for CMA represented by equation (29), this is not a true twodimensional cost function. Rather, it is the sum of two independent onedimensional cost functions CFI and CFQ . The application of the MMA algorithm in the context of the three illustrative types of equalizers (described above) will now be described.
For the phasesplitting equalizer structure shown FIG. 2, the gradients of the cost function in equation (37) with respect to the tap vectors cn and dn are equal to:
Vc (CF) = 2L X E[(IYnlL ~ RL)IYnlL~2Ynrn]~ and (38a) Vd (CF) = 2L x E[(l Yn IL ~ RL)I Yn IL2 Yn rn] (38b) Assuming a perfectly equalized channel, the following value for RL results:
E[a 2 L ~
E[lanlL ]
The best colllpl~ lise between cost and performance is achieved with L=2, in which case the tap updating algorithms become cn+l = cn  a(ynR2)ynrn, and (40a) dn+l = dn  a (yn2 _ R2 ) Yn rn. (40b) Turning now to the crosscoupled FSLE structure illustrated by FIG. 4, the gradient of the cost function represented by equation (37) with respect to the complex tap vector Cn is given by:
Vc (CF) = 2L x ElK Rn]~ (41) where, K [(¦ ¦L RL)¦y ¦L~2y ] +i[(l Yn ¦  R )¦ Yn ¦ Yn ] (42) Assuming a perfectly equalized channel, the value for RL is:
R = E[l IL+ Ib IL ] ' which reduces to equation (39) for the usual case where the symbols an and bn have the same statistics. For L=2, the tap updating algorithm for the complex tap vector Cn becomes:
Cn+l= Cn~ aK Rn~ ( ) where, K=(y2 R2)y+j(y2 _R2) y Turning now to the fourfilter FSLE structure illustrated by FIG. 5, the gradients of the cost function represented by equation (37) with respect to the four tap vectors are similar to the ones given in equations (6.5). For L=2, the tap updating algorithms become:
c/ n+l = cI n  a (yn2 _ R2 )Ynrn~ (46a) dl n+l = dl n + a (yn2 _ R2 ) Yn rn~ (46b) C2 n+l = C2 n ~ a (yn2 _ R2 ) Yn rn, and (46c) d2n+~ = d2n a(YnR ) Ynrn (46d) The constant R is the same as in equation (39).
The abovementioned twostep blind equalization procedure lltili7ing the MMA
al~,o~ l is graphically illustrated by FIGs. 6, 7, 16, and 17 for equalizer 100. The output signal of equalizer 100, before any form of convergence, is shown in FIG. 6. As 20 noted above, FIG. 6 represents ~lhllental data obtained for a 64CAP receiver using a phasesplitting FSLE as represented by FIG. 2. FIG. 7 illu~ tes the beginning of the MMA process convergence. As shown in FIG. 16, the MMA technique converges the equalizer enough to clearly illustrate the 64symbol signal space as 64 noisy clusters.
Although these noisy clusters would, typically, not be acceptable for steadystate 25 operationthe eye is open enough to allow the receiver to switch to a 64point slicer and a decisiondirected LMS algorithm. The end result is a much cleaner constellation, as shown in FIG. 17. Typically, a clean transition can be made between the two modes of adaptation, MMA and decision directed, when the symbol error rate is better than 102, although successful transitions have been observed for worse symbol error rates. It 30 should be pointed out that the noisy clusters in FIG. 16 could be further reduced by decreasing the step size in the MMA tap adjustment algorithm. Indeed, in some applications it may be possible to elimin~te the switching to a decision directed tap adaptation algorithm. However, it should be noted that this would increase the startup time and the required amount of digital precision.
The MMA algorithm for square constellations can be used without modification for nonsquare constellations. In this case, caution has to be exercised in the computation 5 of the constant R, because the discrete levels for the symbols an and bn do not all have the same probability of occurrence (described below). However, it has been found through computer simulations that convergence of the MMA algorithm is somewhat less reliable for nonsquare constellations than for square constellations. This can be corrected by using the modified MMA discussed in the following section.
10 Multimodulus Al~orithm (MM~)  NonSquare Constellations The principle of the modified MMA is illustrated in FIGs. 13, 14, and 15, with respect to a 128CAP signal constellation. (A 128point signal constellation is obtained in the following way. First define a 144point signal constellation using the symbol levels+l,+3,+5,+7,+9,+11,and then remove the four comer points in each quadrant.) 15 Minimi7~tion of the dispersion of the equalizer output samples Yn and ynis now done around piecewise straight lines. Again, this is done independently for Yn and Yn . The inphase cost functions derived from equation (37) are:
CFQ=E[(~R,L)2] if Lynl<K,and (47a) CFQ = E[(ynLR2L)2] if IYnl > K (47b) The quadrature cost functions derived from equation (37) are:
CFI=E[(ynLRIL)2] if lYnl <K,and (47c) CF = E[(YnLR2L)2] if IYnl > K (47d) The constant K is a function of the signal constellation under consideration and is determined empirically. In computer simulations for 128CAP, a suggested value is K =
25 8. Two different moduli R, and R2 are used in equations (47) because the symbols an and bn used in the 128point constellation have two sets of levels {+1,+3,+5,+7} and { + 9, + 113 which have a different probability of occurrence. More moduli can be used if there are more than two sets of symbol levels with dirrelelll statistics.
The moduli Rl and R2 in equations (47) are computed from equation (39) by 30 evaluating the moments of the symbols over the set of symbol levels to which a given modulus applies (additionally described below). As an example, consider FIG. 13, which illustrates the Moduli for the inphase dimension and which applies to the real symbols an of a 128CAP signal constellation. The moments of the symbols can be computed by considering the first quadrant only. Consider the subset of 24 symbols in this quadrant that applies to R,. For these symbols an = 1~ 3, 5, 7, 9, 11; and bn = 1, 3, 5, 7; so that each value of an occurs with probability 4/24=1/6. Similarly, the R2 subset has 8 symbols for which an= 1, 3, 5, 7 and bn = 9~ 11, so that each value of an occurs with probability~ 2/8=1/4. Thus, the variance of the symbols becomes:
r R, symbols, E[an2 ] = 6 (12 + 32 + 52 + 72 + 92 + 112) 4 (48a) f R b l E[ 2] l(12+ 32+ 52+ 72~ 21 (48b) Other moments of the symbols are computed in a similar fashion and then used in equation (39) to evaluate the values of the various moduli.
The tap updating algorithms for the modified MMA algorithm are the same as the ones given in equations (40), (44), and (46), except that the constant R is replaced by either Rl or R2 depending on which equalizer output sample Yn is received. FIG. 14 illustrates the Moduli for the quadrature dimension and which applies to the symbols bn of the 128CAP signal constellation. It should be appal~,~l from FIG. 15, which 15 represents the union of FIGs. 13 and 14, that the inphase and quadrature tap updating algorithms need not use the same moduli R, or R2 in a given symbol period.
Moments of Data Synlhols The following description discusses the concept of "moments of data symbols."
In particular, the closedform expressions for the moments E[lanlL], E[lbnlL], and E[lAnlL]
20 when the symbols an and bn take values proportional to the odd integers ~ 3,i5,_7, are presented. These expressions are then used to get closedform expressions for the constants R used in the three blind equalization algorithms and illustrated in the table of FIG. 19 (described below).
First, it is assumed that the symbols an and bn have the same statistics, so that 25 E[anlL] = E[lbnlL]. Consider first the following known summations of powers of integers:
m ~k= 2 m(m+1), (49a) k l k2 = 6 m(m + 1)(2m + 1), (49b) k3 = 4 [m(m + 1)]2, and (49c) CA 02217093 1997~09~30 m ~ k4 = 30 m(m + 1)(2m + 1)(3m2 + 3m  1) . (49d) k=l These summations can be used to find closedform expressions for sums of powers of odd integers. For example, for power one:
(1 +3+5+7)=(1 +2+3+4+5+6+7)2(1 +2+3) S (50) m 2m1 ml ~(2k1)=~k2 ~k=m2, k=l k=l k=l where the two summations in the middle have been evaluated by using the closedform expression of equation (49a). Similar series manipulations can be used for other sums of powers of odd integers.
Now, consider square signal constellations which use symbols an and bn with values _1,_3,_5,_7,_(2m1), where m is the number of different symbol levels (in magnitude). As an example, for the 4CAP, 16CAP, 64CAP, and 256CAP square signal constellations, m = 1, 2, 4, and 8, respectively. It is also assumed that all the symbol values are equiprobable. As a result, the moments of the symbols an are:
E[l anl ] = ~ (2k1) = m, (51) m k=l E[an2] = ~(2k  1)2 =(4m2 _ 1) (52) E[lan13] = ~(2k1)3 = m(2m2 1), and (53) m k=l E[an4] = ~ (2k  1)4 = l 5 (4m2  1)(12m2  7) . (54) m k=l Next, consider the complex symbols An = an + jbn . Assuming that the symbols an 20 and bn are uncorrelated, the following expressions for the even moments of the complex symbols result:
E[1 An12] = 2E[an2], and (55a) E[1 An 14] = 2E[an4] + 2[EI an2 l ]2. (55b) Using equations (52) and (54) in equation (55b), results in:
E[l Anl4] = 45 (4m2  1)(28m2  13) (56) The above results can now be used to get closedform expressions for the constants R used in the various blind equalization algorithms. The following (remarkably simple) expressions for these constants result:
E[ a"l ] 3m E[a~] 12m2  7 (58) R2 E[l An I ] = 56m 26 (59) With respect to nonsquare signal constellations, the various symbol levels 2k  1 for an and bn have a different probability of occurrence, even when all the complex symbols An are equiprobable. This should be appa,ellt from the 128point constellation 10 illustrated by FIG. 15. In this case, the moments of the symbols have to be computed according to the general formula:
m~ m2 m3 E[lanl ]= P~ ~(2k1) +P2 ~(2k1) +P3 ~(2k1) + (60) k=l m~+l m2+1 where P, is the probability of occurrence of the symbol levels appearing in the corresponding summation. For typical 32CAP and 128CAP constellations the 15 expression in (60) is restricted to two different probabilities P~ and P2.
Everything else being equal (i.e. symbol rate, shaping filters, etc.), it is possible to guarantee a constant average power at the output of a CAP transmitter if E[ an2 ] = E[ bn2 ] = constant, independently of the type of signal constellation that is used. Of course, different signal constellations will have to use different symbol values if 20 the average power constraint has to be satisfied. Thus, in general, a signal constellation will use symbol values ~(2k  1) where ~ is chosen in such a way that the average power constraint is satisfied. For simplicity, it is assumed that E[ an2 ] = 1. For square constellations, the value of ~ can then be determined from equation (52), to result in:
E[a2]= 1 ~[~.(2k1)]2=~(4m 1)=1 ~ ~2 = 4 2 1 (61) Using this expression of ~ in equations (57), (58), and (59), the following expressions for the norm~li7ecl constants R result:
R = ~ E[a,, I = J4m 1 (62) R2 = ~,2 E[a" ] = 3 1 2m  7 and (63) Rcma = ~,2 [¦ A I ] = 1 56m226 (64) Similar ~x~les~ions can be obtained for nonsquare constellations in a similar 5 fashion. When the number of points in the signal constellation becomes very large, the following asymptotic values for the norm~ d constants result:
m~ oo R ~ 1155 Rmma ~ 1.342 RCma ~ 1.673 . (65) S~mm~y of RCA CMA~ and MMA Al~orithms A general comparison of the RCA, CMA, and MMA techniques is shown in the 10 table of FIC~. 18. In addition, the table shown in FIG. 19 shows illustrative values, for signal constellations of different sizes, of the constants R, Rl, and R2, which are used in the tap updating algorithrns of the RCA, CMA, and MMA, blind equalization techniques described above. The data shown in FIG. 19 assumes that the symbols an and bn take the discrete values +1,+3,+5,+7, . The closedforrn expressions for these constants are 15 derived as described above.
Generally speaking, the RCA algorithm has less reliable convergence than either the CMA or MMA algorithms. As between the CMA and MMA algorithms, these algorithms have both benefits and drawbacks. For example, the CMA algorithm provides reliable convergence  thus avoiding incorrect diagonal solutions  but the CMA
20 algorithm requires an expensive rotator. In comparison, the MMA algorithm does not require an expensive rotator but is more susceptible than the CMA algorithm to incorrect convergence.
Transitio~ orithm~
In accordance with the inventive concept, we have realized a blind equalization technique 25 that uses both the CMA and MMA algorithms to achieve complementary results. This approach provides the basis for a "transition algorithm" that yields more reliable blind convergence without creating diagonal solutions, and avoids the expense of a rotator.
CMAMMA Transition Algorithm A first illustrative transition algorithm is called the "CMAMMA transition algorithm." An illustrative twofilter structure is shown in FIG. 21 and a corresponding flow chart is shown in FIG. 22. A blind startup procedure using a transition algorithm 5 can be scheduledriven, eventdriven, or both. With a scheduledriven approach, the switch between two different tap updating algorithms occurs after some fixed number, M, of iterations (which can be determined by a counter, for example). With an eventdriven approach, the switch occurs when a certain quality of eye opening is achieved. This can be done, for example, by continuously monitoring the MSE and making the switch when 10 the MSE is below some threshold T. FIG. 22 shows an exarnple of a startup procedure that uses both approaches. Values for M and T depend on the application and are determined ~x~elilllentally. During startup, the first blind equalization algorithm used is the CMA algorithm as shown in step 805. In this step, CMA elements 6201 and 6301, of FIG. 21, are used to converge the tap coefficients of adaptive filters 610 and 615, 15 respectively. As shown in step 810, the CMA algorithm is used until the number of iterations, n, is greater than M. Once this condition is reached, the transition algorithm switches to using the MMA algorithm in step 815. In this step, MMA elements 6202 and 6203, of FIG. 21 are used to converge the tap coefficients of adaptive filters 610 and 615, respectively. Finally, when the eye opens even more, e.g., to an MSE less than or 20 equal to T, the receiver switches to a decision directed tap adaptation algorithm as shown in steps 820 and 825. In this step, LMS elements 6203 and 6303 are used to further adapt the tap coefficients of adaptive filters 610 and 615, respectively. (Illustrative values of M and T, based on a computer simulation of a 64CAP constellation, areM=25,000 and T= 12dB.) Thus, in the CMAMMA transition algorithm, the receiver 25simply switches from CMA to MMA during blind startup.
From equation (29), the cost function for the CMA algorithm is given by:
CF=E[(Yn~Rc)2]~ (66) where L = 2.
The CMA tap updating algorithms, from equations (32a) and (32b) are:
30Cn+l= Cn~ a (Yn + yn2 _ R2 )Ynrn~ and (67a) dn+/ = dn ~ a (Yn + yn2 _ Rc ) Yn rn (67b) where a is the step size, and R2 = E[l A 12 ] .
From equation (36), the cost function for the MMA algorithm is:
CF = E [(y2 _ R2)2 +(y2 _ R2)2] (68) where R,2n = E[ ~]. The MMA tap updating algorithms, from equations (40a) and (40b) are:
cn+l = cn  a(yn2R,2n)ynrn, and (69a) dn+l = dn~ a(yn2R~n) Yn rn (69b) Unfortunately, computer simulations have indicated that a simple switch from theCMA algorithm to the MMA algorithm does not provide a smooth rotation of the constellation. As a result, additional examples of transition algorithms are described 1 0 below.
Constant Rotation CMA (CRCMA) Transition Al~orithm An analysis of the CMA algorithm shows that rotated constellations are generatedbecause only magnitude information of the complex symbols Yn is used during convergence. With CMA, the cost function minimi7es the errors as follows:
CF = 2 E[(l Ynl2R2)2] = E[(y2 + y2 _ R2~2~
This cost function minimi7~s the dispersion of the magnitude yn2 = yn2 + yn2 of the complex output samples of the equalizer around a circle, and does not take advantage of the phase information of Yn. In accordance with the inventive concept, the CMA
algorithrn is modified by adding a weighting factor ~ to the complex symbol Yn. In the 20 inphase dimension, yn2 becomes:
Yn + E~n Yn (71) and in the quadrature phase dimension, yn2 changes to:
~Yn + Yn Yn (72) Note that yn2 = yn2 + ~jn2, which has phase information, is not invariant to a 25 rotation. The proposed algorithm is called the Constant Rotation CMA (CRCMA) transition algorithm. The cost functions of CRCMA are:
CF E[( 2 + ~~2 _ R2)2] d (73a) CFQ = E[( ~3yn2 + yn2 _ Rr2 )2]. (73b) By using the weighting factor ~, the equalizer has convergence properties which are intermediate between CMA and MMA.
An illustrative twofilter structure is shown in FIG. 23 and a corresponding flow chart is shown in FIG. 24 for the CRCMA transition algorithm. During startup, the first blind equalization algorithm used is the CMA algorithm as shown in step 850. In this step, CMA elements 7201 and 7301, of FIG. 23, are used to converge the tap coefficients of adaptive filters 710 and 715, respectively. As shown in step 855, the CMA algorithm is used until the number of iterations, n, is greater than Ml. Once this condition is reached, the transition algorithm switches to using the CRCMA algorithm in 10 step 860 and n is reset to zero. In this step, CRCMA elements 7202 and 7302, of FIG.
23 are used to converge the tap coefficients of adaptive filters 710 and 715, respectively.
As the eye continues to open, the receiver switches to using the MMA algorithm when n is greater than M2. This is indicated by steps 865 and 870. When the MMA algorithm is in use, elements 7203 and 7303, of FIG. 23 are used to converge the tap coefficients of 15 adaptive filters 710 and 715, respectively. Finally, when the eye opens even more, e.g., to an MSE less than or equal to T2, the receiver switches to a decision directed tap adaptation algorithm as shown in steps 875 and 880. At this point, LMS elements 7204 and 7304 are used to further adapt the tap coefficients of adaptive filters 710 and 715, respectively. (Illustrative values of Ml, M2, and T2, based on a computer simulation of a 20 64CAP constellation, are Ml=25,000, M2=24,000 and T2 = 12dB).
As such, in this approach, the equalizer achieves blind convergence by using three blind algorithms, which are CMA, CRCMA and MMA. The value of ~ satisfies 0 < ~ < 1. Illustratively, ~ = 0.5. In this case, the effect of Yn on Yn is reduced by one half.
As noted above, the cost functions of CRCMA are:
CF, = E[(yn + ~n2 _ R,2)2] (74a) CFQ = E[( ~Yn + yn2 _ Rr2 )2] (74b) where ,~ is a scaling factor, which satisfies 0 c ,~ < 1. Constellation rotation is obtained with nonzero ~. When ~ = 1, the cost function in equation (74a) becomes the 30 same as that of CMA. When ,B = 0, the sum of the two cost functions in equations (74a) and (74b) result in the cost function of MMA. The gradients of the two cost functions are derived as follows:
Vc (CF, ) = E[(yn2 + ,Byn2Rr2)ynrn ] Vd (CFQ ) = E[(,By2 + yn2 _ Rr2)ynrn ] (75) The constant Rr is computed as:
E[¦Anl2 ] (76) = E[(an + ~bn ) ] (77) = (1 + ~ )E[an ] + 2~E [an ] (78) When ,B = 1, R = RC and when ~ = 0, R = Rm. The results are consistent with 5 those obtained for the cost functions in equations (74a) and (74b). The tap coefficients of the filter are updated in the opposite direction of the gradients. So that the tap updating algorithms are:
Cn+l = Cn~l(yn2 + ,~yn2 _ Rr2~nrn (79) dn+l = dn,u(~yn2 + yn2 _ Rr2)~nrn (80) Because of the use of the constant factor ~, a transition from CMA to MMA that keeps the eye open during the rotation cannot be done with CRCMA. Normally, MMAis required to achieve final blind convergence as described above.
Dynamic Rotation CMA (DRCMA) Transition Algorithm In the abovedescribed CRCMA transition algorithm, the constellation gradually 15 converges to its final position. However, the eye does not remain open during the rotation. Therefore, in order to keep the eye open during the rotation, a Dynamic Rotation CMA (DRCMA) Transition Algorithm is presented. In this approach, a timevarying ,B is used in equation (73). To reduce the mutual effects between the two dimensions, a declining error factor ¦er nl is introduced as a weighting factor into the cost 20 function of CMA. Thus, equation (73) becomes:
CF, = E[(yn2 + ~BIe~nlyn2(R2 + ~BIernl(Rc2R~2n))2] (81a) CFQ = E[(yn2 + ,BIernlyn2(R,2n + ~BIernl(Rc2R,2n))2] (81b) where RC refers to the constant of CMA and Rm refers to the constant of MMA.
Note that the timevarying factor has two functions. In the case of the inphase25 dimension, ern is used to reduce gradually the effect of Yn. In addition, it should be noted that the constants R are dirrelen~ for CMA and MMA. Normally, RC > Rm. So that ¦er nl is also used to make the change from RC to Rm dynamically. When the error ¦er nl goes to zero, the contribution of the two quantities Yn and (RC2 Rnn) converge to zero as well. Eventually, with a perfect convergence, the cost function of CMA converges to the one of MMA. Thus, a smooth transition from CMA to MMA that keeps the eye open can be done during filter adaptation without using a rotator. Note that ,B in equation (81) functions as a step size to control the scale of the timevarying weighting factor. A
S smooth equalizer rotation can be obtained when the step size ~ is properly chosen.
An illustrative twofilter structure is shown in FIG. 25 and a corresponding flow chart is shown in FIG. 26 for the DRCMA transition algorithm. During startup, the first blind equalization algoliLl~ used is the CMA algorithm as shown in step 950. In this step, CMA elements 9201 and 9301, of FIG. 25, are used to converge the tap coefficients of adaptive filters 910 and 915, respectively. As shown in step 955, the CMA algorithm is used until the number of iterations, n, is greater than M. Once this condition is reached, the transition algorithm switches to using the DRCMA algorithm in step 960. In this step, DRCMA elements 9202 and 9302, of FIG. 25 are used to converge the tap coefficients of adaptive filters 910 and 915, respectively. As the eye continues to open, the receiver switches to using the MMA algorithm when an MSE less than or equal to Tl is reached. This is indicated by steps 965 and 970. When the MMA
algorithm is in use, elements 9203 and 9303, of FIG. 25 are used to converge the tap coefficients of adaptive filters 910 and 915, respectively. Finally, when the eye opens even more, e.g., to an MSE less than or equal to T2' (where T2' < Tl'), the receiver switches to a decision directed tap adaptation algorithm as shown in steps 975 and 980.
At this point, LMS elements 9204 and 9304 are used to further adapt the tap coefficients of adaptive filters 910 and 915, respectively. Illustrative values of Ml, M2 and T2, based on a computer simulation of a 64CAP constellation, are M = 25,000, Tl = lOdB, and T2 = 12dB.
As with the CRCMA algorithm, the cost functions for DRCMA are written separately for the inphase and quadrature phase dimensions:
CF~ = E[(yn + ,BIernlyn2(R2 + ,BIernl(Rc2R2))2] (82) CFQ = E[(yn2 + ,BIernlyn(R2 + ~lernl(Rc2Rm))2] (83) In the case of the inphase dimension, when ¦,Ber nl = 1, the cost function CF~ is equivalent to the one used for CMA, and for ¦,Ber nl = ~~ the cost function CFI reduces to the one used for MMA. Note that the transition cost function between CMA to MMA is dynamic because of the factor ¦er nl ~ The gradients of the cost functions are derived as:
VC(CF) = E[(Yn + ~lernlYn2(Rm + ~BIernl(Rc2Rm))Ynrn] (84) Vd(CF) = E[(yn2 + ~lernlYn2(Rn2, + ~lernl(R2R"2,))y"rn] (85) The timevarying weighting factor is treated as a constant when deriving the algorithms for tap updating. Tap updating is done in the opposite direction of the gradients, so that:
cn+l = cn~l[Yn2 + ~lernlYn2(R,2n + ~lernl(R2R,2n)]ynrn (86) dn+l = dn~l[y 2 + ~ 1 er n I Yn2 _ ( R~2n + ~i l er n I ( RC2 _ R~2n )]Ynrn (87) With dynamic adaptation, in the case of inphase dimension, the tap updating starts in the CMA mode with ¦~er n ¦ = 1, and it is smoothly transferred to the MMA mode with gradually reduced ¦~ern¦. A complete transition from CMA to MMA can be l 0 obtained when ¦~er n ¦ = ~ and ¦~er.n ¦ = ~
FIGs. 27  28 show illustrative signal constellations for comparison purposes only. FIG. 27 illustrates a signal constellation obtained by using CRDMA. FIG. 28 illustrates a signal constellation obtained by using DRCMA. Obviously, a smoother rotation can be achieved with DRCMA than with CRCMA. It should be noted that experiments show that for some applications, complete blind equalization can be achieved by using DRCMA with a carefully chosen ~ so that the MMA algorithm need not be used, and the equalizer can directly switch to the LMS algorithm to achieve steadystate equalization thus bypassing steps 970 and 975 of FIG. 26. In comparison to CRCMA, DRCMA is more complicated to implement. However, better convergence performance can be achieved with DRCMA than with CRCMA in most cases.
Illustrative embodiments of the inventive concept are shown in FIGs. 11 and 12.
FIG. 11 illustrates an embodiment representative of a digital signal processor 400 that is programmed to implement an FSLE in accordance with the principles of the invention.
Digital signal processor 400 comprises a central processing unit (processor) 405 and memory 410. A portion of memory 410 is used to store program instructions that, when executed by processor 405, implement the combined CMAMMA type algorithms. This portion of memory is shown as 411. Another portion of memory, 412, is used to store tap coefficient values that are updated by processor 405 in accordance with the inventive concept. It is assumed that a received signal 404 is applied to processor 405, which equalizes this signal in accordance with the inventive concept to provide a output signal 406. For the purposes of example only, it is assumed that output signal 406 represents a sequence of output samples of an equalizer. (As known in the art, a digital signal processor may, additionally, further process received signal 404 before deriving output signal 406.) An illustrative software program is not described herein since, after learning Vd(CF) = E[(yn2 + ~lernlYn2(R,2n + ,BIernl(Rc2R,2n))ynrn] (85) The timevarying weighting factor is treated as a constant when deriving the algorithms for tap updating. Tap updating is done in the opposite direction of the gradients, so that:
cn+l = cn,u[yn2 + ~lernlYn(R,2n + ~lernl(Rc2R,2n)]ynrn (86) dn+l = dn _~l[y2 +~BIernlyn2(R~2n +~BIernl(Rc2R2)]ynrn (87) With dynamic adaptation, in the case of inphase dimension, the tap updating starts in the CMA mode with ¦,Be, nl = 1~ and it is smoothly transferred to the MMA mode with gradually reduced ¦,Bern¦. A complete transition from CMA to MMA can be obtained when ¦~er n l = O and ¦~er " ¦ = 0.
FIGs. 27  28 show illustrative signal constellations for comparison purposes only. FIG. 27 illustrates a signal constellation obtained by using CRDMA. FIG. 28 illustrates a signal constellation obtained by using DRCMA. Obviously, a smoother rotation can be achieved with DRCMA than with CRCMA. It should be noted that experiments show that for some applications, complete blind equalization can be achieved by using DRCMA with a carefully chosen ~ so that the MMA algorithm need not be used, and the equalizer can directly switch to the LMS algorithm to achieve steadystate equalization thus bypassing steps 970 and 975 of FIG. 26. In comparison to CRCMA, DRCMA is more complicated to implement. However, better convergence performance can be achieved with DRCMA than with CRCMA in most cases.
Illustrative embodiments of the inventive concept are shown in FIGs. 11 and 12.
FIG. 11 illustrates an embodiment representative of a digital signal processor 400 that is programmed to implement an FSLE in accordance with the principles of the invention.
Digital signal processor 400 comprises a central processing unit (processor) 405 and memory 410. A portion of memory 410is used to store program instructions that, when executed by processor 405, implement the combined CMAMMA type algorithms. This portion of memory is shown as 411. Another portion of memory, 412, is used to store tap coefficient values that are updated by processor 405 in accordance with the inventive concept. It is assumed that a received signal 404 is applied to processor 405, which equalizes this signal in accordance with the inventive concept to provide a output signal 406. For the purposes of example only, it is assumed that output signal 406 represents a sequence of output samples of an equalizer. (As known in the art, a digital signal processor may, additionally, further process received signal 404 before deriving output signal 406.) An illustrative software program is not described herein since, after learning of the combined CMAMMA type algorithms as described herein, such a program is within the capability of one skilled in the art. Also, it should be noted that any equalizer structures, such as those described earlier, can be implemented by digital signal processor 400 in accordance with the inventive concept.
FIG. 12 illustrates another alternative embodiment of the inventive concept.
Circuitry 500 comprises a central processing unit (processor) 505, and an equalizer 510.
The latter is illustratively assumed to be a phasesplitting FSLE as described above. It is assumed that equalizer 510 includes at least one tapcoefficient register for storing values for corresponding tap coefficient vectors (e.g., as shown in FIG. 3). Processor 505 10 includes memory, not shown, similar to memory 410 of FIG. 11 for implementing the combined CMAMMA type algorithms. Equalizer output signal 511, which represents a sequence of equalizer output samples, is applied to processor 505. The latter analyzes equalizer output signal 511, in accordance with the inventive concept, to adapt values of the tap coefficients in such a way as to converge to a correct solution.
The foregoing merely illustrates the principles of the invention and it will thus be appreciated that those skilled in the art will be able to devise numerous alternative arrangements which, although not explicitly described herein, embody the principles of the invention and are within its spirit and scope.
For example, although the invention is illustrated herein as being implemented 20 with discrete functional building blocks, e.g., an equalizer, etc., the functions of any one or more of those building blocks can be carried out using one or more appropfiate programmed processors.
In addition, although the inventive concept was described in the context of an FSLE, the inventive concept is applicable to other forms of adaptive filters, such as, but 25 not limited to, a decision feedback equalizer (DFE). The inventive concept is applicable to all forms of communications systems, e.g., broadcast networks, e.g., highdefinition television (HDTV), pointtomultipoint Networks like fiber to the curb (mentioned above), signal identification, or classification, applications like wiretapping, etc.
Claims (15)
using an adaptive filter structure for processing a received signal, the adaptive filter structure including a corresponding set of tap coefficient values; and using at least two different convergence algorithms for blindly converging the set of tap coefficient values.
using a constant modulus algorithm (CMA) to initially adapt the set of coefficient values of the adaptive filter; and using a multimodulus algorithm (MMA) to finally adapt the set of coefficient values of the adaptive filter.
using a constant modulus algorithm (CMA) to initially adapt the set of coefficient values of the adaptive filter;
changing to a Constant Rotation CMAMMA algorithm to continue to adapt the set of coefficient values of the adaptive filter; and changing to a multimodulus algorithm (MMA) to finally adapt the set of coefficient values of the adaptive filter.
using a constant modulus algorithm (CMA) to initially adapt the set of coefficient values of the adaptive filter;
changing to a Dynamic Rotation CMAMMA algorithm to continue to adapt the set of coefficient values of the adaptive filter; and changing to a multimodulus algorithm (MMA) to finally adapt the set of coefficient values of the adaptive filter.
the improvement comprising:
a processor that adapts coefficients of the equalizer by using at least two different blind equalization algorithms.
Priority Applications (2)
Application Number  Priority Date  Filing Date  Title 

US744,908  19961108  
US08744908 US5809074A (en)  19961108  19961108  Technique for improving the blind convergence of an adaptive equalizer using a transition algorithm 
Publications (1)
Publication Number  Publication Date 

CA2217093A1 true true CA2217093A1 (en)  19980508 
Family
ID=24994428
Family Applications (1)
Application Number  Title  Priority Date  Filing Date 

CA 2217093 Abandoned CA2217093A1 (en)  19961108  19970930  Technique for improving the blind convergence of an adaptive equalizer using a transition algorithm 
Country Status (4)
Country  Link 

US (1)  US5809074A (en) 
JP (1)  JPH10163928A (en) 
CA (1)  CA2217093A1 (en) 
EP (1)  EP0841786A3 (en) 
Families Citing this family (25)
Publication number  Priority date  Publication date  Assignee  Title 

US6178201B1 (en) *  19980311  20010123  Agilent Technologies Inc.  Controlling an adaptive equalizer in a demodulator 
US6314134B1 (en) *  19980424  20011106  Lucent Technologies Inc.  Blind equalization algorithm with joint use of the constant modulus algorithm and the multimodulus algorithm 
EP0963085A1 (en) *  19980604  19991208  Siemens Aktiengesellschaft  Method of setting adaptive filters in a QAM/CAP system 
US6259743B1 (en) *  19980702  20010710  Lucent Technologies Inc.  Automatic constellation phase recovery in blind startup of a dual mode CAPQAM receiver 
JP2000078060A (en) *  19980827  20000314  Mitsubishi Electric Corp  Adaptive equalizer device and its control method 
FR2785747B1 (en) *  19981109  20040213  Commissariat Energie Atomique  Parallel digital filter architecture and signal receiver spread spectrum using such a filter 
US6452987B1 (en)  19981125  20020917  Lucent Technologies Inc.  Fast startup in discrete multitone (DMT) based communications system 
US6418164B1 (en) *  19990114  20020709  Nxtwave Communications, Inc.  Adaptive equalizer with enhanced error quantization 
US6490318B1 (en)  19990624  20021203  Agere Systems Inc.  Phasecompensating constant modulus algorithm 
US6493381B1 (en) *  19990624  20021210  Agere Systems Inc.  Blind equalization algorithm with the joint use of the constant R and the sliced symbols 
US7143013B2 (en) *  19991104  20061128  Verticalband, Limited  Reliable symbols as a means of improving the performance of information transmission systems 
US7110923B2 (en) *  19991104  20060919  Verticalband, Limited  Fast, blind equalization techniques using reliable symbols 
US7085691B2 (en) *  19991104  20060801  Verticalband, Limited  Reliable symbols as a means of improving the performance of information transmission systems 
GB9930716D0 (en)  19991224  20000216  Nokia Networks Oy  A method and an arrangement relating to digital communication systems 
US7016405B2 (en) *  20020214  20060321  Mitsubishi Electric Research Laboratories, Inc.  Signal circulation for adaptive signal equalization in digital communication systems 
US7894513B2 (en) *  20020319  20110222  Thomson Licensing  Slicing algorithm for multilevel modulation equalizing schemes 
US7330068B2 (en) *  20020416  20080212  Bose Corporation  Adjusting coefficients of a filter 
US7492818B2 (en) *  20020417  20090217  Thomson Licensing  Equalizer mode switch 
US6904087B2 (en) *  20020917  20050607  Via Technologies, Inc.  Adaptive multimodulus algorithm method for blind equalization 
CN1176524C (en) *  20021016  20041117  威盛电子股份有限公司  Method and circuit for blind equalizer 
ES2221568B2 (en) *  20030526  20050716  Diseño De Sistemas En Silicio, S.A.  Method of reducing the variance of the estimate of the signal to noise ratio of a signal with phase differential modulation and coherent in amplitude. 
GB2408398B (en) *  20031118  20051123  Motorola Inc  Calibration device for a phased locked loop synthesiser 
US7773668B1 (en) *  20040121  20100810  Altera Corporation  Adaptive equalization methods and apparatus for programmable logic devices 
WO2005091583A1 (en) *  20040220  20050929  Nokia Corporation  Channel equalization 
US9559875B2 (en) *  20120509  20170131  Northrop Grumman Systems Corporation  Blind equalization in a single carrier wideband channel 
Family Cites Families (4)
Publication number  Priority date  Publication date  Assignee  Title 

FR2428946B1 (en) *  19780613  19801024  Ibm France  
US5263033A (en) *  19900622  19931116  At&T Bell Laboratories  Joint data and channel estimation using fast blind trellis search 
US5432794A (en) *  19910129  19950711  Canon Kabushiki Kaisha  Automatic Equalizer 
CA2073944C (en) *  19910726  20000919  Woo H. Paik  Carrier phase recovery for an adaptive equalizer 
Also Published As
Publication number  Publication date  Type 

JPH10163928A (en)  19980619  application 
EP0841786A2 (en)  19980513  application 
EP0841786A3 (en)  20010228  application 
US5809074A (en)  19980915  grant 
Similar Documents
Publication  Publication Date  Title 

Proakis  Adaptive equalization for TDMA digital mobile radio  
US5263033A (en)  Joint data and channel estimation using fast blind trellis search  
US5604769A (en)  Hybrid equalizer arrangement for use in data communications equipment  
US4550415A (en)  Fractionally spaced equalization using nyquistrate coefficient updating  
US5267266A (en)  Fast converging adaptive equalizer using pilot adaptive filters  
He et al.  A hybrid adaptive blind equalization algorithm for QAM signals in wireless communications  
US5844951A (en)  Method and apparatus for simultaneous beamforming and equalization  
US5353307A (en)  Automatic simulcast alignment  
US6337878B1 (en)  Adaptive equalizer with decision directed constant modulus algorithm  
US5311546A (en)  Carrier phase recovery for an adaptive equalizer  
US5661528A (en)  Apparatus and method for controlling operation of a high defination television adaptive equalizer  
US5692011A (en)  Dual decision equalization apparatus  
Yeh et al.  Adaptive minimum biterror rate equalization for binary signaling  
US6693958B1 (en)  Adaptive channel equalizer having a training mode  
US20020150155A1 (en)  Convergence speed, lowering the excess noise and power consumption of equalizers  
US5111484A (en)  Adaptive distortion canceller  
US20070133672A1 (en)  Apparatus and method for stable DEF using selective FBF  
US5446758A (en)  Device and method for precoding  
US5835532A (en)  Blind equalizer for a vestigial sideband signal  
US6011813A (en)  Blind equalization method and apparatus having reduced complexity  
Chen  Low complexity concurrent constant modulus algorithm and soft decision directed scheme for blind equalisation  
Ready et al.  Blind equalization based on radius directed adaptation  
US20040008765A1 (en)  Joint adaptive optimization of soft decision device and feedback equalizer  
US6426972B1 (en)  Reduced complexity equalizer for multi mode signaling  
US4412341A (en)  Interference cancellation method and apparatus 
Legal Events
Date  Code  Title  Description 

EEER  Examination request  
FZDE  Dead 