Communication system employing transmit macrodiversity

0Associated
Cases 
0Associated
Defendants 
0Accused
Products 
99Forward
Citations 
0
Petitions 
1
Assignment
First Claim
1. A method of employing coherent transmit diversity in a wireless communication network, the method comprising:
 forming a plurality of transmit signals, each comprising a combination of information signals intended for different ones of a plurality of wireless receivers, such that, at each one of said wireless receivers, the intended information signals in the plurality of transmit signals add while the other information signals cancel; and
coherently transmitting said transmit signals from said plurality of transmitters over different propagation channels to said plurality of receivers.
1 Assignment
0 Petitions
Accused Products
Abstract
Coherent transmit macrodiversity involves transmitting simultaneous signals from one or more transmitters to one or more wireless receivers. Transmit signals are formed as weighted combinations of the individual information signals for each of the wireless receivers. By properly weighting these receiverspecific signals, the combined signals combine at each receiver to cancel all but the desired information signal for that receiver. This improves reception performance and allows multiple receivers to operate on the same communication channel in the same area if desired. The weighting factors used to form the combined signals are based on propagation path characteristics between each of the transmit antennas and each of the receivers. This information is generally referred organized as a channel estimate matrix and is used to filter the information signals such that they may be combined in the required weighted combinations.
132 Citations
View as Search Results
Advanced signal processors for Interference Cancellation in baseband receivers  
Patent #
US 20110019656A1
Filed 09/28/2010

Current Assignee
III Holdings 1 LLC

Sponsoring Entity
Rambus Inc.

Advanced Signal Processors for Interference Suppression in Baseband Receivers  
Patent #
US 20110069796A1
Filed 12/01/2010

Current Assignee
III Holdings 1 LLC

Sponsoring Entity
Rambus Inc.

Scheduling multiuser transmission in the downlink of a multiantenna wireless communication system  
Patent #
US 7,907,911 B2
Filed 08/16/2005

Current Assignee
Provenance Asset Group LLC

Sponsoring Entity
AlcatelLucent USA Inc.

CODE DIVISION MULTIPLE ACCESS TRANSMISSION ANTENNA WEIGHTING  
Patent #
US 20110026496A1
Filed 10/11/2010

Current Assignee
Interdigital Technology Corporation

Sponsoring Entity
Interdigital Technology Corporation

ENABLING PHASE TRACKING FOR A COMMUNICATION DEVICE  
Patent #
US 20110134899A1
Filed 12/06/2010

Current Assignee
Qualcomm Inc.

Sponsoring Entity
Qualcomm Inc.

Interference Suppression for CDMA Systems  
Patent #
US 20110080923A1
Filed 12/13/2010

Current Assignee
III Holdings 1 LLC

Sponsoring Entity
Rambus Inc.

Interference cancellation in a spread spectrum communication system  
Patent #
US 7,953,139 B2
Filed 06/08/2009

Current Assignee
Interdigital Technology Corporation

Sponsoring Entity
Interdigital Technology Corporation

MULTISITE MIMO COOPERATION IN CELLULAR NETWORKS  
Patent #
US 20110194540A1
Filed 09/21/2010

Current Assignee
Mohammad Hadi Baligh, Hua Xu, Jianglei Ma

Sponsoring Entity
Mohammad Hadi Baligh, Hua Xu, Jianglei Ma

INTERFERENCE CANCELLATION IN A SPREAD SPECTRUM COMMUNICATION SYSTEM  
Patent #
US 20110228710A1
Filed 05/27/2011

Current Assignee
Interdigital Technology Corporation

Sponsoring Entity
Interdigital Technology Corporation

Advanced Signal Processors for Interference Cancellation in Baseband Receivers  
Patent #
US 20100329402A1
Filed 08/30/2010

Current Assignee
III Holdings 1 LLC

Sponsoring Entity
III Holdings 1 LLC

RESOURCE ALLOCATION IN A COMMUNICATION SYSTEM  
Patent #
US 20100177670A1
Filed 01/13/2010

Current Assignee
Nokia Corporation

Sponsoring Entity
Nokia Corporation

Antenna array communication using spreading codes  
Patent #
US 7,813,413 B2
Filed 12/22/2005

Current Assignee
Interdigital Technology Corporation

Sponsoring Entity
Interdigital Technology Corporation

Code division multiple access transmission antenna weighting  
Patent #
US 7,684,469 B2
Filed 12/06/2006

Current Assignee
Interdigital Technology Corporation

Sponsoring Entity
Interdigital Technology Corporation

Methods and systems for tracking signals with diverse polarization properties  
Patent #
US 7,518,551 B2
Filed 06/29/2007

Current Assignee
Aerospace Corporation

Sponsoring Entity
Aerospace Corporation

METHOD AND SYSTEM FOR CODE REUSE AND CAPACITY ENHANCEMENT USING NULL STEERING  
Patent #
US 20090135889A1
Filed 12/05/2008

Current Assignee
InterDigital Patent Holdings Inc.

Sponsoring Entity
InterDigital Patent Holdings Inc.

Methods and systems for tracking signals with diverse polarization properties  
Patent #
US 7,551,134 B2
Filed 06/29/2007

Current Assignee
Aerospace Corporation

Sponsoring Entity
Aerospace Corporation

ROWVECTOR NORM COMPARISON METHOD AND ROWVECTOR NORM COMPARISON APPARATUS FOR INVERSE MATRIX  
Patent #
US 20090157787A1
Filed 06/30/2008

Current Assignee
Samsung Electronics Co. Ltd., Electronics and Telecommunications Research Institute

Sponsoring Entity
Samsung Electronics Co. Ltd., Electronics and Telecommunications Research Institute

Interference cancellation in a spread spectrum communication system  
Patent #
US 7,545,846 B2
Filed 12/13/2005

Current Assignee
Interdigital Technology Corporation

Sponsoring Entity
Interdigital Technology Corporation

METHOD AND APPARATUS FOR TRANSFERRING SIGNAL DATA  
Patent #
US 20090177295A1
Filed 12/18/2008

Current Assignee
ABB Schweiz AG

Sponsoring Entity
Abb Oy

Interference matrix construction  
Patent #
US 7,577,186 B2
Filed 09/07/2004

Current Assignee
III Holdings 1 LLC

Sponsoring Entity
TensorComm Inc.

INTERFERENCE CANCELLATION IN A SPREAD SPECTRUM COMMUNICATION SYSTEM  
Patent #
US 20090257472A1
Filed 06/08/2009

Current Assignee
Interdigital Technology Corporation

Sponsoring Entity
Interdigital Technology Corporation

Methods and Systems for Tracking Signals with Diverse Polarization Properties  
Patent #
US 20080001820A1
Filed 06/29/2007

Current Assignee
Aerospace Corporation

Sponsoring Entity
Aerospace Corporation

Methods and Systems for Tracking Signals with Diverse Polarization Properties  
Patent #
US 20080012763A1
Filed 06/29/2007

Current Assignee
Aerospace Corporation

Sponsoring Entity
Aerospace Corporation

Generalized mrank beamformers for MIMO systems using successive quantization  
Patent #
US 7,336,727 B2
Filed 08/19/2004

Current Assignee
Sisvel International S.A.

Sponsoring Entity
Nokia Corporation

METHOD AND SYSTEM FOR DATA RETRIEVAL USING PUSH TO TALK  
Patent #
US 20080085688A1
Filed 10/06/2006

Current Assignee
Motorola Inc.

Sponsoring Entity
Motorola Inc.

Adaptive channel estimation by means of variation of the integration length during the despreading of spreadcoded training symbol sequences  
Patent #
US 7,369,602 B2
Filed 08/12/2004

Current Assignee
Intel Deutschland GmbH

Sponsoring Entity
Infineon Technologies AG

Telescoping window based equalization  
Patent #
US 7,400,692 B2
Filed 12/20/2004

Current Assignee
Interdigital Technology Corporation

Sponsoring Entity
Interdigital Technology Corporation

Joint channel equalizer interference canceller advanced receiver  
Patent #
US 7,437,135 B2
Filed 10/29/2004

Current Assignee
Interdigital Technology Corporation

Sponsoring Entity
Interdigital Technology Corporation

MultiCarrier CQI Feedback Method and Apparatus  
Patent #
US 20080253336A1
Filed 11/30/2007

Current Assignee
Telefonaktiebolaget LM Ericsson

Sponsoring Entity
Telefonaktiebolaget LM Ericsson

Distributed Antenna Wlan AccessPoint System and Method  
Patent #
US 20080267142A1
Filed 06/19/2005

Current Assignee
Stellaris Ltd.

Sponsoring Entity
Stellaris Ltd.

MIMO WLAN SYSTEM  
Patent #
US 20080285669A1
Filed 04/30/2008

Current Assignee
Qualcomm Inc.

Sponsoring Entity
Qualcomm Inc.

Scheduling multiuser transmission in the downlink of a multiantenna wireless communication system  
Patent #
US 20070042717A1
Filed 08/16/2005

Current Assignee
Provenance Asset Group LLC

Sponsoring Entity
AlcatelLucent USA Inc.

CODE DIVISION MULTIPLE ACCESS TRANSMISSION ANTENNA WEIGHTING  
Patent #
US 20070091985A1
Filed 12/06/2006

Current Assignee
Interdigital Technology Corporation

Sponsoring Entity
Interdigital Technology Corporation

Construction of an interference matrix for a coded signal processing engine  
Patent #
US 7,200,183 B2
Filed 11/15/2002

Current Assignee
III Holdings 1 LLC

Sponsoring Entity
TensorComm Inc.

Ubiquitous coverage OFDM hybrid system  
Patent #
US 20070121743A1
Filed 11/22/2006

Current Assignee
Go Net Systems Ltd.

Sponsoring Entity
Go Net Systems Ltd.

Frequency domain training of prefilters for receivers  
Patent #
US 7,248,849 B1
Filed 06/03/2003

Current Assignee
Radia Communications Inc.

Sponsoring Entity
Radia Communications Inc.

Methods and systems for tracking signals with diverse polarization properties  
Patent #
US 7,239,275 B2
Filed 03/22/2004

Current Assignee
Aerospace Corporation

Sponsoring Entity
Aerospace Corporation

Interference cancellation in a spread spectrum communication system  
Patent #
US 6,985,515 B2
Filed 08/23/2004

Current Assignee
Interdigital Technology Corporation

Sponsoring Entity
Interdigital Technology Corporation

Base station for use in a CDMA communication system using an antenna array  
Patent #
US 6,983,008 B2
Filed 02/06/2002

Current Assignee
Interdigital Technology Corporation

Sponsoring Entity
Interdigital Technology Corporation

Generalized mrank beamformers for mimo systems using successive quantization  
Patent #
US 20060039493A1
Filed 08/19/2004

Current Assignee
Sisvel International S.A.

Sponsoring Entity
Nokia Corporation

Antenna array communication using spreading codes  
Patent #
US 20060098720A1
Filed 12/22/2005

Current Assignee
Interdigital Technology Corporation

Sponsoring Entity
Interdigital Technology Corporation

Interference cancellation in a spread spectrum communication system  
Patent #
US 20060093020A1
Filed 12/13/2005

Current Assignee
Interdigital Technology Corporation

Sponsoring Entity
Interdigital Technology Corporation

Power control for partial channelstate information (CSI) multipleinput, multipleoutput (MIMO) systems  
Patent #
US 7,076,263 B2
Filed 02/19/2002

Current Assignee
Qualcomm Inc.

Sponsoring Entity
Qualcomm Inc.

Adaptive channel estimation by means of variation of the integration length during the despreading of spreadcoded training symbol sequences  
Patent #
US 20050036539A1
Filed 08/12/2004

Current Assignee
Intel Deutschland GmbH

Sponsoring Entity
Infineon Technologies AG

Interference matrix construction  
Patent #
US 20050031060A1
Filed 09/07/2004

Current Assignee
III Holdings 1 LLC

Sponsoring Entity
TensorComm Inc.

Interference cancellation in a spread spectrum communication system  
Patent #
US 20050025224A1
Filed 08/23/2004

Current Assignee
Interdigital Technology Corporation

Sponsoring Entity
Interdigital Technology Corporation

Forward link handoff for wireless communication systems with OFDM forward link and CDMA reverse link  
Patent #
US 20050083888A1
Filed 12/22/2003

Current Assignee
Qualcomm Inc.

Sponsoring Entity
Qualcomm Inc.

Methods and systems for tracking signals with diverse polarization properties  
Patent #
US 20050227659A1
Filed 03/22/2004

Current Assignee
Aerospace Corporation

Sponsoring Entity
Aerospace Corporation

Construction of an interference matrix for a coded signal processing engine  
Patent #
US 20040052305A1
Filed 11/15/2002

Current Assignee
III Holdings 1 LLC

Sponsoring Entity
TensorComm Inc.

Data detection and demodulation for wireless communication systems  
Patent #
US 20040137863A1
Filed 10/23/2003

Current Assignee
Qualcomm Inc.

Sponsoring Entity
Qualcomm Inc.

Power control for partial channelstate information (CSI) multipleinput, multipleoutput (MIMO) systems  
Patent #
US 20030157954A1
Filed 02/19/2002

Current Assignee
Qualcomm Inc.

Sponsoring Entity
Qualcomm Inc.

Method for connecting a terminal over an access network to the core part of a radio communication network and corresponding gateway  
Patent #
US 20030227940A1
Filed 05/28/2003

Current Assignee
Evolium SAS

Sponsoring Entity
Evolium SAS

Base station for use in a CDMA communication system using an antenna array  
Patent #
US 20020094016A1
Filed 02/06/2002

Current Assignee
Interdigital Technology Corporation

Sponsoring Entity
Interdigital Technology Corporation

Method and system for code reuse and capacity enhancement using null steering  
Patent #
US 8,149,895 B2
Filed 12/05/2008

Current Assignee
Interdigital Technology Corporation

Sponsoring Entity
Interdigital Technology Corporation

Data detection and demodulation for wireless communication systems  
Patent #
US 8,145,179 B2
Filed 10/31/2007

Current Assignee
Qualcomm Inc.

Sponsoring Entity
Qualcomm Inc.

Data detection and demodulation for wireless communication systems  
Patent #
US 8,170,513 B2
Filed 10/23/2003

Current Assignee
Qualcomm Inc.

Sponsoring Entity
Qualcomm Inc.

Closedloop rate control for a multichannel communication system  
Patent #
US 8,218,609 B2
Filed 05/31/2003

Current Assignee
Qualcomm Inc.

Sponsoring Entity
Qualcomm Inc.

Method and apparatus for transferring signal data  
Patent #
US 8,253,354 B2
Filed 12/18/2008

Current Assignee
ABB Schweiz AG

Sponsoring Entity
Abb Oy

MIMO WLAN system  
Patent #
US 8,320,301 B2
Filed 10/23/2003

Current Assignee
Qualcomm Inc.

Sponsoring Entity
Qualcomm Inc.

MULTISITE MIMO COOPERATION IN CELLULAR NETWORKS  
Patent #
US 20130029677A1
Filed 09/14/2012

Current Assignee
Blackberry Limited

Sponsoring Entity
Blackberry Limited

Forward link handoff for wireless communication systems with OFDM forward link and CDMA reverse link  
Patent #
US 8,400,979 B2
Filed 12/22/2003

Current Assignee
Qualcomm Inc.

Sponsoring Entity
Qualcomm Inc.

MIMO WLAN system  
Patent #
US 8,462,643 B2
Filed 07/15/2008

Current Assignee
Qualcomm Inc.

Sponsoring Entity
Qualcomm Inc.

Rowvector norm comparison method and rowvector norm comparison apparatus for inverse matrix  
Patent #
US 8,521,799 B2
Filed 06/30/2008

Current Assignee
Samsung Electronics Co. Ltd., Electronics and Telecommunications Research Institute

Sponsoring Entity
Samsung Electronics Co. Ltd., Electronics and Telecommunications Research Institute

PRESIB2 CHANNEL ESTIMATION AND SIGNAL PROCESSING IN THE PRESENCE OF MBSFN FOR LTE  
Patent #
US 20130242766A1
Filed 06/13/2012

Current Assignee
Qualcomm Inc.

Sponsoring Entity
Qualcomm Inc.

Method and system for code reuse and capacity enhancement using null steering  
Patent #
US 8,553,746 B2
Filed 04/03/2012

Current Assignee
Interdigital Technology Corporation

Sponsoring Entity
Interdigital Technology Corporation

Advanced signal processors for interference cancellation in baseband receivers  
Patent #
US 8,654,689 B2
Filed 09/28/2010

Current Assignee
III Holdings 1 LLC

Sponsoring Entity
Rambus Inc.

Random access for wireless multipleaccess communication systems  
Patent #
US 8,711,763 B2
Filed 01/19/2012

Current Assignee
Qualcomm Inc.

Sponsoring Entity
Qualcomm Inc.

Multicarrier CQI feedback method and apparatus  
Patent #
US 8,797,889 B2
Filed 11/30/2007

Current Assignee
Telefonaktiebolaget LM Ericsson

Sponsoring Entity
Telefonaktiebolaget LM Ericsson

Transmit diversity processing for a multiantenna communication system  
Patent #
US 8,873,365 B2
Filed 06/24/2009

Current Assignee
Qualcomm Inc.

Sponsoring Entity
Qualcomm Inc.

MIMO WLAN system  
Patent #
US 8,913,529 B2
Filed 04/30/2008

Current Assignee
Qualcomm Inc.

Sponsoring Entity
Qualcomm Inc.

MULTISITE MIMO COOPERATION IN CELLULAR NETWORKS  
Patent #
US 20150003431A9
Filed 09/21/2010

Current Assignee
Blackberry Limited

Sponsoring Entity
Mohammad Hadi Baligh, Hua Xu, Jianglei Ma

Transmit diversity processing for a multiantenna communication system  
Patent #
US 8,934,329 B2
Filed 10/13/2005

Current Assignee
Qualcomm Inc.

Sponsoring Entity
Qualcomm Inc.

Interference cancellation in a spread spectrum communication system  
Patent #
US 9,036,680 B2
Filed 05/27/2011

Current Assignee
Interdigital Technology Corporation

Sponsoring Entity
Interdigital Technology Corporation

MIMO system with multiple spatial multiplexing modes  
Patent #
US 9,048,892 B2
Filed 04/04/2013

Current Assignee
Qualcomm Inc.

Sponsoring Entity
Qualcomm Inc.

Advanced signal processors for interference cancellation in baseband receivers  
Patent #
US 9,172,411 B2
Filed 12/16/2013

Current Assignee
III Holdings 1 LLC

Sponsoring Entity
III Holdings 1 LLC

Multicarrier CQI feedback method and apparatus  
Patent #
US 9,191,847 B2
Filed 06/18/2014

Current Assignee
Telefonaktiebolaget LM Ericsson

Sponsoring Entity
Telefonaktiebolaget LM Ericsson

Code division multiple access transmission antenna weighting  
Patent #
US 9,219,522 B2
Filed 10/11/2010

Current Assignee
Interdigital Technology Corporation

Sponsoring Entity
Interdigital Technology Corporation

MIMO WLAN system  
Patent #
US 9,240,871 B2
Filed 04/30/2008

Current Assignee
Qualcomm Inc.

Sponsoring Entity
Qualcomm Inc.

Interference cancellation in a spread spectrum communication system  
Patent #
US 9,270,327 B2
Filed 05/12/2015

Current Assignee
Interdigital Technology Corporation

Sponsoring Entity
Interdigital Technology Corporation

Enabling phase tracking for a communication device  
Patent #
US 9,288,096 B2
Filed 12/06/2010

Current Assignee
Qualcomm Inc.

Sponsoring Entity
Qualcomm Inc.

Two stage radio frequency interference cancellation system and method  
Patent #
US 9,312,895 B1
Filed 03/09/2015

Current Assignee
Hypres Inc.

Sponsoring Entity
Hypres Inc.

ENABLING PHASE TRACKING FOR A COMMUNICATION DEVICE  
Patent #
US 20160197710A1
Filed 03/14/2016

Current Assignee
Qualcomm Inc.

Sponsoring Entity
Qualcomm Inc.

Method and apparatus for providing an efficient control channel structure in a wireless communication system  
Patent #
US 9,473,269 B2
Filed 12/01/2003

Current Assignee
Qualcomm Inc.

Sponsoring Entity
Qualcomm Inc.

Systems and methods for parallel signal cancellation  
Patent #
US 9,490,857 B2
Filed 10/28/2015

Current Assignee
III Holdings 1 LLC

Sponsoring Entity
III Holdings 1 LLC

Systems and methods for parallel signal cancellation  
Patent #
US 9,544,044 B2
Filed 12/13/2010

Current Assignee
III Holdings 1 LLC

Sponsoring Entity
III Holdings 1 LLC

Multicarrier CQI feedback method and apparatus  
Patent #
US 9,585,042 B2
Filed 10/15/2015

Current Assignee
Telefonaktiebolaget LM Ericsson

Sponsoring Entity
Telefonaktiebolaget LM Ericsson

Advanced signal processors for interference cancellation in baseband receivers  
Patent #
US 9,647,708 B2
Filed 08/30/2010

Current Assignee
III Holdings 1 LLC

Sponsoring Entity
III Holdings 1 LLC

Interference suppression for CDMA systems  
Patent #
US 9,735,816 B2
Filed 05/15/2015

Current Assignee
III Holdings 1 LLC

Sponsoring Entity
III Holdings 1 LLC

Channel state feedback for multicell MIMO  
Patent #
US 9,801,167 B2
Filed 02/12/2013

Current Assignee
Fujitsu Limited

Sponsoring Entity
Fujitsu Limited

Two stage radio frequency interference cancellation system and method  
Patent #
US 9,838,051 B1
Filed 04/11/2016

Current Assignee
Hypres Inc.

Sponsoring Entity
Hypres Inc.

Method and apparatus for providing an efficient control channel structure in a wireless communication system  
Patent #
US 9,876,609 B2
Filed 05/18/2016

Current Assignee
Qualcomm Inc.

Sponsoring Entity
Qualcomm Inc.

Pilots for MIMO communication systems  
Patent #
US 9,967,005 B2
Filed 08/26/2015

Current Assignee
Qualcomm Inc.

Sponsoring Entity
Qualcomm Inc.

Advanced signal processors for interference cancellation in baseband receivers  
Patent #
US 10,050,733 B2
Filed 10/27/2015

Current Assignee
III Holdings 1 LLC

Sponsoring Entity
III Holdings 1 LLC

Enabling phase tracking for a communication device  
Patent #
US 10,057,026 B2
Filed 03/14/2016

Current Assignee
Qualcomm Inc.

Sponsoring Entity
Qualcomm Inc.

Monitoring rotating machinery using radio frequency probes  
Patent #
US 10,280,787 B2
Filed 11/09/2015

Current Assignee
University of Notre DAME Du Lac

Sponsoring Entity
University of Notre DAME Du Lac

Apparatus and method for transmitting and receiving information and power in wireless communication system  
Patent #
US 10,411,764 B2
Filed 10/31/2018

Current Assignee
Industry Academic Cooperation Foundation

Sponsoring Entity
Industry Academic Cooperation Foundation

Determining base station fronthaul transmission control based on loopback signal  
Patent #
US 10,455,496 B2
Filed 10/06/2016

Current Assignee
AlcatelLucent SA

Sponsoring Entity
AlcatelLucent SA

Coherent signal analyzer  
Patent #
US 10,605,841 B2
Filed 11/09/2015

Current Assignee
University of Notre DAME Du Lac

Sponsoring Entity
University of Notre DAME Du Lac

Multisite MIMO cooperation in cellular networks  
Patent #
US 10,615,851 B2
Filed 09/21/2010

Current Assignee
Blackberry Limited

Sponsoring Entity
Blackberry Limited

Method and apparatus for measuring reporting channel state information in a high efficiency, high performance communications system  
Patent #
US 6,473,467 B1
Filed 03/30/2000

Current Assignee
Qualcomm Inc.

Sponsoring Entity
Qualcomm Inc.

Simulcasting system with diversity reception  
Patent #
US 5,845,199 A
Filed 12/05/1996

Current Assignee
Telefonaktiebolaget LM Ericsson

Sponsoring Entity
Telefonaktiebolaget LM Ericsson

Method and apparatus for transmitting communication signals using transmission space diversity and frequency diversity  
Patent #
US 6,006,075 A
Filed 06/18/1996

Current Assignee
Telefonaktiebolaget LM Ericsson

Sponsoring Entity
Telefonaktiebolaget LM Ericsson

Wireless system with transmitter having multiple transmit antennas and combining open loop and closed loop transmit diversities  
Patent #
US 6,594,473 B1
Filed 05/24/2000

Current Assignee
Texas Instruments Inc.

Sponsoring Entity
Texas Instruments Inc.

Seamless soft handoff in a CDMA cellular communications system  
Patent #
US 5,883,888 A
Filed 12/03/1996

Current Assignee
Unwired Planet LLC

Sponsoring Entity
Telefonaktiebolaget LM Ericsson

Cellular/satellite communications system with improved frequency reuse  
Patent #
US 5,848,060 A
Filed 08/21/1995

Current Assignee
Telefonaktiebolaget LM Ericsson

Sponsoring Entity
Telefonaktiebolaget LM Ericsson

Method of upstream channel modeling for PCM modem  
Patent #
US 5,887,027 A
Filed 12/24/1997

Current Assignee
WSOU Investments LLC

Sponsoring Entity
Lucent Technologies Inc.

Receiver  
Patent #
US 6,577,686 B1
Filed 10/12/1999

Current Assignee
Intertechnology Global LLC

Sponsoring Entity
Matsushita Electric Industrial Company Limited

Initial power control for spreadspectrum communications  
Patent #
US 20020036996A1
Filed 02/28/2001

Current Assignee
Interdigital Technology Corporation

Sponsoring Entity
Interdigital Technology Corporation

Method and system for estimating adaptive array weights used to transmit a signal to a receiver in a wireless communication system  
Patent #
US 6,434,366 B1
Filed 05/31/2000

Current Assignee
Google Technology Holdings LLC

Sponsoring Entity
Motorola Inc.

Array antenna system of wireless base station  
Patent #
US 6,064,338 A
Filed 08/27/1998

Current Assignee
Fujitsu Limited

Sponsoring Entity
Fujitsu Limited

Method and system for transmitting and demodulating a communications signal using an adaptive antenna array in a wireless communication system  
Patent #
US 6,067,324 A
Filed 06/30/1998

Current Assignee
Google Technology Holdings LLC

Sponsoring Entity
Motorola Inc.

Method and apparatus for interference rejection with different beams, polarizations, and phase references  
Patent #
US 6,081,566 A
Filed 04/19/1996

Current Assignee
Telefonaktiebolaget LM Ericsson

Sponsoring Entity
Telefonaktiebolaget LM Ericsson

Twoway paging system and apparatus  
Patent #
US 6,088,593 A
Filed 12/04/1998

Current Assignee
Telefonaktiebolaget LM Ericsson

Sponsoring Entity
Telefonaktiebolaget LM Ericsson

Method and apparatus for control of base stations in macro diversity radio systems  
Patent #
US 6,104,933 A
Filed 06/23/1997

Current Assignee
Telefonaktiebolaget LM Ericsson

Sponsoring Entity
Telefonaktiebolaget LM Ericsson

Spatiotemporal processing for communication  
Patent #
US 6,144,711 A
Filed 08/27/1997

Current Assignee
Cisco Systems Inc.

Sponsoring Entity
Cisco Systems Inc.

Up link macro diversity method and apparatus in a digital mobile radio communication system  
Patent #
US 5,867,791 A
Filed 05/23/1997

Current Assignee
Unwired Planet LLC

Sponsoring Entity
Telefonaktiebolaget LM Ericsson

Efficient apparatus for simultaneous modulation and digital beamforming for an antenna array  
Patent #
US 5,909,460 A
Filed 12/07/1995

Current Assignee
Telefonaktiebolaget LM Ericsson

Sponsoring Entity
Telefonaktiebolaget LM Ericsson

Twoway paging system and apparatus  
Patent #
US 5,940,742 A
Filed 09/19/1997

Current Assignee
Telefonaktiebolaget LM Ericsson

Sponsoring Entity
Telefonaktiebolaget LM Ericsson

Method and apparatus arranged to reduce traffic on a network channel  
Patent #
US 5,940,741 A
Filed 12/20/1996

Current Assignee
Google Technology Holdings LLC

Sponsoring Entity
Motorola Inc.

Radio station arranged for spacediversity and polarization diversity reception  
Patent #
US 5,963,874 A
Filed 09/26/1996

Current Assignee
Telefonaktiebolaget LM Ericsson

Sponsoring Entity
Telefonaktiebolaget LM Ericsson

Radio communication system with diversity reception on a timeslot by timeslot basis  
Patent #
US 5,991,282 A
Filed 05/28/1997

Current Assignee
Telefonaktiebolaget LM Ericsson

Sponsoring Entity
Telefonaktiebolaget LM Ericsson

Polarization diversity phased array cellular base station and associated methods  
Patent #
US 5,724,666 A
Filed 05/11/1995

Current Assignee
Unwired Planet LLC

Sponsoring Entity
Telefonaktiebolaget LM Ericsson

Cellular system employing base station transmit diversity according to transmission quality level  
Patent #
US 5,812,935 A
Filed 03/03/1995

Current Assignee
Hughes Electronics Corporation

Sponsoring Entity
Hughes Electronics Corporation

Cellular/satellite communications systems with improved frequency reuse  
Patent #
US 5,812,947 A
Filed 12/02/1996

Current Assignee
Telefonaktiebolaget LM Ericsson

Sponsoring Entity
Telefonaktiebolaget LM Ericsson

A cellular/satellite communications system with generation of a plurality of sets of intersecting antenna beams  
Patent #
US 5,594,941 A
Filed 05/16/1995

Current Assignee
Telefonaktiebolaget LM Ericsson

Sponsoring Entity
Telefonaktiebolaget LM Ericsson

Cellular/satellite communications system with improved frequency reuse  
Patent #
US 5,619,503 A
Filed 01/11/1994

Current Assignee
Telefonaktiebolaget LM Ericsson

Sponsoring Entity
Telefonaktiebolaget LM Ericsson

Cellular/satellite communications system with improved frequency reuse  
Patent #
US 5,631,898 A
Filed 05/16/1995

Current Assignee
Telefonaktiebolaget LM Ericsson

Sponsoring Entity
Telefonaktiebolaget LM Ericsson

Multiple beamwidth phased array  
Patent #
US 5,642,358 A
Filed 04/08/1994

Current Assignee
Telefonaktiebolaget LM Ericsson

Sponsoring Entity
Telefonaktiebolaget LM Ericsson

Method of and apparatus for reducing the complexitiy of a diversity combining and sequence estimation receiver  
Patent #
US 5,481,572 A
Filed 08/02/1994

Current Assignee
Unwired Planet LLC

Sponsoring Entity
Telefonaktiebolaget LM Ericsson

Diversity receiver for signals with multipath time dispersion  
Patent #
US 5,499,272 A
Filed 05/31/1994

Current Assignee
Telefonaktiebolaget LM Ericsson

Sponsoring Entity
Telefonaktiebolaget LM Ericsson

Technique for achieving the theoretical coding gain of digital signals incorporating error correction  
Patent #
US 4,995,057 A
Filed 01/30/1989

Current Assignee
ATT Information Systems Inc., American Telephone Telegraph

Sponsoring Entity
ATT Information Systems Inc., American Telephone Telegraph

Snagging hook  
Patent #
US 4,930,248 A
Filed 05/03/1989

Current Assignee
John B. Owens

Sponsoring Entity
John B. Owens

64 Claims
 1. A method of employing coherent transmit diversity in a wireless communication network, the method comprising:
forming a plurality of transmit signals, each comprising a combination of information signals intended for different ones of a plurality of wireless receivers, such that, at each one of said wireless receivers, the intended information signals in the plurality of transmit signals add while the other information signals cancel; and
coherently transmitting said transmit signals from said plurality of transmitters over different propagation channels to said plurality of receivers.  View Dependent Claims (2, 3, 4, 5, 6, 7, 8, 9, 10, 11, 12, 13, 14)
 15. A method of transmitting signals in a wireless communication network from one or more transmitters to a plurality of receivers, the method comprising:
receiving a plurality of information signals at a transmit processor, each information signal intended for a different receiver;
generating a plurality of transmit signals by forming weighted combinations of said information signals based on channel estimates for propagation paths between said one or more transmitters and said plurality of receivers; and
coherently transmitting said transmit signals from said one or more transmitters to said plurality of receivers.  View Dependent Claims (16, 17, 18, 19, 20, 21, 22, 23, 24, 25, 26, 27, 28, 29, 30, 31, 32, 33, 34, 35, 36, 37, 38, 39, 40)
 41. A method of reducing interference at one or more receivers, the method comprising:
prefiltering information signals for one or more receivers using channel estimates representing propagation channels between two or more transmitters and said one or more receivers to form two or more transmit signals, such that said transmit signals combine at said one or more receivers to reduce interference between information signals; and
transmitting a different one of said transmit signals from each one of said two or more transmitters.  View Dependent Claims (42, 43, 44, 45)
 46. A method of transmitting signals in a wireless communication network from a plurality of transmitting stations to a plurality of receivers, the method comprising:
forming a matrix of elements, each element being a multivalued element describing the propagation path from one of said transmitting stations to one said receivers;
forming an inverse of said matrix comprising an adjoint matrix and a determinant;
determining the frequency response from a signal input in said communications network to a receiver for which the signal applied to said signal input is intended, assuming said inverse matrix is used for filtering and combining said applied signals to obtain signals for transmission from respective transmitting stations;
forming said applied signal based on information to be transmitted to said intended receiver and said frequency response so that the information is coded in said applied signal to produce a spectrum of the applied signal that is efficient for transmitting said information to said intended receiver.  View Dependent Claims (47, 48, 49, 50)
 51. A wireless communication network comprising:
a transmitter operative to transmit a plurality of transmit signals to a plurality of receivers; and
a transmit processor operative to form said transmit signals as weighted combinations of individual information signals intended for respective ones of said plurality of receivers by weighting said information signals using channel estimates representing propagation channels between said transmitter and said receivers.  View Dependent Claims (52, 53, 54)
 55. A wireless communication network comprising:
a plurality of transmitters operative to transmit a plurality of transmit signals over different propagation channels to one or more receivers; and
a transmit processor operative to form said plurality of transmit signals by prefiltering one or more information signals intended for said one or more receivers using channel estimates representing propagation channels between said transmitters and said one or more receivers.  View Dependent Claims (56, 57, 58, 59, 60, 61, 62, 63, 64)
1 Specification
[0001] The present invention generally relates to wireless communication networks, and particularly relates to inventive coherent transmit diversity techniques that improve communication performance and efficiency.
[0002] Wireless communication networks employ various techniques to maximize their communication efficiency, performance, and economy. Economy and efficiency may be measured from the service provider'"'"'s perspective, and generally relate to supporting the greatest possible number of users for lowest possible service and equipment costs. Performance may be measured from the perspective of both service providers and subscribers, and centers on providing the highest possible reliability at whatever data rate is being used.
[0003] Approaches to improving communication reliability are at least as numerous as the factors influencing that reliability. Thus, one sees a broad range of technologies and methods in modern communication systems, all meant to improve the underlying reliability of the communication services provided. Of these varied approaches, diversity communications itself represents a varied mix of techniques.
[0004] In wireless communication, successful reception depends on a relatively unobstructed signal path between a receiver and a supporting transmitter. If the path is obstructed, fading may occur. Fading may be particularly problematic when the wireless receiver is moving with respect to the transmitter. In this scenario, the signal path is ever changing and fading becomes a highly dynamic factor influencing reception quality.
[0005] Diversity combats fading by providing additional communication elements capable of transmitting or receiving on different signal paths in the hope that at least one of these signal paths will remain unfaded at any particular instant in time. Diversity may be implemented on the transmit side using different frequencies, antennas, polarizations, or combinations thereof to transmit to a given receiver.
[0006] Implementing diversity on the receive side, which is a more common practice, usually entails providing multiple receiving elements, such as two or more receive antennas. With multiple receiving elements, the receiver may receive its signal of interest on multiple signal paths, thus increasing the likelihood that at least one of the signal paths will be unobstructed at any instant in time.
[0007] With transmit macro diversity multiple transmitting antennas may be used to transmit signals to a given wireless receiver. When the antennas are individual elements in an antenna array, coherent transmit beamforming may be used to advantage. With this technique, reception performance is improved by the summation of transmit powers received from the multiple transmit antenna elements, as well as by the focusing of the transmit signal towards the receiver.
[0008] If the transmit antenna elements are not colocated, spatial transmit diversity may be achieved by transmitting the same signal from multiple, spacedapart transmit antennas. However, noncoherent (e.g., timeoffset) diversity signals transmitted in this manner increase the geographical spread of interference between different wireless receivers using the same communication channel.
[0009] The present invention relates to a method of providing coherent transmit macro diversity in a wireless communication network, such as a mobile cellular communications network. Multiple transmitters transmit signals to one or more wireless receivers. Signals transmitted from each transmitter are preprocessed before transmission using dynamic channel estimates characterizing the downlink propagation paths between that transmitter and each one of the one or more wireless receivers. Preprocessing in this manner attempts to compensate for the expected influence of the downlink propagation channels on the transmitted signal so as to minimize interference at the wireless receivers.
[0010] One of the many advantages arising from the above technique is that the same communication channel may be used to transmit to multiple wireless receivers operating in the same service area. This applies even where the receivers are colocated. In a cellular communication network, the ability to commonly assign one communication channel between multiple users, even if those users are immediately adjacent one another, represents a clear gain in system efficiency. For TimeDivision Multiple Access (TDMA) systems, using the same communication channel for multiple users might entail assigning the same TDMA time slot on the same frequency channel to a group of wireless receivers operating in the same general area. For Code Division Multiple Access (CDMA) systems, using the same communication channel for multiple users might entail assigning the CDMA code channel for the group of wireless receivers.
[0011] Preferably, N transmitters transmit to M wireless receivers, where N is greater than or equal to M, although in some implementations, there may be more receivers than transmitters (M>N). Individual information signals intended for different ones of the wireless receivers are combined for each transmitter, based on the overall set of channel estimates characterizing the propagation paths between each transmitter and each wireless receiver. The goal of this preprocessing is to adjust the transmission characteristics of each information signal as a component in each combined transmit signal such that, at each wireless receiver, the desired information signal adds between the different transmit signals, while the unwanted information signals cancel. For any one wireless receiver, the desired information signal is the one intended for it, while the unwanted information signals are the information signals corresponding to the remaining wireless receivers.
[0012] The above goal may be realized through implementation of a central transmit processor or other like system that has access to the different information signals corresponding to the wireless receivers and to the propagation path information relating transmitters to receivers. This processor may receive the information signals from, for example, a mobile switching center handling call management for the involved wireless receivers, or from the involved base station controllers. Similarly, the processor may receive channel estimate information from elsewhere in the communication network, or may generate it locally, based on incoming data, such as loop back information from the wireless receivers that may be used to generate downlink channel estimates.
[0013] In any case, the transmit processor combines the information signals using the channel estimate information to form the transmit signals, such that the transmit signals, as received at each wireless receiver, combine to minimize interference at that receiver. Each transmit signal travels to the receivers through a different propagation channel, which may be a multipath channel, and this allows the transmit processor to fashion the set of transmit signals to achieve the desired interference cancellation at each wireless receiver.
[0014] Realizing the different propagation channels may involve transmitting each transmit signal from one in a set of spaced apart transmit antennas or antenna elements. In a cellular environment, this might involve distributing the transmit signals to a set of spaced apart radio base stations with the desired timealignment between each of the transmit signals. In other instances, a single antenna assembly having a number of different polarizations may be used to transmit the different transmit signals, each one assigned to a different polarization. In all cases, coherency (e.g., relative time alignment) is preserved between the plurality of transmit signals.
[0015] Flexibility also exists in terms of forming the transmit signals. Conventionally, an information signal intended for a given wireless receiver is transmitted from a transmitter to the wireless receiver. Propagation path characteristics between the transmitter and receiver determine to what extent the received signal approximates the signal transmitted. With knowledge of these propagation path characteristics, the transmitted signal may be precompensated to offset channel effects in the received signal at the wireless receiver. How well the precompensation eliminates channel distortion at the receiver depends on the accuracy of the channel estimate.
[0016] The present invention uses dynamically updated channel estimate information to precompensate the transmit signal to effect the desired interference cancellation at each receiver. For example, current channel estimate information may be held in a memory accessible to the transmit processor. Because transmit diversity is employed, transmit signal precompensation entails forming a different weighted or filtered combination of the individual information signals for transmission for each transmitter using the channel estimate information.
[0017] One approach bases formation on matrixoriented manipulations. A channel estimate matrix comprises matrix elements, each one of which characterizes the propagation channel between one transmitter (e.g., separate transmit antenna or different antenna polarization) and one of the wireless receivers. The transmit processor uses the channel estimate matrix to generate filters through which the information signals are passed and then combined. Information signals are passed through a different set of filters for each transmitter, as each transmitter has different propagation paths with respect to the various receivers.
[0018] In an exemplary approach, the transmit processor essentially applies the inverse of the channel estimate matrix to the set of information signals to generate the desired combinations of information signals which form the transmit signals. Forming an inverse matrix generally entails dividing the adjoint matrix by the determinant of the original matrix. Here, each element of the channel state matrix may be expressed in terms of the digital delay operator “z.” If multipath exists, the elements are expressed as z polynomials.
[0019] One approach for realizing this is to form a logical array of Finite Impulse Response (FIR) filters that corresponds to the adjoint of the channel estimate matrix. Thus, each FIR filter in the array corresponds to the similarly positioned adjoint matrix element, which as noted may be expressed as z polynomial. Each information signal feeds into one row of the filter array, and sums are taken down the array columns to form the individual transmit signals. These filtering and summing operations create a set of transmit signals which, when passed through the involved propagation channels, will combine at each receiver in a manner that ideally cancels out all but that receiver'"'"'s desired information signal.
[0020] In advance of passing the information signals into the FIR array, each information signal may be passed through an Infinite Impulse Response (IIR) filter based on the determinant of the channel estimate matrix. This is essentially equivalent to the “divide by” step noted in the inverse matrix algorithm above. Bringing in this determinantbased filtering process provides InterSymbol Interference (ISI) cancellation within the desired information signals received by the corresponding wireless receivers. Since the wireless receivers may themselves include ISIcompensation, considerable flexibility exists with regard to determinantbased filtering and root elimination.
[0021] As a variation on the above matrix processing, the channel estimate matrix may be expressed in terms of frequency response. That is, the propagation channels between transmitters and receivers are expressed in terms of their frequency response, and these expressions are evaluated over a range of frequencies. This evaluation allows the system to perform power spectral shaping, which allows the communication network to reduce or eliminate transmit power in portions of the spectrum where it is most attenuated. In this way, total transmit power is more efficiently used in transmission.
[0022]FIG. 1 is a diagram of an exemplary wireless communication network for practicing the present invention.
[0023]FIG. 2 is a diagram of the transmit processor of FIG. 1.
[0024]FIG. 3 is a diagram of exemplary transmit signal preprocessing.
[0025]FIG. 4 is a diagram of alternate exemplary transmit signal preprocessing.
[0026]FIG. 5 is a diagram of exemplary wireless receiver grouping.
[0027]FIG. 6 is a graph of the frequency response of an exemplary channel estimate matrix determinant before and after pole deletion.
[0028]FIG. 7A is a graph of the frequency responses for a set of adjoint matrix polynomials determined from an exemplary channel estimate matrix.
[0029]FIG. 7B is a graph of the frequency responses of the adjoint matrix polynomials divided by the associated reduced channel estimate matrix determinant.
[0030]FIG. 7C is a graph of the spectral shape of total transmit power to each one the group of wireless receivers illustrated in FIG. 5.
[0031]FIG. 8A is a graph of net power gain/loss for one through five deleted determinant poles for an exemplary three receiver/three transmitter configuration.
[0032]FIG. 8B is a gain/loss histogram associated with FIG. 8A for an optimal number of deleted determinant poles.
[0033]FIG. 9 is a diagram of exemplary loop back signal provisions in a wireless receiver operating in the network of FIG. 1.
[0034]FIG. 10 is a gain/loss histogram for one through five deleted determinant poles in a three receiver/three dualpolarization transmitter configuration.
[0035]FIG. 11 is a diagram of an exemplary RAKE receiver as might be used in some configurations of the wireless receivers in FIG. 1.
[0036]FIG. 12 is a graph of the probability distribution of the energy lost in restricting the number of RAKE taps in the receiver of FIG. 11.
[0037]FIG. 13 is a graph of the probability distribution of the gain for coherent polarization diversity in a threetap RAKE receiver.
[0038]FIG. 14 is a diagram of an alternate embodiment of the present invention, which is exemplary for use in CDMAbased wireless networks.
[0039]FIG. 15 is a diagram of InterSymbol Interference (ISI) in a CDMA signal.
[0040] Turning now to the drawings, FIG. 1 illustrates an exemplary wireless communication network for practicing the present invention, generally referred to by the numeral 10. A plurality of base stations 12 transmit via corresponding transmit antennas 14 to a plurality of wireless receivers, shown here as mobile terminals 16. A central processing system, referred to herein as a transmit processor 18, provides transmit signals to each base station 12 in baseband form. In turn, each base station 12 converts its input transmit signal into a form suitable for transmission from its associated antenna 14. The transmit processor 18 may, for example, be associated with or be part of a mobile switching center (MSC) 19 that interfaces the network 10 with one or more external networks 21. Such external networks 21 may include the Public Switched Telephone Network (PSTN), and one or more packet data networks (PDNs), such as the Internet.
[0041] Although illustrated as using multiple, spacedapart base stations 12 and antennas 14 to achieve transmit diversity to the mobile terminals 16, the network 10 may also use one or more antenna assemblies having multiple polarizations, wherein each transmit signal is transmitted using a different polarization. These polarized transmitting elements may be colocated. With the present invention, it is enough that the different transmit signals each be transmitted through an independently fading propagation channel. Obviously, this may be achieved in any number of different transmit configurations.
[0042] As illustrated, there is a onetoone correspondence between the number of antennas 14 (e.g., transmitting elements or stations) and the number of mobile terminals 16. However, the present invention may be varied such that there are more or fewer transmit stations than there are wireless receivers.
[0043] Referring again to the illustration, base station 12A broadcasts transmit signal T_{1 }via antenna 14A, and likewise, base stations 12B and 12C broadcast transmit signals T_{2 }and T_{3 }via antennas 14B and 14C, respectively. Each transmit signal T_{x }represents a weighted combination of different information signals intended for respective ones of the mobile terminals 16. Each mobile terminal 16 receives each of the transmit signals, albeit through different downlink propagation channels. That is, each mobile terminal 16 receives the combination of T_{1}, T_{2}, and T_{3}, as affected by the propagation channel between that mobile terminal 16 and point of transmission for T_{1}, T_{2}, and T_{3}. As used herein, the term “channel” refers to the radio signal propagation path or paths between a wireless transmitter and a wireless receiver, unless specifically noted otherwise.
[0044] The set of downlink channels between each transmit antenna 14 and each mobile terminal 16 may be represented using the nomenclature C_{jk}. This symbol denotes the channel between the jth wireless receiver (e.g., a given one of the mobile terminals 16) and the kth transmitter (e.g., a given one of the transmit antennas 14). Thus, in the illustration, C_{11 }denotes the channel between mobile terminal 16A and transmit antenna 14A, C_{12 }denotes the channel between mobile terminal 16A and transmit antenna 14B, and so on.
[0045] Typically, the propagation channel between a given antenna 14 and a given mobile terminal 16 comprises a number of downlink propagation paths. These multiple propagation paths, referred to as multipaths, each have characteristic attenuation, phase, and delay attributes, which may be expressed as a complex coefficient representing magnitude and phase, and a corresponding delay attribute. Thus, channel coefficient C_{jk }may be represented by the polynomial C_{0}+C_{1}_{z}^{−1}+C_{2}_{z}^{−2}+. . . +C_{n−1}_{z}^{n−1}, where C_{n }represents the channel coefficient associated with a single multipath and z^{x }is a delay operator that represents the unit delay of the various multipaths relative to the first received multipath. The time delay operator could be expressed relative to a multipath other than the first received multipath, in which case the above expression might include channel coefficients with positive delay elements (e.g., C_{x}z^{+4},C_{x−1}z^{+3}, and so on).
[0046] In any case, the above expressions demonstrate that the multipath channel between any transmit antenna 14 and a mobile terminal 16 may be expressed as a polynomial in z, based on the channel coefficients and corresponding path delays associated with the multipaths involved. The complete set of channel coefficients from all antennas to all receivers forms a channel estimate matrix and may be expressed as follows:
[0047] where each matrix element C_{jk }is a polynomial that corresponds to one multipath channel between a given transmit station and a given mobile terminal. Again, transmit station broadly refers to the transmitting element associated with one of the transmit signals.
[0048] In this example, the present invention makes use of a channel estimate matrix to generate transmit signals T_{1}, T_{2}, and T_{3 }in such a manner as to allow the same downlink communication channel to be used by multiple mobile terminals 16 operating within the same service area. The transmit signals T_{1}, T_{2}, and T_{3 }comprise weighted combinations of information signals S_{1}, S_{2}, and S_{3}, which are intended for three different mobile terminals. Information signals S_{1}, S_{2}, and S_{3 }are combined such that at each mobile terminal 16 receives only its wanted signal, with the unwanted signals (e.g., those intended for the other mobile terminals 16) canceling.
[0049] To better understand the approach to the unwanted signal cancellation, it is helpful to examine the received signal R at any one of the receivers (e.g., mobile terminals 16). The signal received by a mobile terminal 16 during a given symbol period may be described by the following equation,
R(i)=C_{0}S(i)+C_{1}S(i−1)+C_{2}S(i−2)+. . . +C_{L−1}S(i−L+1) (Eq. 2)
[0050] where R(i) is the complex value of the received signal sample (with carrier frequency removed) at time instant i, S(i) is the symbol transmitted at time i, C_{j }is the complex number describing the phase and amplitude of the propagation path with jsymbol periods of delay (relative to the shortest path, C_{0}), and L represents the number of multipaths between the antenna 14 and the mobile terminal 16.
[0051] The combination of a signal with itself delayed in steps of one symbol period may also be described by the use of the Ztransform. A discretetime signal S that is delayed by one sample is denoted by z^{−1}·S, where “z” is the time advance operator and its reciprocal is the time delay operator. Thus the polynomial a+bz^{−1}+cz^{−2 }times S means aS(i)+bS(i−1)+cS(i−2).
[0052] Therefore, equation (1) can be written as:
R=C(z^{−1})S (Eq. 3)
[0053] where “C” is now a polynomial, in z^{−1}, with coefficients C_{0 }. . . C_{L−1}, and R stands for a received signal sample stream, while S stands for the corresponding transmitted symbol stream.
[0054] If R_{j }denotes the signal received at receiver j and T_{k }denotes the signal transmitted by transmit antenna k, the entire picture of propagation from all antennas 14 to all mobile terminals 16 may be described by the matrix equation,
[R]=[C]·[T] (Eq. 4)
[0055] where C is the channel estimate matrix from Eq. 1. For the 3×3 example of FIG. 1, Eq. (4) is expressed as,
[0056] Here, R_{1 }is the composite signal received at mobile terminal 16A, R_{2 }is the composite signal received at mobile terminal 16B, and R_{3 }is the composite signal received at mobile terminal 16C. Similarly, T_{1 }is the transmit signal from antenna 14A, T_{2 }is the transmit signal from antenna 14B, and T_{3 }is the transmit signal from antenna 14C. Thus, each R_{x }is a composite of the three transmit signals T_{1}, T_{2}, and T_{3}, each one received through a different propagation path relative to the receiver receiving R_{x}. In turn, each of the transmit signals T_{k }comprises differently weighted components of each of the desired information signals S_{1}, S_{2}, and S_{3}.
[0057] If we desire each receiver j (e.g., a given mobile terminal 16) to receive only its intended symbol stream S_{j}, then the transmit signals T_{k }needed to achieve that are given by,
[T]=[C]^{−1}·[S] (Eq. 6)
[0058] where [C]^{−1 }is the inverse of the channel estimate matrix [C], and [S] is the matrix of desired information symbol streams intended for the set of mobile terminals 16. In (Eq. 6), each T_{k }in [T] comprises a weighted combination of the information symbol streams S_{j }intended for each receiver. Assuming that the channel coefficients for the propagation paths are estimated correctly, multiplication of the symbol stream matrix [S] by the inverse of the channel estimate matrix [C]^{−1 }yields the transmit symbol matrix [T] which results in complete cancellation of the unwanted signals at each mobile terminal 16. Thus the problem at hand concerns how to invert the channel estimate matrix [C] of z−polynomials, and whether such an entity exists and is numerically well or ill conditioned.
[0059] Matrix theory describes the procedure for computing the inverse of a matrix as follows. First, transpose the matrix then replace every element by its cofactor to obtain a matrix known as the adjoint matrix. Then divide each element of the adjoint matrix by the determinant of the original matrix to obtain the inverse matrix.
[0060] The determinant of a matrix is given by sums of products of its elements and is computable in a deterministic fashion. For example, assuming the following 3×3 channel estimate matrix,
[0061] the determinant may be calculated as:
C_{11}(C_{22}·C_{33}−C_{32}·C_{23})−C_{12}(C_{21}·C_{33}−C_{31}·C_{23}) (Eq. 8) +C_{13}(C_{21}·C_{32}−C_{31}·C_{22})
[0062] The cofactor of element C_{11 }is (C_{22}·C_{33}−C_{32}·C_{23}), which is therefore the first element of the adjoint matrix. The first element of the inverse matrix is thus,
[0063] When each matrix element C_{jk }in the above expression is a z−polynomial, it may be deduced that the expression'"'"'s numerator is a z−polynomial of twice the order of the C_{jk}, while the denominator is a polynomial of three times the order of C_{jk}. Therefore, the inverse matrix comprises elements that have both a numerator z−polynomial and a denominator z−polynomial. According to Eq. 6, information bearing signals S_{j }to be transmitted to the mobile terminals 16 must be processed by multiplication with respective inverse matrix elements and summed to produce the transmit signals T_{k}.
[0064] Processing a signal S_{j }by multiplication with a numerator polynomial is not problematic, and only involves delaying the signal by multiples of a symbol period, along with proper weighting and adding. A finite impulse response (FIR) filter may be used to implement these steps. The signal might conveniently be assembled into blocks, such as TDMA bursts, and a whole burst processed at once within digital signal processing memory before being output to transmit modulator circuits. The processed burst is longer than the original by the length of the impulse response of the processing zfunction, but can be prevented from overlapping adjacent bursts by allowing a guard time between bursts, as is common in TDMA and GSM cellular systems. In the case of a continuous signal, such as in CDMA for example, the signal can be segmented into blocks, processed in the same way as described above for a TDMA burst, and then the processed segments, which are now longer and overlap, may be linearly superposed so that the extended tails add to the neighboring segments. It is acceptable for the impulse response tails to overlap neighboring segments so long as the intended receiver positions are the same for the neighboring segments. When the intended receiver positions are different for neighboring segments, the tails of a previous segment overlapping a new segment would cause interference to be transmitted in the direction of the new receivers. In this case, adequate guard times should be used, or alternatively a means to allocate receivers to timeslots based on position can be used so that the spatial pattern of receivers served in adjacent timeslots is similar.
[0065] Processing a signal by a denominator z−polynomial is more problematic however. A denominator z−polynomial may be realized as an Infinite Impulse Response (IIR) filter. The impulse response tails off exponentially as a function of the successive powers of the poles of the z−polynomial, i.e. the roots of the denominator polynomial. Thus, if a root of the denominator polynomial has a magnitude less than 1, the impulse response decays; if the root has a magnitude equal to one, the impulse response rings forever; and if the root has a magnitude greater than one, the impulse response grows. The latter case may be dealt with by processing the hypothesized signal block backwards in time to apply those roots that are greater than unity. Thus, the signal block may be passed backwards through an IIR filter comprised of the reciprocal roots of the denominator polynomial, which are less than unity, and therefore give a decaying timereversed impulse response. The roots that were already less than unity form an IIR filter, which is applied in the forward time direction. It remains to determine what to do about roots that lie close to unity. Due to their slow decay, excessive latency in processing the signal exactly would be entailed. If the signal is not processed accurately, by truncating the IIR response for example, there is a risk of causing excessive interference overlap with adjacent blocks because the exact inverse Cmatrix would not have been applied.
[0066] If the exact inverse Cmatrix (downlink channel coefficient matrix) is applied to the information signals S_{j }to form the transmit signals T_{k }from each of the antennas 14, the mobile terminals 16 will receive only their intended information signal S_{j }with multipath interference already cancelled at the transmitter, since the resulting channel from the transmitter to the receiver is [C]·[C]^{−1}, which is unity.
[0067] Omitting division by the troublesome denominator polynomial is equivalent to multiplying each information signal S_{J }by the omitted denominator polynomial, so the resultant channel from each antenna 14 to each mobile terminal 16 would be equal in length to that omitted polynomial, which in the above 3×3 example was three times the length of the individual channel coefficients C_{jk}. The mobile terminals 16 may have an equalizer that can decode symbols in the presence of normal amounts of multipath propagation delay, but perhaps not three times the normal amount. Therefore omitting the entire denominator polynomial may produce a net channel whose impulse response length exceeds the receiving equalizer'"'"'s capability. However, one third of the factors in the denominator polynomial may be omitted and this omission is equivalent to multiplying by a polynomial of ⅓rd the order of the denominator polynomial, thereby creating an artificial multipath channel with the same impulse response length as the original multipath channel.
[0068] According to one aspect of this invention therefore, ⅓rd of the denominator factors are omitted. The omitted factors are those corresponding to the roots of magnitude closest to unity, which are the terms that cause the most slowly decaying impulse responses, as well as peaks in the frequency spectrum of transmitted energy. Preferably, the roots having logmagnitude closest to zero may be selected as the roots of magnitude closest to unity.
[0069] Referring still to FIG. 1, information symbol streams S_{1}, S_{2}, and S_{3 }intended respectively for mobile terminals 16A, 16B, and 16C are input to a central transmit processor 18 that serves a plurality of base stations 12A, 12B and 12C. Transmit processor 18 includes a numerical processor 20 to perform the matrix operations on the information signals as described above. The transmit processor 18 may comprise one or more processors, such as digital signal processors, or may be comprise a portion of another network processing system, such as the MSC 19 mentioned earlier.
[0070] Using knowledge of the multipath channels C_{jk}, the numerical processor 20 filters the information signals S_{1}, S_{2 }and S_{3 }and forms weighted combinations of the information signals S_{1}, S_{2 }and S_{3 }to generate transmit signals T_{k}. The information signals S_{1 }S_{2 }and S_{3 }may be processed by the numerical processor 20 at baseband, which equates to symbolrate processing. Processing by the numerical processor 20 is described in greater detail below.
[0071] The transmit signals T_{k }are relayed from the transmit processor 18 to respective base stations 12A, 12B, and 12C, where they are converted to continuous time waveforms by digitaltoanalog converters (not shown) and upconverted to a desired radio frequency channel by modulators 22 in respective base stations 12A, 12B, and 12C. The upconverted signals are then amplified to desired transmit power levels by PAs 26 in each of the base stations 12 and transmitted by respective antennas 14A, 14B, and 14C. The transmit signals T_{k }(e.g., T_{1}, T_{2}, and T_{3}) propagate to the mobile terminals 16A, 16B, and 16C by way of the multipath channels C_{jk}. With the inventive generation of the transmit signals T_{k}, each mobile terminal 16 receives only the information signal intended for that mobile terminal 16 with the unwanted information signals canceling.
[0072] In FIG. 1, the antennas 14 may be on the same mast or tower on the same site (e.g., base station 12) or on different towers at different sites, as shown. In the latter case, the connection from the outputs of transmit processor 18 to the inputs of modulators 26 may be made by auxiliary intranetwork communications means, such as optical fiber links, wireline or microwave links, always observing the necessary precautions to maintain the phase coherency between the transmit signals T_{k }across all antennas 14. U.S. Pat. No. 5,594,941, entitled “A cellular/satellite communications system with generation of a plurality of sets of intersecting antenna beams,” discloses various means to accomplish this coherency requirement and is incorporated in its entirety herein by reference.
[0073]FIG. 2 illustrates the transmit processor 18 and numerical processor 20 in greater detail. The illustration assumes N transmit antennas 14 transmitting N transmit signals to N mobile terminals 16. Numerical processor 20, as previously described, processes the information signals S_{1}, S_{2 }and S_{3 }to form the set of transmit signals, T_{1 }T_{2 }and T_{3 }that results in unwanted signal cancellation at each mobile terminal 16. Numerical processor 20 comprises a set of IIR filters 30 and an array 32 of FIR filters 34. Transmit processor 18 includes, in addition to the numerical processor 20, a channel estimator 36 and a matrix calculator 38.
[0074] The channel estimator 36 derives the set of channel coefficients characterizing the multipath channels C_{jk }between each of the antennas 14 and each of the mobile terminals 16. The set of coefficients output from channel estimator 36 is referred to herein as channel state information (CSI). The channel estimator 36 outputs these channel coefficients to the matrix calculator 38, which calculates the coefficients of the IIR and FIR filters as previously described. More particularly, the channel estimator 36 provides N×N×L channel coefficients describing the N×N multipath channels C_{jk }from each one of the N antennas 14 to each one of the N mobile terminals 16. L is the impulse response length of the multipath channels C_{jk}, which equates to the number of multipaths in each multipath channel C_{jk}.
[0075] Treating each set of channel coefficients as a z−polynomial of order L and as an element of an N×N square matrix (i.e., the channel estimate matrix above), the matrix calculator 38 forms the N×N adjoint matrix, whose elements are polynomials of order (N−1)L, and a determinant polynomial of order NL. Matrix calculator 38 factors the determinant polynomial to determine the NL roots of the determinant polynomial and deletes the L roots having smallest absolute value of logmagnitude, leaving (N−1)L roots forming a reduced determinant polynomial of order (N−1)L. The reduced determinant polynomial determines the IIR filter as previously described and the adjoint matrix determines the coefficients of the FIR filters 34 in filter array 32.
[0076] The information bearing signals S_{j }(e.g., symbol streams S_{1}, S_{2}, and S_{3}) carrying desired information for mobile terminals 16A, 16B, and 16C, respectively are input to numerical processor 20. Within the numerical processor 20, signals S_{j }are processed through IIR filters 30 of order (N−1)L corresponding to the reduced denominator polynomial and through FIR filters 34 of order (N−1)L corresponding to the adjoint matrix polynomials as follows.
[0077] Information blocks intended to be transmitted to different mobile terminals 16 are assembled in the transmit processor 18. First the information signals are IIR filtered by the respective IIR filters 30A30N, which compensates for anticipated intersymbol interference (ISI) caused by delayed versions of the same signal S_{j}. In this exemplary implementation, the IIR filters 30 are formed from the determinant z−polynomial of the channel estimate matrix [C], and may be reduced in order by omission of factors corresponding to roots closest to the unit circle as previously described. The number of roots omitted may be anything from none up to the equalizer capability of the respective mobile terminals 16. The IIR filters 30 comprise one or more filter elements corresponding to the reciprocals of the roots of the reduced determinant z−polynomial. Each information block is passed in the reverse time direction through those filter elements formed from the reciprocals of roots of the reduced determinant that have a magnitude greater than unity and in the forward time direction through filter elements formed from the reduced determinant roots of magnitude less than unity to obtain IIRprocessed blocks.
[0078] Each omitted root requires the mobile terminal receiver to deal with one extra symbol period worth of multipath time dispersion. If the respective mobile terminals 16 have different receiver equalizer capabilities, the corresponding IIR filters 30A . . . 30N could be different, with different numbers of factors omitted in each filter 30. The IIR filters 30 may also differ by factors intended to optimize communication efficiency separately for each mobile terminal 16, as will be further discussed below. IIR filter coefficients in each of the IIR filters 30 are updated as needed based on changing CSI.
[0079] The IIR processed blocks are then FIR processed by matrix multiplication with the adjoint matrix polynomials to obtain transmit signal blocks. Filter array 32 comprising FIR filters 34 process the IIRfiltered signals to compensate for interference between signals S_{1}, S_{2}, and S_{3 }at the mobile terminals 16. Each signal is processed by a corresponding row of FIR filters 34 in the FIR filter array 32. The output signals from FIR filters 34 are summed down filter array columns, indicated by the + sign at the junction of the line from one output to another. These summed outputs represent the baseband combined transmit signals relayed by the transmit processor 18 to the modulators 22 used to generate transmit signals T_{1 }. . . T_{N}, which are in turn transmitted by transmit antennas 14A . . . 14N.
[0080] Note that each of the FIR filters 34 is updated with new coefficients as needed based on changing propagation path channel coefficients. Also, note that the FIR filters 34 in the filter array 32 correspond to elements in the adjoint of the channel estimate matrix C. Thus, each signal input into the filter array 32 is processed by a row of filters 34 corresponding to a respective row in the adjoint matrix discussed above.
[0081] The transmit signal may be further filtered to restrict transmission bandwidth and upsampled and converted to continuous time signals. The transmit signals, denoted as T_{1}, T_{2}, and T_{3 }in the illustration, are output by the transmit processor 18 and coupled to corresponding base stations 12, which may be base transceiver stations (BTSs). A modulator 22 in each BTS modulates the input transmit signal onto an RF carrier signal, which is amplified by a suitable RF power amplifier (PA) 26 to make the transmit signals (i.e., T_{1}, T_{2}, or T_{3}) discussed above suitable for transmission. Since IIR, FIR and bandwidthrestriction filtering are all linear processes, they may be carried out in any order. Generating T_{1}, T_{2}, and T_{3 }in this manner causes the combination of transmit signals to cancel differently at each mobile terminal 16, such that each mobile terminal 16 receives only its desired signal.
[0082] The abovementioned polynomial lengths apply to the case where the three transmitting antennas 14 are on the same sites or nearby sites. The base stations 12 (or, more generally, RF generation resources) and corresponding antennas 14 need not be on the same site. Rather, the base stations 12 with their corresponding transmit antennas 14 can be located apart from one another, with each linked to the transmit processor 18 by optical fiber landline for example. Different sites (e.g., base stations 12) must however synchronize the frequencies of their radio frequency carrier generators 24 to the same stable reference, using GPS for example.
[0083] The relative phases of the transmit signals T_{K }must be at least as stable as the changes in multipath propagation channel coefficients, which means that their relative frequency accuracy should be better than the Doppler frequency caused by relative motion between the transmit antennas 14 and the mobile terminals 16. This ensures that the mechanism used to keep the transmit processor 18 provided with uptodate and accurate CSI is also sufficient to compensate for any phase difference or phase drift between sites. When the three transmitting antennas 14A, 14B, and 14C are on different sites far apart, the polynomial orders may be somewhat higher, owing to the potentially greater relative path delays, as will be further discussed below.
[0084]FIG. 3 is a flowchart illustrating the operation of the transmit processor 18 located in the network 10 for determining the coefficients of the IIR filters 30 and FIR filters 34 based on channel state information (CSI). Processing begins (block 200) with updating the CSI information is updated to reflect latest estimates of the downlink channel z−polynomials comprising the channel estimate matrix C.
[0085] Once the CSI values are updated, the transmit processor 18 determines whether the network 10 is transmitting to a greater, lesser or equal number M of mobile terminals 16 compared to the number N of antennas 14 at its disposal (block 202). This test entails an evaluation of the number of mobile terminals 16 as compared to the number of transmit antennas 14 for a commonly assigned communications channel.
[0086] In some implementations, the network 10 and particularly the transmit processor 18 are not configured for transmit preprocessing when M>N. In these implementations, M>N constitutes an error condition. Error recovery procedures (block 204) can for example comprise selecting to serve only the first N mobile terminals 16 and not the remaining M−N mobile terminals 16.
[0087] If N=M, the matrix C is square and it is effectively inverted by computing the adjoint matrix of polynomials already noted and the determinant polynomial (block 206). The determinant polynomial is then factored, i.e. its roots are found, and the roots are sorted in order of magnitude (block 208). Then, the L roots and factors that are closest in magnitude of “1” are optionally discarded (block 210). The number L of roots and factors discarded for each mobile terminal 16 may be different if the mobile terminals 16 have different equalizer capabilities, which the network 10 would be informed about in advance.
[0088] If the number M of mobile terminals 16 is less than the number N of transmitting antennas 14 (transmit sites), then the extra degrees of freedom are used to optimize performance when computing the matrix C·C^{#} of polynomials of length 2L (block 212). Then the adjoint matrix of this matrix is computed, as is its determinant (block 212). The adjoint matrix polynomials are of length 2L(N−1) and the determinant polynomial is of length 2NL. Apart from this, however, the procedure is the same as for N=M (block 206). Thus, the roots are magnitude sorted (block 208), and the L roots closest to a magnitude of “1” are deleted (block 210).
[0089] In both the M<N and the N=M cases, the determinant polynomial is now of length (2N−1)L after deleting the L roots closest to the unit circle. The results of the above processing are then used to set the IIR filtering (e.g., filters 30) applied to the input information symbol streams (block 214). Likewise, the FIR filtering is configured based on the above processing and applied to the postIIR filtered signals for additional filtering, combining, and outputting as illustrated in FIG. 2 (block 216).
[0090] Where the mobile terminals 16 are the same and IIR filters 30 are all the same, the net channels (paths) by which respective signals reach respective mobile terminals 16 are identical, despite different propagation coefficients. In effect, the transmit preprocessing of the information signals compensates exactly for the different channels.
[0091] When practicing the present invention, it is desirable to group mobile terminals 16 using the same communication channel relative to a group of three neighboring base stations 12. FIG. 11 illustrates the difference between desirable and less desirable groupings. Forming group 1 as comprising the three mobile terminals 16 which are nearest to their respective base stations 12 obtains the most desirable groupings. This can be considered as producing a propagation loss matrix with the least loss along the diagonal. Then group 2 comprises the three mobiles with the second least loss to a respective one of the three base stations; while group 3 comprises the three mobiles with the third lowest loss to their respective base stations, and so on.
[0092] The least desirable groupings would comprise grouping together the three mobile terminals 16 nearest to a single base station 12, for example. The reason for favoring the former grouping method is that it obtains the highest possible wantedtounwanted signal ratio even without employing coherent transmit macrodiversity, such that the additional signals transmitted to cancel unwanted cochannel interference when practicing the present invention are a minimum. With the least desirable grouping, three mobile terminals 16 receive cochannel signals from the same, nearest base station 12, and the other, more distant base stations 12 must transmit at significant interferencecanceling signal power due to their greater distance.
[0093] Power spectral shaping amounts to waterfilling. In the trivial case of N=1 (one antenna 14 transmitting to one mobile terminal 16), precompensation as above for the propagation channel at the transmitter is not optimum. In this instance, such precompensation would use more transmit power to transport frequency components lying in a bad frequency region of the channel having high path attenuation, and less power to transport frequency components in a low attenuation region. An optimum prefiltering would do the reverse, transmitting most power in frequencies where the propagation is good and less power in frequencies where the propagation is bad. An optimum solution for a single transmit antenna 14 transmitting to a single mobile terminal 16 is called the “waterfilling” solution.
[0094] Waterfilling looks at the frequency response of the channel (path) and figuratively “pours” transmit signal power across frequency/attenuation curve, filling in the low points in the curve as would water poured into a container with an irregularly curved bottom. The waterfilling solution has the characteristic that no power is transmitted in regions of the spectrum where the received signal is instantaneously below a threshold signal to noise ratio. Producing a spectral null over a broad spectral region however implies high order bandstop filtering, which may be impractical or produce time dispersion in excess of the receiver equalizer'"'"'s capability to compensate. The current invention avoids these disadvantages, as is made evident in the following discussion.
[0095] When received signal power varies with time instead of with frequency, i.e. socalled flat fading, conventional techniques for CDMA systems, such as IS95, teach that rapid power control should be used to attempt to keep the received signal levels more or less constant. However, the waterfilling solution for determining how transmitted energy should be distributed over the frequency spectrum may be applied to time varying channels, showing that it is better to transmit less power when the channel is bad and more when the channel is good. Temporal waterfilling is the opposite of the conventional power control technique propounded for existing CDMA systems.
[0096] When received signal energy is a two dimensional function of both frequency and time, twodimensional waterfilling should ideally be employed to determine the energy to be transmitted in each frequency subband at each instant. The principles for optimum prefiltering according to the present invention for the multiple antenna and multiple receiver case will now be derived.
[0097] Instead of describing the propagation channels between the transmit antennas 14 and the mobile terminals 16 by z−polynomials, they may be described by frequency response functions C_{jk}(ω). The frequency response function can be obtained from the z−polynomial simply be substituting exp(jωt) for z or exp(−jωt) for z^{−1}. The resultant frequency response functions can be evaluated at a spot frequency to obtain a channel estimate matrix of complex numbers at each spot frequency ω. This matrix of complex numbers can be inverted by computing its adjoint matrix and its determinant to form the baseband processing implemented in the numerical processor 20 to transmit that spot frequency component of each signal to respective mobile terminals 16 in an optimum manner. This can be repeated for each spot frequency component to obtain the baseband processing required for each spot frequency component, thereby determining a set of inverse frequency response functions which form a matrix for preprocessing the information signals S_{1 }. . . S_{N }to generate T_{1 }. . . T_{N }such that desired cancellation is achieved at each mobile terminal 16.
[0098] The '"'"'941 patent to Applicant, which is incorporated herein by reference, discloses a method of prefiltering using spot frequency analysis where the number N of transmit antennas 14 exceeds the number M of mobile terminals 16. This M×N scenario is referred to as the “overdimensioned case.”
[0099] If [C (ω)] is an M×N matrix of complex numbers describing the phase and amplitude of the propagation path from each antenna k to each mobile terminal j at a spot frequency ω, and S_{j}(ω) is the frequency component ω of the information symbol stream S_{j }intended for mobile terminal j, then the transmit signals T_{k}(ω) that should be transmitted from antenna k in order to ensure that each mobile terminal 16 receives only its intended signal, while minimizing the total transmit power needed, are given by:
[T(ω)]=[C]^{#}·[[C]·[C]^{190}]^{−1}·[S(ω)] (Eq. 10)
[0100] where the symbol “#” for matrices of z−polynomials means the timereversed (endtoend reversed) conjugate transpose of the matrix. An example of the correct way to form the timereversed conjugate transpose of a matrix of z−polynomials is given below. Assuming an exemplary matrix,
[0101] it is desired to find the matrix that would have the conjugate frequency response and also be the transpose. The z−polynomial matrix with the conjugate frequency response is given as,
[0102] Now, to eliminate the negative powers of z in order to obtain polynomials in the same variable as the original matrix, it is necessary to multiply the whole matrix by the highest power of z in the entire matrix, in this case z^{3}. This operation results in,
[0103] where it may be seen that the polynomials have extra powers of z as compared with merely reversing the order of the original coefficients.
[0104] Returning focus to the spot frequency analysis, by evaluating C[C·C^{#}]^{−1 }at each spot frequency ω, a matrix of frequency response functions for transmit preprocessing is found. These frequency response functions can be converted to a set of impulse responses by finding the Fourier transform of the frequency response function if desired, so that the preprocessing of the signals S_{j }can take place in the time domain with FIR filters. However, it is possible that, in attempting to invert [C·C^{#}]^{−1 }at certain spot frequencies ω, it is found that its determinant is zero or near zero. This would signify one of two problem situations. There is either an unavoidable null in the frequency response of the channel from all antennas 14 to at least one of the mobile terminals 16, or that the propagation channels from each of the antennas 14 to two or more mobile terminals 16 are identical, causing the matrix to be rankdeficient.
[0105] As in the z−polynomial formulation, zeros in the determinant of a matrix may be avoided by omitting that factor, corresponding now to a zero root, from the determinant. More specifically, roots in the frequency domain that lie on or close to the imaginary axis represent “high−Q” poles and may be removed by multiplying the frequency response functions [C·C^{#}]^{−1}C^{#} with a zero term in the numerator having the same root. While this is in practice close to optimum, as in the timedomain formulation, this would result in all mobile terminals 16 experiencing the same net propagation channel, which may be nonoptimum.
[0106] Instead therefore, the following procedure below is used. When the spotfrequency inverse matrix has been computed (the adjoint matrix elements and the determinant can be kept separate to avoid dividing by a zero determinant), the net spotfrequency power attenuation factor from the N transmit antennas 14, which may be considered a diversity transmitter, to each mobile terminal 16 may be computed by summing the squares of the magnitudes of the adjoint matrix elements down columns. The squared magnitude of the determinant is then divided by the column sum to obtain a net power attenuation factor from the transmitting system to the mobile terminal 16.
[0107] Repeating these two operations for all spot frequencies gives the power attenuation frequency response function from the transmitting system (the set of transmit antennas 14) to each mobile terminal 16. Now the N=1 theory can be applied to determine how an information signal S_{i }for a given mobile terminal 16 should be prefiltered for transmission through the associated propagation channel having that determined net power/frequency response.
[0108] Conventional theory for N=1 gives the optimum transmit power spectral shaping P(ω) for transmitting information through a propagation channel with a given power spectral shape H(ω) as “the waterfilling solution”. As noted earlier, the waterfilling solution considers pouring water onto an object of shape 1/H(ω) (with walls at the band edges) until the total amount of “water” contained is equal to the total power available. The function P(ω) then equals 1−H_{0}/H(ω) where H_{0 }is the reciprocal of the level of the water.
[0109] No power is transmitted in regions of the band still not submerged, i.e. where H(ω)<H_{0}. The solution thus depends on the total power available relative to the receiver noise and the mean path attenuation, i.e. on the mean signaltonoise ratio (SNR) that can be created at a given mobile terminal 16. Given a desired data rate and a channel bandwidth, the mean SNR needed to sustain that data rate can be determined, and the waterpouring solution for that SNR used.
[0110] The waterpouring solution results in zero power transmission in regions of the spectrum where the SNR is less than a threshold. It is complicated to realize a transmit prefilter with abrupt cutoff at the boundaries of such regions, and these prefilters moreover have long impulse responses, requiring complex equalizers in the receivers of mobile terminals 16. A compromise filter that centers single zeros of the form 1+az^{−1 }on the nullregions can be used instead, which is nearly equivalent to the aboveproposed timedomain solution of deleting determinant polynomial roots close to the unit circle. The difference is that, in the justdescribed approach, the position “a” of the zeros may be adjusted slightly away from the true determinant zeros in order to flatten the net frequency response from the transmitting system to each mobile terminal 16 separately, taking account therefore of the effect of each column of adjoint matrix elements, thereby satisfying the intuitive expectation that the net channels should not turn out to be identical for each mobile terminal 16.
[0111] The frequencydomain solution and the timedomain solution are in fact identical as the matrix C(ω)^{#}[C(ω).C(ω)^{#}]^{−1 }is just the matrix C(z^{−1})^{#}[C(z^{−1}).C(z^{−1})^{#}]^{−1 }with z^{−1 }set equal to exp(−jωt), where the symbol # for matrices of z−polynomials means time reversed conjugate transpose as discussed in detail above. Therefore the procedure can simply comprise determining the matrix C of z−polynomials as before, and then forming the matrix of doublelength (2L) z−polynomials: [C(z^{−1})·C(z^{−1})^{#}] where # means timereversed, conjugate transpose matrix.
[0112] The above matrix is inverted as before by computing the adjoint matrix of z−polynomials of length 2L(N−1) and the determinant z−polynomial (now of length 2LN), and then multiplying the adjoint matrix by C(z^{−1})^{#} to obtain a matrix of length (2N−1)L polynomials. The attenuation/frequency curve from the transmitter to each mobile terminal 16 can now be determined by letting z=exp(jωt) and summing down matrix columns the squares of the modules of the elements, and repeating for each spot frequency ω. The square of the determinant modulus is then divided by the row sums. For each mobile terminal 16, the resulting attenuation/frequency curve may be examined to locate the highest L attenuation peaks. FIR zeros of the form 1+az^{−1 }are then placed on the attenuation peaks to either flatten the frequency response, which is close to optimum, or to overflatten the response by turning the peaks into troughs, which is even closer to the optimum waterpouring solution.
[0113] An approximate ruleofthumb for overflattening would be to choose a zero to turn a peak originally×decibels above the mean to a trough×decibels below the mean. The use of no more than L zeros ensures that the mobile terminals 16 do not need to equalize more than L symbol periods of net time dispersion. The result of applying the zeros to the signal will not however flatten the transmitterreceiver frequency response, as that is flat already when the zeros are not used; rather, it is the curve of transmit power versus frequency which is flattened across the spectrum, resulting in notches in the frequency response from the transmitting system (e.g., transmit antennas 14) to the mobile terminal 16. These notches appear to the mobile terminal 16 to have been caused by multipath propagation with a channel equal to the product of the applied “flattening” zeros.
[0114] Implementation of the above frequencybased waterfilling solution is similar to that illustrated in FIG. 2. Indeed, there are substantially no differences in implementation save that the IIR filters 30 (e.g., 30A . . . 30N) include FIR filter components composed of the abovementioned flattening zeros, if they do not exactly annihilate factors of the determinant. Further, the FIR matrix polynomials and the IIR determinant polynomials are different owing to the different coefficient values arising from adoption of the spot frequency approach.
[0115] Note that if a determinant polynomial has a pole exactly on the unit circle, giving an infinite attenuation peak, then the pole would be annihilated by a flattening zero also exactly on the unit circle, i.e. a=1. Likewise poles close to the unit circle giving high attenuation peaks would be flattened by placing a zero nearby. If the zero was placed exactly on the pole, thereby annihilating it, the solution is identical to deleting the L poles of the determinant closest to the unit circle.
[0116] In accordance with the above waterfilling discussion, which in part adopted a frequencybased approach to polynomial determination, the flow of FIG. 3 may be modified if it is desired to perform optimum transmit spectral shaping for each mobile terminal 16. FIG. 4 illustrates this waterfilling approach.
[0117] Processing begins with updating the CSI (block 220). Once the CSI values are updated, the transmit processor 18 determines whether the network 10 is transmitting to a greater, lesser or equal number N of mobile terminals 16 compared to the number N of antennas 14 at its disposal (block 222). This test entails an evaluation of the number of mobile terminals 16 as compared to the number of transmit antennas 14 for a commonly assigned communications channel.
[0118] In some implementations, the network 10 and particularly the transmit processor 18 are not configured for transmit preprocessing when the M>N. In these implementations, M>N constitutes an error condition. Error recovery procedures (block 224) can for example comprise selecting to serve only the first N receivers and not the remaining M−N.
[0119] If N=M, the matrix C is square and it is effectively inverted by computing the adjoint matrix of polynomials of order (N−1)L and the determinant polynomial of order NL (block 226). If M<N, then, as before, the extra degrees of freedom are used to optimize performance when computing the matrix C·C^{#} of polynomials of length 2L (block 232). Then the adjoint matrix of this matrix is computed, as is its determinant (block 232). The adjoint matrix polynomials are of length 2L(N−1) and the determinant polynomial is of length 2NL.
[0120] Rather than perform the root determination and processing as in FIG. 3 (i.e., blocks 208210), the net power attenuation as a function of frequency is computed for each mobile terminal 16 (block 228). The determined frequency response is then flattened by adding FIR zeros to each signal processing path in the transmit processor 18. More particularly, FIR zeros are added to the signal processing path of input IIR filters 30, which are applied to the input information symbol streams representing the desired information signals for the associated set of mobile terminals 16. Thus, in the flow of FIG. 4, the root determination operations of FIG. 3 (blocks 208210) are replaced by net power attenuation determination (blocks 228230).
[0121] The best achievable net attenuation versus frequency response from the transmitting network 10 to each mobile terminal 16 is computed (block 228). It is then desired to warp the transmit spectrum for each mobile terminal 16 to direct more power to regions of the spectrum with more favorable attenuation and that would use less transmitter power and direct less power to regions of the spectrum that suffer greater attenuation that would use more transmitter power. The waterpouring algorithm explained earlier is optimum and thus could be used here (blocks 228 and 230) if not considered too complicated.
[0122] Alternatively, the simpler approach of adding flattening or over flattening zeros can be used. In the case of high−Q determinant poles, i.e. roots very close to the unit circle, a flattening zero may be placed exactly over the pole to annihilate it. Instead of adding a zero then, a pole is annihilated from the determinant instead. On the other hand if one of the L poles closest to the unit circle is a low−Q pole, the attenuation frequency response may not show a peak exactly on the pole frequency but will be displaced due to the influence of the adjoint matrix FIR polynomials. In that case a zero is centered on the displaced peak and does not annihilate the nearby pole.
[0123]FIG. 5 illustrates one example of the inventive method applied in a three transmitterthree receiver scenario, where the transmitting antennas 14 are relatively far apart. The three base stations 12, for simplicity, have been sited at the corners of an equilateral triangle with 10 Km sides. The three transmit sites (e.g., base stations 12 and corresponding transmit antennas 14) are jointly supporting communications with multiple mobile terminals 16, of which three are located at positions m_{1}, m_{2 }and m_{3 }as shown. It is desired to use the same communications channel (e.g., frequency time slot), simultaneously if possible, for all three mobile terminals 16 in order to increase capacity according to this invention. In this example, the base stations 12 are assumed to transmit at an exemplary symbol rate of 13 MHz/48, which is the symbol rate of a standard GSM/EDGE cellular system. Dividing the symbol rate by the speed of light yields the spatial spread, which is calculated at 1107 meters. The distances between the mobile terminals 16 and the base stations 12 expressed in terms of symbol wavelengths is shown in Table 1 below:
[0124] where TX1 . . . TX3 correspond to base stations 12A . . . 12C and RX1 . . . RX3 correspond to mobile terminals 16A . . . 16C, as illustrated in FIG. 1.
[0125] If the channels for each of the nine paths involved in this example are of equal amplitude and phase and are all “line of sight” with no multipaths, the zmatrix relating received signals R_{1}, R_{2}, and R_{3 }to transmitted signals T_{1}, T_{2}, and T_{3 }would thus be:
[0126] In Eq. 14, the common fractional symbol delay parts of each row and a common integer power of z have been transferred to the left into R, such that the powers of z remaining are simplified to the lowest integers, representing whole symbol periods of delay.
[0127] Maintaining the simplifying assumption of equal phase and amplitude on all nine paths, the adjoint of this matrix is:
[0128] and the determinant polynomial is −1+z^{−3}+z^{−4}−z^{−7}=−(1−z^{−4})(1−Z^{−3}). The determinant has all seven roots on the unit circle at:
[0129] z=1 (two roots)
[0130] z=−1
[0131] z=j
[0132] z=−j
[0133] z=exp(j120°)
[0134] z=exp(j240°)
[0135] Each root represents a frequency at which infinite attenuation can arise between the transmitting system and the mobile terminal 16, so it is inefficient to attempt to convey energy at those frequencies to the mobile terminals 16.
[0136] To avoid this problem, all seven roots on the unit circle are optionally annihilated by zeros in the numerator, which is the same as deleting the roots of the determinant. Annihilating all seven roots could cause the equalizers in the mobile terminals 16 to have to deal with an effective channel length (delay) equal to seven symbol periods of delay—keeping in mind that not dividing by one or more factors in the determinant polynomial is the equivalent of multiplying in the numerator by those omitted factors. All the adjoint matrix elements however share at least one root with the determinant that can be annihilated. Canceling the factor −1+z^{−1 }from both the adjoint matrix and the determinant polynomial leaves:
[0137] The determinant is now being (1−z^{−4})(1+z^{−1}+z^{−2}). Not dividing by the 6th order reduced determinant means that the mobile terminals 16 will receive their signals modified by a 6th order FIR filter, and their equalizers must be able to deal with 7 symbol periods of delay.
[0138] The abovesimplified example gave rise to many determinant roots on the unit circle due to the assumption of equal amplitude paths. If the paths are not equal, but exhibit a single propagation path of amplitude and phase given by the complex numbers C_{jk}, then the channel estimate matrix becomes instead:
[0139] Each C_{jk }in the matrix C may comprise complex numbers that are changing due to Rayleigh fading. With multipath phenomenon between the transmit antennas 14 and the mobile terminals 16, the C_{jk }may be expressed as z−polynomials as in the earlier propagation path matrix examples. The additional powers of z^{−1 }that are now attached to the C_{jk }account for the extra propagation delay due to the substantial path differences that exist in the case of widely separated transmit sites.
[0140] As an example, the above C_{ij }were each chosen to be second order polynomials with three random Gaussian complex coefficients, representing multipath channels of three, symbolspaced rays. A typical frequency response plot of the determinant polynomial is shown in FIG. 6.
[0141] Other trials indicated the same trend, that there were in general only three or four attenuation peaks indicative of roots close to the unit circle. Therefore in general, only three or four roots need to be deleted from the determinant. The roots can be estimated approximately from the position and depth of the attenuation peaks, but it is better to use a root finding program.
[0142] A fast algorithm which iteratively finds all roots at the same time of a polynomial A(z) is given below. Since α(z)=(z−R1)(z−R2) . . . (z−R(i)) . . . (z−Rn), then
z−R(i)=A(z)/[product of all factors except (z−R(i))] (Eq. 18)
[0143] Substituting the current approximation for the root z=R′(i), obtain
dR(i)=A(R′(i))/[product of all factors (R′(i)−R′(j))except i=j] (Eq. 19)
[0144] The corrected root is then R′(i)−dR(i). It can be desirable to move the approximation only a fraction ALPHA towards the above value, where ALPHA starts small when the initial approximation is bad, and increases to unity. It is also desirable to continue the iteration for several cycles after convergence where the lease significant bits are undergoing a limit cycle due to rounding errors, remembering the solution for which the largest error was a minimum. A test for convergence is that the largest error, i.e., the largest value of A(R′(i)) over all roots (which should all ideally be zero) reduces from cycle to cycle. If it increases, then the limit cycle due to rounding error has been reached.
[0145] Repeating the above sequentially for i=1 to n and then iterating, exhibits rapid convergence from substantially arbitrary starting points to the desired roots and “polishes” their final values to high accuracy.
[0146] The FORTRAN subroutine below implements the algorithm.
[0147]FIG. 6 plots the determinant polynomial and the flattened polynomial obtained by deleting the four roots closest to the unit circle. The coefficients of the example polynomial are given in Table 2 below:
[0148] The roots in Z of the above polynomial were found by the above computer analysis to be:
[0149] The four roots of magnitude closest to unity were determined by comparing the values of ABS(REAL(CLOG(ROOT(I)))), where the complex logarithm function “CLOG” returns a real part equal to the logmagnitude. Thus, the preceding expression returns the absolute value of the real portion equal to the logmagnitude of the complex value. The roots with the smallest absolute value of this logmagnitude are ROOT(5), ROOT(2), ROOT(11) and ROOT(10) and were deleted to produce the flattened curve of the reduced determinant.
[0150] The adjoint matrix polynomials before dividing by the reduced determinant have the nine frequency responses shown in FIG. 7A. When the reduced determinant is used as the denominator and the adjoint matrix polynomials are used in the numerator to obtain the ninefilter transmit preconditioning used in the transmit processor 18, the nine frequency responses of FIG. 7B are obtained. When combined with the actual propagation paths between transmit antennas 14 and mobile terminals 16, these responses contrive jointly to ensure that each mobile terminal 16 receives only its own intended signal through an effective channel comprised of the product of the four deleted determinant factors. Due to selecting the deleted roots to be those closest to the unit circle, this effective channel comprises four rays that are exactly symbol spaced and contain the maximum possible energy.
[0151] The nine frequency responses of FIG. 7B can also be combined in threes by adding their power responses to determine how much power in total is being used to transmit to each mobile terminal 16, as shown in FIG. 8C. The integral of the power spectral curves yields the total power used for transmitting the intended signals to each mobile terminal 16. These powers can be compared to the powers that would have been necessary to communicate the same total signal power to one mobile terminal 16 at a time from the best base station 12 with the same channel state to obtain the comparative gain or loss for the macrodiversity network 10. In the above case, the results were:
[0152] Mobile 1: +4.46 dB more power needed (a loss)
[0153] Mobile 2: +0.98 dB more power needed (a loss)
[0154] Mobile 3: +2.50 dB more power needed (a loss)
[0155] The net loss is less than appears, because the mobile terminals 16 that are relatively nearer their supporting base stations 12 receive the significantly increased power while those further away receive smaller power increases.
[0156] After carrying out many trials using random propagation channels each comprising the exemplary three, symbolspaced, delayed multipath rays, a histogram was obtained of the total transmitter power used compared to the transmitter power that would have been used for each base station 12 talking to one mobile terminal 16 independently, ignoring mutual interference. FIG. 8A shows the histograms of gain/loss for this threebythree system, with the different curves corresponding respectively to the deletion of one, two, three, four, or five poles from the determinant. FIG. 8B shows the gain/loss histogram for the case where the number of poles deleted was selected in each trial to provided maximum communications efficiency. The mean loss in the threebythree case is about 1.5 dB in this case, for which the benefit of a tripling in capacity is obtained. It will be shown below that overprovision of base station transmitters in relation to mobile terminals 16 can turn this net power loss into a net power gain.
[0157] An implementation of the present invention can comprise deleting only those determinant polynomial roots with absolute logmagnitude less than a threshold, and thus not always the maximum number that the various mobile terminal equalizers can handle, with the intention of approximating the performance of FIG. 8B for deleting the optimum number of roots. For example, the logmagnitude of a root of the determinant may be compared to a threshold to determine whether to delete the root. The logmagnitude of a root of the determinant polynomial indicates the rate of decay of the impulse response due to the associated root, in Nepers per symbol period. It could be decided that it is unnecessary to delete roots representing greater than 3 dB (0.35 Nepers) decay per symbol period, for example. If only three roots had logmagnitudes less than 0.35, then only those need be deleted.
[0158] Other strategies for slating determinant poles for deletion other than selecting the poles nearest to the unit circle can also be used. If for example the adjoint matrix polynomials in the same column all have a factor equal to or close to a determinant factor, then it is unnecessary to delete that determinant factor since it cancels with the same factor in the numerator. Likewise, when the numerators contain a zero close to an offending pole, they reduce the undesirable effect of the offending pole. A systematic approach to detecting and exploiting such occurrences to select the poles for deletion to obtain the best effect is desired. One such systematic approach is provided using Cauchy'"'"'s residue theorem. Cauchy'"'"'s residue theorem potentially provides a much faster method to determine the total power transmitted when a “white” signal input is filtered by one or more numerator (FIR) z−polynomials and a denominator (IIR) polynomial to obtain the transmit signals (e.g., the signals transmitted by antennas 14).
[0159] Oftentimes it is desired to compute the total power output from a filter driven at its input with a serially uncorrelated (white) sample or symbol stream. This is the same computation as determining the noise bandwidth of a receiver filter. The expression to be computed is:
[0160] If this value is needed in real time during signal processing for example, it is too cumbersome to compute the integral by summing many points across the frequency domain. When the filter is a discrete time filter with an FIR numerator polynomial N(z) and a denominator D(z) representing an IIR filter, the integral is the same with z set equal to EXP(−j2π.f/B). The variable of integration may be changed from f to z by using z=exp(−j2π.F/B), therefore df=−B dz/z 2j π. The limits of integration will now be from +π, to − , i.e. going around the unit circle backwards. To change to the forward direction, drop the minus sign obtaining the following contour integral:
[0161] Cauchy'"'"'s residue theorem says that such integrals may be computed as the sum of residues at the poles. However, attempts to apply Cauchy'"'"'s residue theorem to expressions containing such squared terms will fail to give the right answer, as the  ^{2 }function does not meet the requirement of analyticity that Cauchy'"'"'s theorem requires. However, the modulus function can be converted to an analytic form that is valid on the unit circle contour of integration as follows:
[0162] Assume N(z) and D(z) are polynomials with roots P and corresponding factors such as (z−P).
[0163] Therefore every root of the N(z) polynomial gives rise to a factor 1/z and every root of the denominator D(z) gives rise to a factor z on the numerator. If the order of the numerator is greater or equal to the order of the denominator, there would be net addition on the denominator of some power of z which would give rise to residues at z=0 in addition to those from D(z). If on the other hand the numerator order is at least one less than the denominator (one less because we already have dz/z), then there are no powers of z on the denominator to cause the need to consider additional residues.
[0164] With this assumption Np>Nz therefore, we can write the integral as,
[0165] The analytic form of the numerator, which is valid on the contour of integration, must be used rather than just staying with N(z)^{2}, as when the value of a pole is substituted to get the residue, the results are quite different. N(z) is always real and positive, while the analytic form is complex. The latter is necessary to ensure that the result of the integral turns out to be real and positive, which it would not always be if N(z)^{2 }were used.
[0166] Cauchy'"'"'s theorem now says that the value of the above integral is 2jπ times the sum of the residues at the poles within the contour of integration. Now either p_{i }or 1/p_{i}^{* }will lie within the unit circle. Therefore one residue arises from each pole, which is the result of substituting either z=p_{i}, or z=1/p_{i}^{* }in the integrand, after deleting the corresponding factor (which becomes zero) in the denominator. The result is,
[0167] assuming that we have switched p_{i }with 1/p_{i}^{* }where the former was outside the unit circle in order to obtain a set of p_{i}'"'"'s all within the unit circle.
[0168] This expression will always yield a real, positive result as required, and the values have been confirmed to be the same as obtained by laborious pointbypoint integration along the frequency axis. Moreover, by examining the amount contributed by each pole residue to the real part of the result, we can associate a corresponding proportion of the total spectral energy with each pole, which is useful in determining optimum filters in some signal processing problems. The above may be computed without finding the roots of the numerator. The product of the numerator factors (p_{k}−q_{i}) is simply N(p_{k}). The product of the factors (1/p_{k}−q_{i}^{*}) is N^{* }(1/p_{k}^{*}).
[0169] The method outlined above may be extended by those skilled in the art using the same principles to cover all practical cases. Thus, a version of Cauchy'"'"'s residue theorem is developed for integrating the modulus squared of a complex numerator frequency function divided by a complex denominator frequency function. In the case where the numerator is a z−polynomial at least one order less than the denominator z−polynomial, the result is a sum of terms, each corresponding to the contribution from a respective denominator pole. If the numerator is small at that pole, then the contribution (the residue at the pole) to the total power integral will be low. Therefore the systematic method desired above is to select for deletion poles that have the largest real part to their residues, including the numerator term when computing the residues.
[0170] It can be argued that the selection process should first select the pole having the largest real part to its residue for deletion, and then reevaluate the residues of the remaining poles having deleted the pole. This process should continue until the desired number of poles have been deleted. The remaining denominator order must be at least one higher than the numerator order. If the process shall be continued to denominators of lower order, then an extra residue appears due extra powers of z on the denominator.
[0171] The extra residue can be determined by making a partial fraction expansion to extract a term having that power of z on the denominator, the other term now having a numerator of lower order than the denominator. The extra residue is then equal to the coefficient of the highest power of z in the numerator above the extracted term. The extra term is needed to determine the power after deleting a pole causing the numerator order to equal or exceed the denominator order.
[0172] In the application to this invention, the numerator can comprise the sum of the moduli of all the adjoint matrix polynomials down a column, where the modulus of a z−polynomial means the product of itself with its timereversed conjugate. The residue at a pole therefore comprises a part from each adjoint matrix numerator polynomial, so only if all adjoint matrix polynomials in the same column are small at the denominator pole with that pole have a small residue, and thus avoid being slated for deletion.
[0173] In light of the above discussion, it is evident that maintaining current CSI is useful in generating the appropriate filter coefficients used by the transmit processor 18. This may entail periodically updating or modifying the channel estimate matrix C. Updating the CSI requires continuously updated knowledge of the downlink channel conditions between the various transmit antennas 14 and mobile terminals 16. Many different approaches are available for obtaining the needed downlink channel information, and some of these approaches are discussed below for illustration.
[0174] Possible approaches to obtaining CSI include but are not limited to these items:
[0175] (i) Using the same channel frequency for both the downlink and uplink (from the mobile terminals 16 to the base stations 12) channels alternately in quick succession, thus implementing a socalled timeduplex or pingpong system. Then the transmitting base stations 12 may assume that the downlink channels are the same as they measure on the uplink when decoding the signals received back from the mobile terminals 16.
[0176] (ii) Measuring downlink channelrelated information in the receivers at the mobile terminals 16, and then transmitting these measurements back to the network 10 with a small turnaround delay. For example, the Universal Mobile Telecommunications System (UMTS) Wideband CDMA system (WCDMA) has the ability to serve up to 200 voice users per frequency channel per cell, or a proportionally lower number of high bitrate users such as mobile webbrowsers.
[0177] Therefore, for mobile webbrowsers desirous of receiving a high instantaneous data rate, it is acceptable to use the whole capacity of a voice channel or more on the uplink to feedback CSIrelated data.
[0178] (iii) Determining relative mobile terminal position in a mobile satellite communications system, where the relative coupling from transmit antenna elements to mobile terminals 16 is almost static.
[0179] (iv) Implementing a wirelessinthelocalloop system for transmitting Internet or voice services wirelessly to the home, where the receive antenna is fixed.
[0180] (v) Implementing a mobile system wherein the mobile terminal 16 is likely to be stationary when high bitrate services are invoked.
[0181] The above approaches can all provide feedback of CSI, but the case of fastmoving mobile terminals 16 is the most challenging as the CSI changes rapidly, and lowdelay, highrate feedback of CSI is required. Some solutions for rapid feedback of changing CSI will now be described.
[0182] In a “loopback” approach, let C′ denote the current CSI assumed by the transmitting system (e.g., network 10), which is in error from the correct CSI C by an error matrix E so that in matrix equation form:
[C′]=[C]+[E], or conversely [C]=[C′]−[E]. (Eq. 24)
[0183] The transmitter (antenna T_{k}) transmits [C′]^{−1}P_{j}S_{j}, where P_{j }is the effective net channel for signal S_{j}. P_{j }is the factor by which prefilters 30A, 30B, . . . and 30N of FIG. 2 differ from the true determinant polynomial.
[0184] Reciever R receives
R(i)=[C]_{ij}[C′]_{kj}^{−1}P_{j}S_{j}, (Eq. 25)
[0185] where summation over the common index k is implied. The above expression reduces as follows:
=[C′−E]_{ik}[C]_{kj}^{−1}P_{j}S_{j}, (Eq. 26)
[0186]=P_{i}S_{i}−[E]_{ik}[C′]_{kj}^{−1}P_{j}S_{j}, (Eq. 27)
[0187] since [C′]_{ik}[C′]_{kj}^{−1}−d^{ij}^{1 }if i=j, else 0.
[0188] Thus a given mobile terminal 16 as receiver R correlates its received signal R(i) with known symbols embedded in the transmission S_{j }to receivers) (e.g., another of the mobile terminals 16), the error polynomial term [E]_{ik}[C′]_{kj}^{−1}P_{j }summed over index k will be obtained.
[0189] If all mobile terminals 16 do this for all j, including their own, and return the results to the network 10, the channel estimator 36 can compute E_{ij }and hence correct C′_{ij }towards the actual or changing C_{ij}, thereby tracking changes in the CSI. This is possible because the channel estimator 36 already knows or has access to the S_{j }it transmitted, the prefilters P_{j }it used, and the assumed CSI represented by C′_{ij}.
[0190] From these interference correlations, the channel estimator 36 deduces how its CSI must have been in error, and corrects it. Specifically, receiver R_{1 }reports the polynomials determined by correlation with shifts of respective known symbol patterns as follows:
[0191] This is a set of N equations for the N unknown polynomials E_{11}, E_{12}, E_{13 }. . . E_{IN}. Likewise, receiver R_{2 }reports,
[0192] and this is a set of N equations for the N unknown polynomials E_{21}, E_{22}, E_{23 }. . . E_{2N}. Similarly, receiver R_{N }reports
X_{N1}(z)=−sigma(E_{Nk}C^{−1}_{kl})P_{1 }
X_{N2}(z)=−sigma(E_{Nk}C^{−}_{k2})P_{2 }
X_{NN}(z)=P_{N}−sigma(E_{Nk}C^{−1}_{kN})P_{N}. (Eq. 30)
[0193] which represents a set of equations for E_{N1}, E_{N2 }. . . E_{NN}.
[0194] The solution of each of such sets of equations for one row of [E] is [C][P^{−1}]·X, where [P^{−1}] is a diagonal matrix of the reciprocals of the prefilters used in the numerical processor 20. If the reported measurements X were exact, the X polynomials would contain P as a factor, which would cancel. The remaining factors would give a solution for E that was entirely FIR, i.e. no denominator polynomials, as required.
[0195] However, due to noise, the reported X polynomials probably will not have this exact property. A solution is to find the pure FIR solution of order L for E that best matches the frequency responses given by Equation 25 for E. For example, denominator roots from P can be paired with the closest numerator roots from C or X for annihilation until only numerator roots remain. These then yield the “best” pure FIR solution for E.
[0196]FIG. 9 shows another method for providing channel state feedback from the mobile terminals 16 to the network 10. Here the transmit processor 18 additionally performs mobile terminal feedback correlation operations. For simplicity, only two base station/antenna sites are depicted (i.e., 12A/14A and 12B/14B). As before, the mobile terminal 16 receives transmit signals T_{1 }and T_{2 }from the transmit antennas 14A and 14B, respectively.
[0197] In an exemplary, simplified arrangement, the mobile terminal 16 comprises a transmit/receive antenna 101 coupled via a duplexer 102 to receive circuits 103 which filter, amplify and convert the composite received signal to signal samples, preferably in digital form, i.e. using an AtoD converter. The signal samples from the receiver 103 are then added in summer 105 to a pilot code and fed to transmitter circuits 104. The transmitter circuits 104 convert the signal samples to an continuous signal using a DtoA converter for digital samples, and the continuous signal is upconverted to a transmit frequency, amplified to an appropriate transmit power level, and transmitted via antenna 101 back to the transmitting base stations 12.
[0198] The base stations 12 receive the transmitted loopback signal from various mobile terminals 16. The loopback signals from different mobile terminals 16 may be separated by interference rejection combining of the signals from the different sites in the channel estimator 36. Also, the channel estimator 36 computes correlations between the received signals and the pilot code inserted by a mobile terminal 16 to determine the uplink channel. Correlations are also computed between the received signal from a mobile terminal 16 and the signals the network 10 transmitted from each of its sites in order to determine the total loopback channel, which is the product of the downlink and uplink channels.
[0199] The uplink channel is then divided out to reveal the downlink channel. If necessary, the network sites (e.g., BTSs 12) can also each add a different, lowlevel pilot code to their transmissions, which would be chosen to assist in this loopback channel determination. Using this method, the mobile terminals 16 are relieved of the complexity of performing channel determination.
[0200] Generally, it is desirable to simplify mobile terminals 16 due to their high production volumes, and place complexity instead in the networks 10, which are much less numerous. Thus a simplified method by which the mobile terminals 16 can feedback downlink channel information to the transmitting network 10 would be useful. For example, the signal received at each mobile terminal 16 could be simply turned around and retransmitted with minimum delay back to the network, as shown already in FIG. 9.
[0201] If all mobile terminals 16 do so on the same channel, the network 10 must separate them by uplink beamforming/interference cancellation, which implies knowledge of uplink CSI. Uplink CSI is also needed to divide out the effect of the uplink channel polynomials on the retransmitted signal so that it reflects only the effect of the downlink channel.
[0202] In a CDMA system, the mobile terminals 16 can retransmit the signal received on the downlink with the addition of an uncorrelated pilot code sequence that the channel estimator 36 can use to derive uplink CSI. In a nonCDMA system that would not tolerate an overlapping pilot sequence, the feedback signal can instead be periodically interrupted at known times to insert pilot symbols that the network 10 can use to derive uplink CSI. Thereby the onus for analyzing what the mobile terminals 16 have received is placed back on the network 10.
[0203] The network 10 has the great advantage of knowing every symbol that was transmitted to every mobile terminal 16 and what prefilters (e.g., 30 and 32) were used in the generation of all the transmit signals. The network 10 can therefore perform correlations using the entire symbol sequence, or a portion thereof, transmitted to each mobile terminal 16, including data symbols and not just known pilot symbols.
[0204] Thus a modification to FIG. 9 can comprise interrupting the loopback signal to insert pilot symbols, replacing the additive combination of pilot and loopback signals formed by adder 105. In general, any suitable combination of the loopback signals with mobilespecific pilot symbols or mobilediscriminating information can be used.
[0205] Many variations of the above principle of “mirror reflection” of the received signals back to the network 10 can be devised. For example in a CDMA system, the received signal at each mobile terminal 16 can be despread using the codes of each mobile terminal 16 to obtain despread symbols. Then, the despread symbols can be respread using corresponding uplink codes and added. The multicode uplink signal is then mirrored from each mobile terminal 16 to the network 10.
[0206] Interference correlations (the X polynomials in the above notation—see 2426 for example) can also be digitally coded of course, and transmitted as a data stream protected by error correction coding. For high symbol rates giving long channel polynomials (large L) or for large N (e.g., greater than 3) the amount of digital information to be transmitted may exceed the uplink capacity available, presumed to be for example the capacity of one voice channel, or about 4 to 12 kilobits per second.
[0207] The information to be sent to the network 10 could be selectively reduced by including in the reports only the X polynomial or polynomials having the greatest coefficient magnitudes, including only polynomial coefficients that had changed by more than a threshold amount from a predicted value, or by some other means of downselecting.
[0208] Reporting only the coefficient with the greatest magnitude will cause the network 10 to correct its transmitted signals to reduce only that largest interference component. However, if this action is repeated sequentially, it will reduce multiple interference components in order of strongest components first.
[0209] The mathematics given above include the derivation of the optimum transmit signals when the number of transmit antennas 14 exceeds the number of mobile terminals 16. With such overdimensioning, more degrees of freedom exist to find a way of overcoming fading on any particular path, and/or of canceling interference at each mobile terminal 16. This should translate to fewer problematic or pathological determinant roots close to the unit circle, and therefore to greater transmit power efficiency. To obtain the benefits of overdimensioning it is only necessary for the number of independently fading paths of the same delay to be greater than the number of mobile terminals 16.
[0210] For meeting this criterion, mean delay from an additional base station 12 can always be made the same as the delay from existing base stations 12 artificially, which occurs automatically when the abovedescribed mathematical operations are performed (e.g., Right Hand Circular and Left Hand Circular). However, this criterion can also be met by providing dual polarization transmitters at one or more base stations. Field measurements confirm that the propagation channel for a transmitted wave of one polarization reaching a mobile terminal 16 over a landmobile propagation path shows little correlation to the propagation channel of the other, orthogonally polarized wave. Thus if each base station 12 and antenna 14 are equipped to transmit power with both polarizations, each base station 12 counts as two independent base stations 12 as regards the mathematical formulation, even though the RHC and LHC transmitters are colocated.
[0211]FIG. 10 shows a gain/loss histogram computed from many trials of a 3base/3mobile system wherein each base station is equipped to transmit dual polarized signals. The channel estimate matrix is then a 3×6 matrix of polynomials, and the transmit processor 18 computes six waveforms to be transmitted from respective antennas 14 and with respective polarizations, based on the three information signals to be transmitted to respective mobile terminals 16.
[0212] The different curves again correspond to the deletion of one, two, three, four, or five of the “worst” determinant poles, and now show a net gain in total power efficiency of 4 to 4.5 dB in the mean. That means the sum of the power transmitted at all three sites and in both polarizations can be significantly less than needed in conventional systems not using the invention at the same time as allowing the same communication channel to be reused three times at each location. Moreover, the effect of using both polarizations to thereby overdimension the number of transmitters relative to the number of receivers is to virtually eliminate determinant poles close to the unit circle, so no determinant poles need be deleted. If no determinant poles are deleted, then the network 10 provides a constant net propagation channel for information signals from the network 10 to the mobile terminals 16, which is free of fading and multipath distortion.
[0213] Another aspect of transmit macrodiversity considers a degenerate case of the Nantenna, Mreceiver solution using two antennas 14, for example two colocated RHC and LHC transmitting antennas 14, to transmit a signal to a single mobile terminal 16. When the two antennas 14 are colocated and differ only by polarization, the channel coefficients for the multipath channels are substantially uncorrelated for the two polarizations; however, the delays of the fading, multipath rays are the same. That is, if the channel for RHC is expressible as,
C^{RHC}(z)−C_{0}^{RHC}+C_{1}^{RHC}z−^{P1}+C_{2}^{RHC}z^{−P2}. (Eq. 31)
[0214] then the LHC channel is expressible as,
C^{LHC}(z)−C_{0}^{LHC}+C_{1}^{LHC}z−^{P1}+C_{2}^{LHC}z^{−P2}. (Eq. 32)
[0215] The powers of z are the same for both polarizations, expressing equality of the multipath delays, but the coefficients are different and fade in an uncorrelated fashion.
[0216] The optimum way to transmit a wanted signal to the mobile terminal 16 is then given as in the earlier discussion of unwanted signal cancellation by C^{#}[C·C^{#]}^{−1 }times the wanted signal waveform, where C is now the 2×1 matrix of polynomials,
[0217] Thus ∩ C·C^{#}]^{−1}−1/(C^{RHC}·C^{#RHC}+C^{LHC}·C^{#LHC}) and there is no adjoint matrix. The implementation then depends only on the location of the poles of this denominator/determinant polynomial. It may be appreciated that, when P1 and P2 are large integers, this determinant has a prodigious number of roots even though it has a small number of coefficients.
[0218] The above channels, which have three significant multipaths of relative delays 0, P1 and P2 symbol periods, are typical of a CDMA channel. A known CDMA receiver, such as might be found in some implementation of mobile terminals 16, would conventionally employ a form of multipath equalizer known as a RAKE receiver.
[0219]FIG. 11 depicts an exemplary RAKE receiver 110 as might be employed within a mobile terminal 16. It comprises three RAKE fingers or “taps” 112A, 112B, and 112C, with each finger including a correlator 114 and channel compensator 116. The outputs from each finger 112 are coherently combined in the summing circuit 118.
[0220] While not essential to understanding the coherent transmit macrodiversity techniques of the present invention, it may be helpful to review the basic operation of the RAKE receiver 110. The received signal r(t) comprises the entirety of information streams for all mobile terminals 16 operating on the same communications channel as the given mobile terminal 16. By correlating r(t) with its uniquely assigned psuedonoise (PN) sequence, the receiver 110 recovers only its intended information stream from r(t).
[0221] Because r(t) generally comprises multipath components, each having a determinable delay or offset, the receiver 110 can timealign each of its fingers 112 to a given one of the multipath signals. One technique for synchronizing multipaths is to control the code offset of the PN code fed to each finger 112. While each finger 112 receives the same PN code sequence, the code offset of the sequence received by a given finger 112 is matched to the relative path delay of the multipath signal operated on by that finger 112. The differing code offsets are denoted as PN, PN′, and PN″ in the illustration and correspond to relative symbol delays of 0, P1, and P2, respectively.
[0222] Each multipath component of the received signal experiences channel distortions unique to its associated propagation path, so each finger includes a channel compensator 116 to compensate for path attenuation and phase characteristics of the propagation path. After channel compensation, the finger outputs are coherently combined in combiner 118 to produce an overall RAKE output signal. In this manner, the signaltonoiseratio (SNR) of the received signal may actually be improved with multipath since the final output signal is a combination of the RAKEreceived multipath components.
[0223] Thus, the RAKE taps 112 in the present example are placed at delays of 0, P1 and P2 symbols or chips. The RAKE receiver 110 is not limited so much by the amounts P1, P2 of delay, but in the total number of taps 112.
[0224] Conventionally, RAKE receivers do not include a substantial number of taps owing to the understandable desire to limit complexity and expense, as well as the “diminishing return” aspects of adding more than a basic complement of taps. A threetap RAKE receiver represents a common configuration and it is desirable that the coherent transmit macrodiversity techniques of the present invention be amenable to use with mobile terminals 16 that include such threetap RAKE receivers.
[0225] Recall from earlier discussion that the path delay requirements for receivers within the mobile terminals 16 is affected by the number of determinant polynomial roots used or omitted in the IIR filtering performed in the transmit processor (e.g., prefilters 30). In the extreme, omitting the whole of the above determinant polynomial would result in a net transmission channel having seven delays and needing in principle a seven tap RAKE receiver with tap delays of −P2, −P1, −(P2−P1), 0, (P2−P1), P1 and P2 chip periods.
[0226] If the alreadydescribed poledeletion procedure is used to improve the conditioning of the denominator for implementation, only two roots may be deleted while preserving only a 3tap channel for the RAKE receiver 110 to deal with. Moreover, the threetap channel would always comprise delays of exactly 0, 1 and 2 chips with this method. Thus the three RAKE taps 112 would be adjacent instead of at 0, P1 and P2 symbols delay. However, as noted, RAKE receivers 110 are generally limited only in the number of taps available to combine, and not in tap spacing. Therefore any threecoefficient numerator polynomial of the form
A_{0}+A_{1}_{z}^{Q1}+A_{2}_{z}^{Q2 } (Eq. 33)
[0227] could be used as an FIR prefilter 34 in the transmit processor 18, and would preferably share roots with the denominator that it is desired to cancel. Alternatively, a polynomial in which most of the energy was contained in three taps could be used.
[0228] A denominator polynomial of the form C_{1}(z)C_{1}^{#}(z)+C_{2}(z)C_{2}^{#}(z)+ . . . has the property that half of its roots are the conjugate reciprocals of the other half. Those roots having magnitude greater than unity (and which would therefore require timereverse IIR processing) may be slated for deletion. The remaining poles form a causal IIR filter (e.g., filters 32) that may be implemented entirely by forwardtime processing, and the deleted poles give rise to a net FIR channel.
[0229] This is one way of factoring the denominator polynomial into the form C_{equ}(z)C_{equ}^{#}(z), which factoring is always possible by selecting one root to form C_{equ }and its conjugate reciprocal to form C_{equ}^{#}. The choice of the larger or smaller root for C_{equ }can be made in 2^{N/2 }different ways, given N/2 pairs of roots, so the number of such factorizations that can be made is large.
[0230] Trial computations showed that factoring into a purely causal and a purely anticausal part gives a channel that, although not limited to the same number of taps as the original channels C_{1 }and C_{2}, nevertheless contains most of its energy in relatively few coefficients, allowing a RAKE receiver 110 with a small number of taps 112 to be used. When fewer taps are used than the number of multipaths containing signal energy, a loss of energy is entailed.
[0231]FIG. 12 plots the probability distribution function (PDF) of this energy loss for using one to six taps 112 to receive the signal through the abovementioned anticausal factor. The RHC and LHC channels were both of the form C_{0}+C_{1}_{z}^{5}+C_{2}_{z}^{13 }and thus the determinant is of order 26. The causal and anticausal factors are then both of order 13, but do not have nonzero zpowers of only 5 and 13. Nevertheless, 80% of the energy is most often contained in the tap delay 14, 93% is contained in tap delay 14 and tap delay 9 and 98% is most often contained in tap delays 14, 9 and 1. Thus, a RAKE receiver 110 with no greater than an ordinary number of taps 112 is suitable for use in a network 10 employing the coherent transmit macrodiversity techniques of the present invention.
[0232]FIG. 13 plots a histogram of the net gain in communications efficiency for this method, when the number of RAKE taps 112 is fixed at three, the same as the number of multipaths in the multipath channels. The gain of 2 dB equates to a 60% increase in communications capacity in terms of number of users or data rate when practicing coherent transmit macrodiversity in accordance with the present invention.
[0233]FIG. 14 shows another application of the invention to a CDMA system. The simplified network 10 comprises a matrix precombiner 300 and a spread spectrum coder 302. The matrix precombiner 300 is used in this implementation much like the earlier example based on the transmit processor 18.
[0234] In this example, only a single transmit antenna 14 of a single polarization is available to transmit to N mobile terminals 16. The signal to each mobile terminal 16 is spreadspectrum coded using a unique spreading code, and the set of spreading codes may be mutually orthogonal, at least when the codes are timealigned. When the codes are not timealigned, such codes will show some degree of correlation or mutual interference.
[0235] A conventional CDMA system lacks the precombiner 300 and simply comprises inputting symbol streams S_{1 }. . . S_{N }to spread spectrum encoder 302, where S_{i }is the coded information stream intended for a particular mobile terminal 16. Stream S_{i }is multiplied with spreading code PN_{i }and transmitted to mobile terminal 16 through a propagation channel described by downlink channel polynomial P_{i}(z).
[0236] The ith mobile terminal 16, denoted as mi, where i is an index, correlates the received signal with a first shift of code PN_{i }and obtains P_{0}(i)S_{i }where P_{0}(i) is the first coefficient of the propagation polynomial P_{i}(z) and likewise obtains P_{i}(i)S_{i }for other code shifts. These are then RAKE combined into a single sample stream whose dominant component is then, hopefully, Q_{0ii}S_{i }where Q_{0ii}−P_{0}^{2}+P_{1}^{2 }. . . P_{L−1}^{2 }and L is the length (e.g., number of paths or delay spread) of the downlink channel to mobile(i).
[0237] However, the combined RAKE output signal will contain other, undesirable terms.
[0238] There will be some intersymbol interference (ISI) from the previous and the next symbol of stream S_{i }due to correlation of the spreading code PN(i) shifted with itself. If the channel time dispersion is less than one information symbol in duration however, only one previous and one future symbol will contribute to the RAKE output. This an be accounted for by coefficients Q_{−1 }and Q_{+1 }of a 3coefficient z−polynomial Q_{ii}(z) and the combined RAKE output can be written as Q_{ii}(z)−S_{l}(z).
[0239] In addition, there are correlations between spreading code PN_{1 }of a first mobile terminal M_{1 }and code PN_{2 }of second mobile terminal M_{2 }shifted, which give terms Q_{ij}(z)S_{j}(z). The order of Q_{ij }may, as already seen in the above simplified example, be extended when signal S_{j }is received from a transmit antenna 14 lying at a greater or lesser distance from mobile terminal M_{1}, but not when all signals are received from the same transmit antenna 14.
[0240] Thus all Q_{ij }in this example have three coefficients. This matrix of z−polynomials describes the coupling from other mobile terminals'"'"' intended information symbols to a given mobile terminals'"'"' RAKE output from its RAKE receiver 110. Conventional networks 10 do not work to prevent these undesirable terms in each mobile terminal'"'"'s RAKE output. With regard to this aspect of the present invention, precompensation of the transmit signal within the matrix precombiner 301 operates to cancel the unwanted signals at each mobile terminal 16. Particularly, transmit preprocessing works to cancel ISI at each mobile terminal 16 that is caused by the signals intended for the other mobile terminals 16.
[0241] The matrix of z−polynomials used by the matrix precombiner 301 may be inverted as before by forming the adjoint matrix and the determinant polynomial. If the determinant polynomial has no unpleasant roots, then using the inverted matrix in the precombiner 300 cancels all interference at the RAKE receiver output of each mobile terminal 16 caused by both other mobile'"'"'s symbols and from that mobile'"'"'s own previous and future symbols.
[0242] If unpleasant roots of the determinant polynomial have to be deleted, and the remaining, reduced determinant used as the IIR part of precombiner 300, then all interference from other mobiles'"'"' signals will still be cancelled, but there will be a net multipath channel created from the product of the omitted determinant factors through which each mobile terminal 16 will receive its intended symbol stream.
[0243] This can be completely compensated at the mobile terminals 16 by following the RAKE combiner 18 in the mobile terminals'"'"' RAKE receivers 110 with a 3tap, Viterbi Maximum Likelihood Sequence Estimator (not shown) or other known form of symbolspaced equalizer running only at the information symbol rate and not at the spread spectrum chip rate.
[0244] Thus, FIG. 14 contemplates the addition of an informationrate matrix precombiner 300 to combine information symbol streams S_{1}, S_{2 }. . . S_{n }to form informationrate sample streams S_{1}, S_{2 }. . . S_{N}, which are input to the spread spectrum coder 302.
[0245] The precombiner 300 has a structure similar to that of the numerical processor 20 in the transmit processor 18, as shown in FIG. 2. The polynomials used in precombiner 300 are computed using knowledge both of the downlink channels P_{i}(z) and of the crosscorrelations between and the autocorrelations of the different spreading codes used in coder 302, and using the assumption that the mobile terminals 16 employ conventional RAKE combining with weighting coefficients adapted to the chiplevel channel impulse responses (i.e., the attenuation and phase characteristics of the multipath channels).
[0246] In the discussion immediately above, the matrix C included crosscorrelation interference between the different information signals being transmitted to the receivers, and may be thought of as a coupling or interference matrix Cij. FIG. 15 illustrates how a rake output sample depends typically on three adjacent symbols, due to delayed multipath rays. When the sample corresponding to symbol (i) is calculated by the rake receiver, it first correlates the received signal with the chip sequence used to spread symbol (i), with the chip sequence aligned in time with ray 1 to obtain a value U1, then with the chip sequence time aligned with ray 2, ray 3, ray 4 . . . to obtain values U2, U3, U4 . . . correspondingly. Correlation means multiplying each received chip sample by the complex conjugate of the spreading code symbol, and summing over the number of chips used to spread the symbol to obtain each Uvalue. The Uvalues are then weighted by the complex conjugates of the separately estimated channel coefficients C1, C2, C3, C4 . . . for rays 1 . . . 4, giving the rake receiver output R(i) as
R(i)=C1^{#}U1+C2^{#}U2+C3^{#}U3+C4^{#}U4 . . . (Eq. 34)
[0247] However, as may be seen from FIG. 16a, when correlating with the chip sequence timealigned with ray 1, there is an overlap with the end of the chip sequence for symbol (i−1) due to delayed rays 2, 3 and 4, so that there is a contribution to U1 from symbol (i−1). This contribution is
S(i−1)[C2.p2+C3.p3+C4.p4 . . . ] (Eq. 35)
[0248] where p2 is the partial correlation between the lefthand end of the chip sequence for symbol i and the overlap with the righthand end of the chip sequence for symbol (i−1) due to delayed ray 2; p3 is the partial correlation of the chip sequence for symbol i for the slightly greater overlap with the end of the chip sequence for symbol (i−1) due to delayed ray 3, and so on.
[0249] Likewise, U2 has a contribution from S(i−1) due to delayed rays 3 and 4 of
S(i−1)[C3.p2+C4.p3 . . . ] (Eq. 36)
[0250] and U3 has a contribution from S(i−1) due to delayed ray 4 of
S(i−1)[C4.p2 . . . ] (Eq. 37)
[0251] Thus the total contribution to R(i) from S(i−1) is
C1^{#}.∩ C2.p2+C3.p3+C4.p4 . . . ]+C2^{#}.[C3.p2+C4.p3 . . . ]+C3^{#}.[C4.p2 . . . ] (Eq. 38)
[0252] which is therefore the first coefficient C11^{−1 }of the channel polynomial C11 that describes interference between a mobile symbol and the preceding symbol for the same mobile, i.e. ISI at the output of the rake receiver. Likewise, C11^{+1 }describes ISI from the succeeding symbol S(i+1) and is given by
C4^{#}.[C3.q3+C2.q2+C1.q1 . . . ]+C3^{#}.[C2.q3+C1.q2. . . ]+C2^{#}.[C1.q3 . . . ] (Eq. 39)
[0253] where now q3 is the partial correlation between the overlap of the righthand end of the code sequence for symbol (i) with the lefthand end of the code sequence for symbol (i+1). The middle coefficient C11° of polynomial C11 is given by,
[0254] and the value of w1 is unity, as it is the correlation of a code with itself fully aligned.
[0255] The above exemplary expressions are for the case of four delayed rays, but may be extended to any number of rays in an analogous manner. Likewise, if ray delays of greater than one symbol period are encountered, the channel polynomials such as C11 may lengthen to 5 coefficients to incorporate dependence of the rake output sample on S(i−2) and S(i+2).
[0256] The entire matrix of polynomials Cij is defined in the same manner as above, except that Cij means the dependence of a sample from rake receiver (i) (which uses mobile (i)'"'"'s code, code (i), for correlation) on the symbols of mobile (j) which are spread with code (j). The values of the partial correlations above are then between overlapping segments of different codes, and in particular, the value of w1 is no longer unity, but the correlation between a symbollength segment of code (i) and a fully aligned segment of code (j), which is zero when codes (i) and (j) are orthogonal.
[0257] If the multipath delay spread is less than an information symbol in duration, each sample stream will depend only on current information symbols, one symbol period delayed symbols and one symbol period advanced symbols, making each Z−polynomial of order three.
[0258]FIG. 15 is for the case of transmitting from one base station to ‘N’ mobile stations. The matrix C is then square and of size N×N.
[0259] For the case of transmitting from N diversity base stations to M mobile stations, the total number of inputs to the N base station transmitters would be MN, so the corresponding interference or coupling matrix C, which describes the coupling from each transmitter input to each mobile rake receiver output, would be of dimension MN×N. Since this is an overdimensioned case, the optimum transmit conditioning is then given by,
T=C^{#}[C.C^{#}]^{−1}S (Eq. 41)
[0260] where C is the abovedefined matrix of MN×N coupling polynomials, S is a vector of information symbol sequences intended for the receivers, and T is a vector of as yet unspread sample sequences for transmission. Each of the unspread sample sequences comprises a sequence of samples at the information symbol rate, which samples however are not necessarily of the same values as information symbols, but are multivalued samples which are additive combinations of many symbols. Each sample, being of symbolperiod duration, is then multiplied by the spreading code segment for that symbol corresponding to the transmitter input to which it is applied.
[0261] The results of all inputs of the same transmitter are added and transmitted as a composite spreadspectrum (CDMA) signal. When all N transmitters collaborate in this way, the rake output of receiver (i) contains the intended information symbol stream S(i) only, with neither ISI nor interference from symbol streams intended for other receivers. If instead of [C.C^{#]}^{adj}, [C.C^{#]}^{adj }is used instead, the rake output will be free of interference from nonintended symbol streams, but will contain ISI. The ISI delay spread will depend on whether all of the determinant of C.C^{#} is omitted from the inverse, or whether only illconditioned poles are omitted. The advantage of the overdimensioned case is that all the poles of the determinant lie further from the unit circle, and therefore are more wellconditioned.
[0262] The present invention may, of course, be carried out in other specific ways than those herein set forth without departing from the spirit and essential characteristics of the invention. The present embodiments are, therefore, to be considered in all respects as illustrative and not restrictive, and all changes coming within the meaning and equivalency range of the appended claims are intended to be embraced therein.