-5*4- 77/3 ;intellectual property office of n z. ;- 9 MAY 2008 ;received ;COMPLETE SPECIFICATION ;WIRELESS COMMUNICATION SYSTEM WITH IMPROVED ;CAPACITY ;I, ADRIAN DAVID BUSCH, a New Zealand citizen, of 22 Jean Batten Place, Christchurch, New Zealand on behalf of TAIT ELECTRONICS LIMITED, a New Zealand company, of 558 Wairakei Road, Burnside, Christchurch, New Zealand ;HEREBY declare this invention, for which I pray that a patent may be granted to me and the method by which it is to be performed, to be particularly described in and by the following statement: ;-1 - ;FIELD OF THE INVENTION ;The present invention relates to wireless communication systems and to a system and method for increasing the data transfer capacity of such systems. ;BACKGROUND TO THE INVENTION ;Many radio systems take the form of a central station and a number of remote stations, with the central station controlling access to the radio media. In centrally polled radio systems the central station polls the remote stations, querying them as to whether they have any data to be forwarded to the central station. Examples of this are IEEE 802.11 wireless local area networks using the point co-ordination function [1]. The central station polls each remote station, which responds with its data status. If the remote station has data to forward to the central station, the central station may instruct the remote station to forward the data. Otherwise title central station may poll another outstation in accordance with some polling policy. ;A figure of merit for a centrally polled radio system is the time it takes for the central station to poll a remote station and for this remote station to then respond with its status. If this poll-response cycle may be made briefer, the time it takes for the central station to poll the complete set of remote stations will be made shorter. Consequently the average time that data must wait at a remote station before the central station polls this remote station is reduced. Additionally, since data is subjected to a lower delay at the remote stations, it is transmitted earlier, in time that would otherwise be used for polling. As a result such a communication system spends a greater proportion of time transferring data and a lesser proportion doing poll-response cycles, thereby increasing the net data transfer capacity of the system. ;One factor affecting the duration of the poll-response cycle is the overhead imposed by physical framing on the transmission of poll and response messages. In time division duplex systems transmission bursts often commence with a synchronisation sequence to enable the receiving station to detect the start of the transmission frame, and to set its automatic gain control to an initial value. ;-2- ;intellectual property office of n.z. ;"9 MAY 2008 ;received ;This is often followed by a training sequence to enable the receiving station to produce a mathematical estimate of the distortion in the received signal, a process commonly known as estimating the channel. Digital radio receivers need accurate estimates of the distortion in the received signal to successfully decode received data. The remainder of the physical frame generally consists of a header and the data to be exchanged. An example of a system employing such physical framing is DEEE 802.11a [2]. The physical frame of IEEE 802.11a begins with a physical layer convergence procedure (PLCP) preamble, which consists of ten short symbols, equivalent to a synchronisation sequence, followed by two long symbols, equivalent to a training sequence. ;While the synchronisation and training sequences are being sent, data cannot be transmitted. Therefore a system that allows for the reduction or elimination of the synchronisation and training sequences is desirable, as it will reduce the duration of poll-response cycles and thereby increase the net data transfer capacity of the system. ;The problem of the overhead imposed by synchronisation and training sequences in physical framing applies to radio systems other than centrally polled radio systems. For instance IEEE 802.11 systems using the distributed coordination function to manage channel access [1] use the same frame format as those using the point co-ordination function described above. Another example is the Japanese Personal Digital Cellular (PDC) system [3], which uses time division multiple access for communication between base stations and mobile stations. In PDC traffic channel transmission frames, a combined synchronisation and training sequence is sent in the middle of the frame. In a similar way, Global System for Mobile Communications (GSM) cellular systems [4] use a combined synchronisation and training sequence in the middle of normal burst transmissions, whereas the synchronisation sequence is placed at the start of access burst transmissions. IEEE 802.16 [5] requires that each uplink burst transmission begin with an uplink preamble for synchronisation and training. In each of these systems synchronisation and training sequences occupy time that could otherwise be used for data transmission. Indeed the problem of the intellectual property office of n z ;"9 MAY 2008 ;RECEIV E D ;overhead of synchronisation and training sequences is common to some wireless LANs (e.g. IEEE 802.11), cellular systems (e.g. PDC and GSM) and wireless MANs (e.g. EEEE 802.16). The reduction or elimination of synchronisation and training sequences would increase the net data transfer capacity of each of these systems. ;SUMMARY OF THE INVENTION ;In many situations a radio station commencing a transmission burst will have recently transmitted a similar burst. If the previous transmission was sufficiently recent then the receiving stations may still have valid estimates of the distortion of the received signal induced by the channel. In this case the present invention requires that the transmitting station omits synchronisation and training sequences from the transmission and transmits a burst consisting only of header and data. Alternatively the quality of the estimates of the channel distortion may too low to allow a modulation scheme with a dense constellation to be sent (such as QAM-256) but it may still allow some data to be sent using a modulation scheme with a less-dense constellation (such as QAM-16). In this case the present invention requires that the transmitting station omits synchronisation and training sequences from the burst and commences transmission using a modulation scheme with a less-dense constellation. Receiving stations that receive this transmitted data can improve their estimates of the channel distortion during this time using the technique of blind equalisation. After some time the present invention requires that the transmitting station changes to a modulation scheme with a dense constellation for sending the balance of the transmission burst. The blind equalisation performed at the receiving stations will have improved their estimates of the channel distortion so they will be able to successfully receive and decode the data sent using the modulation scheme with a dense constellation. The data sent using the modulation scheme with the less-dense constellation is sent during the time normally used for synchronisation and training. This data is sent at a time when data would not normally be sent and ;-4- ;intellectual property office of n.z. ;"9 MAY 2008 ;RE C EIV E D ;therefore results in a net data transfer capacity improvement over systems that incur synchronisation and training sequence overheads on all transmission bursts. ;As time passes during the reception of a burst transmission the receiving stations use the technique of technique of blind equalisation to improve their estimates of the channel distortion up to a limit set by factors such as noise and finite numerical precision in signal processing. The present invention allows the transmitting station to transmit a burst using a number of modulation schemes with successively dense constellations. ;In some cases the amount of data to be sent in the transmission burst will be small enough that it may be sent in its entirety using a modulation scheme with a less-dense constellation. The present invention allows the transmitting station to end the transmission burst once all data has been sent even though the elapsed time from the start of transmission has not yet reached the time at which the transmitting station would normally change to a modulation scheme with a dense constellation. ;In some cases the elapsed time since the reception of the last transmission burst will be short enough that the receiver's estimate of channel distortion will be accurate enough that the transmitter may transmit with a denser constellation than would otherwise be the case. Shorter intervals of elapsed time from the reception of the last transmission burst allows the receiver to correctly receive and decode denser constellations. The present invention allows the transmitting station to commence a transmission burst using the most dense constellation possible with such transmissions subsequently employing a number of modulation schemes with successively dense constellations and concluding once all data has been sent. ;Some multiple-input multiple-output (MIMO) systems employ the technique of blind equalisation to improve the receiver's channel estimates as a transmission proceeds. The present invention allows for such MIMO systems to employ the techniques as described for single-transmitter and single-receiver systems. ;-5- ;intellectual property office of n.z. ;"9 MAY 2008 ;receiv e d ;BRIEF DESCRIPTION OF THE FIGURES ;Preferred forms of the invention will now be described with reference to the accompanying figures in which: ;FIG. 1 illustrates a typical arrangement of a centrally polled radio system; FIG. 2 illustrates a portion of a typical polling cycle; ;FIG. 3 illustrates the form of the transmission bursts commonly used in wireless communication systems; ;FIG. 4 illustrates the form of the new transmission bursts introduced with the present invention; ;FIG. 5 depicts a representative graph of the channel estimation error experienced by a receiving station; ;FIG. 6 illustrates an algorithm to be used by transmitting stations; FIG. 7 illustrates another algorithm to be used by transmitting stations; FIG. 8 illustrates an algorithm to be used by receiving stations to receive the transmissions of the type illustrated in FIG. 7; and ;FIG. 9 illustrates a central station and a remote station using adaptive multivariate decision feedback equalised multiple-input multiple-output transmission. ;DETAILED DESCRIPTION OF THE PREFERRED FORM ;FIG. 1 illustrates a typical arrangement of a centrally polled radio system. A central station 101 communicates with a number of remote stations 102,103 and 104. Transmissions by the central station 101 to the remote stations 102,103 and 104 are referred to as the downlink, represented by the transmission paths 105, 107 and 109. Transmissions by the remote stations 102, 103 and 104 to the central station 101 are referred to as the uplink, represented by the transmission paths 106,108 and 110. ;FIG. 2 illustrates a portion of a typical polling cycle, where a central station polls remote stations 102, 103 and 104. Transmissions by the central station are made on the downlink 201, and transmissions by remote stations are made on the uplink 202. The polling cycle starts with the central station sending intellectual property office of n.2. ;~6~ "9 MAY 2008 ;RECEIVED ;a transmission burst 203 on the downlink to poll remote station 102. The remote station 102 responds with a transmission burst 204 on the uplink to indicate that it has no data to forward to the central station. The central station then transmits burst 205 to poll remote station 103. The remote station 103 transmits burst 206 on the uplink to indicate that it has some data to forward to the central station. The central station then transmits burst 207 to demand that remote station 103 transmit its data. The remote station 103 then transmits its data 208 on the uplink. Having received the data from remote station 103 the central station then transmits burst 209 to poll remote station 104. The remote station 104 transmits burst 210 to indicate that it has no data to forward. ;FIG. 3 illustrates the form of the transmission bursts commonly used in wireless communications in systems such as the centrally polled radio system described in FIG. 1. Transmission burst 301 is comprised of a synchronisation sequence 302, a training sequence 303 and a payload of header and data 304. ;Stations receiving the transmission use the synchronisation sequence 302 to detect the start of the transmission burst. Stations receiving the transmission use the training sequence 303 to estimate the distortion of the received signal induced by the channel. The receiver's estimate the distortion of the received signal induced by the transmission channel is commonly called the channel estimate and is used by the receiver to compensate for the distorting effects of the channel. The payload 304 of the burst consists of the header and data that make up the message that the transmitting station wishes to forward to receiving stations. ;FIG. 4 illustrates the form of the new transmission bursts introduced with the present invention. Transmission burst 401 omits synchronisation and training sequences, and sends only the header and data message. The first part 402 of the header and data message is sent using a low constellation density modulation, ;such as QAM-16. The second part 403 of the header and data message is sent using a high constellation density modulation, such as QAM-256. According to the present invention transmission burst 401 is to be sent in situations where receiving stations have a sufficiently good channel estimate to receive data ;"7" 1 -9 MAY 2008 ;fteceivfni ;transmitted using the low constellation density modulation but a channel estimate that is inadequate for receiving data transmitted using the high constellation density modulation. The length of the part of the transmission burst that is sent using the low constellation density modulation 402 is such that receiving stations employing blind equalisation to improve their channel estimate over this period will have improved their estimates such that the part of the transmission burst that is sent using the high constellation density modulation 403 can be successfully decoded. ;Transmission burst 404 omits synchronisation and training sequences, and carries only the header and data message. It is comprised of a first part 405 sent using a low constellation density modulation, a second part 406 sent using a medium constellation density modulation and a second part 407 sent using a high constellation density modulation. The length of the first part 405 is such that receiving stations employing blind equalisation to improve their channel estimate over this period will have improved their estimates such that the second part 406 sent using the medium constellation density modulation can be successfully decoded. Similarly the length of the second part 406 is such that receiving stations employing blind equalisation to improve their channel estimate over this period will have further improved their estimates such that the third part 407 sent using the high constellation density modulation can be successfully decoded. ;Transmission burst 408 omits synchronisation and training sequences, and carries only the header and data message. It is comprised of a first part 409 sent using a medium constellation density modulation and a second part 410 sent using a high constellation density modulation. According to the present invention transmission burst 408 is to be sent in situations where receiving stations have a sufficiently good channel estimate to receive data transmitted using the medium constellation density modulation but a channel estimate that is inadequate for receiving data transmitted using the high constellation density modulation. ;FIG. 5 depicts a representative graph of the channel estimation error experienced by a receiving station adapted to receive the transmission bursts illustrated in FIG. 4 according to the present invention. The graph trace 501 ;" 8 " intellectual property office of n.z. ;- 9 MAY 2008 ;received ;illustrates how the error in the receiver's channel estimate 502 varies with time expressed in received symbols 503. If the graph trace is lower than the QAM-16 threshold 504 then the receiver can correctly decode data sent using QAM-16 modulation. If the graph trace is lower than the QAM-256 threshold 505 then the receiver can correctly decode data sent using QAM-256 modulation. ;During the periods A 506 and B 507 the receiving station receives a common transmission burst 301. The synchronisation 302 and training sequences 303 are received by the receiving station during period A 506. It is the job of the training sequence 303 to allow the calculation of the channel estimate at the receiving station. It can be seen that by the end of the period A 506 the graph trace 501 is lower than the QAM-256 threshold 505 indicating that the channel estimate is sufficient for the receiver to correctly decode data sent using QAM-256 modulation. During period B 507 the receiving station receives the header and data payload 304 of the transmission burst. During this period the graph trace 501 indicates that with the operation of blind equalisation the channel estimate first improves then settles around a limit set by factors such as noise and finite numerical precision in signal processing. No transmission is made during period C 508. During this period the channel estimate of the receiving station steadily deteriorates due to factors such as clock drift and a changing radio propagation environment. At the end of period C 508 the graph trace 501 indicates that the receiving station's channel estimate is no longer adequate to correctly decode data sent using QAM-256 modulation but it is sufficient for the receiver to correctly decode data sent using QAM-16 modulation. The transmitting station transmits a transmission burst of the form 401 introduced with the present invention in periods D 509 and E 510. The first part of the transmission burst 402 is transmitted using QAM-16 in period D 509. During period D 509 the receiving station uses blind equalisation to improve its channel estimate so that by the end of the period the channel estimate has improved sufficiently for the receiver to correctly decode data sent using QAM-256 modulation. The transmitting station then transmits the remainder of burst 403 using QAM-256 in period E 510. During this period the graph trace 501 indicates inteuectualproperty office of N.Z. ;"9 MAY 2008 ;I I# r* p\
that with the operation of blind equalisation the channel estimate remains adequate for the receiver to correctly decode the part of the transmission burst 403 sent using QAM-256 modulation. No transmission is made during period F 511, and the channel estimate of the receiving station steadily deteriorates as it did in period C 508.
FIG. 6 illustrates an algorithm to be used by transmitting stations for transmitting bursts in the manner described above. Upon commencing a transmission 601, a station determines the time elapsed since its last transmission 602. If the time since the transmitting station's last transmission exceeds some threshold value ti then the station transmits a standard transmission burst 603 in the form of burst 301. If the time since the transmitting station's last transmission does not exceed the threshold ti then the station transmits ni bytes of data 605 using low constellation density modulation in the form of burst 402. Following this, if the transmitting station has further data to send 606 then it transmits the remaining data 607 using high constellation density modulation in the form of burst 403. Otherwise the transmission burst ends 604.
The present invention allows for further improvement of net data transfer capacity by recognising that some time receiving that part of a burst transmitted at a low constellation density modulation the operation of blind equalisation improves the channel estimate to the point where the receiving station can successfully decode medium constellation density modulation. The time elapsed from the start of the transmission burst at which medium constellation density modulation can be successfully decoded will be less than the time elapsed from the start of the transmission burst at which high constellation density modulation can be successfully decoded. The present invention allows for the transmitting station to transmit data using medium constellation density modulation between these elapsed times. This is demonstrated in transmission burst 404, in which the first part 405 and the second part 406 are sent during the period at which the data sent solely low constellation density modulation would otherwise be sent 402 in transmission burst 401. The further improvement of net data transfer capacity comes about during the time at which data is sent using medium constellation
,NTEnISInAKf-PR0PERTY office of n.z.
"9 MAY 2008 received
density modulation when data would otherwise be sent using low constellation density modulation.
The present invention also allows for still further improvement of net data transfer capacity by allowing transmitting stations to commence transmission of data using medium constellation density modulation if the time elapsed since the completion of the most recent transmission is such that the deterioration in the receiving stations' channel estimates is low enough to allow data sent using this modulation to be correctly decoded by the receiving station. To achieve this the transmitting station checks whether the time elapsed since the completion of the most recent transmission is less than some threshold value t2, where t2<ti. If so, the transmitting station transmits data using transmission burst 408, which is similar to transmission burst 404 except that it omits the first part transmitted using low constellation density modulation. The further improvement of net data transfer capacity comes about at the start of the transmission burst when data is sent using medium constellation density modulation during the time at which data would otherwise be sent using low constellation density modulation.
Additionally the present invention also allows for further still improvement of net data transfer capacity by allowing transmitting stations to commence transmission of data using high constellation density modulation if the time elapsed since the completion of the most recent transmission is such that the deterioration in the receiving stations' channel estimates is low enough to allow data sent using this modulation to be correctly decoded by the receiving station. To achieve this the transmitting station checks whether the time elapsed since the completion of the most recent transmission is less than some threshold value t3, where t3<t2. If so, the transmitting station transmits data using transmission burst 411, which consists solely of data transmitted using high constellation density modulation. The further improvement of net data transfer capacity comes about as all data is sent using high constellation density modulation which transmits data at a higher rate than the alternative modulation types.
intellectual property office of n.z.
-9 MAY 2008
RECEIVED
FIG. 7 illustrates an algorithm to be used by transmitting stations employing these further adaptations to improve net data transfer capacity. Upon starting transmission 701, the transmitting station checks the time elapsed since its last transmission against thresholds ti 702, t2 705 and t3 711 where ti>t2 and t2>t3- If the time elapsed since its last transmission is greater than ti then the transmitting station commences transmission 703 in the form of burst 301. Otherwise, if the time elapsed since its greater transmission is less than t2 then the transmitting station commences transmission 706 in the form of burst 404. Otherwise, if the time elapsed since its last transmission is greater than t3 then the transmitting station commences transmission 712 in the form of burst 408. Otherwise the transmitting station commences transmission 715 in the form of burst 411.
The transmitting station then proceeds with the transmission. If the transmitting station transmits a standard transmission burst 703 then this burst will contain the entirety of the header and data to be sent. Otherwise, if the transmitting station transmits a transmission burst 715 then this burst will contain the entirety of the header and data to be sent. Otherwise, if the transmitting station transmits a transmission burst 712 then this burst will contain n2 bytes of the header and data to be sent, with any remaining header and data being sent using high constellation density modulation 714. Otherwise, if the transmitting station transmits a transmission burst 706 then this burst will contain ni bytes of the header and data to be sent, with any remaining header and data being sent with the next n2 bytes being transmitted using medium constellation density modulation 708 and any further remaining header and data transmitted using high constellation density modulation 710.
FIG. 8 illustrates an algorithm to be used by receiving stations to receive the transmissions of the type illustrated in FIG. 7. The receiving station first calculates the time elapsed since the end of the last transmission of the transmitting station. Having calculated the time elapsed since the last transmission the receiving station compares this time to thresholds ti 802, t2 805 and t3 811 where ti>t2 and t2>t3. In order to do this the receiving station needs to intellectual property 12 . office of n.z.
- 9 MAY 2008 received
know both the time at which the transmission commences and the identity of the transmitting station in cases where a system comprises a number of transmitting stations. The dependence of the modulation solely on elapsed time distinguishes the present invention from those of Acampora [6] and Webb [7]. Acampora's system uses a feedback channel from the receiving station to the transmitting station for controlling the constellation density of the transmitting station's modulation. In Webb's system the transmitting station uses the first part of the transmission burst to announce the constellation density of the modulation that it will use to transmit the remainder of the transmission burst. Hence in Acampora's system the receiving station controls the transmitting station's modulation whereas in Webb's system the transmitting station controls the transmitting station's modulation. In the present invention neither the transmitting nor receiving station controls the type of modulation used in the transmission. Instead type of modulation used in the transmission is simply a function of time elapsed since the previous transmission by the transmitting station.
Having determined the type of transmission by examining the elapsed time since the previous transmission, the receiving station then proceeds to receive and decode the transmission. If the receiving station commences receiving a standard transmission burst 803 then this burst will contain the entirety of the header and data to be received. Otherwise, if the receiving commences receiving a transmission burst 815 then this burst will contain the entirety of the header and data to be sent. Otherwise, if the receiving station commences receiving a transmission burst 812 then this burst will contain n2 bytes of the header and data to be sent, with any remaining header and data being encoded in the remainder of the burst using high constellation density modulation 814. Otherwise, if the receiving station commences receiving a transmission burst 806 then this burst will contain ni bytes of the header and data, with any remaining header and data being found with the next n2 bytes in the form of medium constellation density modulation 808 and any further remaining header and data in the form of high constellation density modulation 810.
intellectual property office of n.2.
13" -9 MAY 2008
RECEIVED
The present invention also allows for the application of these algorithms to systems using multiple-input multiple-output (MIMO) transmission. Systems that use MIMO transmission have multiple antennas at sending and receiving stations. An example of a MIMO system is the adaptive multivariate decision feedback equalised (AMV-DFE) system described in [8]. The AMV-DFE MIMO system employs four antennas at both the transmitting and receiving stations. The transmitting station transmits using 7t/4-differential quadrature phase shift keying (tc/4-DQPSK) on each transmitting antenna. This system uses a training sequence in the similar manner as the systems using single transmit and receive antennas discussed above to estimate the channel. Similar systems have been constructed using the same principles of AMV-DFE MIMO transmission that use higher constellation density modulation such as QAM-16 or QAM-256.
FIG. 9 illustrates a central station 901 and a single remote station 906 in a centrally polled radio system using 4x4 AMV-DFE MIMO transmission. For clarity only the uplink is shown; the downlink operates in a comparable manner. The centrally polled radio system may include other remote stations but for clarity these are omitted from FIG. 9. The remote station 906 transmits on the uplink using antennas 907, 908, 909 and 910. The central station 901 receives these transmissions using antennas 902, 903, 904 and 905. More specifically, the central station 901 antenna 902 receives the transmission from antenna 907 through path 911, the transmission from antenna 908 through path 915, the transmission from antenna 909 through path 919, and the transmission from antenna 910 through path 923. Similarly central station 901 antenna 903 receives a composite signal from antennas 907, 908, 909 and 910 through paths 912, 916,
920 and 924 respectively. Similarly central station 901 antenna 904 receives a composite signal from antennas 907, 908, 909 and 910 through paths 913, 917,
921 and 925 respectively. In a similar manner central station 901 antenna 905 receives a composite signal from antennas 907, 908, 909 and 910 through paths 914, 918, 922 and 926 respectively. In total this system has a total of sixteen paths between the transmitting station's antennas and the receiving station's antennas. To correctly receive and decode the transmission the receiving station
intellectual property office of n.z.
- 9 MAY 2008
RECEIVED
needs channel estimates for each of these sixteen paths. As the transmission proceeds the receiving station's channel estimates for each of these sixteen paths in this system improves through the operation of blind equalisation. This allows MIMO systems to employ the techniques as described above and as illustrated in FIG. 6, FIG. 7 and FIG. 8. In particular the transmission bursts of FIG. 4 and FIG. 5 can be used in MIMO systems where it is understood that in such systems the transmitting station encodes the burst such that a portion of the burst is transmitted by each transmit antenna and the receiving station operates to reconstruct the burst by appropriate processing of the signals received by its multiple antennas. In the case of AMV-DFE MIMO transmission these processes of decoding and reconstruction are described in [8].
The foregoing describes the invention including preferred forms thereof. Alterations and modifications as will be obvious to those skilled in the art are intended to be incorporated within the scope hereof.
REFERENCES
[1] B. O'Hara and A. Petrick, The 802.11 Handbook: A Designer's Companion, 2nd. ed. New York: IEEE Press, 2005.
[2] IEEE-SA Standards Board, IEEE Std 802.1la-1999(R2003): Wireless LAN Medium Access Control (MAC) and Physical Layer (PHY) specifications. Piscataway, NJ: IEEE, 1999.
[3] K. Kinoshita, M. Kuramoto, and M. Nakajima, "Development of a TDM A digital cellular system based on Japanese standard," presented at 41st IEEE Veh. Technol. Conf., 1991.
[4] 3GPP, ETSI TS 100 573 V8.9.0 Digital cellular telecommunications system (phase 2+); Physical layer on the radio path (general description), Sophia Antipolis Cedex, France: ETSI, 2004.
[5] IEEE-SA Standards Board, IEEE Standard for local and metropolitan area networks, Part 16: Air interface for fixed broadband wireless access systems. New York: IEEE, 2004.
intellectual property office of n.z.
-9 MAY 2008
RECEIVED
[7]
[8]
intellectual property office of n.z.
- 9 MAY 2008
RECEIVED
A. Acampora, "Transmitter and receivers using resource sharing and coding for increased capacity." United States Patent 4,495,619: AT&T Bell Laboratories, 1984.
W. T. Webb, "QAM system in which the constellation is modified in accordance with channel quality." United States Patent 5,828,695: British Telecommunications, 1998.
S. H. Kuo, J. Dowle, and I. V. McLoughlin, "A reconfigurable platform for MIMO research realtime implementation of 4x4 adaptive multivariate DFE," in Proceedings of Virginia Tech's 14th Symposium on Wireless Personal Communications, Blacksburg, Va, USA, June 2004.