JP2004135099A - Multi-user receiver, radio base station device having the same and mobile machine - Google Patents

Abstract

<P>PROBLEM TO BE SOLVED: To provide a multi-user receiver in a DS-CDMA system with a small scale circuit. <P>SOLUTION: A reception signal Y(t) becomes matching filter output by diffusion signals f<SB>1</SB>(t) to f<SB>L</SB>(t) created by an extended diffusion signal creation part 12. A result obtained by linearly coupling a plurality (K-pieces) of outputs y<SB>1</SB>to y<SB>K</SB>in them and K-pieces of weight coefficients x<SB>1</SB>to x<SB>K</SB>created in a weight coefficient creation part 9 from a deformed diffusion signal created in the extended diffusion signal creation part 12 in a linear coupling part 6 becomes output whose interference for one bit is suppressed. The weight coefficients x<SB>1</SB>to x<SB>K</SB>can be obtained as elements of an inverse matrix of correlation coefficient matrix being KxK matrix between the diffusion signals. The weight coefficient creation part in a structure of a diagram 37 realizes a process for obtaining the correlation coefficient matrix and the inverse matrix by one process. Consequently, a circuit scale proportional to a cube of K in a conventional case is reduced to that proportional to a square of K. <P>COPYRIGHT: (C)2004,JPO

Description

【００３８】
ここで，
【００３９】
【数２】

【００４０】
と置くと（ｊ＝１〜Ｋ），ｙ_ｊ，　ｚ_ｊおよびａ_ｉｊは，それぞれ，整合フィルタ（ｍａｔｃｈｅｄ　ｆｉｌｔｅｒ）の出力信号，整合フィルタ出力に含まれる雑音成分および相関係数を表す。
【００４１】
ここで，積分区間を−∞から＋∞としているが，拡散信号をバースト信号として取り扱うので，積分は，そのバースト信号のある，有限の区間で行われる。
【００４２】
式（１）と式（２）から，整合フィルタの出力信号ｙ_ｊは以下の式（３）により表される。
【００４３】
【数３】

【００４４】
式（３）を行列で表せば，以下の式（４）となる。
【００４５】
【数４】

【００４６】
式（４）をさらにベクトルで表記すると，以下の式（５）となる。
【００４７】
【数５】

【００４８】
ここに，ベクトルｙ，Ｒ，Ｗ，ｄ，およびｚはそれぞれ以下の通りである。
【００４９】
【数６】

【００５０】
行列Ｒは相関係数行列と呼ばれ，Ｋ×Ｋ行列である。また，ａ_ｉｊ＝ａ_ｊｉであるので，行列Ｒは対称行列である。
【００５１】
ここで，希望信号を取り出す手法としては，デコリレータ（Ｄｅｃｏｒｒｅｌａｔｏｒ）または復相関検出器（Ｄｅｃｏｒｒｅｌａｔｉｎｇ　Ｄｅｔｅｃｔｏｒ）のように，判定前の信号に含まれる干渉信号成分を零にする第１の方法と，ＭＭＳＥ検出器（Ｍｉｎｉｍｕｍ　Ｍｅａｎ−ＳｑｕａｒｅＥｒｒｏｒ　Ｄｅｔｅｃｔｏｒ）のように，背景雑音も考慮して，干渉雑音および背景雑音の電力の和を最小にする第２の方法とがある。以下，これら第１および第２の方法の各原理について説明する。
【００５２】
（１）第１の手法
第１の方法は干渉雑音を零にする方法である。干渉雑音を零にするためには，以下の式（６）に示すように，上記式（４）（または式（５））の両辺に相関係数行列Ｒの逆行列Ｒ^−１を乗算すればよい。これにより，右辺の第２項で表される背景雑音を含んだ，希望信号Ｗ_ｉｄ_ｉが得られる。
【００５３】
【数７】

【００５４】
（２）第２の手法
第１の方法では，干渉信号成分を零にするが，背景雑音がある場合には，その成分を増大させてしまう雑音増大（Ｎｏｉｓｅ　　Ｅｎｈａｎｃｅｍｅｎｔ）と呼ばれる現象がある。第２の方法は，上記の欠点を補うために，干渉雑音および背景雑音の合計を最小にする方法で，相関行列の対角要素に，背景雑音の電力密度σ^２とビット当たりの干渉信号電力Ｗ_ｋ ^２の比σ^２Ｗ_ｋ ^−２を加えた行列Ｒ＋σ^２Ｗ^−２（＝Ｒ´とする）を行列Ｒに代わって用いればよいことが知られている。すなわち，式（６）の代わりに以下の式（７）を用いるのが第２の方法（ＭＭＳＥ検出法）となる。
【００５５】
【数８】

【００５６】
以上は，復相関検出器もＭＭＳＥ検出機も，Ｋ×Ｋ行列の逆行列を求め，それを整合フィルタ出力に乗じることによって，Ｋビットのデータを全て復号できることを示している。
【００５７】
（３）希望信号の１ビットの復号に必要なＫ個の要素
相関検出器もＭＭＳＥ検出器も，Ｋ×Ｋ行列の逆行列を求め，それを整合フィルタの出力に乗じることで，干渉信号成分を抑圧した，Ｋビットのデータを復号するが，希望信号の１ビット（希望ビット）の復号に関与するのは，前記逆行列の特定の１行（または１列）である。従って，復号したいビットが，前記Ｋビットの全部ではなく，ある特定のビットだけの場合（移動機の場合，これが当てはまる）には，逆行列を求める代わりに，逆行列の１行（または，１列）を求める方が便利である。ある行（または列）に対応するＫ個の要素は，Ｋ元連立方程式の解として得られることは，以下の様に示される。
【００５８】
先ず，一般性を失わないので，ｋ＝１を希望信号ビットとすると，逆行列の第１行が，このビットの復号に関与する行（または列）となる。式（６）の第１行表すと，以下の式（８）となる。
【００５９】
【数９】

【００６０】
ここに，ｒ_ｉｊは，以下の式（９）で示される行列Ｒ^−１の要素である。
【００６１】
【数１０】

【００６２】
式（９）の左辺に現れる逆行列Ｒ^−１の第１行の要素ｒ_１１〜ｒ_１Ｋは，以下の式（１０）に示すように，相関係数行列Ｒとその逆行列Ｒ^−１との積が単位行列Ｅとなることから，Ｋ元連立方程式（１１）の解となることが分かる。
【００６３】
【数１１】

【００６４】
【数１２】

【００６５】
この式は，復相関検出器に関するものであるが，ＭＭＳＥ検出器とする場合は，式（７）に示す様に相関行列の対角要素に，σ^２Ｗ_ｋ ^−２を加えればよいことを考慮して，式（１１）を両検出器を含む形に拡張して表記すると，下記の式（１２）となる。
【００６６】
【数１３】

【００６７】
ここに，
【００６８】
【数１４】

【００６９】
今後，式（１３）で定義したα_ｋ（ｋ＝１，２，…，Ｋ）をＭＭＳＥ補正値，行列Ｒ＋σ^２Ｗ^−２を変形相関行列Ｒ´と呼び，ＲおよびＲ´の逆行列を，行列Ｒ^＋と略記することにする。
【００７０】
＜無線基地局装置の構成および動作＞
先ずは従来技術に基づく無線基地局装置の構成および動作について説明し，次に，本発明を同装置に適用した場合の，実施の形態を，動作原理と併せて説明する。
【００７１】
図５は，無線基地局装置１２０として，特にマルチユーザ受信部の構成を示すブロック図で，アンテナ１，周波数変換部２，整合フィルタバンク３，拡散信号生成部４，相関係数生成部１０，前述したＭＭＳＥ検出法を適用するためのＭＭＳＥ補正部８，Ｋ×Ｋ行列の逆行列を演算する逆行列演算部１４，線形結合部６および判定回路７から構成される。
【００７２】
移動機２００からの無線信号（ＲＦ信号）は，アンテナ１により受信され，周波数変換部２に与えられる。周波数変換部２は，アンテナ１からの無線信号を増幅する増幅器，周波数変換を行う乗算器（ミキサ），ＲＦ信号に同期した発振器から成り，入力されたＲＦ帯域の信号をベースバンド信号に変換する。このベースバンド信号が，前述した式（１）により表される受信信号Ｙ（ｔ）である。この信号Ｙ（ｔ）は，整合フィルタバンク３，拡散信号生成部４およびＭＭＳＥ補正部に入力される。
【００７３】
拡散信号生成部４（詳細は，後述する）は，受信信号から，複数（Ｌ個とする）の拡散信号ｆ_１（ｔ）〜ｆ_Ｌ（ｔ）と，それらに対して，すべて一定時間シフトしているＬ個の拡散信号ｇ_１（ｔ）〜ｇ_Ｌ（ｔ）を生成する。そして，生成した拡散信号ｆ_１（ｔ）〜ｆ_Ｌ（ｔ）を整合フィルタバンク３に与えるとともに，拡散信号ｇ_１（ｔ）〜ｇ_Ｌ（ｔ）を相関係数生成部１０に与える。
【００７４】
整合フィルタバンク３（詳細は，後述する）は，ベースバンド信号Ｙ（ｔ）を，Ｌ個の拡散信号ｆ_１（ｔ）〜ｆ_Ｌ（ｔ）によりそれぞれ逆拡散した後，逆拡散した信号を整合フィルタによりフィルタリングする。フィルタリング後の信号は，各信号ビットが整合フィルタに入力し終わるタイミングでサンプリングされ，前述した式（３）に示す，Ｋ個の信号ｙ_１〜ｙ_Ｋとなり，線形結合部６に与えられる。
【００７５】
相関係数生成部１０は，入力したＬ個の拡散信号ｇ_１（ｔ）〜ｇ_Ｌ（ｔ）、または、Ｋ個の信号ｙ_１〜ｙ_Ｋに対応する部分である，拡散信号ｓ_１（ｔ）〜ｓ_Ｋ（ｔ）から，Ｋ×Ｋ行列の相関係数行列を生成する。
【００７６】
ＭＭＳＥ補正部８は，Ｋ個の信号ｙ_１〜ｙ_Ｋに対応する受信信号の受信電力と背景雑音電力を推定し，式（７）に示したように，相関係数生成部１０で生成される相関係数行列の対角要素を補正して，変形相関係数行列を生成する。なお，背景雑音電力の推定値を零と置いて，ＭＭＳＥ補正を行わない場合は，復相関受信機（デコリレータ，Ｄｅｃｏｒｒｅｌａｔｉｎｇ　Ｄｅｔｅｃｔｏｒ）として機能する。
【００７７】
逆行列演算部１４は，相関係数生成部１０で生成された，Ｋ×Ｋ行列の相関係数行列Ｒ，または変形相関係数行列Ｒ´の逆行列Ｒ^＋を演算する。
【００７８】
線形結合部６では，式（６）または（７）に示される演算が，入力された信号ｙ_１〜ｙ_Ｋと，逆行列演算部１４から出力される行列Ｒ^＋の間で行われ，（信号ｙ_１〜ｙ_Ｋの各々を行列Ｒ^＋の各行のＫ個の要素と乗算し，それらを加え合わせる演算を，以下，線形結合と呼ぶことにする），線形結合後の信号を判定回路７に与える。
【００７９】
線形結合部６_ｉにより，行列Ｒ^＋のｉ行と線形結合された結果は，式（６）または，式（７）の右辺の信号となる。
【００８０】
判定回路７は，線形結合部６から与えられた信号を所定の閾値と比較し，変調信号ｄ_１〜ｄ_Ｋから，離散的値または符号を判定して出力する。変調信号ｄ_１〜ｄ_Ｋの値は，変調方式としてＢＰＳＫ（Ｂｉｎａｒｙ　Ｐｈａｓｅ　Ｓｈｉｆｔ　Ｋｅｙｉｎｇ）が使用されている場合には−１または＋１となる。判定回路７_１〜７_Ｋにより得られた復号信号ｄ_１〜ｄ_Ｋは，その後，誤り訂正等の所定の処理が施されて，無線ネットワーク制御装置１１０_１または１１０_２に伝送されるか，あるいは，同じ基地局装置１２０のサービスエリア内の他の移動機に送信される。
【００８１】
図７は，整合フィルタバンク３の詳細な構成図である。整合フィルタバンク３は，Ｌ個の乗算器３ａ_１〜３ａ_Ｌ，Ｌ個の整合フィルタ（ＭＦ：Ｍａｔｃｈｅｄ　Ｆｉｌｔｅｒ）３ｂ_１〜３ｂ_Ｌ，およびＬ個のサンプルホールド回路（ＳＨＣ：Ｓａｍｐｌｅ　Ｈｏｌｄ　Ｃｉｒｃｕｉｔ）３ｃ_１〜３ｃ_Ｌを有する。
【００８２】
ベースバンド信号Ｙ（ｔ）は，Ｌ個の乗算器３ａ_１〜３ａ_Ｌの第１入力端子に入力される。Ｌ個の乗算器３ａ_１〜３ａ_Ｌの第２入力端子には，拡散信号生成部４により生成された拡散信号ｆ_１（ｔ）〜ｆ_Ｌ（ｔ）がそれぞれ入力される。　ベースバンド信号Ｙ（ｔ）と拡散信号ｆ_１（ｔ）〜ｆ_Ｌ（ｔ）とが乗算器３ａ_１〜３ａ_Ｌによってそれぞれ乗算されることを，逆拡散と呼ぶが，乗算器３ａ_１〜３ａ_Ｌによってそれぞれ逆拡散された信号は，ＭＦ３ｂ_１〜３ｂ_Ｌにそれぞれ入力され，ＭＦ３ｂ_１〜３ｂ_Ｌは，各信号の１ビット長Ｔの間，入力された信号をそれぞれ積分する。
【００８３】
ＭＦ３ｂ_１〜３ｂ_Ｌからそれぞれ出力された信号は，ＳＨＣ３ｃ_１〜３ｃ_Ｌにそれぞれ入力される。ＳＨＣ３ｃ_１〜３ｃ_Ｌは，ＭＦ３ｂ_１〜３ｂ_Ｌが各受信信号の１ビット分の積分を完了した時点の出力信号の値をサンプリングし，その値を保持して，出力する。
【００８４】
ＳＨＣ３ｃ_１〜３ｃ_Ｌのそれぞれからは，連続してサンプリング値が出力されるが，各整合フィルタ毎の複数のサンプリング値が，希望信号の１ビットを復号するのに利用される。図４（Ｂ）に示した単一レートの非同期系のモデルを例にとれば，図７に示したように，各整合フィルタから各々出力される３個×ユーザ数すなわち，３Ｌ（＝Ｋ）個のデータが利用される。
【００８５】
整合フィルタバンク３は，図５に示すように受信信号に対して作用するだけではなく，相関係数を求める際に使用される。図８に，相関係数を求めるための整合フィルタバンクの構成を示すが，前記整合フィルタバンク３と区別するために，相関係数整合フィルタバンクと呼ぶことにする。同じ拡散信号同士の相関係数，すなわち相関係数行列の対角要素は，伝送レートが決まれば定まる値で，求める必要がないので，対応する部分を省略する（点線で表示）。
【００８６】
図６は，拡散信号ｇ_１（ｔ）〜ｇ_Ｌ（ｔ）と（バースト）拡散信号ｓ_１（ｔ）〜ｓ_Ｋ（ｔ）の関係を示したものである。
【００８７】
相関係数整合フィルタバンクは，そこに使用されるスイッチＳＷの状態を明示しないと，入力と出力の関係が定まらないので，さらに図９に示す，変形整合フィルタバンクを導入する。この変形整合フィルタバンクの入力は，Ｋ個の信号ｙ_１〜ｙ_Ｋに対応する拡散信号ｓ_１（ｔ）〜ｓ_Ｋ（ｔ）で，その出力である相関係数が明示的に示されるので，今後の説明は，相関係数整合フィルタバンクではなく，それと等価な変形整合フィルタバンクを使って行う。観測窓長Ｎ＝２の場合には，整合フィルタの数において，両者の違いは無いので，実際に変形整合フィルタバンクを使っても実用上問題はない。但し，Ｎ＞２では，整合フィルタの数の点で，実際には，相関係数整合フィルタバンクを使用する方が有利である。
【００８８】
図１０は，拡散信号生成部４の構成例で，受信信号に対して，同期捕捉（ａｃｑｕｉｓｉｔｉｏｎ）および同期保持（ｔｒａｃｋｉｎｇ）を行って，受信信号を整合フィルタバンク３で逆拡散するための拡散信号ｆ_１（ｔ）〜ｆ_Ｌ（ｔ）を生成する。それと同時に，相関係数行列生成部で相関係数行列を生成するための拡散信号ｇ_１（ｔ）〜ｇ_Ｌ（ｔ）を生成する。拡散信号ｇ_１（ｔ）〜ｇ_Ｌ（ｔ）は，拡散信号ｆ_１（ｔ）〜ｆ_Ｌ（ｔ）に対して，一定時間シフトしているだけで，同じ拡散コードから生成される。拡散信号の時間を任意の時間シフトさせた拡散信号を生成するには，例えば，図１０に示す例（周期３１のＭ系列）の場合，線形帰還シフトレジスタの内容を，遅れまたは進めたい時間だけ基準時刻とずれた時刻で初期値を設定することにより，任意の時間だけ時間シフトした拡散信号を生成することができる。または，基準時刻におけるシフトレジスタの内容と時間シフト量との関係はあらかじめ分かっているので，基準時刻でシフトレジスタの内容を適当に書き換えても，所望の拡散信号を得ることができる。シフトされる時間の量は，復号遅延（観測窓長Ｎのシステムでは，平均Ｎ／２の復号遅延が生じる）と相関係数行列の逆行列の演算に要する時間を考慮して決められる。
【００８９】
（１）第１の形態
図１１は，第１の形態による無線基地局装置１２０の，特にマルチユーザ受信部の構成を示すブロック図である。この図は，図５に示した無線基地局受信装置と基本的に同じ構成であるが，図５に示した，相関係数生成部１０と逆行列演算部１４とが一体化され，希望信号の１ビットを復号するのに関与する，希望信号と干渉信号のビット総数をＫとすると，Ｋ個の重み係数生成部５_１〜５_Ｋ，で構成され，またＫ個の線形結合部（または内積部）６_１〜６_Ｋ，およびＫ個の判定回路７_１〜７_Ｋを有する。
【００９０】
図１１には，重み係数生成部，線形結合部，および判定回路をそれぞれＫ個ずつ設けて，Ｋ個のビットのすべてを復号する構成を示しているが，Ｋ個のビットのうちＭ個（１≦Ｍ＜Ｋ）を復号する場合には，重み係数生成部，線形結合部，および判定回路はそれぞれＭ個ずつとなる。
【００９１】
尚，今後は，特別に断らない限り，希望信号のビット番号を１として説明する。
【００９２】
重み係数生成部５_１〜５_Ｋは，入力されたＬ個の拡散信号ｇ_１（ｔ）〜ｇ_Ｌ（ｔ）のＫ個の受信信号ビットに対応する部分である，Ｋ個の拡散信号ｓ_１（ｔ）〜ｓ_Ｋ（ｔ）に基づいて行列Ｒ^＋の各要素ｒ_ｉｊ（ｉ＝１〜Ｋ，ｊ＝１〜Ｋ）を生成し，出力する。具体的には，第ｉ番目の重み係数生成部５_ｉは，行列Ｒ^＋の第ｉ行のＫ個の要素ｒ_ｉ１〜ｒ_ｉＫを生成し，生成した要素ｒ_ｉ１〜ｒ_ｉＫを第ｉ番目の線形結合部６_ｉに与える。
【００９３】
図１２は，重み係数生成部５_１〜５_Ｋのうち，重み係数生成部５_１の構成を示すブロック図である。重み係数生成部５_１は，第１係数生成部５０_１から第Ｋ重み係数生成部５０_ＫまでのＫ個の重み係数生成部を有する。
【００９４】
各重み係数生成部には，Ｋ個の拡散信号ｓ_１（ｔ）〜ｓ_Ｋ（ｔ）が共通に入力される。また，各重み係数生成部には，全ての重み係数生成部の各出力信号が入力される。但し，それぞれの出力は，それぞれの重み係数生成部の内部で既に結合している（入力元の無い矢印がそのことを示している）。すなわち，第ｊ重み係数生成部５０_ｊ（ｊ＝１〜Ｋ）には，第ｊ重み係数生成部５０_ｊの出力信号ｒ_１ｊを除く出力信号ｒ_１１〜ｒ_１ｊ−１，ｒ_１ｊ＋１〜ｒ_１Ｋが入力される。
【００９５】
第ｊ重み係数生成部５０_ｊは，入力されるこれらの信号と拡散信号ｓ_１（ｔ）〜ｓ_Ｋ（ｔ）に基づいて，行列Ｒ^＋の第１行第ｊ列の要素ｒ_１ｊを生成し，出力する。
【００９６】
ＭＭＳＥ補正部８は，受信信号から背景雑音電力密度σ^２，と整合フィルタバンク３の出力ｙ_１〜ｙ_Ｋに対応するビット当たりの推定受信電力の比である，上記式（１３）の補正値α_ｉを求め，重み係数生成部５_１〜５_Ｋに与える。
【００９７】
図１３は，重み係数生成部５_１の第１係数生成部５０_１の詳細な構成を示すブロック図である。第１係数生成部５０_１は，相関係数生成部５１１，線形結合部５１２，および帰還部５１３からなる。
【００９８】
相関係数生成部５１１は，前述した変形整合フィルタバンク（図９参照）と同じ機能を有し，Ｋ個の乗算器５１ａ_１〜５１ａ_Ｋ，Ｋ個のＭＦ（整合フィルタ）５１ｂ_１〜５１ｂ_Ｋ，およびＫ個のＳＨＣ（サンプルホールド回路）５１ｃ_１〜５１ｃ_Ｋを有する。線形結合部５１２は，線形結合部６_１〜６_Ｋ（後に詳述）のそれぞれと同じ構成を有し，Ｋ個の乗算器５１ｄ_１〜５１ｄ_Ｋおよび合成器（合成器）５１ｅを有する。
【００９９】
帰還部５１３は，減算器５１ｆ，反転増幅器５１ｇ，帰還ループフィルタ（低域フィルタ，ＬＰＦ）５１ｈ，および合成器５１ｍを有する。
【０１００】
相関係数生成部５１１の乗算器５１ａ_１〜５１ａ_Ｋのそれぞれの第１入力端子には，拡散信号ｓ_１（ｔ）が共通に入力される一方，第２入力端子には，拡散信号ｓ_１（ｔ）〜ｓ_Ｋ（ｔ）がそれぞれ入力される。したがって，乗算器５１ａ_ｊは，拡散信号ｓ_１（ｔ）とｓ_ｊ（ｔ）とを乗算し，積ｓ_１（ｔ）ｓ_ｊ（ｔ）を出力する。
【０１０１】
ＭＦ５１ｂ_１〜５１ｂ_Ｋは，乗算器５１ａ_１〜５１ａ_Ｋの出力信号を１ビットの長さＴに亘って，それぞれ積分（フィルタリング）し，積分結果をＳＨＣ５１ｃ_１〜５１ｃ_Ｋにそれぞれ出力する。ＳＨＣ５１ｃ_１〜５１ｃ_Ｋは，ＭＦ５１ｂ_１〜５１ｂ_Ｋが１ビットの長さの積分をそれぞれ完了した時点の出力信号の値をサンプリングしその値を保持して出力する。したがって，ＳＨＣ５１ｃ_ｉの出力信号は，上記式（２）から，相関係数行列Ｒの第ｉ行第１列の要素（相関係数）ａ_ｉ１となる。すなわち，ＳＨＣ５１ｃ_１〜５１ｃ_Ｋからは，相関係数行列Ｒの第１列の相関係数ａ_１１〜ａ_Ｋ１が出力される。ＳＨＣ５１ｃ_１の出力ａ_１１は，伝送レートが決まれば定まる定数で，ビット毎に変化することが無いので乗算器５１ａ_１から，ＳＨＣ５１ｃ_１までの回路は省略してもよい。これらの相関係数ａ_１１〜ａ_Ｋ１は，線形結合部５１２の乗算器５１ｄ_１〜５１ｄ_Ｋの第１入力端子にそれぞれ入力される。ＨＣ５２ｃ_１の出力は，ＭＭＳＥ補正値をα_１とすると（１＋α_１）倍されて出力される。乗算器５１ｄ_１〜５１ｄ_Ｋの第２入力端子には，行列Ｒ^＋の第１行の要素に対応するｒ_１１〜ｒ_１Ｋ（これらの値を，今後，重み係数ｘ_１〜ｘ_Ｋとも呼ぶ）がそれぞれ入力される。
【０１０２】
要素ｒ_１１は，第１係数生成部５０_１の帰還部５１３の出力信号がフィードバックされたものである。要素ｒ_１２〜ｒ_１Ｋは，図６に示すように，第２係数生成部５０_２から第Ｋ重み係数生成部５０_Ｋの各出力信号である。
【０１０３】
乗算器５１ｄ_１〜５１ｄ_Ｋは，第１入力端子および第２入力端子からそれぞれ入力された信号を乗算し，乗算結果を合成器５１ｅに与える。合成器５１ｅは，乗算器５１ｄ_１〜５１ｄ_Ｋから与えられたＫ個の乗算結果を合成し，合成結果を帰還部５１３に与える。乗算器５１ｄ_１〜５１ｄ_Ｋの乗算結果として，Ｋ個の乗算結果ａ_１１（１＋α_１）ｒ_１１，ａ_２１ｒ_１２，…，ａ_Ｋ１ｒ_１Ｋが得られ，合成器５１ｅによる合成結果として，ａ_１１（１＋α_１）ｒ_１１＋ａ_２１ｒ_１２＋…＋ａ_Ｋ１ｒ_１Ｋが得られる。この合成結果は，上記連立方程式（１２）の第１式の左辺に対応する。
【０１０４】
合成器５１ｅの出力信号は，減算器５１ｆによって，定数ａ_１１が減じられて，減算結果を反転増幅器５１ｇに入力する。反転増幅器５１ｇは，入力信号を利得−Ｇで増幅（反転増幅）して出力する。
【０１０５】
反転増幅器５１ｇの出力信号は，ＬＰＦ５１ｈを介して合成器５１ｍ入力され，帰還ループの初期値である，定数１／（１＋α_１）が加えられる。この合成結果が，第１係数生成部５０_１の出力となるとともに，乗算器５１ｄ_１の第２入力端子に入力さる。
【０１０６】
この結果，線形結合部５１２と帰還部５１３とは，帰還ループを形成することになり，帰還部５１３の出力（すなわち第１係数生成部５０_１の出力）は，帰還ループの収れん状態で，要素ｒ_１１に収れんし，式（１２）で表せる連立方程式の解の１つ（要素ｒ_１１）が求められる（帰還ループの動作原理は，後述する）。
【０１０７】
図１４は，重み係数生成部５_１の第２係数生成部５０_２の詳細な構成を示すブロック図である。第２係数生成部５０_２は，相関係数生成部５２１，線形結合部５２２，および帰還部５２３を有する。
【０１０８】
相関係数生成部５２１は，前述した変形整合フィルタバンク（図９参照）と同じ機能を有し，Ｋ個の乗算５２ａ_１〜５２ａ_Ｋ，Ｋ個のＭＦ５２ｂ_１〜５２ｂ_Ｋ，およびＫ個のＳＨＣ５２ｃ_１〜５１ｃ_Ｋを有する。線形結合部５２２は，線形結合部６_１〜６_Ｋ（後に詳述）のそれぞれと同じ機能を有し，Ｋ個の乗算器５２ｄ_１〜５２ｄ_Ｋおよび合成器５２ｅを有する。
【０１０９】
帰還部５１３は，反転増幅器５２ｇおよび帰還ループフィルタ（低域フィルタ，ＬＰＦ）５２ｈを有する。
【０１１０】
相関係数生成部５２１の乗算器５２ａ_１〜５２ａ_Ｋのそれぞれの第１入力端子には，拡散信号ｓ_２（ｔ）が共通に入力される一方，第２入力端子には，拡散信号ｓ_１（ｔ）〜ｓ_Ｋ（ｔ）がそれぞれ入力される。したがって，乗算器５２ａ_ｊは，拡散信号ｓ_２（ｔ）とｓ_ｊ（ｔ）との乗算結果である積ｓ_１（ｔ）ｓ_ｊ（ｔ）を出力する。
【０１１１】
ＭＦ５２ｂ_１〜５２ｂ_Ｋは，乗算器５２ａ_１〜５２ａ_Ｋの出力信号を１ビット長Ｔの間，それぞれ積分（フィルタリング）し，積分結果をＳＨＣ５２ｃ_１〜５２ｃ_Ｋにそれぞれ出力する。ＳＨＣ５２ｃ_１〜５２ｃ_Ｋは，ＭＦ５２ｂ_１〜５２ｂ_Ｋが１ビットの積分をそれぞれ完了した時点の出力信号の値をサンプリングし，保持して出力する。したがって，ＳＨＣ５２ｃ_ｉの出力信号は，上記式（２）から，相関係数行列Ｒの第ｉ行第２列の要素（相関係数）ａ_ｉ２となる。すなわち，ＳＨＣ５２ｃ_１〜５２ｃ_Ｋからは，相関係数行列Ｒの第２列の相関係数ａ_１２〜ａ_Ｋ２が出力される。ＳＨＣ５２ｃ_１の出力ａ_２２は，伝送レートが決まれば定まる定数で，ビット毎に変化することが無いので乗算器５２ａ_２から，ＳＨＣ５２ｃ_２までの回路は省略してもよい。但し，ＨＣ５２ｃ_２の出力は，ＭＭＳＥ補正値をα_２とすると（１＋α_２）倍されて出力される。
【０１１２】
これらの相関係数ａ_１２〜ａ_Ｋ２は，線形結合部５２２の乗算器５２ｄ_１〜５２ｄ_Ｋの第１入力端子にそれぞれ入力される。乗算器５２ｄ_１〜５２ｄ_Ｋの第２入力端子には，行列Ｒ^＋の第１行の要素に対応するｒ_１１〜ｒ_１Ｋがそれぞれ入力される。
【０１１３】
要素ｒ_１２（重み係数ｘ_２）は，第２係数生成部５０_２の帰還部５２３の出力信号がフィードバックされたものである。要素ｒ_１１，ｒ_１３〜ｒ_１Ｋは，図１２に示すように，第１係数生成部５０_１および第３重み係数生成部５０_３から第Ｋ重み係数生成部５０_Ｋの各出力信号である。
【０１１４】
乗算器５２ｄ_１〜５２ｄ_Ｋは，第１入力端子および第２入力端子からそれぞれ入力された信号を乗算し，乗算結果を合成器５２ｅに与える。合成器５２ｅは，乗算器５２ｄ_１〜５２ｄ_Ｋから与えられたＫ個の乗算結果を合成し，合成結果を帰還部５２３に出力する。乗算器５２ｄ_１〜５２ｄ_Ｋの乗算結果として，Ｋ個の乗算結果ａ_１２ｒ_１１，ａ_１２（１＋α_２）ｒ_１２，…，ａ_Ｋ２ｒ_１Ｋが得られ，合成器５２ｅによる合成結果として，ａ_１２ｒ_１１＋ａ_２２（１＋α_２）ｒ_１２＋…＋ａ_Ｋ２ｒ_１Ｋが得られる。この合成結果は，上記連立方程式（１２）の第２式の左辺に対応する。
【０１１５】
合成器５２ｅの出力信号は，反転増幅器５２ｇに入力され，反転増幅された後，ＬＰＦ５２ｈを介して乗算器５２ｄ_２の第２入力端子に入力される。また，ＬＰＦ５２ｈの出力が第２係数生成部５０_２の出力ｒ_１２となる。
【０１１６】
この結果，線形結合部５２２と帰還部５２３とは，帰還ループを形成することになり，帰還部５２３の出力値（すなわち第２係数生成部５０_２の出力）は，帰還ループの収れん状態で，要素ｒ_１２に収れんし，式（１２）で表せる連立方程式の解の１つ（要素ｒ_１２）が求められる。
【０１１７】
第３重み係数生成部５０_３から第Ｋ重み係数生成部５０_Ｋも第２係数生成部５０_２とほぼ同様の構成を有し，相関係数生成部の乗算器の第１入力端子に入力される拡散信号がそれぞれｓ_３（ｔ）〜ｓ_Ｋ（ｔ）となる点，および，自己の出力信号がフィードバックされる線形結合部の乗算器がそれぞれ第３番目から第Ｋ番目の乗算器になる点が異なるだけで，それぞれ要素ｒ_１３〜ｒ_１Ｋを出力する（これらの出力を，重み係数ｘ_３〜ｘ_Ｋとも呼ぶ）。
【０１１８】
したがって，これら第３重み係数生成部５０_３から第Ｋ重み係数生成部５０_Ｋの説明は省略する。また，重み係数生成部５_２〜５_Ｋについても，重み係数生成部５_１と同様の構成を有するので，その説明を省略する。
【０１１９】
図１１に戻って，整合フィルタバンク３の出力ｙ_１〜ｙ_Ｋは，線形結合部６_１〜６_Ｋに共通に入力される。重み係数生成部５_１〜５_Ｋの各出力は，線形結合部６_１〜６_Ｋのそれぞれに入力される。すなわち，重み係数生成部５_ｉの出力ｒ_ｉ１〜ｒ_ｉＫは，線形結合部６_ｉに入力される。線形結合部６_１〜６_Ｋは，これら入力された信号を線形結合する。
【０１２０】
図１５は，線形結合部６_１の詳細な構成を示すブロック図である。線形結合部６_１は，前述したように，第１係数生成部５０_１の線形結合部５１２と同じ機能を有し，Ｋ個の乗算器６１ａ_１〜６１ａ_Ｋおよび合成器６１ｂを有する。
【０１２１】
乗算器６１ａ_１〜６１ａ_Ｋの第１入力端子には，整合フィルタバンク３の出力信号ｙ_１〜ｙ_Ｋがそれぞれ入力され，第２入力端子には，重み係数生成部５_１の出力信号ｒ_１１〜ｒ_１Ｋがそれぞれ入力される。乗算器６１ａ_１〜６１ａ_Ｋは，第１入力端子および第２入力端子に入力された信号をそれぞれ乗算し，乗算結果を合成器６１ｂにそれぞれ出力する。
【０１２２】
合成器６１ｂは，乗算器６１ａ_１〜６１ａ_Ｋの出力信号を合成する。この合成器６１ｂの合成結果はｙ_１ｒ_１１＋ｙ_２ｒ_１２＋…＋ｙ_Ｋｒ_１Ｋとなり，この合成結果は上記式（８）の右辺であり，その値はＷ_１ｄ_１と雑音成分の和となる。この合成結果は，判定回路７_１に入力され，判定回路７_１からｄ_１の判定結果Ｄ_１が出力される。
【０１２３】
他の線形結合部６_２〜６_Ｋも，線形結合部６_１と同じ構成で，Ｗ_２ｄ_２〜Ｗ_Ｋｄ_Ｋとそれぞれの雑音成分の和をそれぞれ出力する。そして，これらの出力は，判定回路７_２〜７_Ｋにそれぞれ入力され，判定回路７_２〜７_Ｋから判定結果としてＤ_２〜Ｄ_Ｋ（ｄ_２〜ｄ_Ｋの判定結果の値）がそれぞれ出力され，復相関形受信が行われる。
【０１２４】
以上，第１の形態における，無線基地局装置１２０の装置の構成を説明したが，次に，この様な構成で，なぜマルチユーザ受信が可能となるかについて，原理の説明を行い，シミュレーションによって動作の確認を行った結果についても言及する。
【０１２５】
合成器５１ｅの出力が，ａ_１１ｒ_１１＋ａ_２１ｒ_１２＋…＋ａ_Ｋ１ｒ_１Ｋに，また合成器５２ｅの出力が，ａ_１２ｒ_１１＋ａ_２２ｒ_１２＋…＋ａ_Ｋ２ｒ_１Ｋになることは，既に述べた。合成器５１ｅの出力は，帰還部５１３において，第１の定数（ここでは，ａ_１１とする）が引かれ，反転増幅器，帰還ループフィルタを経て，第２の定数（ここでは，１＋α_１の逆数）が合成されて，線形結合部に戻され，全体で帰還ループを形成している。初期値として第２の定数を与えてループを起動し，一定時間が経過した後に，定常状態に達するが，このとき，線形結合部５１２と帰還部５１３とで形成される帰還ループの利得を−Ｇ_１とすると，次の式（１４）が成り立つ。
【０１２６】
−Ｇ_１［ａ_１１（１＋α_１）ｒ_１１＋ａ_２１ｒ_１２＋…＋ａ_Ｋ１ｒ_１Ｋ−ａ_１１］
＝ｒ_１１−１／（１＋α_１）　　（１４）
【０１２７】
同様に，線形結合部５２２と帰還部５２３とで形成される帰還ループの利得を−Ｇ_２とすると，次の式（１５）が成り立つ。
【０１２８】
−Ｇ_２［ａ_１２ｒ_１１＋ａ_２２（１＋α_２）ｒ_１２＋…＋ａ_Ｋ２ｒ_１Ｋ］＝ｒ_１２　　（１５）
【０１２９】
式（１４）と式（１５）の両辺をそれぞれの帰還ループ利得−Ｇ_１，　−Ｇ_２で割って，右辺を左辺に移項すれば，次の式（１６）と式（１７）になる。
【０１３０】
［ａ_１１（１＋α_１）＋１／Ｇ_１］［ｒ_１１−１／（１＋α_１）］＋ａ_２１ｒ_１２＋…
＋ａ_Ｋ１ｒ_１Ｋ＝０　　　　　　　　　　　　　　　　　　　　　（１６）
ａ_１２ｒ_１１＋［ａ_２２（１＋α_２）＋１／Ｇ_１］ｒ_１２＋…＋ａ_Ｋ２ｒ_１Ｋ＝０　（１７）
【０１３１】
式（１７）を，他の第２係数生成部に関するものを含めて，一般化すると，式（１８）となる。
【０１３２】
ａ_１ｊｒ_１１＋ａ_２ｊｒ_１２＋…＋［ａ_ｊｊ（１＋α_ｊ）＋１／Ｇ_１］ｒ_１ｊ＋…＋ａ_Ｋ１ｒ_１Ｋ＝０
（ｊ＝２，３，…，Ｋ）　　　　　　　　（１８）
【０１３３】
式（１６）および式（１８）において，Ｇ_ｊ＝∞（ｊ＝１，２，…，Ｋ）と置くと，これらの式は，次の様になる。
【０１３４】
ａ_１１（１＋α_１）ｒ_１１＋ａ_２１ｒ_１２＋…＋ａ_Ｋ１ｒ_１Ｋ＝ａ_１１（１９）
ａ_１ｊｒ_１１＋ａ_２ｊｒ_１２＋…＋ａ_ｊｊ（１＋α_ｊ）ｒ_１ｊ＋…＋ａ_Ｋ１ｒ_１Ｋ＝０
（ｊ＝２，３，…，Ｋ）（２０）
【０１３５】
受信信号の整合フィルタバンク３の出力である，Ｋ個の値（ｙ_１，ｙ_２，…，ｙ_Ｋ）と，重み係数生成部の出力である（ｒ_１１，ｒ_１２，…，ｒ_１Ｋ）とを線形結合部６_１で線形結合した結果を，Ｂ＝ｙ_１ｒ_１１＋ｙ_２ｒ_１２＋…＋ｙ_Ｋｒ_１Ｋとすると，Ｂは，干渉信号成分が零となる（復相関受信）か，抑圧されている（ＭＭＳＥ受信）かになることは，既に述べたがが，その理由は次の通りである。Ｋ個の値（ｙ_１，ｙ_２，…，ｙ_Ｋ）を，式（３）を使って展開し，Ｂを求めると，式（２１）となる。
【０１３６】
Ｂ＝Ｗ_１ｄ_１［ａ_１１（１＋α_１）ｒ_１１＋ａ_２１ｒ_１２＋…＋ａ_Ｋ１ｒ_１Ｋ］
＋Ｗ_２ｄ_２［ａ_１２ｒ_１１＋ａ_２２（１＋α_２）ｒ_１２＋…＋ａ_Ｋ２ｒ_１Ｋ］＋…
＋Ｗ_Ｋｄ_Ｋ［ａ_１Ｋｒ_１１＋ａ_２Ｋｒ_１２＋…＋ａ_ＫＫ（１＋α_Ｋ）ｒ_１Ｋ］＋…
＋ｚ_１ｒ_１１＋ｚ_２ｒ_１２＋…＋ｚ_Ｋｒ_１Ｋ（２１）
【０１３７】
式（２１）に，式（１９）と式（２０）を代入すると，Ｂは，次の式（２２）になる。
【０１３８】
Ｂ＝Ｗ_１ｄ_１ａ_１１（１−α_１ｒ_１１）−Ｗ_２ｄ_２ａ_２２α_２ｒ_１２−…−Ｗ_Ｋａ_ＫＫα_Ｋｒ_１Ｋｄ_Ｋ＋ｚ_１ｒ_１１＋ｚ_２ｒ_１２＋…＋ｚ_Ｋｒ_１Ｋ（２２）
【０１３９】
式（２２）から，α_１＝α_２＝…＝α_Ｋ＝０，すなわち，背景雑音電力を零とおくと，式（２２）の右辺の第１項の，Ｗ_１ｄ_１ａ_１１，すなわち，希望信号と雑音成分だけとなって，復相関受信が行われることが分かる。α_ｊを，背景雑音電力密度とビット番号ｊのビット当たりの推定受信信号電力とすると，干渉信号電力と背景雑音電力の和が最小となる，ＭＭＳＥ受信が行われることは，既に述べた。
【０１４０】
また，式（１９）の右辺のａ_１１を１と置いて，式（２０）と合わせて，式（１２）と比較すると，両者が一致することからも，これまで述べてきた受信機で，復相関形受信が行われることが分かる。
【０１４１】
ａ_１１を１と置いてよいことは，次のようにして，説明される。まず，相関係数行列の対角要素は，式（２）から，次の式（２３）で表される。
【０１４２】
【数１５】

【０１４３】
ここでは，希望信号ビットの始まりをｔ＝０とした。
【０１４４】
式（２３）から分かるように，ａ_ｊｊすなわち，相関係数行列の対角要素は，ビットレート（１／Ｔ）によって決まる。従って，単一レートのシステムでは，対角要素が全て同じになり，その値は，線形結合部６のレベルを相対的に変えるだけなので，通常はその値を１として取り扱う場合が多い。本発明は，単一レートのみならず，マルチレートにも適用できるので，相関係数行列の対角要素がすべて同じとは限らないので，明細書の記載は，対角要素を変数表示のまま残してある。
【０１４５】
相関係数行列の対角要素ａ_ｊｊが，ビットレート（１／Ｔ_ｊ）によって決まる定数なので，ａ_ｊｊを求めるための乗算器と整合フィルタを省略できる。また，対角要素ａ_ｊｊまたは，それをＭＭＳＥ補正した，ａ_ｊｊ（１＋α_ｊ）とｒ_１ｊとの乗算を行う線形結合部における乗算器も，レベル変換器と置き換える（あるいは，低速で動作する乗算器と解釈する）ことが可能である。
【０１４６】
以上は，帰還ループ利得Ｇ_ｊ＝∞（ｊ＝１，２，…，Ｋ）として，話を進めた。しかし，帰還ループ利得が小さくなると，演算誤差が生じるので，次に，この誤差が生じない方法（利得補正）について述べる。
【０１４７】
図１６（Ａ）は，第１係数生成部に，利得補正を施したもので，合成器５１ｎの出力をｒ_１１´とし，これを元の線形結合部に戻して，帰還ループを形成するとともに，ＬＰＦ５１ｈの出力（ｒ_１１´−１／（１＋α_１））をｃ_１倍し，それに１／（１＋α_１）を加えたものを，ｒ_１１として，他のＫ−１個の第２係数生成部に接続している。ｒ_１１´とｒ_１１の関係は，次の式（２４）で表される。
【０１４８】
［ｒ_１１´−１／（１＋α_１）］ｃ_１＝ｒ_１１−１／（１＋α_１）（２４）
【０１４９】
ここに，ｃ_１は，次の式（２５）で表される。
【０１５０】
ｃ_１＝１＋１／ａ_１１／（１＋α_１）／Ｇ_１　　　　　　　　　　　（２５）
【０１５１】
ｒ_１１´に関しては，式（１６）において，ｒ_１１と置き替えたものであるから，次の式（２６）が成り立つ。
【０１５２】
［ａ_１１（１＋α_１）＋１／Ｇ_１］［ｒ_１１´−１／（１＋α_１）］
＋ａ_２１ｒ_１２＋…＋ａ_Ｋ１ｒ_１Ｋ＝０　　　　　　　　　　　　　　　　（２６）
【０１５３】
式（２６）に，式（２４）を代入すると，次の式（２７）になる。
ａ_１１（１＋α_１）ｒ_１１＋ａ_２１ｒ_１２＋…＋ａ_Ｋ１ｒ_１Ｋ＝ａ_１１　　　　　　（２７）
【０１５４】
式（２７）では，Ｇ_１に関する項が相殺された結果，Ｇ_１が無限大である場合と同じく，誤差の無い出力が得られることが分かる。図１６（Ｂ）は，図１６（Ａ）と等価な回路で，帰還部の出力をそのまま，他のＫ−１個の重み係数生成部に接続し，帰還ループ内の元の線形結合部へ戻す際に，そのレベルを１／ｃ_１倍したものである。
【０１５５】
利得補正の他の方法を図１７に示す。式（１６）に戻ると，利得Ｇ_１の影響を除くには，この式の左辺に，−［ｒ_１１−１／（１＋α_１）］／Ｇ_１を加えるのでもよいことが分かる。但し，加える場所は，合成器の入力から反転増幅器の入力の間とするか，または，−［ｒ_１１−１／（１＋α_１）］／Ｇ_１を−Ｇ_１倍して反転増幅器の出力に加えても同じであって，図１７は，この考えに基づいて構成したものである。尚，ＬＰＦ５１ｈの入出力で局所的にループが形成されるので，各所のレベル合わせを行って，帰還利得を調整する必要がある。増幅器５１ｑは，ＬＰＦには損失があって，それを補うために挿入したが，ＬＰＦがアクティブ素子で構成されていて利得があれば，増幅器５１ｑは，減衰器となる場合もある。
【０１５６】
以上で，帰還ループ利得が小さい場合にも，逆行列要素の演算誤差が生じない方法について述べたが，次は，線形結合部に関する，いくつかの変形例について説明する。
【０１５７】
図１８（Ａ）と図１３では，ａ_１１（１＋α_１）ｒ_１１を乗算器５１ｄ_１で作り出すのに，（１＋α_１）の乗算をどちらの入力側で行うか，が違っている。図１３では，相関係数行列生成部側で行い，図１８（Ａ）では，重み係数生成部側で行っているが，何れも，式で表せば同じであるので，等価回路と言える。
【０１５８】
図１８（Ａ）と（Ｂ）とは，ＭＭＳＥ補正を，帰還ループの中で行うか，求まった重み係数を，他のＫ−１個の重み係数部に分配する側で行うかの違いであるが，これも式で表せば同じなので，互いに等価である。
【０１５９】
図１９も，ＭＭＳＥ補正に関する変形例で，ａ_１１（１＋α_１）ｒ_１１を作り出すのに，ａ_１１ｒ_１１とａ_１１α_１ｒ_１１とを加え合わせる（図１９）か，ａ_１１ｒ_１１と乗じるか（図１３）の相違であって，式で表せば同じなので，互いに等価である。
【０１６０】
図２０は，図１３に示した，第１係数生成部の変形例である。図１３においては，線形結合部の合成器５１ｅの出力から，減算器５１ｆで，第１の定数ａ_１１を減じてから，ＬＰＦ５１ｈの出力で，第２の定数を加える構成になっているが，図２０においては，合成器５１ｅの出力から，第１の定数を減じないで，帰還ループを形成し，第２の定数の合成を，帰還ループの外で行った結果を，他のＫ−１この，重み係数生成部に結合している。両者が等価であることは，次のように説明できる。
【０１６１】
先ず，ＬＰＦ５１ｈの出力を，ｒ_１１´とすると，次の式（２８）が成り立つ。
ａ_１１（１＋α_１）ｒ_１１´＋ａ_２１ｒ_１２＋…＋ａ_Ｋ１ｒ_１Ｋ＝０　　　　　　　（２８）
ｒ_１１´とｒ_１１との差は，合成器５１ｍによる，第２の定数１／（１＋α_１）の違いであるから，次の式（２９）が成り立つ。
ｒ_１１´＝ｒ_１１−１／（１＋α_１）　　　　　　　　　　　　　　　　　（２９）
式（２９）式を式（２８）に代入し，左辺に生じたａ_１１を右辺に移項すれば，式（１９）となるので，図１３と図２０の回路は，互いに等価である。
図１６，図１７，図１８および図１９を使った説明は，希望信号ビットに対する重み係数ｒ_１１（ｘ_１）を求める回路に付いてであったが，干渉信号ビットに対応する重み係数を求める回路に付いては，減算器５１ｆと合成器５１ｍに相当する部分がないだけで，同じ手法が適用できるので，説明と図示を省略した。
【０１６２】
次に，第１の形態で説明した構成によって，回路が正しく動作するかどうかをシミュレーションによって確認した結果について，図２１，図２２，図２３，図２４および図２５を使って説明する。
【０１６３】
モデルは，観測窓長Ｎ＝２，　　ユーザ数Ｌ＝２，　Ｋ＝ＮＬ＝４（ビットｄ_１〜ｄ_４），帰還ループ利得補正は，図２０に示した手法を用いた，簡単な場合で，逆行列の要素ｒ_１１，ｒ_１２，ｒ_１３およびｒ_１３（重み係数ｘ_１〜ｘ_４）の収れんの様子を調べたものである。
【０１６４】
図２１は，ビットｄ_１〜ｄ_４とそれに対応する拡散信号ｓ_１（ｔ）〜ｓ_４（ｔ）ならびに相関係数を求めるのに必要な，拡散信号の積ｓ_１（ｔ）ｓ_２（ｔ），ｓ_２（ｔ）ｓ_３（ｔ），ｓ_３（ｔ）ｓ_４（ｔ）の波形を示している。図２２（ａ）は，拡散信号ｓ_１（ｔ）〜ｓ_４（ｔ）を，図２２（ｂ）は，に拡散信号の積ｓ_１（ｔ）ｓ_２（ｔ），ｓ_２（ｔ）ｓ_３（ｔ）およびｓ_２（ｔ）ｓ_３（ｔ）をそれぞれ数値により示している。図２２（ｃ）は，あらかじめコンピュータで求めた相関係数行列Ｒとその逆行列Ｒ^−１の具体的な数値を示している。
【０１６５】
ここでは，希望信号のビット番号を３とし，時刻ｔ＝０で，重み係数ｘ_３の初期値を１，他の重み係数の初期値を０と置いて帰還ループを動作したときの，重み係数ｘ_１〜ｘ_４（ｒ_１１，ｒ_１２，ｒ_１３およびｒ_１４）の時間変化を，シミュレーションにより求めると，図２３に示すグラフとなる。横軸の時間τは帰還ループフィルタの時定数により正規化した時間を示している。モデルが簡単なので，収れん値と初期値の差が小さく，収れん状況が分かりにくいので，重み係数ｘ_１〜ｘ_４そのそれぞれの厳密解からの偏差を示したのが，図２４である。この図から，帰還ループフィルタの時定数の数倍の時間をかければ，全ての重み係数が厳密解に収れんすることが分かる。図２５は，重み係数を，帰還ループの外，すなわち他の重み係数回路に与える場合に，一部を帰還ループフィルタの入力から，取り出して，高速化を図った例である。図２４および図２５から，帰還ループフィルタの時定数を充分小さい値を選べば，重み係数を求める時間を１ビット長以下にすることが可能なことが分かる。
【０１６６】
（２）第２の形態
重み係数生成部５_１〜５_Ｋのそれぞれは，第１係数生成部５０_１から第Ｋ重み係数生成部５０_Ｋを有する（図１２参照）。これら第１係数生成部５０_１から第Ｋ重み係数生成部５０_Ｋのそれぞれは，相関係数生成部５１１または５２１を有し，相関係数行列の第１列の要素ａ_１１〜ａ_Ｋ１から第Ｋ列の要素ａ_１１〜ａ_Ｋ１をそれぞれ出力する。すなわち，重み係数生成部５_１〜５_Ｋのいずれもが，相関係数行列の第１列の要素ａ_１１〜ａ_Ｋ１から第Ｋ列の要素ａ_１１〜ａ_Ｋ１（すなわち相関係数行列のすべての相関係数）を生成する。
【０１６７】
したがって，重み係数生成部５_１〜５_Ｋにおける相関係数生成部を１つの回路にまとめることができるので，重み係数生成部５_１〜５_Ｋに代えて，相関係数生成部１０およびＫ個の重み係数演算部１１_１〜１１_Ｋで構成したのが図２６に示した第２の形態である。尚，整合フィルタバンクについては，これまで，変形整合フィルタバンク（図９）を使って説明したが，これからは，相関係数整合フィルタバンク（図８）を使って説明する。
【０１６８】
相関係数生成部１０は，Ｋ×Ｋ個の相関係数ａ_１１〜ａ_ＫＫを生成し，生成した相関係数ａ_１１〜ａ_ＫＫを重み係数演算部１１_１〜１１_Ｋに共通に与える。重み係数演算部１１_１〜１１_Ｋは，相関係数生成部１０から与えられた相関係数ａ_１１〜ａ_ＫＫに基づいて行列Ｒ^＋の第１行の要素ｒ_１１〜ｒ_１Ｋから第Ｋ行の要素ｒ_Ｋ１〜ｒ_ＫＫをそれぞれ求め，線形結合部６_１〜６_Ｋにそれぞれ与える。
【０１６９】
図２７は，相関係数生成部１０の詳細な構成を示すブロック図である。相関係数生成部１０は，Ｌ個の相関係数整合フィルタバンク１０ａ_１〜１０ａ_Ｌを有する。
【０１７０】
これらの相関係数整合フィルタバンク１０ａ_１〜１０ａ_Ｌには，拡散信号ｇ_１（ｔ）〜ｇ_Ｌ（ｔ）が入力され，相関係数整合フィルタバンク１０ａ_１〜１０ａは，入力された拡散信号に基づいて第１列の相関係数ａ_１１〜ａ_Ｋ１から第Ｋ列の相関係数ａ_１Ｋ〜ａ_ＫＫまでを出力する。回路１０ｓは，これらの出力を並べ替えて出力する。
【０１７１】
図２８は，重み係数演算部１１_１の詳細な構成を示すブロック図である。重み係数演算部１１_１は，Ｋ個の線形結合部１１ａ_１〜１１ａ_ＫおよびＫ個の帰還部１１ｂ_１〜１１ｂ_Ｋを有する。
【０１７２】
線形結合部１１ａ_１は図１３に示す第１係数生成部５０_１の線形結合部５１２と同じ構成を有し，帰還部１１ｂ_１は図１２に示す帰還部５１３と同じ構成を有する。また，線形結合部１１ａ_１に入力される信号も，線形結合部５１２に入力される信号と同じである。したがって，帰還部１１ｂ_１からは，行列Ｒ^＋の要素ｒ_１１が出力される。
【０１７３】
線形結合部１１ａ_２および帰還部１１ｂ_２は，図１４に示す線形結合部５２２および帰還部５２３とそれぞれ同じ構成を有し，線形結合部１１ａ_２に入力される信号も線形結合部５２２に入力されるものと同じである。したがって，帰還部１１ｂ２からは，要素ｒ_１２が出力される。
【０１７４】
同様にして，線形結合部１１ａ_３（図２８には図示略）〜１１ａ_Ｋはいずれも，線形結合部５２２と同じ構成を有し，帰還部１１ｂ_３〜１１ｂ_Ｋはいずれも，帰還部５２３と同じ構成を有する。したがって，帰還部１１ｂ_３〜１１ｂ_ｋからは，要素ｒ_１３〜ｒ_１Ｋがそれぞれ出力される。
【０１７５】
（３）第３の形態
第３の形態における，無線基地局の構成は，図１１に示した第１の形態による無線基地局装置の構成と基本的に同じであるが，重み係数生成部５_１〜５_Ｋを変換し，回路規模を更に小さくしたものである。
【０１７６】
まず，第１の形態における重み係数生成部と，第３の形態における重み係数生成部の違いを理解するために，前者から後者への回路の変換の過程を順を追って説明し，後に式を使って動作原理の説明を行う。
【０１７７】
図２９から図３２は，重み係数生成部５_１の第１係数生成部５０_１の変換のステップを示している。
【０１７８】
まず，第１ステップとして，図１３に示すＳＨＣ５１ｃ_１〜５１ｃ_Ｋを除去し，第１係数生成部５０_１を図２９に示す構成に変形する。ＳＨＣ５１ｃ_１〜５１ｃ_Ｋの除去によっても除去前と等価な動作を行わせるためには，ＭＦ５１ｂ_１〜５１ｂ_Ｋの積分時間が，ＳＨＣ５１ｃ_１〜５１ｃ_Ｋによって積分結果が保持される時間だけ延長すればよい。
【０１７９】
続いて，第２ステップとして，図３０に示すように，ＭＦ５１ｂ_１〜５１ｂ_Ｋと乗算器５１ｄ_１〜５１ｄ_Ｋとの順序を入れ替える。重み係数ｘ_１（ｒ_１１）の値が，ＭＦ５１ｂ_１〜５１ｂ_Ｋの積分時間（０≦ｔ≦Ｔ）内で一定であれば，この変換は等価変換であるが，実際には，ｘ_１の値は，時間と共に変化するので，この変換は，等価でなく近似変換である。但し，後述する手法を用いることで，発生する近似誤差は，実用上問題ない範囲に抑えることができる。ＭＦ５１ｂ_１０
は，図２９には無かったものであるが，図３０では，減算器５１ｆの入力を作り出すために追加した。
【０１８０】
続いて，第３ステップとして，図３１に示すように，加算器５１ｅの前段にあるＭＦ５１ｂ_１〜５１ｂ_Ｋを加算器５１ｅの後段に移動し，後段にあるＬＰＦ５１ｈと置き換える。このことによって，Ｋ個のＭＦ５１ｂ_１〜５１ｂ_Ｋを１つのＭＦ５１ｂに削減することができる。ＬＰＦ５１ｈをＭＦ５１ｂに置き換えることによって，帰還ループの処理時間が長くなるが，後述するように，このことも問題とならない。
【０１８１】
続いて，第４ステップとして，図３２に示すように，乗算器５１ａ_１〜５１ａ_Ｋが加算器５１ｅの後段に移動される。移動の結果，Ｋ個の乗算器５１ａ_１〜５１ａ_Ｋを１つの乗算器５１ａに削減することができる。これにより，第１係数生成部５０_１の変換が終了する。
【０１８２】
第２係数生成部５０_２についても同様の変換を行うことにより，図３３に示す回路が得られ，第３係数生成部５０_３から第Ｋ係数生成部５０_Ｋについても，図３３と同様の回路が得られる。
【０１８３】
図３２と図３３を比較すると，第１係数生成部５０_１の乗算器５１ｄ_１〜５１ｄ_Ｋおよび加算器５１ｅと，第２係数生成部５０_２の乗算器５２ｄ_１〜５２ｄ_Ｋおよび加算器５２ｅは共通となるので，この共通部分については，全体で１つ使うだけでよいことになる。また，図示を省略するが，第３係数生成部５０_３から第Ｋ係数生成部５０_Ｋも図３３に示す第２係数生成部５０_２と同じ構成を有するので，これらの乗算器および加算器の部分は共通にできる。
【０１８４】
したがって，第１係数生成部５０_１から第Ｋ係数生成部５０_ＫのそれぞれにおけるＫ個の乗算器および１つの加算器は１つにまとめることができる。図３４は，これらを１つにまとめた第３の形態による重み係数生成部５_１の詳細な構成を示すブロック図である。
【０１８５】
この重み係数生成部は，各係数生成部の乗算器および加算器が１つにまとめられた線形結合部５００と，重み係数ｘ_１〜ｘ_Ｋ（ｒ_１１〜ｒ_１Ｋ）をそれぞれ出力する帰還部５０１_１〜５０１_Ｋを有する。
【０１８６】
帰還部５０１_１は，図３２の加算器５１ｅの後段に配置された加算器５１ｐから加算器５１ｍまでの構成要素を有する。帰還部５０２_２は，図３３の加算器５２ｅの後段の加算器５２ｐからＭＦ５２までの構成要素を有する。他の帰還部５０１_３〜５０１_Ｋも帰還部５０１_２と同じ構成を有する。
【０１８７】
他の重み係数生成部５_２〜５_Ｋについても，同様の構成とすることができる。ただし，第ｊ係数生成部５_ｊでは，帰還部５０１_１と同じ構成を有する帰還部がｊ番目の要素ｒ_ｊｊを出力する部分に配置され，それ以外の要素を出力する部分に帰還部５０１_２と同じ構成を有する帰還部が配置される。また，この帰還部では，定数１／（１＋α_１）　に代えて定数１／（１＋α_ｊ）　が加算器に入力される。
【０１８８】
図３３に示す重み係数生成部は，帰還部５０１_１〜５０１_Ｋおよび線形結合部５００の左右の配置を入れ替えることにより，図３４にように表すこともできる。式（１２）において，ｒ_１１＝ｘ_１，ｒ_１２＝ｘ_２，…，ｒ_１Ｋ＝ｘ_Ｋと置きかえ，相関係数を積分表示にもどしてみると，次の式（３０）になる。
【０１８９】
【数１６】

【０１９０】
積分区間は，となっているが，拡散信号ｓ_１（ｔ）〜ｓ_Ｋ（ｔ）は，ある有限な区間で−１か＋１（非零）であるが，その前後で零であるから，積分はある有限な区間で行われる。
式（１００）において，重み係数ｘ_１，ｘ_２，…，ｘ_Ｋが，時間によって変化しない場合は，積分の内側にいれて見ると，次の式（３１）になる。
【０１９１】
【数１７】

【０１９２】
次に，拡散信号と重み係数を線形結合したものを，拡張拡散信号と定義すると，拡張拡散信号は，次の式（３２）で表せる。
【０１９３】
【数１８】

【０１９４】
式（３２）を使って，式（３１）を表すと，次の式（３３）になる。
【０１９５】
【数１９】

【０１９６】
第１の形態で用いた手法では，式（１２）を，図１３および図１４に示す回路形式によって，式（１４）および式（１５）の様に変形し，ｒ_１１（ｘ_１），ｒ_１２（ｘ_２），…，ｒ_１Ｋ（ｘ_Ｋ）を求めた。同じ考えを，式（３３）に適用すると，次の式（３４）となる。
【０１９７】
【数２０】

【０１９８】
式（３４）は，図３５に対応していて，線形結合部５００が拡張拡散信号ｑ_１（ｔ）を出力し，帰還部５０１_１の出力が，式（３４）の第１行の左辺と対応している。帰還部５０１_１〜５０１_Ｋの出力は，式（３４）の第２行の左辺に対応している。
【０１９９】
図３５の回路が，最初は，整合フィルタの出力が零の状態から起動し，定常状態に達するまでの間，重み係数ｘ_１，ｘ_２，…，ｘ_Ｋの値は，初期値から収れん値まで，変化する。
【０２００】
式（３０）を式（３１）に変形（図２９から図３０への変換に対応）できたのは，重み係数ｘ_１，ｘ_２，…，ｘ_Ｋすが，時間によって変化しないと仮定したからである。
【０２０１】
従って，図１３および図１４に示す回路形式によって求められる重み係数ｘ_１，ｘ_２，…，ｘ_Ｋが，厳密解であるのに対して，図３４または図３５に示す回路形式から求まる重み係数ｘ_１，ｘ_２，…，ｘ_Ｋは，近似解である。
近似解であっても，それらの値を用いることで，復相関受信が可能なことと，近似解を厳密解に近づける方法について説明する。
【０２０２】
式（１）で表される受信信号Ｙ（ｔ）に，式（３２）で表される拡張拡散信号ｑ_１（ｔ）を乗じて整合フィルタに通した結果をＳとすると，Ｓは，次の式（３５）で表せる。
【０２０３】
【数２１】

【０２０４】
式（３４）は，Ｇ＝∞とすると，式（３３）になるが，Ｇが小さい場合にも，利得補正を行えば，式（３３）で表せるので，式（３５）に，式（３３）を代入すると，次の式（３６）になる。
【０２０５】
【数２２】

【０２０６】
式（３６）から，α_１＝α_２＝…＝α_Ｋ＝０の場合には，整合フィルタ出力には，干渉信号成分が現れないことが分かる。但し，雑音成分については，積分期間中（すなわち，整合フィルタを通過中），重み係数が変化するので，従来方式によるものとは，異なってくる。
【０２０７】
α_１，α_２，…，α_Ｋが零で無い場合には，同様の理由で，従来のＭＭＳＥ受信機によるものと雑音成分が，変わってくる。
【０２０８】
次は，近似的に求まる重み係数を，厳密解に近づける方法について説明する。第１の方法は，観測窓長をＮとすると，整合フィルタの積分時間を（Ｎ＋１）Ｔとし，第１回の演算を，拡散信号ｇ_ｊ（ｔ）（ｊ＝１，２，…，Ｌ）を使って０≦ｔ≦（Ｎ＋１）Ｔで行った後，第１回目の収れん値を第２回目の初期値とし，拡散信号として，第１回目で使った拡散信号よりも，（Ｎ＋１）Ｔ遅れた拡散信号ｇ_ｊ（ｔ−（Ｎ＋１）Ｔ）（ｊ＝１，２，…，Ｌ）を使って，（Ｎ＋１）Ｔ≦ｔ≦２（Ｎ＋１）Ｔの間，第２回目の演算を行う。得られた重み係数の近似度が充分であれば，ここで演算を終えるが，近似度が充分でなければ，同じように，拡散信号ｇ_ｊ（ｔ−２（Ｎ＋１）Ｔ）（ｊ＝１，２，…，Ｌ）を使って，第３回目の演算を行う。このようにすると，初期値として，前回よりもより厳密解に近い値を使うので，収れん値としては，前回よりもより厳密解に近いものが得られる。反復回数を増やせば，より厳密解に近い値を得ることができる。ただし，この方法だと，演算に要する時間は（Ｎ＋１）Ｔ×反復数となる。
【０２０９】
第２の方法は，整合フィルタとして，積分時間をビット長と同じＴとし，第一回の演算に要する時間帯を，長さＴの複数の区間，区間１，区間２，区間Ｐに分割し，区間毎に拡張拡散信号を生成した上，区間ｐの拡散信号の時間区間ｐと区間ｐ_０の時間差だけシフトして，１回目の演算を区間ｐ_０で行った後，周期Ｔで，演算を反復する方法である。この方法だと，１回の演算に要する時間はＴであり，全体の時間はＴ×反復数となり，第１の方法に比べて，演算時間が（Ｎ＋１）分の１に短縮される。第１の方法と，第２の方法に要する時間を比較したのが，図３６である。尚，この図においては，数字はビット番号ではなく，区間番号を表す。
【０２１０】
（４）第４の形態
第４の形態の無線基地局１２０の受信装置の構成図を図３７に示す。
【０２１１】
この装置の構成は，基本的に，第１の形態および第２の形態と同様であるが，変形拡散信号生成部１２を備え，前記２つの形態に使用した技術を基に，更に重み係数生成部の高速演算と装置の小型化を達成するものである。
【０２１２】
変形拡散信号生成部１２では，拡散信号生成部で生成される，相関係数生成用の拡散信号ｇ_１（ｔ），ｇ_２（ｔ），…，ｇ_Ｌ（ｔ）の他に，それらの拡散信号に対して，所定の時間だけ時間シフトした複数の拡散信号信号を生成する。時間シフトした拡散信号の生成方法に付いては，図１０を使って説明した方法と同様の方法が用いられる。これらの複数の拡散信号を，変形拡散信号Σｇ_ｊ（ｔ）と呼ぶことにする。
【０２１３】
図４を，再び用いて説明すると，この図は，単一レートで，ユーザ数Ｌの受信信号間の非同期状態を図示したもので，受信信号の受信時刻は特定の時刻に集中することなく，一様に分布している場合のモデルで，受信機への到達時刻の早い順にビット番号が付与されている。図４（Ａ）は，観測窓長（Ｏｂｓｅｒｖａｔｉｏｎ　Ｗｉｎｄｏｗ　Ｓｉｚｅ）Ｎ＝２で，Ｋ＝２Ｌの場合を，図４（Ｂ）は，観測窓長Ｎ＝３で，Ｋ＝３Ｌの場合を示している。
【０２１４】
ここで，図４（Ａ）において，希望信号ビットを第（Ｌ＋１）番目のビット（以下「ビットＬ＋１」と表す）とすると，観測窓長Ｎ＝２のシステムでは，希望ビットを復号するのに，ビット１の先頭からビット２Ｌの末尾まで，合計３ビットの時間を要することとなる。
【０２１５】
したがって，ビット１，Ｌ＋１，２Ｌ＋１，３Ｌ＋１というように，ユーザ信号１のビット列を連続的に復号するには，前述した無線基地局装置１２０における重み係数生成部の規模が３倍必要となる。同様に，観測窓長Ｎのシステムでは，希望信号ビットを復調するのに，Ｎ＋１ビットの時間を要するので，あるユーザ信号のビット列を連続的に復号するには，重み係数生成部の規模をＮ＋１倍必要になる。
【０２１６】
一方，受信される各ユーザ信号は，Ｎ＋１ビットの区間に広がっているものの，重み係数の演算時間帯は必ずしも受信信号と同期している必要がないことに着目し，変形拡散信号Σｇ_ｊ（ｔ）を利用して重み係数の演算時間を特定の時間帯に集約して，１ビットの復号に要する時間を短縮した所に，第４の形態の特徴がある。
【０２１７】
第４の形態について，一例として，第３の形態で説明した技術に基づいて実施の形態を説明するが，第１，第２の形態で説明した相関係数生成にも同じように適用できる。
【０２１８】
まず，第４の形態の原理について，まず簡単なモデルで説明し，後に一般化する。
【０２１９】
図３８（ａ）は，一例として，非同期に受信されるシングルレートの２つのユーザ信号１および２のそれぞれ２つのビット（ビット１および３ならびにビット２および４）に対応する拡散信号を示している。ユーザ信号１のビットとユーザ信号２のビットとは，時間τだけずれている。すなわち，ビット２はビット１の開始から時間τ遅れて開始し，ビット３はビット２の開始から時間τ遅れて開始する。
【０２２０】
ビット１〜４のそれぞれの拡散信号をｓ_１（ｔ）〜ｓ_４（ｔ）とし，重み係数をｒ_１１（ｘ_１）〜ｒ_１Ｋ（ｘ_４）とすると，前述した式（１０）から，希望信号のビット番号を３として，以下の式（３７）が得られる。
【０２２１】
【数２３】

【０２２２】
式（１８）を相関係数を求める積分表示のまま表すと，以下の式（３８）になる。
【０２２３】
【数２４】

【０２２４】
式（３８）では，積分区間が−ＴからＴ＋τに亘っている。このことは，受信信号に同期した拡散信号ｓ_１（ｔ）〜ｓ_４（ｔ）を使用して重み係数を求めようとすると，積分時間２Ｔ＋τ（τがＴに近い値を有する場合には約３Ｔ）を要することを意味する。また，ユーザ信号１のビット列を連続的に復号するには，前述したように，重み係数生成部としては，３倍の規模の回路が必要となる。
【０２２５】
一方，被積分関数（すなわち拡散信号）を時間的にシフトさせても，積分区間の始点と終点をそのシフト量だけ変化させれば，積分結果は同じになることを利用すると，積分の時間を特定の時間帯に集約させれば，全体の処理時間を短縮することができる。
【０２２６】
たとえば上記式（３８）の第１式左辺の第２項は次のようになる。
【０２２７】
【数２５】

【０２２８】
同ようにして，式（３８）の第４式左辺の第２項は次のようになる。
【０２２９】
【数２６】

【０２３０】
これら２つの式は拡散信号を１ビット時間Ｔだけ進めるか遅らせるかして，積分時間を０≦ｔ≦Ｔにシフトした例である。式（３８）のすべてについて，積分区間を０≦ｔ≦Ｔにすると，式（３８）は以下の式（３９）になる。
【０２３１】
【数２７】

【０２３２】
このような時間のシフトを図３８によって説明すると，−Ｔ≦ｔ≦０の範囲にある拡散信号ｓ_１（ｔ）のすべておよび拡散信号ｓ_２（ｔ）の一部分は，時間Ｔだけ遅らされ，図３８（ｃ）の位置にシフトする。０≦ｔ≦Ｔの範囲にある拡散信号ｓ_３（ｔ）のすべて，拡散信号ｓ_２（ｔ）の残り部分，および拡散信号ｓ_４（ｔ）の一部分は，図３８（ｂ）に示すようにシフトされず，そのままとされる。Ｔ≦ｔの範囲にある拡散信号ｓ_４（ｔ）の残りの部分は，時間Ｔだけ進められ，図３８（ｄ）に示すようにシフトされる。尚，説明は，時間Ｔだけ“進められ”たり“遅らされ”たりと言う表現をしたが，実際の回路で行うのは，変形拡散信号生成部より，連続信号として送られてくる，変形拡散信号Σｇ_ｊ（ｔ）の，対称ビットと区間に応じて特定の部分を，スイッチで選択することである。
【０２３３】
このように，積分区間を０≦ｔ≦Ｔ内にシフトする変換をより一般化するために，積分区間のシフトを，図４（Ａ）に示す観測窓長Ｎ＝２，ユーザ信号数Ｌの受信信号に適用すると，以下の式（４０）から（４２）になる。（図３９（ａ）〜（ｄ）参照）
１≦ｋ≦Ｌについては，
【０２３４】
【数２８】

【０２３５】
Ｌ＋２≦ｋ≦２Ｌについては，
【０２３６】
【数２９】

【０２３７】
希望ビットＬ＋１については，
【０２３８】
【数３０】

【０２３９】
ここで，ｔ_ｋは，ビット番号ｋのビットの０≦ｔ≦Ｔにおける変換点の時刻である。
【０２４０】
次に，このような原理を図３４または図３５に示す重み係数生成部に適用した実施の形態について説明する。受信信号は，図３９（ａ）に示すものとする。
【０２４１】
上記式（４０）〜（４２）から，図３４（または，図３５）に示す帰還部５０１_２〜５０１_Ｋ（干渉ビットに関する帰還部）の乗算器５２ａ〜５Ｋａのそれぞれの第１入力端子および第２入力端子に入力される拡散信号を，０≦ｔ≦ｔ_ｋとｔ_ｋ≦ｔ≦Ｔとで元のままのを入力するか，または，遅れた，もしくは進んだものを入力させるかで切り替えることにより，シフト処理を行うことができる。
【０２４２】
図４０は，干渉ビットに関する帰還部の１部を図示したもので，その乗算器の第１入力端子（入力１）および第２入力端子（入力２）に入力される拡散信号の対応関係を表に示したものである。
【０２４３】
図４１は，図４０に基づいて構成した重み係数生成部の構成を示すブロック図で，図３５に示す重み係数生成部を基に，必要な機能追加を行ったものである。但し，図４１では，図が煩雑になるのを避けるために，ＭＭＳＥ補正部，帰還ループ利得補正の表示を省略している。
【０２４４】
この重み係数生成部では，干渉ビット１〜Ｌ，Ｌ＋２〜Ｌをそれぞれ処理する帰還部の乗算器の前段に，スイッチＶ_１〜Ｖ_Ｌ，Ｖ_Ｌ＋２〜Ｖ_２Ｌがそれぞれ設けられる。
【０２４５】
ｋ＝１〜ＬのスイッチＶ_ｋは，時刻ｔ_ｋにおいて，破線で示す第１入力端子から実線で示す第２入力端子に切り替えられ，０≦ｔ≦ｔ_ｋでは，分配器８０１の出力信号を後段の乗算器に出力し，ｔ_ｋ≦ｔ≦Ｔでは，分配器８０２の出力信号を後段の乗算器に出力する。
【０２４６】
ｋ＝Ｌ＋２〜２ＬのスイッチＶ_ｋは，時刻ｔ_ｋにおいて，破線で示す第２入力端子から実線で示す第１入力端子に切り替えられ，０≦ｔ≦ｔ_ｋでは，分配器８０３の出力信号を後段の乗算器に出力し，ｔ_ｋ≦ｔ≦Ｔでは，分配器８０１の出力信号を後段の乗算器に出力する。
【０２４７】
干渉ビット１〜Ｌ，Ｌ＋２〜Ｌをそれぞれ処理する帰還部の後段に配置された乗算器（線形結合部の乗算器）の出力側には，スイッチＷ_１〜Ｗ_Ｌ，Ｗ_Ｌ＋２〜Ｗ２Ｌがそれぞれ設けられる。
【０２４８】
ｋ＝１〜ＬのスイッチＷ_ｋは，時刻ｔ_ｋにおいて，破線で示す第１出力端子から実線で示す第２入力端子に切り替えられ，０≦ｔ≦ｔ_ｋでは，前段の乗算器の出力信号を合成器（加算器）７０１に出力し，ｔ_ｋ≦ｔ≦Ｔでは，前段の乗算器の出力信号を合成器（加算器）７０２に出力する。
【０２４９】
ｋ＝Ｌ＋２〜２ＬのスイッチＷ_ｋは，時刻ｔ_ｋにおいて，破線で示す第２出力端子から実線で示す第１出力端子に切り替えられ，０≦ｔ≦ｔ_ｋでは，前段の乗算器の出力信号を合成器７０３に出力し，ｔ_ｋ≦ｔ≦Ｔでは，前段の乗算器の出力信号を合成器７０１に出力する。
【０２５０】
合成器（加算器）７０１〜７０３が入力された信号を合成し，合成結果を分配器８０１〜８０３にそれぞれフィードバックする。合成器７０１は図３９（ａ）の区間２に対応し，合成器７０２は区間１に対応し，合成器７０３は区間３に対応する。このように合成器の個数および区間数はＮ＋１となる。
【０２５１】
合成器７０１〜７０３のそれぞれの出力信号は以下の式により表される。
【０２５２】
【数３１】

【０２５３】
分配器８０１〜８０３は，入力された信号を減算器またはスイッチに分配する。
【０２５４】
干渉ビット１〜Ｌに関する帰還部の２つの乗算器の入力２には，拡散信号ｓ_ｋ（ｔ）の他に，ｓ_ｋ（ｔ−Ｔ）が入力される。また，干渉ビットＬ＋２〜２Ｌに関する帰還部の２つの乗算器の入力２には，拡散信号ｓ_ｋ（ｔ）の他に，ｓ_ｋ（ｔ＋Ｔ）が入力される。また，各帰還部のＭＦの積分時間は０〜Ｔまでとされる。
【０２５５】
以上により，第２の形態では，３Ｔ（Ｔはビット長）掛かっていた重み係数演算の１周期（前述したように，重み係数の演算精度を上げるためには，演算を複数周期反復する必要がある）を，第３側面ではＴに縮めることができ，その結果，回路規模をさらに縮小することができる。
【０２５６】
以上は，処理区間をビット長（Ｔ）単位でシフトし，処理時間を短縮する原理と実施の形態について述べたが，処理区間をＴ／Ｑ　（Ｑは整数）とすると，Ｑ＞１とすることで，処理時間をさらに短縮することができる。
【０２５７】
まずは，Ｑ＝２の場合を説明し，Ｑ＝１と異なる側面について説明し，後に，Ｑ＞２である任意の整数に対しての実施の形態について説明する。
【０２５８】
図４２（ａ）は，図３８（ａ）と同じ拡散信号の配置を示しているが，区間の長さＴ／２ごとに６区間に分割している点が異なる。また，図４４（ｂ）〜（ｄ）は，全ての拡散信号を区間３にシフトされる様子を示している。
【０２５９】
希望信号のビット番号を３とし，対応する拡散信号をｓ_３（ｔ）とする。干渉ビットに対応する拡散信号ｓ_２（ｔ）とｓ_４（ｔ）の境界，すなわち干渉信号のビットの変換点が，区間３にあるか，区間４にあるかにより，取り扱いが異なる。したがって，ビット番号ｋが１≦ｋ≦Ｍのときに，変換が区間３に存在し，ビット番号ｋが，Ｍ＋１≦ｋ≦Ｌのときに，変換点が，区間４に存在するものとすると，ビット番号ｋがＬ＋２≦ｋ≦Ｌ＋Ｍの場合の変換点も区間３に，また，ビット番号ｋＬ＋Ｍ＋１≦ｋ≦２Ｌのときも，変換点が区間４に存在することになる。
【０２６０】
図３９に示すユーザ数Ｌ，観測窓長Ｎ＝２の場合における，上記式（４０）〜（４２）に相当する式を求めると，以下の式（４３）〜（４７）となる。
【０２６１】
１≦ｋ≦Ｍについては，
【０２６２】
【数３２】

【０２６３】
Ｍ＋１≦ｋ≦Ｌについては，
【０２６４】
【数３３】

【０２６５】
Ｌ＋２≦ｋ≦Ｌ＋Ｍについては，
【０２６６】
【数３４】

【０２６７】
Ｌ＋Ｍ＋１≦ｋ≦２Ｌについては，
【０２６８】
【数３５】

【０２６９】
希望ビットについては，
【０２７０】
【数３６】

【０２７１】
ここで，ｔ_ｋは，ビット番号ｋのビットの変換点の時刻であり，０≦ｔ_ｋ≦Ｔである。
【０２７２】
上記式（２４）〜（２８）から明らかなように，すべての積分（すなわちＭＦによる処理）は，０からＴ／２の間で完了する。
【０２７３】
同様にして，Ｑ＞２である整数Ｑに対して，１／Ｑビット長の区間に拡散信号をシフトさせ，Ｔ／Ｑの時間に処理を短縮することもできる。以下，この手法を１／Ｑビット法と名付けて，その手法の実施の形態について説明する。
【０２７４】
図４５は，１／Ｑビット法による重み係数生成部における干渉ビットに関する帰還部および線形結合部の乗算器の部分を示すブロック図である。既に説明した１ビット法との比較を容易にするために，図４６（ａ）に，１ビット法による重み係数生成部（図４３参照）における干渉ビットに関する帰還部および線形結合部の乗算器の部分を示す。
【０２７５】
１／Ｑビット法では，反転増幅器の前段にＱ個の乗算器が必要となる。また，ＭＦの後段にも同じくＱ個の乗算器が必要となる。一方，１ビット法では，図４６（ａ）に示すように，ＭＦの前後にスイッチＶ_ｉおよびＷ_ｉを設けることで，ＭＦの前後の乗算器を１つにまとめることができる。
【０２７６】
この理由は，１／Ｑビット法では，異なる拡散信号同士の乗算結果を同じ時刻で，ＭＦに入力する必要があることによる。たとえば，式（２６）の左辺の第２項および第３項において，ｔ_ｋ≦ｔ≦Ｔ／２では拡散信号ｓ_ｋ（ｔ）と，このｔ_ｋ≦ｔ≦Ｔ／２と一部重なる０≦ｔ≦Ｔ／２の時間帯では，拡散信号ｓ_ｋ（ｔ）を時間Ｔ／２進ませた拡散信号ｓ_ｋ（ｔ＋Ｔ／２）の２つの拡散信号を同時にＭＦに入力する必要がある。一方，１ビット法では，ある拡散信号とそれを時間的にシフトさせた拡散信号とを，同じ時間帯にＭＦに入力することがない。
【０２７７】
図４４を用い，Ｌ＋２≦ｋ≦Ｌ＋Ｍの場合を例にとってさらに具体的に説明する。拡散信号ｓ_４（ｔ）を区間３にシフトするために，拡散信号ｓ_４（ｔ）の区間４の部分をＴ／２進ませ，同信号の区間５の部分の時間をＴ進ませる必要がある。これにより，拡散信号ｓ_４（ｔ）は，符号▲１▼〜▲３▼で示す３つの部分に分割される。部分▲１▼および▲２▼は，時間帯が重ならないので，ＭＦの前段および後段の乗算器を共用でき，したがって，乗算器は前段および後段のそれぞれに１つずつで足りる。
【０２７８】
一方，部分▲２▼は，０≦ｔ≦Ｔ／２の時間帯を占有するので，他の部分▲１▼および▲３▼により使用される乗算器を共用できない。したがって，部分▲２▼の乗算を行う乗算器が別途必要となる。その結果，部分▲１▼および▲３▼により共用される乗算器と，部分▲２▼により使用される乗算器との，計２個の乗算器が必要となる。
【０２７９】
１／Ｑビット法に一般化すると，１／Ｑビット法では，１ビットに対応する拡散信号が（Ｑ＋１）個の部分に分割される。これらの部分のうち，（Ｑ−１）個の部分の長さ（時間）はＴ／Ｑである。残りの２つの部分は，時間的にオーバラップせず，両者を合わせて調度Ｔ／Ｑの長さになるので，これら２つの部分を処理する乗算器が，図４５における最上段の乗算器７００_１および８００_１に対応する。乗算器７００_１の前段および８００_１の後段には，スイッチがそれぞれ設けられ，２つの部分が切り替えられて処理される。乗算器７００_２〜７００_Ｑおよび８００_２〜８００_Ｑは，（Ｑ−１）個の部分を処理する乗算器である。このようにして，１／Ｑビット法では，乗算器が合計２Ｑ個必要となる。
【０２８０】
なお，２つのスイッチをなくして，時間がオーバラップしない２つの部分を個別に処理する２つの乗算器を設けることもできる。この場合は，乗算器の数は合計２Ｑ＋２個となる。また，１ビット法においても同ようにして，図４６（ｂ）に示すように，スイッチをなくして，２つの乗算器を設けることもできる。
【０２８１】
図４５の破線で囲まれた回路を回路Ｊとすると，１／Ｑビット法による重み係数生成部は，図４７に示す構成となる。この重み係数生成部は，Ｋ個の回路Ｊ_１〜Ｊ_Ｋ，Ｐ個の合成器９００_１〜９００_Ｐ，およびＰ個の分配器９０１_１〜９０１_Ｐを有する。図面が煩雑になるのを避けるために，回路Ｊ_１〜Ｊ_Ｋと合成器９００_１〜９００_Ｐとの結線を接続網Ｒとして示している。分配器９０１_１〜９０１_Ｐと回路Ｊ_１〜Ｊ_Ｋとの続接も続接網Ｌとして示している。
【０２８２】
ここで，Ｐは観測区間（観測窓長Ｎの場合にはＮ＋１ビットの範囲）の分割数であり，図４４の例では，Ｐ＝６である。
【０２８３】
回路Ｊ_１〜Ｊ_Ｋの接続網Ｒへの出力信号は，図４５から明らかなように，シフトされた拡散信号と重み係数との乗算結果である。この出力信号は，シフト前の拡散信号が属する区間に対応する合成器に接続される。したがって，接続網Ｒの結線は，拡散信号の配置が決まれば，一意に定まる。
【０２８４】
分配器９０１_１〜９０１_Ｋには，合成器９００_１〜９００_Ｐの出力信号がフィードバック信号をして入力される。そして，分配器９０１_１〜９０１_Ｋは，合成器の出力を接続網Ｌを介して回路Ｊ_１〜Ｊ_Ｋに分配する。右側の合成器と左側の分配器は，同じ区間同士のものが１対１に接続され，系全体が帰還ループを形成する。接続網Ｌの結線構造は，接続網Ｒと対称となる。
【０２８５】
変形拡散信号部１２からは、拡散信号ｇ_１（ｔ）〜ｇ_Ｌ（ｔ）の他に、それらに対して、整合フィルタの積分時間の１／Ｑの整数倍だけ進んだ、または遅れた複数の信号（変形拡散信号ｇ_１（ｔ）〜ｇ_Ｌ（ｔ）または、変形拡散信号Σｇ_ｊ（ｔ））が出力され、回路網Ｓに入力する。回路網Ｓの出力は、重み係数生成部の乗算器（図４３では、乗算器７００_１〜７００_Ｑおよび乗算器８００_１〜８００_Ｑ）に接続される。前記乗算器に入力する拡散信号の長さは、１／Ｑビットまたはそれ以下なので、変形拡散信号ｓ_１（ｔ）〜ｓ_Ｋ（ｔ）または変形拡散信号Σｓ_ｊ（ｔ）と名付けるが、変形拡散信号Σｓ_ｊ（ｔ）は、回路網Ｓの入出力間に設けられたスイッチの接続のタイミングと長さを変えることで生成される。スイッチの制御は、（図３７には図示されない）外部の制御回路によって行われる。
【０２８６】
非同期系のシステムにおいては、受信信号が特定の区間に集中することがあり得る。従って、合成器への入力数は、対応する区間によって大きかったり、小さかったりする。その場合には、回路網Ｒの入出力の端子数を増やし、入出力の端子間にスイッチを設け、合成器への必要な経路の数を確保するよう（図３７には図示されない）、外部の制御回路によって行われる。回路網Ｌに対しても、同様な処置が施される。
【０２８７】
次に，第４の形態の説明で示した，装置構成に基づいてで，実際の動作をシミュレーションによって確認したので，その結果を説明する。
シミュレーションのモデルは，Ｎ＝２，Ｌ＝２，Ｋ＝Ｎ×Ｌ＝４，ビット配置と拡散信号パターンは，図２１に示した，第１，第２の形態に即して行ったシミュレーションと同じである。第４の形態として説明した，１ビット法を適用して，希望信号ビット（ビット番号３）の区間を区間２，とし，その前の区間である区間１と，後の区間である区間３にある拡散信号を，全て区間２にシフトした。装置構成は，図４１に示すものである。乗算器の入力（図４０で定義した入力２）には，各拡散信号を周期Ｔで繰り返し入力し，帰還ループが定常状態に達し，正しい重み係数が求まるまでの様子を図４６に示す。縦軸は，重み係数の値で，横軸は，ビット長Ｔで規格化した時間で，希望信号ビット（ビット番号３）の始まりをｔ＝０とした。モデルが単純なため，収れんの様子が分かり難いので，重み係数の厳密解（相関係数行列の逆行列として，求まる。図２２（ｃ）参照）からの偏差を示したのが，図４７である。この図から，第３，第４の形態で示した技術によって，相関形受信に必要な重み係数が求まることが分かる。
【０２８８】
＜移動機の構成＞
次に，本発明の実施の形態による移動機２００について説明する（第５の形態）。移動機２００の受信装置の構成は，基本的には無線基地局装置１２０と変わらないが，異なる点は，自局宛の信号のみを復号すればよいので，逆行列要素生成部，線形結合部，および判定回路の数は少なくなる。また，同じ基地局から送られる自局宛と他局宛の信号は，通常，同期して送られ，かつそれらは直交している（逆拡散後の信号に干渉信号として現れない）ので，それらの信号の，伝送経路が異なり，遅延時間が違ってきた場合のみ，干渉信号源となる。
【０２８９】
逆行列要素生成部５_１についても，無線基地局装置１２０について述べた種々の形態の何れも適用することができるが，図３５の線形結合部の出力である拡張拡散信号ｑ_１（ｔ）を，重み付け係数を求めると同時に，受信信号の逆拡散に使用することが，整合フィルタ３の構成を簡略化するので，その方法について説明する。
【０２９０】
図４８は，移動機２００の構成を示すブロック図である。この図では，ユーザ信号数（ユーザ数）Ｌ＝１の復号に必要な部分のみを示している。また，第１〜４の形態による無線基地局装置１２０と同じ構成要素には同じ符号を付し，その詳細な説明を省略する。
【０２９１】
図４８に示す移動機２００には，逆行列要素生成部５_１〜５_Ｋに代えて，（Ｎ＋１）個の重み係数生成部１９_１，１９_Ｌ＋１，１９_２Ｌ＋１，…，１９_ＮＬ＋１が設けられ，整合フィルタバンク３に代えて拡張整合フィルタバンク１３が設けられる。また，線形結合部６_１〜６_Ｋは省略される。
【０２９２】
ここで，Ｎは図４に示す受信信号の観測窓長（図４（Ａ）ではＮ＝２，（Ｂ）ではＮ＝３），Ｌはユーザ信号数（ユーザ数）であり，拡散信号生成部４の符号の下付き符号“１”，“Ｌ＋１”，“ＮＬ＋１”等は，各ユーザ信号の観測窓内におけるビット番号に対応している。
【０２９３】
図４９は，拡張整合フィルタバンク１３の構成を示すブロック図である。合成部１３は，（Ｎ＋１）個の乗算器１３０_１，１３０_Ｌ＋１，１３０_２Ｌ＋１，…，１３０_ＮＬ＋１，（Ｎ＋１）個のＭＦ１３１_１，１３１_Ｌ＋１，１３１_２Ｌ＋１，…，１３１_ＮＬ＋１，（Ｎ＋１）個のＳＨＣ１３２_１，１３２_Ｌ＋１，１３２_２Ｌ＋２，…，１３２_ＮＬ＋１，を有する。
【０２９４】
重み係数生成部１９_１，１９_Ｌ＋１，１９_２Ｌ＋１，…，１９_ＮＬ＋１は，拡張拡散信号ｑ_１（ｔ），ｑ_Ｌ＋１（ｔ），ｑ_２Ｌ＋１（ｔ），…，ｑ_ＮＬ＋１（ｔ）をそれぞれ出力する。ここで，拡張拡散信号ｑ_ｉ（ｔ）は，式（３２）により表される。
【０２９５】
拡張拡散信号ｑ_１（ｔ），ｑ_Ｌ＋１（ｔ），ｑ_２Ｌ＋１（ｔ），…，ｑ_ＮＬ＋１（ｔ）は，合成部１３の乗算器１３０_１，１３０_Ｌ＋１，１３０_２Ｌ＋１，…，１３０_ＮＬ＋１の第２入力端子にそれぞれ入力される。乗算器１３０_１，１３０_Ｌ＋１，１３０_２Ｌ＋１，…，１３０_ＮＬ＋１は，第１入力端子に入力されたベースバンド信号Ｙ（ｔ）と第２入力端子にそれぞれ入力された拡張拡散信号ｑ_１（ｔ）〜ｑ_ＮＬ＋１（ｔ）をそれぞれ乗算し，乗算結果をＭＦ１３１_１〜１３１_ＮＬ＋１にそれぞれ与える。乗算結果は，ＭＦ１３１_１〜１３１_ＮＬ＋１によってそれぞれ積分される。積分時間は，ビット長をＴとすると，（Ｎ＋１）Ｔである。積分結果はＳＨＣ１３２_１〜１３２_ＮＬ＋１に与えられる。
【０２９６】
ＳＨＣ１３２_１〜１３２_ＮＬ＋１は，それぞれの観測窓の終端でサンプリングした値を出力する。
【０２９７】
拡張整合フィルタバンク１３の出力は，判定回路７_１によって，離散的値または，符号が識別される。
【０２９８】
尚，ここで使用する拡散信号は，重み付け係数を求めるためだけではなく，受信信号の逆拡散にも使用するので，受信信号に対して，チップ同期（チップ長の数分の１以下の精度で同期している）している必要がある。重み付け係数の演算精度を高めるために，複数回反復する場合は，最終回を受信信号と同期させる。図３６の例では，（ａ）の場合は，第３回目が受信信号とチップ同期させる。ただし，最終回を除いては，（ｂ）に示した，時間シフトした変形拡散信号を用いて，演算を行い，その収れん値を最終回の初期値として演算を反復することは可能である。
【０２９９】
以上，第１の形態から第５の形態まで説明したが，それらに対して共通の変形例について，説明する。
【０３００】
希望信号のビット番号をｋとし，希望ビットに対する重み係数ｘ_ｋで，他のすべての重み係数を割った値を，新たな重み係数ｘ_ｊ´＝ｘ_ｊ／ｘ_ｋ（ｊ＝１，２，…，Ｋ）として，線形結合部６に入力すると，その出力は，ｘ_１，ｘ_２，…，ｘ_Ｋを入力した場合に比べて，１／ｘ_Ｋ倍されたものとなる。従って，ｘ_ｋをビット毎に求める代わりに，予測される平均値を用いてレベルの変化を無視するか，送信される変調信号に振幅情報がない（例えば，ＢＰＳＫ）場合には，ｘ_ｋを定数（例えば１）とすることができる。この場合，希望信号ビットに対する重み係数を求める装置（図１３に示した第１係数生成部，図３４に示した，希望ビットに対する帰還部５０１_１）を省略することができる。
【０３０１】
図１７に示した帰還ループの利得に応じた誤差の補正（利得補正），は，図３４（または図３５）に示した整合フィルタ（ＭＦ）５１_ｂ，５２_ｂ，…，５Ｋ_ｂに対して適用しても，同じ効果がある。すなわち，整合フィルタ（ＭＦ）５１_ｂ，５２_ｂ，…，５Ｋ_ｂの出力をそれぞれの入力に加える。このとき，フィルタの入出力で形成される局所的帰還ループの利得は，それぞれの整合フィルタの出力から入力へのパスに，増幅器か減衰器を入れて調整する。
【０３０２】
帰還部と線形結合部で形成される帰還ループにおいて，反転増幅器とＬＰＦ（または整合フィルタ）の接続順を変えることができる（図１３の例では、増幅器５１ｇとＬＰＦ５１ｈの接続順を変える。図１４の例では、増幅器５２ｇとＬＰＦ５２ｈの接続順を変える）。また，反転増幅器は，帰還ループ内であれば位置を変えることができる。ただし，帰還ループ内で，加算復は，減算を行う箇所があれば，そのレベルと符号を変える必要がある。
【０３０３】
図３４に示した回路で，合成器５１ｅの出力から，帰還部５０１_１〜５０１_Ｋるの入力までの間に，増幅器または，反転増幅器を挿入し，各帰還部（５０１_１〜５０１_Ｋ）内の反転増幅器の一部または全部を省略することができる。但し，増幅器を入れた場合は，帰還ループ内の何れかの箇所に反転器を挿入して，負帰還ループとして動作させる必要がある。
【０３０４】
帰還ループ内の何れかのコンポーネント（例えば、図１３、図３４の乗算器５１ｄ_１〜５１ｄ_ｋ、図３４の乗算器５１ａ〜５Ｋａ，ＭＦ５１ｂ〜ＭＦ５Ｋｂ）が能動素子で形成され，反転増幅器なしで，帰還ループの利得が確保できる場合は，反転増幅器を反転減衰器に置き換えることができる。
【０３０５】
これまで述べた実施の形態は本発明を無線通信に適用したものであるが，本発明は無線通信だけでなく，有線通信にも適用することができる。
【０３０６】
【発明の効果】
本発明の，特に請求項１〜３に記載の手法を用いた場合には，干渉除去の対象となるビット数をＫとすると，従来Ｋの３乗に比例していた回路規模を，Ｋの２乗に比例した規模に縮小できる。
【０３０７】
また，希望信号１ビットの復号に必要なＫ個の重み係数を求める時間を短縮でき，相関係数を求める整合フィルタの積分時間も短縮できる。
【０３０８】
さらに，Ｋに比べ，復号するデータ列数Ｍが小さい場合（移動機に適用した場合がこれに該当する）には，整合フィルタの数を減らすことができる。
【図面の簡単な説明】
【図１】本発明の実施の形態による移動通信ネットワークシステムの全体構成を示すブロック図である。
【図２】無線基地局装置により受信される非同期かつマルチレートの場合のＤＳ−ＣＤＭＡ方式による信号の構成を示す。
【図３】無線基地局装置により受信されるＤＳ−ＣＤＭＡ方式による信号の構成を示し，（Ａ）は非同期の場合の受信信号の構成を，（Ｂ）は同期している場合の受信信号の構成をそれぞれ示す。
【図４】観測窓の概念を示す図で，（Ａ）は観測窓長Ｎ＝２，ユーザ数Ｌのビット列を示し，（Ｂ）は観測窓長Ｎ＝３，ユーザ数Ｌのビット列を示す。
【図５】無線基地局のマルチユーザ受信装置の構成を示すブロック図である。
【図６】拡散信号とバースト拡散信号の関係を示す。
【図７】整合フィルタバンクの詳細な構成を示すブロック図である。
【図８】相関係数を求めるための整合フィルタバンクの構成を示すブロック図である。
【図９】相関係数を求めるための整合フィルタバンクの他の構成を示すブロック図である。
【図１０】拡散信号生成部の構成を示すブロック図である。
【図１１】第１の形態による無線基地局装置（特にマルチユーザ受信部）の構成を示すブロック図である。
【図１２】第１の形態における重み係数生成部の構成を示すブロック図である。
【図１３】重み係数生成部の第１係数生成部の詳細な構成を示すブロック図である。
【図１４】重み係数生成部の第２係数生成部の詳細な構成を示すブロック図である。
【図１５】線形結合部の構成を示すブロック図である。
【図１６】利得補正を行った第１係数生成部の帰還部の構成を示す図である。
【図１７】利得補正を行った第１係数生成部の帰還部の他の構成を示す図である。
【図１８】ＭＭＳＥ補正を行う箇所を変更した第１係数生成部の帰還部の構成を示す図である。
【図１９】ＭＭＳＥ補正を行う箇所を変更した第１係数生成部の帰還部の他の構成を示す図である。
【図２０】図１３に示した第１係数生成部の帰還部の他の構成を示す図である。
【図２１】シミュレーションに用いたユーザ信号とそれに対応する拡散信号の波形を示す図である。
【図２２】シミュレーションに用いた拡散信号のチップ列と相関係数行列の逆行列の具体的数値を示す図である。
【図２３】シミュレーションの結果である，重み係数の時間変化を示す図である。
【図２４】シミュレーションの結果である，重み係数の厳密解からの偏差の時間変化を示す図である。
【図２５】シミュレーションの結果である，重み係数の厳密解からの偏差の時間変化を示す図である。
【図２６】第１の形態（変形例）による無線基地局装置の構成を示すブロック図である。
【図２７】相関係数生成部の構成を示すブロック図である。
【図２８】重み係数生成部の構成を示すブロック図である。
【図２９】第１係数生成部の変換の第１ステップで，サンプル保持回路を除去した構成を示す。
【図３０】第１係数生成部の変換の第２ステップで，重み係数を乗じる位置をＭＦの出力から入力に変更した構成を示す。
【図３１】第１係数生成部の変換の第３ステップで，ＭＦと帰還ループフィルタを置換した構成を示す。
【図３２】第１係数生成部の変換の第４ステップで，乗算器による乗算の順序を入れ換えた構成を示す。
【図３３】第２係数生成部の変換後の構成を示すブロック図である
【図３４】重み係数生成部（第２の形態）の詳細な構成を示すブロック図である。
【図３５】重み係数生成部（第２の形態）の詳細な構成を示すブロック図である。
【図３６】重み係数生成部における反復演算についての２つの手法を示す図である。
【図３７】変形拡散信号生成部を有する第３の形態による無線基地局装置の構成を示すブロック図である。
【図３８】（ａ）は，非同期に受信されるシングルレートの２つのユーザ信号のそれぞれ２つのビットに対応する拡散信号の位置を示し，（ｂ）〜（ｄ）は，これらの拡散信号を希望信号ビットのある区間（１ビット長）にシフトさせたものを示している。
【図３９】（ａ）は，Ｌ個のユーザ信号の各ビットに対応する拡散信号の位置を示し，（ｂ）〜（ｄ）は，各拡散信号を希望信号ビットのある区間（１ビット長）へシフトした状態を示す。
【図４０】重み係数生成部（第２の形態）の帰還部における乗算器の２つの入力と区間の対応関係を示す。
【図４１】第３の形態における重み係数生成部の構成を示すブロック図である。
【図４２】（ａ）は，図３８（ａ）と同じ拡散信号の配置を示し，区間を１／２ビットごとに６区間に分割している点が異なる。（ｂ）〜（ｄ）は，各区間の拡散信号を希望信号ビットの前半である区間３にシフトされた様子を示す。
【図４３】１／Ｑビット法による重み係数生成部における帰還部および線形結合部の乗算器の構成を示すブロック図である。
【図４４】（ａ）および（ｂ）は１ビット法による帰還部および線形結合部の乗算器の構成を示す。
【図４５】１／Ｑビット法による重み係数生成部の構成を示すブロック図である。
【図４６】１ビット法を検証するために行ったシミュレーションの結果（重み係数の時間変化）を示すグラフである。
【図４７】シミュレーションの結果を重み係数の厳密解からの偏差で示したグラフである。
【図４８】本発明の実施の形態による移動機の構成を示すブロック図である。
【図４９】拡張拡散信号による逆拡散を行うための整合フィルタバンクの構成を示す図である。
【符号の説明】
１２０_１〜１２０_４　無線基地局装置。
２００　移動機
３　整合フィルタバンク
４　拡散信号生成部
５_１〜５_Ｋ，９_１〜９_Ｋ，１８_１〜１８_Ｋ，１９_１〜１９_ＮＬ＋１　重み係数生成部
６_１〜６_Ｋ，５１２，５２２，１１ａ_１〜１１ａ_Ｋ　線形結合部
８　ＭＭＳＥ補正部
１０，５１１，５２１　相関係数生成部
１１　重み係数演算部
１２　変形拡散信号生成部
１３　拡張整合フィルタバンク
１４　逆行列演算部
１５，１０ａ_１〜１０ａ_Ｋ　相関係数整合フィルタバンク
１６　変形整合フィルタバンク
５１３，５２３，１１ｂ_１〜１１ｂ_Ｋ　帰還部
３ａ_１〜３ａ_Ｋ，５１ａ_１〜５１ａ_Ｋ，５１ｄ_１〜５１ｄ_Ｋ　乗算器
３ｂ_１〜３ｂ_Ｋ，５１ｂ_１〜５１ｂ_Ｋ　整合フィルタ（ＭＦ）
５１ｇ，５２ｇ　反転増幅器
５１ｅ，５２ｅ，７０_１〜７０_ｐ　合成器
８０_１〜８０_ｐ　分配器[0001]
TECHNICAL FIELD OF THE INVENTION
The present invention relates to a multi-user receiver in a code division multiple access (CDMA) system in which the same frequency is shared by a plurality of users at the same time, and more particularly to the interference of other signals with a desired signal by using their spreading code information. And a multi-user receiver that removes or suppresses based on Further, the present invention relates to a base station device and a mobile device provided with such a multi-user receiver.
[0002]
[Prior art]
In a code division multiple access (CDMA) system, a plurality of different users use the same frequency at the same time. Therefore, multiple access interference (MAI) occurs in which a plurality of signals interfere with each other.
[0003]
Also, when the same signal is transmitted from the transmitter, the same signal is directly received by the receiver on the one hand, and reflected by a reflector (for example, a building) on the other hand, and then received by the receiver, the multipath interference ( MPI: Multiple {Path} Interference occurs. Due to these interferences, on the receiver side, a signal to be received (hereinafter referred to as a “desired signal”) becomes another noise source.
[0004]
For this reason, in the CDMA system, it is important to remove or suppress the interference signal component from the desired signal in order to decode the desired signal with good quality.
[0005]
Among such receivers for removing or suppressing the interference signal component, those that use the spreading code information of the desired signal and the interference signal are generally called multi-user receivers. Conventionally, a decorrelation detector (decorator) or a minimum mean square error (MMSE) detector is well known (for example, see Non-Patent Documents 1 and 2).
[0006]
[Non-patent document 1]
Verdu, S, Multiuser Detection, Cambridge, UK: CambridgeUniv. Press, 1998.
[0007]
[Non-patent document 2]
Moshave, {S. Author, "Multi-user detection for DS-CDMA communications", IEEE Communications Magazine, magazine, 34th edition, October 10, 1996, pp. 10-28. 124-136
[0008]
[Problems to be solved by the invention]
Assuming that the number of outputs of the matched filter bank (that is, the number of bits to be processed) is K, the decorrelation detector (decorator) and the MMSE detector have a K × K correlation coefficient matrix (or a diagonal of the correlation coefficient matrix). By multiplying the K output signals from the matched filter bank by an inverse matrix of a matrix in which elements are corrected with background noise and interference signal power, interference signal components included in a desired signal are removed or suppressed.
[0009]
However, since the operation of the inverse matrix is generally complicated and time-consuming, if this operation is performed by a normal digital processor, real-time processing that follows the reception speed becomes difficult. Since the circuit scale of a digital processor increases in proportion to the cube of K, when K increases, the same multiplier is repeatedly used using a fixed number of multipliers and memories. Real-time processing according to the speed of the game becomes difficult. As a result, if the value of K to be subjected to interference cancellation or suppression is limited, the performance of interference cancellation or suppression is limited.
[0010]
On the other hand, there is a method of repeatedly using a value of one cycle calculated in advance by using a short-cycle spread signal. However, generally, in an asynchronous system, a code having a long cycle is often used. Is not a valid solution.
[0011]
The present invention has been made in view of such a background, and an object of the present invention is to provide a multi-user reception system capable of removing or suppressing multiple access interference (MAI) in real time regardless of the length of the period of a spread signal. The machine.
[0012]
[Means for Solving the Problems]
In order to achieve the above object, a multi-user receiver according to a first aspect of the present invention comprises a plurality of (L) spread signals f from a received signal of a DS-CDMA system._i(T) a spread signal generator for generating (i = 1, 2,..., L);
A matched filter bank for inputting an output obtained by despreading the received signal with the spread signal to a plurality of matched filters, sampling an output of the matched filter, and outputting the sampled output;
A plurality (K) of values (y) in the output of the matched filter bank₁, Y₂, ..., y_K), K spread signals s_j(T) from (j = 1, 2,..., K), a correlation coefficient generation unit for generating a K × K matrix of correlation coefficients R, using the correlation coefficients between the spread signals as elements, A modified phase for generating a modified correlation coefficient matrix R ', in which a value proportional to the ratio of the background noise power density to the estimated received power per bit of the received signal corresponding to the diagonal element is added to the diagonal element of R. A relation number generator,
The K values (y₁, Y₂, ..., y_K), And K weighting factors (x₁, X₂, ..., x_K), A weighting factor generator,
If k is the bit number of the desired signal,
The inverse matrix of the correlation coefficient matrix R or the modified correlation coefficient matrix R ′ is calculated, and the k-th row or the k-th column of the inverse matrix is replaced with the K weighting factors (x₁, X₂, ..., x_K)age,
The K weighting factors (x₁, X₂, ..., x_K) And the K values (y₁, Y₂, ..., y_K))₁x₁+ Y₂x₂+ ... + y_Kx_KAnd a linear combination for obtaining
In a multi-user receiver that decodes a desired signal bit by determining a discrete level or a sign with respect to the output of the linear combination unit,
The K values (y₁, Y₂, ..., y_K), The observation section including the section in which the received signal is present is equally divided by a predetermined length τ, and numbers are assigned in order from the earliest time zone. Called
The spread signal generation unit generates the spread signal f_i(T) whether a plurality of modified spread signals, or a whole of the modified spread signals, each advanced or delayed by an integer multiple of τ with respect to (i = 1, 2,..., L) for a predetermined time; , Generates a delayed signal,
One of the P sections is referred to as section p₀age,
p <p₀For the interval p, the spread signal s_i(T) The part belonging to the section p of (j = 1, 2,..., K) is represented by the time (p₀-P) A part that coincides between the signal delayed by τ and the modified extended signal is a spread signal s_ip(T),
p> p₀For the interval p, the spread signal s_i(T) The part belonging to the section p of (j = 1, 2,..., K) is represented by time (pp₀) A part that matches the signal advanced by τ and the modified extended signal is represented by a spread signal s_ip(T),
Section p₀For the spread signal s_j(T) (j = 1, 2,..., K) is the section p₀The part belonging to the spread signal s_ip(T),
The weighting factor generator includes:
The calculation of the correlation coefficient or the weight coefficient is performed by using the spread signal s_ip(T) (j = 1, 2,..., K, p = 1, 2,..., P) and a plurality of matched filters whose integration time is τ equal to the section length, and the section p₀Perform in.
[0013]
A multi-user receiver according to a second aspect of the present invention includes a spread signal generation unit that generates a plurality (L) of spread signals from a received signal of a DS-CDMA system.
A matched filter bank for inputting an output obtained by despreading the received signal with the spread signal to a plurality of matched filters and outputting a sampled output of the matched filter;
A plurality (K) of values (y) in the output of the matched filter bank₁, Y₂, ..., y_K) Corresponding to the spread signal s_jFrom (t) (j = 1, 2,..., K), K weighting factors (x₁, X₂, ..., x_K), A weighting factor generator,
The K weighting factors (x₁, X₂, ..., x_K) And the K values (y₁, Y₂, ..., y_K))₁x₁+ Y₂x₂+ ... + y_Kx_KAnd a linear combination for obtaining
In a multi-user receiver that decodes a desired signal bit by determining a discrete level or a sign with respect to the output of the linear combination unit,
The weighting factor generator includes:
The K spread signals s_j(T) (j = 1, 2,..., K) and K weighting factors x_j(J = 1, 2,..., K) and output s_j(T) x_jAnd a linear combination unit composed of K multipliers j (j = 1, 2,..., K) for obtaining
A second multiplier j (j = 1, 2,..., K) for multiplying the output of the linear combination unit by each of the spread signals;
Provided with K matched filters j (j = 1, 2,..., K);
If k is the bit number of the desired signal,
From the output of the linear combination unit, a spread signal s corresponding to a desired signal bit is obtained._kAfter subtracting (t), the spread signal s_k(T) is multiplied by a second multiplier k, the output is inverted by an inverting amplifier, and an output obtained by adding a predetermined constant to the output passed through the matched filter k is used as a weighting factor x_kAnd connected to the multiplier k of the linear combination section,
The output of the linear combination unit includes the spread signal s_kK-1 different spread signals s different from (t)_j(T) is multiplied by a second multiplier j, its output is inverted by an inverting amplifier, and the output passed through a matched filter j is x_jA configuration connected to the multiplier j of the linear combination unit,
Alternatively, in addition to the above configuration, the output s of the multiplier j (j = 1, 2,..., K)_j(T) x_jIs multiplied by a value proportional to the ratio between the background noise power density and the estimated received power per bit of the received signal corresponding to the bit number j, and the input to the second multiplier j (j = 1, 2,..., K) Side.
[0014]
The multi-user receiver according to the third aspect of the present invention comprises a plurality of (L) spread signals f from a received signal of a DS-CDMA system._j(T) a spread signal generator for generating (j = 1, 2,..., L);
A matched filter bank for inputting an output obtained by despreading the received signal with the spread signal to a plurality of matched filters, sampling an output of the matched filter, and outputting the sampled output;
A plurality (K) of values (y) in the output of the matched filter bank₁, Y₂, ..., y_K), K spread signals s_jFrom (t) (j = 1, 2,..., K), K weighting factors (x₁, X₂, ..., x_K), A weighting factor generator,
The K weighting factors (x₁, X₂, ..., x_K) And the K outputs (y₁, Y₂, ..., y_K))₁x₁+ Y₂x₂+ ... + y_Kx_KAnd a linear combination for obtaining
In a multi-user receiver that decodes a desired signal bit by determining a discrete level or a sign with respect to the output of the linear combination unit,
The K values (y₁, Y₂, ..., y_K), The observation section including the section in which the received signal is present is equally divided by a predetermined length τ, and numbers are assigned in order from the earliest time zone. Called
The spread signal generation unit generates the spread signal f_i(T) whether a plurality of modified spread signals, or a whole of the modified spread signals, each advanced or delayed by an integer multiple of τ with respect to (i = 1, 2,..., L) for a predetermined time; , Generates a delayed signal,
The weighting factor generator includes:
K matched filters j (j = 1, 2,..., K) whose integration time is τ equal to the section length;
P synthesizers p (p = 1, 2,..., P) with multiple inputs and one output;
P distributors p (p = 1, 2,..., P) with one input and multiple outputs;
A first connection network L connecting the P distributors p and the K matched filters j,
A second connection network R connecting the K matched filters j and the P synthesizers p,
Including a plurality of multipliers,
The plurality of multipliers are referred to as a first multiplier ip and a second multiplier ip in association with a bit number i and a section p,
One of the P sections is referred to as section p₀age,
p <p₀For the interval p, the spread signal s_i(T) The part belonging to the section p of (j = 1, 2,..., K) is represented by the time (p₀-P) A part that coincides between the signal delayed by τ and the modified extended signal is a spread signal s_ip(T),
p> p₀For the interval p, the spread signal s_i(T) The part belonging to the section p of (j = 1, 2,..., K) is represented by time (pp₀) A part that matches the signal advanced by τ and the modified extended signal is represented by a spread signal s_ip(T),
Section p₀For the spread signal s_j(T) (j = 1, 2,..., K) is the section p₀The part belonging to the spread signal s_ip(T),
Spread signal s_ip(T) (j = 1, 2,..., K, p = 1, 2,..., P) and weight coefficient x_j(J = 1, 2,..., K) and the output obtained by combining the result of multiplication by the first multiplier jp by the combiner p via the second connection network R,
Output through a distributor p and a first connection network L,
If k is the bit number of the desired signal,
If i = k, the modified spread signal s is obtained from the output of the first connection network L._ip(T) minus the spread signal s_ip(T) is multiplied by a second multiplier ip,
If i is different from k, the output of the first connection network L and the spread signal s_ip(T) is multiplied by a second multiplier ip,
The output of the second multiplier is input to a matched filter i via an inverting amplifier, and if i = k, a predetermined constant is added to the output of the matched filter i;
If i is different from k, weight coefficient x_iHas a configuration in which connection to the first multiplier ip is performed for all or some of p.
Or, in addition to the above configuration,
Output s of first multiplier ip_jp(T) x_jIs multiplied by a value proportional to the ratio between the background noise power density and the estimated received power per bit of the received signal corresponding to the bit number j, and is input to the input of the second multiplier jp on the first circuit network L side. There is a configuration in which the addition is performed for all or a part of p.
[0015]
A multi-user receiver according to a fourth aspect of the present invention includes a spread signal generation unit that generates a plurality (L) of spread signals from a received signal of a DS-CDMA system.
A matched filter bank that inputs an output obtained by despreading the received signal with the spread signal to a plurality of matched filters, and outputs a value obtained by sampling the output of the matched filter;
A plurality (K) of values (y) in the output of the matched filter bank₁, Y₂, ..., y_K), K spread signals s_j(T) From (j = 1, 2,..., K), a correlation coefficient generation unit that generates a K × K matrix correlation coefficient matrix R using the correlation coefficients between the spread signals as elements, or R A modified correlation coefficient generation unit that generates a modified correlation coefficient matrix R ′ by adding the ratio of the background noise power density to the estimated received power per bit of the received signal corresponding to the diagonal element ,
The K values (y₁, Y₂, ..., y_K), And K weighting factors (x₁, X₂, ..., x_K), A weighting factor generator,
The K weighting factors (x₁, X₂, ..., x_K) And the K values (y₁, Y₂, ..., y_K))₁x₁+ Y₂x₂+ ... + y_Kx_KAnd a linear combination for obtaining
In a multi-user receiver that decodes a desired signal bit by determining a discrete level or a sign with respect to the output of the linear combination unit,
The weighting factor generator includes:
The jth column or the jth row (a) of the correlation coefficient matrix R or the modified correlation coefficient matrix R ′_1j, A_2j, ..., a_Kj) And weighting factor (x₁, X₂, ..., x_K), And the K multipliers and their outputs are all added to produce a_1jx₁+ A_2jx₂+ ... + a_Kjx_K, And K linear combination units j (j = 1, 2,..., K) configured by a synthesizer that generates
If k is the bit number of the desired signal,
The first constant is subtracted from the output of the linear combination part k, inverted by an inverting amplifier, passed through a low-pass filter, and the output obtained by adding the second constant to the output is x_kConnected to the K linearly coupled sections as
The output of the K−1 linear combination units j, where j ≠ k, is inverted by an inverting amplifier and passed through a low-pass filter, is expressed as x_jAnd a configuration connected to the K linear coupling sections.
[0016]
A wireless base station apparatus according to the present invention is a wireless base station apparatus provided in a mobile communication network system, and includes the multi-user receiver according to any one of claims 1 to 17.
[0017]
A mobile device according to the present invention is a mobile device that performs wireless communication by connecting to a mobile communication network, and includes the multi-user receiver according to any one of claims 1 to 17.
[0018]
BEST MODE FOR CARRYING OUT THE INVENTION
An embodiment in which the present invention is applied to a wireless base station device and a mobile device in a mobile communication network system will be described below.
[0019]
First, after describing the overall configuration of a mobile communication network system, a phase relationship represented by a decorrelating detector (Decorrelating Detector or decorrelator) and an MMSE detector in a code division multiple access (CDMA) system. The principle of a multi-user receiver using a number matrix (collectively referred to as a decorrelation type receiver) and the respective configurations and operations of a radio base station apparatus and a mobile device for realizing this principle will be described.
[0020]
<Configuration of mobile communication network system>
FIG. 1 is an overall configuration diagram of a mobile communication network system according to an embodiment of the present invention. This mobile communication network system is roughly composed of a radio access network (RAN) 100, a mobile station (UE) 200, and a core network (CN) 300.
[0021]
The radio access network 100 includes a plurality (two in the example shown in FIG. 1) of radio network controllers (RNCs) 110.₁And 110₂, And a plurality (four in the example shown in FIG. 1) of the radio base station apparatus (BTS) 120₁~ 120₄Consists of
[0022]
Wireless network controller 110₁Is connected to the core network 300 on the one hand and the radio base station device 120 on the other hand₁And 120₂It is connected to the. Wireless network control 110₂Is connected to the core network 300 on the one hand and the radio base station device 120 on the other hand₃And 120₄And at the same time,₁Is also connected.
[0023]
Radio base station device 120₁~ 120₄Respectively cover one or a plurality of service areas (cells), and perform wireless communication with the mobile device 200 existing in the cell. This wireless communication is performed by a multiple access method because a finite frequency band is shared by a plurality of wireless devices. A CDMA method, particularly a direct spread (DS: Direct Sequence) CDMA (DS-CDMA) method, is a multiple access method. This is a typical method among the methods.
[0024]
Radio base station device 120₁And 120₂Is a wireless network controller 110₁And is controlled by the device. In addition, the radio base station device 120₁And 120₂Transmits communication data from the core network 300 to the radio network controller 110.₁And transmits the received communication data to the destination mobile device 200, and transmits the communication data from the mobile device 200 to the wireless network control device 110 according to the destination.₁Via the core network 300. Similarly, the wireless base station device 120₃And 120₄Is a wireless network controller 110₂And is controlled by the device. In addition, the radio base station device 120₃And 120₄Transmits communication data from the core network 300 to the radio network controller 110.₂And transmits the received communication data to the mobile device 200, and transmits the communication data from the mobile device 200 according to the destination.₂Via the core network 300.
[0025]
In this way, communication between the mobile devices and communication between the mobile device and another terminal device connected to the core network 300 are performed.
[0026]
In the following, the radio base station apparatus 120₁~ 120₄Are collectively referred to as a wireless base station device 120.
[0027]
<Signal Configuration by DS-CDMA System>
In the communication according to the DS-CDMA method, the same frequency is shared by a plurality of users at the same time, and the signal is separated between the users by a spreading code (or a spreading signal) unique to each user. (4) The radio base station apparatus 120 receives a plurality (L) of bit strings subjected to code multiplex modulation. Each individual bit string may be transmitted from a different user (mobile device), or may be transmitted from the same user (mobile device). Further, the same signal transmitted from the same mobile station may be received by the radio base station apparatus 120 by multipath and regarded as a different bit sequence. For this reason, the bit string and the user do not always correspond one-to-one, but in the following description, the bit string will be referred to as a "user signal" according to a common usage.
[0028]
FIG. 2 shows a configuration in the case of asynchronous, multi-rate transmission by the DS-CDMA system received by the radio base station apparatus 120. The user signals 1, 2, and 3 are transmitted at the same transmission rate, The signal 4 is an example transmitted at a transmission rate twice as high as that of the user signals 1 to 3.
[0029]
The fact that the start times of the user signal bits do not match as shown in FIG. 3A is called "asynchronous", and the start times of the user signal bits do not match as shown in FIG. 3B. FIG. 3A shows the structure of a received signal in the case of asynchronous and single rate, and FIG. 3B shows the structure of the received signal in the case of synchronous and single rate. Each is shown.
[0030]
In order to explain the operation principle of the decorrelation type receiver, the concept of observation window length (Observation \ Window \ Size) is necessary, which will be described below. In a decorrelation type receiver, to decode one bit (a desired bit) of a desired signal, mutual interference between a plurality of bits around the bit is used to suppress an interference signal component. The scale that indicates whether to use surrounding bits up to is the observation window length. FIG. 4 shows the example of FIG. 3 (A) in the case of asynchronous and single rate, in which peripheral bits are rearranged in the order of time difference from desired bits, and each bit is numbered. In the example of FIG. 4A, the number of users is L, the observation window length is 2 bits, and the sum of desired bits and peripheral bits used for interference suppression is N × L = 2L, and the desired bits are usually all bits. , The bit number L + 1 or L is the desired bit. FIG. 4B shows an example in which the observation window length is 3 bits. Similarly, for the number of users L, the total number of bits is N × L = 3L. As shown in FIG. 3B, in the case of synchronization, bits other than bits (bit numbers 2, 5,..., K-1) in the same time zone as the desired bit do not contribute to interference suppression. , The observation window length is 1 bit. However, in the synchronous system, since the separation between bits is usually performed by using an orthogonal code as a spreading code, a synchronous system decorrelation receiver having an observation window length of 1 bit is a reference for theoretical performance comparison. Nevertheless, it has no meaning in practical use.
[0031]
Even in the case of asynchronous and multi-rate shown in FIG. 2, each bit (number 1, 2,... 16) of the user signal (1 to 4) included in the predetermined window width is set as an independent burst signal. By handling, a decorrelation type reception algorithm can be applied.
[0032]
<Principle of decorrelation type multi-user receiver>
Normally, user signals from a plurality of mobile stations are not synchronized at the time of being received by the radio base station apparatus 120. Therefore, in the following, interference cancellation in an asynchronous case (including both single-rate and multi-rate) will be described. Alternatively, the principle of suppression will be described (a synchronous system is included as one form of an asynchronous system).
[0033]
The radio base station apparatus 120 shown in FIG. 5 converts a received signal into a baseband signal (including synchronous detection using a synchronous carrier) by the frequency converter 2, and then generates a user signal (or data sequence) from the received signal. The received signal is despread by a plurality of (spread L) spread signals different from one another, passed through L matched filters, and sampled at the optimal sampling time (the time when each bit has been input to the matched filter). A plurality of outputs (this number is assumed to be K) are multiplied by K weighting coefficients, respectively, and a sign determination or a discrete level determination is performed on a result obtained by adding all of them, and decoding is performed. An output is obtained (the former is called a hard decision, and the latter is called a soft decision).
[0034]
The K weighting coefficients are determined based on the condition that the signal before the determination does not include the interference signal component or minimizes the sum of the interference component and the power of the background noise. Since the circuit for obtaining the weighting coefficient is the most complicated and has a large scale, how to reduce the size of this circuit is the most important point in practical use.
[0035]
When the same observation window length is applied to each user (data sequence) at a single rate, as shown in FIG. N × L. The L spread signals are continuous signals as shown in FIG. 6 (A), but for simplicity in the following discussion. As shown in FIG. 6B, the spread signal is also a burst signal s having a length corresponding to each bit.₁(T), s₂(T), ..., s_kLet's proceed with (t).
[0036]
The sum of the K-bit received signal and the background noise is Y (t), and the received amplitude of the k-th (k = 1 to K) bit is W_k, The modulated signal is d_k, And the spread signal_k(T), assuming that the background noise is n (t), Y (t) is represented by the following equation (1).
[0037]
(Equation 1)

[0038]
here,
[0039]
(Equation 2)

[0040]
(J = 1 to K), y_j,Z_jAnd a_ijRepresents the output signal of the matched filter (matched @ filter), the noise component included in the output of the matched filter, and the correlation coefficient, respectively.
[0041]
Here, the integration interval is from -∞ to + ∞, but since the spread signal is handled as a burst signal, the integration is performed in a finite interval where the burst signal exists.
[0042]
From equations (1) and (2), the output signal y of the matched filter_jIs represented by the following equation (3).
[0043]
(Equation 3)

[0044]
If Expression (3) is represented by a matrix, Expression (4) below is obtained.
[0045]
(Equation 4)

[0046]
If the expression (4) is further expressed by a vector, the following expression (5) is obtained.
[0047]
(Equation 5)

[0048]
Here, the vectors y, R, W, d, and z are respectively as follows.
[0049]
(Equation 6)

[0050]
The matrix R is called a correlation coefficient matrix and is a K × K matrix. Also, a_ij= A_jiTherefore, the matrix R is a symmetric matrix.
[0051]
Here, as a method of extracting a desired signal, there are a first method for making an interference signal component included in a signal before determination zero, such as a decorrelator or a decorrelation detector (Decorrelating @ Detector), and MMSE detection. There is a second method for minimizing the sum of the power of the interference noise and the background noise in consideration of the background noise as in the case of a minimum (mean-mean-squareerror @ detector). Hereinafter, each principle of the first and second methods will be described.
[0052]
(1) First method
The first method is to reduce interference noise to zero. In order to reduce the interference noise to zero, as shown in the following equation (6), the inverse matrix R of the correlation coefficient matrix R is added to both sides of the above equation (4) (or equation (5)).^-1May be multiplied. Thereby, the desired signal W including the background noise expressed by the second term on the right side is obtained._id_iIs obtained.
[0053]
(Equation 7)

[0054]
(2) Second method
In the first method, the interference signal component is set to zero. However, when there is background noise, there is a phenomenon called noise increase (Noise @ Enhancement) that increases the component. The second method is to minimize the sum of interference noise and background noise in order to compensate for the above drawbacks.²And the interference signal power per bit W_k ²Ratio σ²W_k ^-2Matrix R + σ²W^-2(= R ′) is known to be used instead of the matrix R. That is, the second method (MMSE detection method) uses the following equation (7) instead of the equation (6).
[0055]
(Equation 8)

[0056]
The above shows that both the decorrelation detector and the MMSE detector can decode all K-bit data by obtaining the inverse matrix of the K × K matrix and multiplying the inverse matrix by the matched filter output.
[0057]
(3) K elements required for 1-bit decoding of the desired signal
Both the correlation detector and the MMSE detector obtain the inverse matrix of the K × K matrix and multiply it by the output of the matched filter to decode the K-bit data in which the interference signal component is suppressed. It is a specific row (or column) of the inverse matrix that is involved in decoding the bits (desired bits). Therefore, if the bits to be decoded are not all of the K bits but only certain bits (this applies to a mobile station), instead of finding the inverse matrix, one row (or 1) of the inverse matrix is used. Column) is more convenient. It is shown as follows that K elements corresponding to a certain row (or column) are obtained as a solution of the K-ary simultaneous equation.
[0058]
First, since generality is not lost, if k = 1 is a desired signal bit, the first row of the inverse matrix is a row (or a column) involved in decoding this bit. Expressing the first line of equation (6) gives equation (8) below.
[0059]
(Equation 9)

[0060]
Where r_ijIs a matrix R expressed by the following equation (9).^-1Is the element.
[0061]
(Equation 10)

[0062]
Inverse matrix R appearing on the left side of equation (9)^-1Element r in the first row of₁₁~ R_1KIs a correlation coefficient matrix R and its inverse matrix R as shown in the following equation (10).^-1Is a unit matrix E, it can be seen that the product is a solution of the K-ary simultaneous equation (11).
[0063]
[Equation 11]

[0064]
(Equation 12)

[0065]
This equation relates to the decorrelation detector. When the MMSE detector is used, the diagonal element of the correlation matrix is represented by σ as shown in equation (7).²W_k ^-2In consideration of the fact that it is necessary to add Equation (11), the expression (11) is extended to include both detectors and expressed as follows.
[0066]
(Equation 13)

[0067]
here,
[0068]
[Equation 14]

[0069]
In the future, α defined by equation (13)_k(K = 1, 2,..., K) is an MMSE correction value, a matrix R + σ²W^-2Is called a modified correlation matrix R ′, and an inverse matrix of R and R ′ is a matrix R⁺Will be abbreviated.
[0070]
<Configuration and operation of wireless base station device>
First, the configuration and operation of a radio base station apparatus based on the prior art will be described, and then an embodiment in which the present invention is applied to the radio base station apparatus will be described together with the operation principle.
[0071]
FIG. 5 is a block diagram showing a configuration of a multi-user receiving unit as the radio base station apparatus 120, in particular, an antenna 1, a frequency converting unit 2, a matched filter bank 3, a spread signal generating unit 4, a correlation coefficient generating unit 10, It comprises an MMSE correction unit 8 for applying the MMSE detection method described above, an inverse matrix operation unit 14 for calculating an inverse matrix of a K × K matrix, a linear combination unit 6, and a determination circuit 7.
[0072]
A radio signal (RF signal) from mobile device 200 is received by antenna 1 and provided to frequency conversion unit 2. The frequency conversion unit 2 includes an amplifier that amplifies a radio signal from the antenna 1, a multiplier (mixer) that performs frequency conversion, and an oscillator that is synchronized with an RF signal, and converts an input RF band signal into a baseband signal. . This baseband signal is the received signal Y (t) represented by the above-described equation (1). This signal Y (t) is input to the matched filter bank 3, the spread signal generator 4, and the MMSE corrector.
[0073]
The spread signal generation unit 4 (details will be described later) converts a plurality (L) of spread signals f from the received signal.₁(T)-f_L(T) and the L spread signals g, all of which are shifted by a certain time, with respect to them.₁(T)-g_L(T) is generated. Then, the generated spread signal f₁(T)-f_L(T) to the matched filter bank 3 and the spread signal g₁(T)-g_L(T) is given to the correlation coefficient generator 10.
[0074]
The matched filter bank 3 (details will be described later) converts the baseband signal Y (t) into L spread signals f₁(T)-f_LAfter despreading according to (t), the despread signal is filtered by a matched filter. The filtered signal is sampled at a timing when each signal bit has been input to the matched filter, and the K signals y shown in the above-mentioned equation (3) are obtained.₁~ Y_KAnd is given to the linear combination unit 6.
[0075]
The correlation coefficient generator 10 receives the L spread signals g₁(T)-g_L(T) or K signals y₁~ Y_KIs the part corresponding to the spread signal s₁(T) -s_KFrom (t), a K × K correlation coefficient matrix is generated.
[0076]
The MMSE correction unit 8 calculates the K signals y₁~ Y_KIs estimated and the diagonal element of the correlation coefficient matrix generated by the correlation coefficient generation unit 10 is corrected as shown in equation (7), and Generate a correlation coefficient matrix. In the case where the estimated value of the background noise power is set to zero and MMSE correction is not performed, the receiver functions as a decorrelation receiver (Decorrelating @ Detector).
[0077]
The inverse matrix calculator 14 calculates the correlation coefficient matrix R of the K × K matrix or the inverse matrix R of the modified correlation coefficient matrix R ′ generated by the correlation coefficient generator 10.⁺Is calculated.
[0078]
In the linear combination unit 6, the operation represented by the expression (6) or (7) is performed by the input signal y₁~ Y_KAnd the matrix R output from the inverse matrix calculator 14⁺And the signal y₁~ Y_KIs a matrix R⁺The operation of multiplying by the K elements in each row and adding them together is hereinafter referred to as linear combination), and the signal after the linear combination is given to the determination circuit 7.
[0079]
Linear combination 6_iGives the matrix R⁺The result of the linear combination with the i-th row is a signal on the right side of Expression (6) or Expression (7).
[0080]
The determination circuit 7 compares the signal provided from the linear combination unit 6 with a predetermined threshold, and₁~ D_K, A discrete value or sign is determined and output. Modulation signal d₁~ D_KIs -1 or +1 when BPSK (Binary Phase Shift Keying) is used as the modulation scheme. Judgment circuit 7₁~ 7_KThe decoded signal d obtained by₁~ D_KAfter that, predetermined processing such as error correction is performed, and the radio network controller 110₁Or 110₂Or transmitted to another mobile station within the service area of the same base station apparatus 120.
[0081]
FIG. 7 is a detailed configuration diagram of the matched filter bank 3. The matched filter bank 3 includes L multipliers 3a₁~ 3a_L, L matched filters (MF: Matched Filter) 3b₁~ 3b_L, And L sample-and-hold circuits (SHC: Sample Hold Circuit) 3c₁~ 3c_LHaving.
[0082]
The baseband signal Y (t) is input to L multipliers 3a.₁~ 3a_LIs input to the first input terminal. L multipliers 3a₁~ 3a_LOf the spread signal f generated by the spread signal generator 4₁(T)-f_L(T) is input. Baseband signal Y (t) and spread signal f₁(T)-f_L(T) is the multiplier 3a₁~ 3a_LIs called despreading, but the multiplier 3a₁~ 3a_LThe signals despread by MF3b₁~ 3b_LMF3b₁~ 3b_LIntegrates the input signals during the 1-bit length T of each signal.
[0083]
MF3b₁~ 3b_LAre output from the SHC3c₁~ 3c_LRespectively. SHC3c₁~ 3c_LIs MF3b₁~ 3b_LSamples the value of the output signal at the time when the integration of one bit of each received signal is completed, holds the value, and outputs it.
[0084]
SHC3c₁~ 3c_L, A sampling value is output continuously, and a plurality of sampling values for each matched filter are used to decode one bit of a desired signal. Taking the model of the single-rate asynchronous system shown in FIG. 4B as an example, as shown in FIG. 7, 3 × the number of users output from each matched filter, that is, 3L (= K) Pieces of data are used.
[0085]
The matched filter bank 3 is used not only to operate on a received signal as shown in FIG. 5, but also to calculate a correlation coefficient. FIG. 8 shows the configuration of a matched filter bank for obtaining a correlation coefficient, which will be referred to as a correlation coefficient matched filter bank to distinguish it from the matched filter bank 3. The correlation coefficient between the same spread signals, that is, the diagonal element of the correlation coefficient matrix is a value determined when the transmission rate is determined, and need not be determined.
[0086]
FIG. 6 shows the spread signal g₁(T)-g_L(T) and (burst) spread signal s₁(T) -s_KIt shows the relationship of (t).
[0087]
Unless the state of the switch SW used in the correlation coefficient matched filter bank is specified, the relationship between the input and the output cannot be determined. Therefore, a modified matched filter bank shown in FIG. 9 is further introduced. The input of this modified matched filter bank is K signals y₁~ Y_KSpread signal s corresponding to₁(T) -s_KIn (t), the output of the correlation coefficient is explicitly shown, so that the description will be made not by using a correlation coefficient matched filter bank but by using a modified matched filter bank equivalent thereto. When the observation window length is N = 2, there is no difference between the two in the number of matched filters. Therefore, there is no practical problem even if the deformed matched filter bank is actually used. However, for N> 2, it is actually more advantageous to use a correlation coefficient matched filter bank in terms of the number of matched filters.
[0088]
FIG. 10 shows a configuration example of the spread signal generation unit 4. The spread signal for performing synchronization acquisition and tracking on the received signal and despreading the received signal by the matched filter bank 3. f₁(T)-f_L(T) is generated. At the same time, the spread signal g for generating the correlation coefficient matrix in the correlation coefficient matrix generation unit₁(T)-g_L(T) is generated. Spread signal g₁(T)-g_L(T) is the spread signal f₁(T)-f_L(T) is generated from the same spreading code only by shifting for a certain period of time. In order to generate a spread signal obtained by shifting the time of the spread signal by an arbitrary time, for example, in the case of the example shown in FIG. By setting an initial value at a time shifted from the reference time, it is possible to generate a spread signal time-shifted by an arbitrary time. Alternatively, since the relationship between the contents of the shift register at the reference time and the time shift amount is known in advance, a desired spread signal can be obtained even if the contents of the shift register are appropriately rewritten at the reference time. The amount of time to be shifted is determined in consideration of a decoding delay (a decoding delay of an average of N / 2 occurs in a system having an observation window length of N) and a time required for calculating an inverse matrix of the correlation coefficient matrix.
[0089]
(1) First form
FIG. 11 is a block diagram showing a configuration of the radio base station apparatus 120 according to the first embodiment, particularly a configuration of a multi-user receiving unit. This figure has basically the same configuration as that of the radio base station receiver shown in FIG. 5, but the correlation coefficient generator 10 and the inverse matrix calculator 14 shown in FIG. Assuming that the total number of bits of the desired signal and the interference signal involved in decoding one bit of K is K, K weighting factor generators 5₁~ 5_K, And K linearly connected parts (or inner product parts) 6₁~ 6_K, And K determination circuits 7₁~ 7_KHaving.
[0090]
FIG. 11 shows a configuration in which each of K weight coefficient generation units, linear combination units, and determination circuits is provided to decode all of the K bits. When decoding (1 ≦ M <K), the number of weighting factor generators, the number of linear combination units, and the number of determination circuits are each M.
[0091]
In the description below, the bit number of the desired signal is assumed to be 1 unless otherwise specified.
[0092]
Weight coefficient generator 5₁~ 5_KIs the input L spread signals g₁(T)-g_LK spread signals s, which are the parts corresponding to the K received signal bits in (t)₁(T) -s_KMatrix R based on (t)⁺Each element r_ij(I = 1 to K, j = 1 to K) are generated and output. Specifically, the i-th weight coefficient generation unit 5_iIs the matrix R⁺K elements r in the i-th row of_i1~ R_iKIs generated, and the generated element r_i1~ R_iKTo the ith linear combination unit 6_iGive to.
[0093]
FIG. 12 shows the weight coefficient generation unit 5₁~ 5_KAmong them, the weighting factor generator 5₁FIG. 3 is a block diagram showing the configuration of FIG. Weight coefficient generator 5₁Is the first coefficient generation unit 50₁To the K-th weight coefficient generation unit 50_KUp to K weighting factor generators.
[0094]
Each weight coefficient generation unit includes K spread signals s₁(T) -s_K(T) is commonly input. In addition, each output signal of all weight coefficient generation units is input to each weight coefficient generation unit. However, the respective outputs have already been combined inside the respective weight coefficient generation units (arrows having no input source indicate this). That is, the j-th weight coefficient generation unit 50_j(J = 1 to K) include a j-th weight coefficient generation unit 50_jOutput signal r_1jOutput signal r excluding₁₁~ R_1j-1, R_{1j + 1}~ R_1KIs entered.
[0095]
J-th weight coefficient generation unit 50_jIs the input signal and the spread signal s₁(T) -s_KBased on (t), the matrix R⁺Element r in the first row and jth column of_1jIs generated and output.
[0096]
The MMSE correction unit 8 calculates the background noise power density σ from the received signal.², And the output y of the matched filter bank 3₁~ Y_KIs the ratio of the estimated received power per bit corresponding to_iAnd a weighting factor generator 5₁~ 5_KGive to.
[0097]
FIG. 13 shows the weight coefficient generation unit 5₁First coefficient generation unit 50₁FIG. 3 is a block diagram showing a detailed configuration of the embodiment. First coefficient generation unit 50₁Is composed of a correlation coefficient generation section 511, a linear combination section 512, and a feedback section 513.
[0098]
The correlation coefficient generation unit 511 has the same function as the above-described deformed matched filter bank (see FIG. 9), and includes K multipliers 51a.₁~ 51a_K, K MF (matched filter) 51b₁~ 51b_K, And K SHCs (sample and hold circuits) 51c₁~ 51c_KHaving. The linear combination unit 512 is connected to the linear combination unit 6.₁~ 6_K(To be described in detail later), and has K multipliers 51d.₁~ 51d_KAnd a synthesizer (synthesizer) 51e.
[0099]
The feedback unit 513 includes a subtractor 51f, an inverting amplifier 51g, a feedback loop filter (low-pass filter, LPF) 51h, and a combiner 51m.
[0100]
Multiplier 51a of correlation coefficient generation section 511₁~ 51a_KAre connected to the first input terminal of the₁(T) is commonly input, while the spread signal s is input to the second input terminal.₁(T) -s_K(T) is input. Therefore, the multiplier 51a_jIs the spread signal s₁(T) and s_j(T) and the product s₁(T) s_j(T) is output.
[0101]
MF51b₁~ 51b_KIs a multiplier 51a₁~ 51a_KAre integrated (filtered) over a 1-bit length T, and the integration result is expressed by the SHC 51c.₁~ 51c_KRespectively. SHC51c₁~ 51c_KIs the MF51b₁~ 51b_KSample the value of the output signal at the time of completion of each integration of the length of one bit, hold the value, and output it. Therefore, SHC51c_iIs an element (correlation coefficient) a in the i-th row and first column of the correlation coefficient matrix R from the above equation (2)._i1Becomes That is, SHC51c₁~ 51c_KFrom the correlation coefficient a in the first column of the correlation coefficient matrix R₁₁~ A_K1Is output. SHC51c₁Output a₁₁Is a constant determined when the transmission rate is determined, and does not change for each bit.₁From, SHC51c₁The circuit up to may be omitted. These correlation coefficients a₁₁~ A_K1Is the multiplier 51d of the linear combination unit 512₁~ 51d_KAre respectively input to the first input terminals. HC52c₁Output the MMSE correction value to α₁Then (1 + α₁) Is multiplied and output. Multiplier 51d₁~ 51d_KOf the matrix R⁺R corresponding to the element of the first row of₁₁~ R_1K(These values are referred to as weighting factors x₁~ X_KAre also input.
[0102]
Element r₁₁Is the first coefficient generation unit 50₁The output signal of the feedback unit 513 is fed back. Element r₁₂~ R_1KIs, as shown in FIG. 6, a second coefficient generator 50.₂To the K-th weight coefficient generation unit 50_KThese are the output signals.
[0103]
Multiplier 51d₁~ 51d_KMultiplies the signals respectively input from the first input terminal and the second input terminal, and gives the multiplication result to the combiner 51e. The synthesizer 51e includes a multiplier 51d₁~ 51d_KAre combined, and the combined result is provided to the feedback unit 513. Multiplier 51d₁~ 51d_KAre the K multiplication results a₁₁(1 + α₁) R₁₁, A₂₁r₁₂, ..., a_K1r_1KIs obtained, and as a result of synthesis by the synthesizer 51e, a₁₁(1 + α₁) R₁₁+ A₂₁r₁₂+ ... + a_K1r_1KIs obtained. This synthesis result corresponds to the left side of the first equation of the simultaneous equation (12).
[0104]
The output signal of the synthesizer 51e is converted into a constant a by a subtractor 51f.₁₁Is subtracted, and the subtraction result is input to the inverting amplifier 51g. The inverting amplifier 51g amplifies (inverts and amplifies) the input signal with a gain of -G and outputs the result.
[0105]
The output signal of the inverting amplifier 51g is input to the synthesizer 51m via the LPF 51h, and is a constant 1 / (1 + α) which is an initial value of a feedback loop.₁) Is added. The result of the synthesis is the first coefficient generation unit 50₁And the output of the multiplier 51d₁To the second input terminal.
[0106]
As a result, the linear combination unit 512 and the feedback unit 513 form a feedback loop, and the output of the feedback unit 513 (that is, the first coefficient generation unit 50).₁Is the convergence state of the feedback loop, and the element r₁₁And one of the solutions of the simultaneous equations expressed by equation (12) (element r₁₁) Is required (the operation principle of the feedback loop will be described later).
[0107]
FIG. 14 shows the weight coefficient generation unit 5₁Second coefficient generation unit 50₂FIG. 3 is a block diagram showing a detailed configuration of the embodiment. Second coefficient generator 50₂Has a correlation coefficient generation unit 521, a linear combination unit 522, and a feedback unit 523.
[0108]
The correlation coefficient generation unit 521 has the same function as that of the above-described modified matched filter bank (see FIG. 9), and includes K multiplications 52a.₁~ 52a_K, K MF52b₁~ 52b_K, And K SHC52c₁~ 51c_KHaving. The linear combination unit 522 is connected to the linear combination unit 6₁~ 6_K(Described later in detail), and has the same functions as the K multipliers 52d.₁~ 52d_KAnd a synthesizer 52e.
[0109]
The feedback unit 513 includes an inverting amplifier 52g and a feedback loop filter (low-pass filter, LPF) 52h.
[0110]
Multiplier 52a of correlation coefficient generation section 521₁~ 52a_KAre connected to the first input terminal of the₂(T) is commonly input, while the spread signal s is input to the second input terminal.₁(T) -s_K(T) is input. Therefore, the multiplier 52a_jIs the spread signal s₂(T) and s_jProduct s that is the result of multiplication with (t)₁(T) s_j(T) is output.
[0111]
MF52b₁~ 52b_KIs a multiplier 52a₁~ 52a_KAre integrated (filtered) during the 1-bit length T, and the integration result is expressed by the SHC 52c.₁~ 52c_KRespectively. SHC52c₁~ 52c_KIs the MF52b₁~ 52b_KSamples the value of the output signal at the time of completion of each 1-bit integration, holds and outputs the value. Therefore, SHC52c_iIs an element (correlation coefficient) a in the i-th row and second column of the correlation coefficient matrix R from the above equation (2)._i2Becomes That is, SHC52c₁~ 52c_KFrom the correlation coefficient a in the second column of the correlation coefficient matrix R₁₂~ A_K2Is output. SHC52c₁Output a₂₂Is a constant determined when the transmission rate is determined, and does not change for each bit.₂From, SHC52c₂The circuit up to may be omitted. However, HC52c₂Output the MMSE correction value to α₂Then (1 + α₂) Is multiplied and output.
[0112]
These correlation coefficients a₁₂~ A_K2Is the multiplier 52d of the linear combination unit 522₁~ 52d_KAre respectively input to the first input terminals. Multiplier 52d₁~ 52d_KOf the matrix R⁺R corresponding to the element of the first row of₁₁~ R_1KAre respectively input.
[0113]
Element r₁₂(Weight coefficient x₂) Is the second coefficient generation unit 50₂The output signal of the feedback unit 523 is fed back. Element r₁₁, R_Thirteen~ R_1KIs, as shown in FIG. 12, the first coefficient generation unit 50₁And the third weight coefficient generation unit 50₃To the K-th weight coefficient generation unit 50_KThese are the output signals.
[0114]
Multiplier 52d₁~ 52d_KMultiplies the signals respectively input from the first input terminal and the second input terminal, and provides the multiplication result to the synthesizer 52e. The synthesizer 52e includes a multiplier 52d₁~ 52d_KAre combined, and the combined result is output to the feedback unit 523. Multiplier 52d₁~ 52d_KAre the K multiplication results a₁₂r₁₁, A₁₂(1 + α₂) R₁₂, ..., a_K2r_1KIs obtained, and as a result of synthesis by the synthesizer 52e, a₁₂r₁₁+ A₂₂(1 + α₂) R₁₂+ ... + a_K2r_1KIs obtained. This synthesis result corresponds to the left side of the second equation of the simultaneous equation (12).
[0115]
The output signal of the synthesizer 52e is input to an inverting amplifier 52g, is inverted and amplified, and is then passed through a LPF 52h to a multiplier 52d.₂Is input to the second input terminal. In addition, the output of the LPF 52h is₂Output r₁₂Becomes
[0116]
As a result, the linear combination unit 522 and the feedback unit 523 form a feedback loop, and the output value of the feedback unit 523 (that is, the second coefficient generation unit 50).₂Is the convergence state of the feedback loop, and the element r₁₂And one of the solutions of the simultaneous equations expressed by equation (12) (element r₁₂) Is required.
[0117]
Third weight coefficient generation unit 50₃To the K-th weight coefficient generation unit 50_KAlso the second coefficient generation unit 50₂And the spread signal input to the first input terminal of the multiplier of the correlation coefficient generator is s.₃(T) -s_K(T) and the multipliers of the linear combination unit to which the own output signal is fed back are changed from the third to the K-th multipliers._Thirteen~ R_1K(These outputs are represented by weighting factors x₃~ X_KAlso called).
[0118]
Therefore, these third weighting factor generators 50₃To the K-th weight coefficient generation unit 50_KIs omitted. In addition, the weight coefficient generation unit 5₂~ 5_K, The weighting factor generator 5₁Since the configuration is the same as that described above, the description thereof is omitted.
[0119]
Returning to FIG. 11, the output y of the matched filter bank 3₁~ Y_KIs the linear combination 6₁~ 6_KIs commonly input to. Weight coefficient generator 5₁~ 5_KAre output from the linear combination unit 6₁~ 6_KIs input to each of. That is, the weight coefficient generation unit 5_iOutput r_i1~ R_iKIs the linear combination 6_iIs entered. Linear combination 6₁~ 6_KPerforms a linear combination of these input signals.
[0120]
FIG. 15 shows the linear combination 6₁FIG. 3 is a block diagram showing a detailed configuration of the embodiment. Linear combination 6₁Is, as described above, the first coefficient generation unit 50₁Has the same function as the linear combination unit 512, and has K multipliers 61a₁~ 61a_KAnd a synthesizer 61b.
[0121]
Multiplier 61a₁~ 61a_KOutput terminal y of the matched filter bank 3₁~ Y_KAre respectively input to the second input terminal,₁Output signal r₁₁~ R_1KAre respectively input. Multiplier 61a₁~ 61a_KMultiplies the signals input to the first input terminal and the second input terminal, respectively, and outputs the multiplication results to the combiner 61b.
[0122]
The synthesizer 61b includes a multiplier 61a₁~ 61a_KAre combined. The synthesis result of this synthesizer 61b is y₁r₁₁+ Y₂r₁₂+ ... + y_Kr_1KThe result of the synthesis is the right side of the above equation (8), and the value is W₁d₁And the noise component. The result of the synthesis is determined by the judgment₁Input to the judgment circuit 7₁From d₁Determination result D₁Is output.
[0123]
Other linear combination 6₂~ 6_KAlso the linear combination 6₁With the same configuration as₂d₂~ W_Kd_KAnd the sum of the respective noise components. These outputs are output to the judgment circuit 7₂~ 7_KAre respectively input to the judgment circuit 7₂~ 7_KFrom the result D₂~ D_K(D₂~ D_KAre output, and decorrelation-type reception is performed.
[0124]
The configuration of the radio base station apparatus 120 according to the first embodiment has been described above. Next, the principle of why such a configuration enables multi-user reception will be described, and simulation will be described. The result of confirming the operation is also mentioned.
[0125]
The output of the synthesizer 51e is a₁₁r₁₁+ A₂₁r₁₂+ ... + a_K1r_1KAnd the output of the synthesizer 52e is a₁₂r₁₁+ A₂₂r₁₂+ ... + a_K2r_1KHas already been mentioned. The output of the synthesizer 51e is supplied to a feedback unit 513 at a first constant (here, a₁₁) Is subtracted, passed through an inverting amplifier and a feedback loop filter, and passed through a second constant (here, 1 + α).₁Are combined and returned to the linear combination to form a feedback loop as a whole. The loop is started by giving the second constant as an initial value, and reaches a steady state after a lapse of a certain time. At this time, the gain of the feedback loop formed by the linear combination unit 512 and the feedback unit 513 is- G₁Then, the following equation (14) holds.
[0126]
-G₁[A₁₁(1 + α₁) R₁₁+ A₂₁r₁₂+ ... + a_K1r_1K-A₁₁]
= R₁₁−1 / (1 + α₁) (14)
[0127]
Similarly, the gain of the feedback loop formed by the linear combination unit 522 and the feedback unit 523 is -G₂Then, the following equation (15) holds.
[0128]
-G₂[A₁₂r₁₁+ A₂₂(1 + α₂) R₁₂+ ... + a_K2r_1K] = R₁₂(15)
[0129]
Both sides of the equations (14) and (15) are respectively represented by the feedback loop gains −G₁, -G₂When the right side is shifted to the left side by dividing by the above, the following equations (16) and (17) are obtained.
[0130]
[A₁₁(1 + α₁) + 1 / G₁] [R₁₁−1 / (1 + α₁)] + A₂₁r₁₂+ ...
+ A_K1r_1K= 0 (16)
a₁₂r₁₁+ [A₂₂(1 + α₂) + 1 / G₁] R₁₂+ ... + a_K2r_1K= 0 (17)
[0131]
Equation (17) is generalized to include equation (18), including those relating to the other second coefficient generation units.
[0132]
a_1jr₁₁+ A_2jr₁₂+ ... + [a_JJ(1 + α_j) + 1 / G₁] R_1j+ ... + a_K1r_1K= 0
(J = 2,3, ..., K) Ｋ (18)
[0133]
In equations (16) and (18), G_j= ∞ (j = 1, 2,..., K), these equations are as follows.
[0134]
a₁₁(1 + α₁) R₁₁+ A₂₁r₁₂+ ... + a_K1r_1K= A₁₁(19)
a_1jr₁₁+ A_2jr₁₂+ ... + a_JJ(1 + α_j) R_1j+ ... + a_K1r_1K= 0
(J = 2, 3,..., K) (20)
[0135]
The output of the matched filter bank 3 of the received signal, K values (y₁, Y₂, ..., y_K) And the output of the weighting factor generator (r₁₁, R₁₂, ..., r_1K) And the linear combination 6₁B = y₁r₁₁+ Y₂r₁₂+ ... + y_Kr_1KThen, B has already been described as to whether the interference signal component becomes zero (decorrelation reception) or suppressed (MMSE reception), for the following reason. K values (y₁, Y₂, ..., y_K) Is expanded using Expression (3), and B is obtained as Expression (21).
[0136]
B = W₁d₁[A₁₁(1 + α₁) R₁₁+ A₂₁r₁₂+ ... + a_K1r_1K]
+ W₂d₂[A₁₂r₁₁+ A₂₂(1 + α₂) R₁₂+ ... + a_K2r_1K] +…
+ W_Kd_K[A_1Kr₁₁+ A_2Kr₁₂+ ... + a_KK(1 + α_K) R_1K] +…
+ Z₁r₁₁+ Z₂r₁₂+ ... + z_Kr_1K(21)
[0137]
By substituting the equations (19) and (20) into the equation (21), B becomes the following equation (22).
[0138]
B = W₁d₁a₁₁(1-α₁r₁₁) -W₂d₂a₂₂α₂r₁₂-...- W_Ka_KKα_Kr_1Kd_K+ Z₁r₁₁+ Z₂r₁₂+ ... + z_Kr_1K(22)
[0139]
From equation (22), α₁= Α₂= ... = α_K= 0, that is, when the background noise power is set to zero, the first term on the right side of the equation (22), W₁d₁a₁₁That is, it is understood that only the desired signal and the noise component are received and the decorrelation reception is performed. α_jIs the background noise power density and the estimated received signal power per bit of the bit number j, and the MMSE reception that minimizes the sum of the interference signal power and the background noise power is described above.
[0140]
Also, a on the right side of equation (19)₁₁When 1 is set as 1 and compared with Expression (20) and Expression (12), it can be seen from the above that the receiver described above performs decorrelation-type reception also from the coincidence.
[0141]
a₁₁The fact that may be set to 1 is explained as follows. First, the diagonal element of the correlation coefficient matrix is expressed by the following equation (23) from the equation (2).
[0142]
[Equation 15]

[0143]
Here, the start of the desired signal bit is set to t = 0.
[0144]
As can be seen from equation (23), a_JJThat is, the diagonal elements of the correlation coefficient matrix are determined by the bit rate (1 / T). Therefore, in a single-rate system, the diagonal elements are all the same, and their values merely change the level of the linear combination unit 6 relatively. Since the present invention can be applied to not only a single rate but also a multi-rate, all the diagonal elements of the correlation coefficient matrix are not necessarily the same. I have left.
[0145]
Diagonal element a of correlation coefficient matrix_JJIs the bit rate (1 / T_j), So a_JJCan be omitted from the multiplier and the matched filter. The diagonal element a_JJOr MMSE corrected it, a_JJ(1 + α_j) And r_1jCan also be replaced with a level converter (or interpreted as a multiplier that operates at a low speed).
[0146]
The above is the feedback loop gain G_j= ∞ (j = 1, 2,..., K). However, when the feedback loop gain becomes small, an arithmetic error occurs. Next, a method (gain correction) that does not cause this error will be described.
[0147]
FIG. 16A is a diagram in which the first coefficient generator is subjected to gain correction, and the output of the synthesizer 51n is represented by r.₁₁', Which is returned to the original linear combination to form a feedback loop, and the output of the LPF 51h (r₁₁'-1 / (1 + α₁)) To c₁Multiply by 1 / (1 + α₁) Is added to r₁₁Are connected to other (K−1) second coefficient generation units. r₁₁'And r₁₁Is expressed by the following equation (24).
[0148]
[R₁₁'-1 / (1 + α₁)] C₁= R₁₁−1 / (1 + α₁)(24)
[0149]
Where c₁Is represented by the following equation (25).
[0150]
c₁= 1 + 1 / a₁₁/ (1 + α₁) / G₁(25)
[0151]
r₁₁', In equation (16), r₁₁Therefore, the following equation (26) holds.
[0152]
[A₁₁(1 + α₁) + 1 / G₁] [R₁₁'-1 / (1 + α₁)]
+ A₂₁r₁₂+ ... + a_K1r_1K= 0 (26)
[0153]
By substituting equation (24) into equation (26), the following equation (27) is obtained.
a₁₁(1 + α₁) R₁₁+ A₂₁r₁₂+ ... + a_K1r_1K= A₁₁(27)
[0154]
In equation (27), G₁As a result of the₁It can be seen that an error-free output can be obtained as in the case where is infinite. FIG. 16 (B) is a circuit equivalent to FIG. 16 (A), in which the output of the feedback unit is directly connected to the other K-1 weight coefficient generation units, and is connected to the original linear combination unit in the feedback loop. When returning, the level is 1 / c₁It is doubled.
[0155]
FIG. 17 shows another method of gain correction. Returning to equation (16), the gain G₁In order to remove the effect of₁₁−1 / (1 + α₁)] / G₁It can be seen that adding However, the place to be added is between the input of the synthesizer and the input of the inverting amplifier, or-[r₁₁−1 / (1 + α₁)] / G₁To -G₁The same applies to the case where the output is multiplied and added to the output of the inverting amplifier. FIG. 17 is based on this idea. Since a loop is locally formed at the input and output of the LPF 51h, it is necessary to adjust the level of each part and adjust the feedback gain. The amplifier 51q has a loss in the LPF and is inserted to compensate for the loss. However, if the LPF is formed of an active element and has a gain, the amplifier 51q may be an attenuator.
[0156]
The method in which the operation error of the inverse matrix element does not occur even when the feedback loop gain is small has been described above. Next, several modified examples regarding the linear combination unit will be described.
[0157]
In FIG. 18 (A) and FIG.₁₁(1 + α₁) R₁₁To the multiplier 51d₁(1 + α)₁) Is different on which input side. In FIG. 13, the calculation is performed on the correlation coefficient matrix generation unit side, and in FIG. 18 (A), the calculation is performed on the weight coefficient generation unit side.
[0158]
FIGS. 18A and 18B are different in whether the MMSE correction is performed in the feedback loop or the obtained weighting coefficient is distributed to the other K−1 weighting coefficient units. There is, however, this is also the same if expressed in an equation, so they are equivalent to each other.
[0159]
FIG. 19 also shows a modified example related to the MMSE correction, in which a₁₁(1 + α₁) R₁₁To create a₁₁r₁₁And a₁₁α₁r₁₁(FIG. 19) or a₁₁r₁₁(FIG. 13), and are the same if expressed by an equation.
[0160]
FIG. 20 is a modification example of the first coefficient generation unit shown in FIG. In FIG. 13, a subtractor 51f calculates a first constant a from the output of the synthesizer 51e of the linear combination unit.₁₁Is reduced and then a second constant is added at the output of the LPF 51h. In FIG. 20, a feedback loop is formed without subtracting the first constant from the output of the combiner 51e. The result of performing the synthesis of the second constant outside the feedback loop is coupled to another K−1 weighting factor generator. The fact that they are equivalent can be explained as follows.
[0161]
First, the output of the LPF 51h is represented by r₁₁Then, the following equation (28) holds.
a₁₁(1 + α₁) R₁₁'+ A₂₁r₁₂+ ... + a_K1r_1K= 0 (28)
r₁₁'And r₁₁Is different from the second constant 1 / (1 + α) by the synthesizer 51m.₁), The following equation (29) holds.
r₁₁'= R₁₁−1 / (1 + α₁) (29)
Substituting equation (29) into equation (28), a₁₁When the term is shifted to the right side, the equation (19) is obtained, and the circuits in FIGS. 13 and 20 are equivalent to each other.
The description using FIG. 16, FIG. 17, FIG. 18 and FIG.₁₁(X₁), But the same method can be applied to the circuit for calculating the weight coefficient corresponding to the interference signal bit, since there is no portion corresponding to the subtractor 51f and the synthesizer 51m. Explanation and illustration are omitted.
[0162]
Next, the result of confirming by simulation whether or not the circuit operates correctly with the configuration described in the first embodiment will be described with reference to FIGS. 21, 22, 23, 24, and 25.
[0163]
The model has an observation window length N = 2, number of users L = 2, K = NL = 4 (bit d₁~ D₄), The feedback loop gain correction is a simple case using the method shown in FIG.₁₁, R₁₂, R_ThirteenAnd r_Thirteen(Weight coefficient x₁~ X₄)).
[0164]
FIG.₁~ D₄And the corresponding spread signal s₁(T) -s₄(T) and the product s of the spread signal required to determine the correlation coefficient₁(T) s₂(T), s₂(T) s₃(T), s₃(T) s₄The waveform of (t) is shown. FIG. 22A shows the spread signal s₁(T) -s₄(T) and FIG. 22 (b) shows the product s of the spread signal.₁(T) s₂(T), s₂(T) s₃(T) and s₂(T) s₃(T) is shown numerically. FIG. 22C shows a correlation coefficient matrix R and an inverse matrix R obtained in advance by a computer.^-1Are shown.
[0165]
Here, the bit number of the desired signal is 3, and at time t = 0, the weight coefficient x₃The weighting factor x when the feedback loop is operated with the initial value of is set to 1 and the initial values of the other weighting factors set to 0₁~ X₄(R₁₁, R₁₂, R_ThirteenAnd r₁₄23) is obtained by simulation to obtain a graph shown in FIG. The time τ on the horizontal axis indicates the time normalized by the time constant of the feedback loop filter. Since the model is simple, the difference between the convergence value and the initial value is small, and the convergence status is difficult to understand.weightCoefficient x₁~ X₄FIG. 24 shows the deviation from the exact solution. From this figure, it can be seen that if a time several times the time constant of the feedback loop filter is applied, all the weighting factors will converge to the exact solution. FIG. 25 shows an example in which when a weight coefficient is provided outside the feedback loop, that is, to another weight coefficient circuit, a part is taken out from the input of the feedback loop filter to increase the speed. From FIGS. 24 and 25, it can be seen that if the time constant of the feedback loop filter is selected to be a sufficiently small value, the time required to obtain the weight coefficient can be reduced to 1 bit or less.
[0166]
(2) Second form
Weight coefficient generator 5₁~ 5_KOf the first coefficient generation unit 50₁To the K-th weight coefficient generation unit 50_K(See FIG. 12). These first coefficient generators 50₁To the K-th weight coefficient generation unit 50_KHas a correlation coefficient generation unit 511 or 521, and the element a in the first column of the correlation coefficient matrix₁₁~ A_K1To the element a in the K-th column₁₁~ A_K1Are output. That is, the weight coefficient generation unit 5₁~ 5_KAre the elements a in the first column of the correlation coefficient matrix.₁₁~ A_K1To the element a in the K-th column₁₁~ A_K1(Ie, all the correlation coefficients of the correlation coefficient matrix).
[0167]
Therefore, the weighting factor generator 5₁~ 5_KCan be integrated into one circuit, so that the weighting coefficient generator 5₁~ 5_KIn place of the correlation coefficient generation unit 10 and the K weight coefficient calculation units 11₁~ 11_KThis is the second embodiment shown in FIG. The matched filter bank has been described using the modified matched filter bank (FIG. 9), but will be described using the correlation coefficient matched filter bank (FIG. 8).
[0168]
The correlation coefficient generation unit 10 calculates K × K correlation coefficients a₁₁~ A_KKAnd the generated correlation coefficient a₁₁~ A_KKTo the weighting coefficient calculation unit 11₁~ 11_KGive to the common. Weight coefficient calculator 11₁~ 11_KIs the correlation coefficient a given by the correlation coefficient generation unit 10.₁₁~ A_KKMatrix R based on⁺Element r in the first row of₁₁~ R_1KTo the element r in the K-th row_K1~ R_KKAre obtained, and the linear combination 6₁~ 6_KGive to each.
[0169]
FIG. 27 is a block diagram illustrating a detailed configuration of the correlation coefficient generation unit 10. The correlation coefficient generator 10 includes L correlation coefficient matched filter banks 10a.₁-10a_LHaving.
[0170]
These correlation coefficient matched filter banks 10a₁-10a_LContains the spread signal g₁(T)-g_L(T) is input and the correlation coefficient matched filter bank 10a₁-10a are the correlation coefficients a in the first column based on the input spread signal.₁₁~ A_K1To the K-th column correlation coefficient a_1K~ A_KKOutput up to. The circuit 10s rearranges these outputs and outputs them.
[0171]
FIG.₁FIG. 3 is a block diagram showing a detailed configuration of the embodiment. Weight coefficient calculator 11₁Is the K linear combination parts 11a₁~ 11a_KAnd K feedback units 11b₁~ 11b_KHaving.
[0172]
Linear combination part 11a₁Is the first coefficient generation unit 50 shown in FIG.₁Has the same configuration as the linear combination unit 512 of the feedback unit 11b.₁Has the same configuration as the feedback unit 513 shown in FIG. Also, the linear combination part 11a₁Is the same as the signal input to the linear combination unit 512. Therefore, the feedback section 11b₁From the matrix R⁺Element r₁₁Is output.
[0173]
Linear combination part 11a₂And return section 11b₂Has the same configuration as the linear combination unit 522 and the feedback unit 523 shown in FIG.₂Are the same as those input to the linear combination unit 522. Therefore, from the feedback section 11b2, the element r₁₂Is output.
[0174]
Similarly, the linear combination unit 11a₃(Not shown in FIG. 28) to 11a_KHave the same configuration as the linear combination unit 522, and the feedback unit 11b₃~ 11b_KHave the same configuration as the feedback unit 523. Therefore, the feedback section 11b₃~ 11b_kFrom the element r_Thirteen~ R_1KAre respectively output.
[0175]
(3) Third form
The configuration of the wireless base station in the third embodiment is basically the same as the configuration of the wireless base station device according to the first embodiment shown in FIG.₁~ 5_KIs converted to further reduce the circuit scale.
[0176]
First, in order to understand the difference between the weighting factor generator in the first embodiment and the weighting factor generator in the third embodiment, the process of circuit conversion from the former to the latter will be described step by step, and the equation will be described later. The principle of operation will be described with reference to FIGS.
[0177]
FIGS. 29 to 32 show the weight coefficient generation unit 5.₁First coefficient generation unit 50₁Shows the conversion steps.
[0178]
First, as a first step, the SHC 51c shown in FIG.₁~ 51c_KIs removed, and the first coefficient generation unit 50₁Is modified to the configuration shown in FIG. SHC51c₁~ 51c_KIn order to perform an operation equivalent to that before the removal by removing the MF 51b₁~ 51b_KIntegration time of SHC51c₁~ 51c_KIs required to be extended by the time during which the integration result is held.
[0179]
Subsequently, as a second step, as shown in FIG.₁~ 51b_KAnd the multiplier 51d₁~ 51d_KAnd swap the order. Weighting factor x₁(R₁₁) Is MF51b₁~ 51b_KIs constant within the integration time (0 ≦ t ≦ T), this conversion is an equivalent conversion.₁Since the value of changes over time, this transformation is not equivalent but an approximate transformation. However, the approximation error that occurs can be suppressed to a practically acceptable range by using the method described later. MF51b₁₀
Is not shown in FIG. 29, but is added in FIG. 30 to create an input of the subtractor 51f.
[0180]
Subsequently, as a third step, as shown in FIG. 31, the MF 51b in the preceding stage of the adder 51e₁~ 51b_KIs moved to the subsequent stage of the adder 51e, and is replaced with the LPF 51h at the subsequent stage. By this, K MF51b₁~ 51b_KCan be reduced to one MF 51b. Replacing the LPF 51h with the MF 51b increases the processing time of the feedback loop, but this does not pose a problem, as will be described later.
[0181]
Subsequently, as a fourth step, as shown in FIG.₁~ 51a_KIs moved to the subsequent stage of the adder 51e. As a result of the movement, the K multipliers 51a₁~ 51a_KCan be reduced to one multiplier 51a. Thereby, the first coefficient generation unit 50₁Is completed.
[0182]
Second coefficient generator 50₂The circuit shown in FIG. 33 is obtained by performing the same conversion for₃To the K-th coefficient generation unit 50_K, A circuit similar to that of FIG. 33 is obtained.
[0183]
Comparing FIG. 32 with FIG. 33, the first coefficient generation unit 50₁Multiplier 51d₁~ 51d_KAnd an adder 51e, and a second coefficient generator 50₂Multiplier 52d₁~ 52d_KSince the adder 52e and the adder 52e are common, it is sufficient to use only one common part as a whole. Although not shown, the third coefficient generation unit 50₃To the K-th coefficient generation unit 50_KThe second coefficient generation unit 50 shown in FIG.₂Since they have the same configuration as those described above, these multipliers and adders can be commonly used.
[0184]
Therefore, the first coefficient generation unit 50₁To the K-th coefficient generation unit 50_KCan be combined into one K multipliers and one adder. FIG. 34 shows a weight coefficient generation unit 5 according to a third mode in which these are combined into one.₁FIG. 3 is a block diagram showing a detailed configuration of the embodiment.
[0185]
The weighting factor generator includes a linear combination unit 500 in which a multiplier and an adder of each coefficient generator are integrated into one, and a weighting factor x₁~ X_K(R₁₁~ R_1K) That output the respective signals₁~ 501_KHaving.
[0186]
Return section 501₁Has components from the adder 51p to the adder 51m arranged at the subsequent stage of the adder 51e in FIG. Return unit 502₂Has components from the adder 52p to the MF 52 at the subsequent stage of the adder 52e in FIG. Other return section 501₃~ 501_KAlso return section 501₂It has the same configuration as
[0187]
Other weight coefficient generation unit 5₂~ 5_KCan have the same configuration. However, the j-th coefficient generation unit 5_jThen, return section 501₁The feedback unit having the same configuration as that of_JJAnd a feedback unit 501 in a part that outputs the other elements.₂A feedback unit having the same configuration as that of the first embodiment is disposed. In this feedback section, the constant 1 / (1 + α₁) In place of 定 数, constant 1 / (1 + α_j) Is input to the adder.
[0188]
The weighting factor generator shown in FIG.₁~ 501_KBy replacing the left and right arrangements of the linear combination section 500, it can also be represented as shown in FIG. In equation (12), r₁₁= X₁, R₁₂= X₂, ..., r_1K= X_KWhen the correlation coefficient is returned to the integral display, the following equation (30) is obtained.
[0189]
(Equation 16)

[0190]
The integration interval is₁(T) -s_K(T) is -1 or +1 (non-zero) in a certain finite section, but is zero before and after that, so that the integration is performed in a certain finite section.
In equation (100), the weighting factor x₁, X₂, ..., x_KHowever, if it does not change with time, the following equation (31) is obtained when viewed inside the integral.
[0191]
[Equation 17]

[0192]
Next, when a signal obtained by linearly combining the spread signal and the weight coefficient is defined as an extended spread signal, the extended spread signal can be expressed by the following equation (32).
[0193]
(Equation 18)

[0194]
When Expression (31) is expressed using Expression (32), the following Expression (33) is obtained.
[0195]
[Equation 19]

[0196]
In the method used in the first embodiment, Equation (12) is transformed into Equations (14) and (15) by the circuit formats shown in FIGS.₁₁(X₁), R₁₂(X₂),…, R_1K(X_K). When the same idea is applied to equation (33), the following equation (34) is obtained.
[0197]
(Equation 20)

[0198]
Equation (34) corresponds to FIG. 35, in which the linear combination unit 500 outputs the extended spread signal q₁(T) is output, and the feedback section 501 is output.₁Corresponds to the left side of the first row of equation (34). Return section 501₁~ 501_KCorresponds to the left side of the second row of equation (34).
[0199]
The circuit shown in FIG. 35 starts from the state where the output of the matched filter is initially zero, and continues until the steady state is reached.₁, X₂, ..., x_KVaries from the initial value to the convergence value.
[0200]
Equation (30) was transformed into equation (31) (corresponding to the conversion from FIG. 29 to FIG. 30) because the weighting factor x₁, X₂, ..., x_KHowever, it is assumed that it does not change with time.
[0201]
Accordingly, the weight coefficient x obtained by the circuit types shown in FIGS.₁, X₂, ..., x_KIs an exact solution, whereas the weighting factor x obtained from the circuit form shown in FIG. 34 or 35 is₁, X₂, ..., x_KIs an approximate solution.
A description will be given of the fact that, using these values, even for an approximate solution, it is possible to perform de-correlation reception and a method of bringing the approximate solution closer to an exact solution.
[0202]
The reception signal Y (t) expressed by the equation (1) is added to the extended spread signal q expressed by the equation (32).₁Assuming that the result of multiplication by (t) and passing through a matched filter is S, S can be expressed by the following equation (35).
[0203]
(Equation 21)

[0204]
The equation (34) becomes the equation (33) when G = ∞. Even when G is small, if the gain is corrected, it can be expressed by the equation (33). ), The following equation (36) is obtained.
[0205]
(Equation 22)

[0206]
From equation (36), α₁= Α₂= ... = α_KWhen = 0, it can be seen that no interference signal component appears in the output of the matched filter. However, the noise component differs from that of the conventional method because the weight coefficient changes during the integration period (that is, while passing through the matched filter).
[0207]
α₁, Α₂,…, Α_KIs not zero, the noise component differs from that of the conventional MMSE receiver for the same reason.
[0208]
Next, a method of approximating the weighting factor to be approximated to an exact solution will be described. In the first method, assuming that the observation window length is N, the integration time of the matched filter is (N + 1) T, and the first calculation is_j(T) (j = 1, 2,..., L) and 0 ≦ t ≦ (N + 1) T, then the first convergence value is set as the second initial value, and the Spread signal g delayed by (N + 1) T from the spread signal used in the first time_jThe second operation is performed using (t− (N + 1) T) (j = 1, 2,..., L) while (N + 1) T ≦ t ≦ 2 (N + 1) T. If the degree of approximation of the obtained weight coefficient is sufficient, the calculation is terminated here. If the degree of approximation is not sufficient, the spread signal g is similarly calculated._jThe third calculation is performed using (t−2 (N + 1) T) (j = 1, 2,..., L). In this way, since a value closer to the exact solution than the previous time is used as the initial value, a value closer to the exact solution than the previous time is obtained as the convergence value. By increasing the number of iterations, a value closer to the exact solution can be obtained. However, in this method, the time required for the operation is (N + 1) T × the number of repetitions.
[0209]
In the second method, as a matched filter, the integration time is set to T, which is the same as the bit length, and the time zone required for the first operation is divided into a plurality of sections of length T, section 1, section 2, and section P. , An extended spread signal is generated for each section, and a time section p and a section p of the spread signal of the section p are generated.₀Is shifted by the time difference of₀After that, the operation is repeated at a period T. According to this method, the time required for one operation is T, and the total time is T × the number of repetitions, and the operation time is reduced to (N + 1) times as compared with the first method. FIG. 36 compares the time required for the first method and the time required for the second method. Note that, in this figure, numbers represent section numbers, not bit numbers.
[0210]
(4) Fourth form
FIG. 37 shows a configuration diagram of a receiving device of the radio base station 120 according to the fourth embodiment.
[0211]
The configuration of this device is basically the same as the first and second embodiments, but includes a modified spread signal generator 12 and further generates a weighting factor based on the technology used in the two embodiments. This achieves high-speed operation of the unit and downsizing of the device.
[0212]
The modified spread signal generation unit 12 generates a spread signal g for correlation coefficient generation, which is generated by the spread signal generation unit.₁(T), g₂(T), ..., g_LIn addition to (t), a plurality of spread signal signals are generated by shifting the spread signals by a predetermined time. As a method of generating a time-shifted spread signal, a method similar to the method described with reference to FIG. 10 is used. The plurality of spread signals are transformed into a modified spread signal Σg_j(T).
[0213]
Referring again to FIG. 4, this figure illustrates an asynchronous state between received signals of the number of users L at a single rate, and the reception time of the received signal does not concentrate on a specific time. In a model in which the signals are uniformly distributed, bit numbers are assigned in ascending order of arrival time at the receiver. FIG. 4A shows the case where the observation window length (Observation Window Size) N = 2 and K = 2L, and FIG. 4B shows the case where the observation window length N = 3 and K = 3L. I have.
[0214]
Here, in FIG. 4A, assuming that the desired signal bit is the (L + 1) th bit (hereinafter, referred to as “bit L + 1”), in a system with an observation window length N = 2, it is necessary to decode the desired bit. , 3 bits from the beginning of bit 1 to the end of bit 2L.
[0215]
Therefore, in order to continuously decode the bit sequence of the user signal 1 such as bits 1, L + 1, 2L + 1, and 3L + 1, the size of the weighting factor generation unit in the wireless base station device 120 needs to be three times as large. Similarly, in a system with an observation window length of N, demodulation of a desired signal bit requires N + 1 bit time. Therefore, in order to continuously decode a bit sequence of a certain user signal, the scale of the weight coefficient generation unit is set to N + 1. Twice as much.
[0216]
On the other hand, although each received user signal spreads in the section of N + 1 bits, attention is paid to the fact that the calculation time zone of the weight coefficient does not necessarily have to be synchronized with the received signal, and the modified spread signal {g_jThe feature of the fourth embodiment resides in that the calculation time of the weight coefficient is consolidated into a specific time zone using (t) to reduce the time required for decoding one bit.
[0219]
The fourth embodiment will be described as an example based on the technology described in the third embodiment, but can be similarly applied to the correlation coefficient generation described in the first and second embodiments.
[0218]
First, the principle of the fourth embodiment will be described using a simple model, and then generalized.
[0219]
FIG. 38A shows, as an example, spread signals corresponding to two bits (bits 1 and 3 and bits 2 and 4) of two single-rate user signals 1 and 2 received asynchronously. . The bits of the user signal 1 and the bits of the user signal 2 are shifted by a time τ. That is, bit 2 starts with a time τ delay from the start of bit 1, and bit 3 starts with a time τ delay from the start of bit 2.
[0220]
Each spread signal of bits 1 to 4 is represented by s₁(T) -s₄(T) and the weighting factor is r₁₁(X₁) -R_1K(X₄), The following expression (37) is obtained from the above expression (10), with the bit number of the desired signal being 3.
[0221]
[Equation 23]

[0222]
When Expression (18) is expressed as an integral expression for obtaining a correlation coefficient, Expression (38) below is obtained.
[0223]
(Equation 24)

[0224]
In equation (38), the integration interval extends from −T to T + τ. This means that the spread signal s synchronized with the received signal₁(T) -s₄Attempting to obtain a weighting factor using (t) means that an integration time of 2T + τ (approximately 3T when τ has a value close to T) is required. Further, in order to decode the bit string of the user signal 1 continuously, as described above, a circuit having a scale three times as large as the weight coefficient generation unit is required.
[0225]
On the other hand, even if the integrand (ie, the spread signal) is shifted in time, if the start and end points of the integration interval are changed by the shift amount, the integration result will be the same. If they are collected in a specific time zone, the overall processing time can be reduced.
[0226]
For example, the second term on the left side of the first equation of the above equation (38) is as follows.
[0227]
(Equation 25)

[0228]
Similarly, the second term on the left-hand side of the fourth equation of the equation (38) is as follows.
[0229]
(Equation 26)

[0230]
These two equations are examples in which the integration time is shifted to 0 ≦ t ≦ T by advancing or delaying the spread signal by one bit time T. Assuming that the integration interval is 0 ≦ t ≦ T for all the expressions (38), the expression (38) becomes the following expression (39).
[0231]
[Equation 27]

[0232]
This time shift will be described with reference to FIG. 38. The spread signal s in the range of −T ≦ t ≦ 0₁All of (t) and the spread signal s₂Part of (t) is delayed by the time T and shifts to the position of FIG. Spread signal s in the range of 0 ≦ t ≦ T₃All of (t), the spread signal s₂(T) and the spread signal s₄Part of (t) is not shifted as shown in FIG. Spread signal s in the range of T ≦ t₄The remainder of (t) is advanced by time T and shifted as shown in FIG. In the description, the expression “advanced” or “delayed” by the time T is used. However, what is actually performed by the circuit is a modified spread signal generation unit that transmits a continuous signal. Spread signal Σg_jIn (t), a specific part is selected by a switch according to the symmetric bit and the section.
[0233]
As described above, in order to generalize the conversion for shifting the integration interval within 0 ≦ t ≦ T, the integration interval is shifted by changing the observation window length N = 2 and the user signal number L shown in FIG. When applied to a received signal, the following equations (40) to (42) are obtained. (See FIGS. 39 (a)-(d))
For 1 ≦ k ≦ L,
[0234]
[Equation 28]

[0235]
For L + 2 ≦ k ≦ 2L,
[0236]
(Equation 29)

[0237]
For the desired bit L + 1,
[0238]
[Equation 30]

[0239]
Where t_kIs the time of the conversion point at 0 ≦ t ≦ T for the bit with bit number k.
[0240]
Next, an embodiment in which such a principle is applied to the weight coefficient generation unit shown in FIG. 34 or 35 will be described. The received signal is as shown in FIG.
[0241]
From the above equations (40) to (42), the feedback section 501 shown in FIG. 34 (or FIG. 35) is obtained.₂~ 501_KThe spread signal input to each of the first input terminal and the second input terminal of each of the multipliers 52a to 5Ka of the (feedback unit relating to interference bits)_kAnd t_kThe shift processing can be performed by switching between inputting the original as ≦ t ≦ T or inputting a delayed or advanced one.
[0242]
FIG. 40 shows a part of the feedback section relating to the interference bit, and shows the correspondence between spread signals input to the first input terminal (input 1) and the second input terminal (input 2) of the multiplier. This is shown in FIG.
[0243]
FIG. 41 is a block diagram showing the configuration of the weighting factor generator configured based on FIG. 40, in which necessary functions have been added based on the weighting factor generator shown in FIG. However, in FIG. 41, the display of the MMSE correction unit and the feedback loop gain correction is omitted to avoid complicating the drawing.
[0244]
In this weight coefficient generation unit, a switch V is provided before a multiplier of a feedback unit that processes interference bits 1 to L and L + 2 to L, respectively.₁~ V_L, V_{L + 2}~ V_2LAre respectively provided.
[0245]
switch V of k = 1 to L_kIs the time t_k, The first input terminal indicated by the broken line is switched to the second input terminal indicated by the solid line, and 0 ≦ t ≦ t_kThen, the output signal of the divider 801 is output to the subsequent multiplier, and t_kIf ≤t≤T, the output signal of the divider 802 is output to the subsequent multiplier.
[0246]
switch V of k = L + 2 to 2L_kIs the time t_k, The second input terminal indicated by the broken line is switched to the first input terminal indicated by the solid line, and 0 ≦ t ≦ t_kThen, the output signal of the distributor 803 is output to the subsequent multiplier, and t_kIf ≤t≤T, the output signal of the distributor 801 is output to the subsequent multiplier.
[0247]
A switch W is provided on the output side of a multiplier (a multiplier of a linear combination unit) disposed at a stage subsequent to the feedback unit for processing the interference bits 1 to L and L + 2 to L, respectively.₁~ W_L, W_{L + 2}To W2L.
[0248]
switch W of k = 1 to L_kIs the time t_k, The first output terminal indicated by the broken line is switched to the second input terminal indicated by the solid line, and 0 ≦ t ≦ t_kThen, the output signal of the multiplier at the previous stage is output to the synthesizer (adder) 701, and t_kIf ≦ t ≦ T, the output signal of the multiplier at the preceding stage is output to the synthesizer (adder) 702.
[0249]
k = L + 2-2L switch W_kIs the time t_k, The output terminal is switched from the second output terminal indicated by the broken line to the first output terminal indicated by the solid line, and 0 ≦ t ≦ t_kThen, the output signal of the multiplier at the previous stage is output to the synthesizer 703, and t_kIf ≦ t ≦ T, the output signal of the multiplier at the preceding stage is output to the combiner 701.
[0250]
The combiners (adders) 701 to 703 combine the input signals and feed back the combined results to the distributors 801 to 803, respectively. The combiner 701 corresponds to section 2 of FIG. 39A, the combiner 702 corresponds to section 1, and the combiner 703 corresponds to section 3. Thus, the number of synthesizers and the number of sections are N + 1.
[0251]
Each output signal of the combiners 701 to 703 is represented by the following equation.
[0252]
(Equation 31)

[0253]
The distributors 801 to 803 distribute the input signal to a subtractor or a switch.
[0254]
The spread signal s is input to the input 2 of the two multipliers of the feedback unit for the interference bits 1 to L._kIn addition to (t), s_k(TT) is input. Also, the spread signal s is input to the input 2 of the two multipliers of the feedback unit for the interference bits L + 2 to 2L._kIn addition to (t), s_k(T + T) is input. The integration time of the MF of each feedback unit is set to 0 to T.
[0255]
As described above, in the second embodiment, one cycle of the weighting coefficient calculation that took 3T (T is a bit length) (as described above, it is necessary to repeat the calculation for a plurality of cycles in order to increase the calculation accuracy of the weighting coefficient. Can be reduced to T in the third aspect, and as a result, the circuit scale can be further reduced.
[0256]
In the above, the principle and the embodiment of shifting the processing section by the bit length (T) to reduce the processing time have been described. If the processing section is T / Q (Q is an integer), then Q> 1. As a result, the processing time can be further reduced.
[0257]
First, the case where Q = 2 will be described, aspects different from Q = 1 will be described, and embodiments for arbitrary integers where Q> 2 will be described later.
[0258]
FIG. 42 (a) shows the same arrangement of spread signals as in FIG. 38 (a), except that it is divided into six sections for each section length T / 2. FIGS. 44 (b) to (d) show how all spread signals are shifted to section 3. FIG.
[0259]
The bit number of the desired signal is 3, and the corresponding spread signal is s.₃(T). Spread signal s corresponding to the interference bit₂(T) and s₄The handling differs depending on whether the boundary of (t), that is, the conversion point of the bit of the interference signal is in section 3 or section 4. Therefore, assuming that when the bit number k is 1 ≦ k ≦ M, the conversion exists in the section 3, and when the bit number k is M + 1 ≦ k ≦ L, the conversion point exists in the section 4. When the bit number k is L + 2 ≦ k ≦ L + M, the conversion point also exists in the section 3, and when the bit number kL + M + 1 ≦ k ≦ 2L, the conversion point exists in the section 4.
[0260]
When equations corresponding to the above equations (40) to (42) in the case of the number of users L and the observation window length N = 2 shown in FIG. 39 are obtained, the following equations (43) to (47) are obtained.
[0261]
For 1 ≦ k ≦ M,
[0262]
(Equation 32)

[0263]
For M + 1 ≦ k ≦ L,
[0264]
[Equation 33]

[0265]
For L + 2 ≦ k ≦ L + M,
[0266]
(Equation 34)

[0267]
For L + M + 1 ≦ k ≦ 2L,
[0268]
(Equation 35)

[0269]
For the desired bit,
[0270]
[Equation 36]

[0271]
Where t_kIs the time of the conversion point of the bit with bit number k, and 0 ≦ t_k≦ T.
[0272]
As is clear from the above equations (24) to (28), all the integrations (that is, the processing by the MF) are completed between 0 and T / 2.
[0273]
Similarly, for an integer Q where Q> 2, the spread signal can be shifted to a section of 1 / Q bit length to shorten the processing to T / Q time. Hereinafter, this technique will be referred to as a 1 / Q bit method, and an embodiment of the technique will be described.
[0274]
FIG. 45 is a block diagram illustrating a multiplier and a feedback unit and a linear combination unit regarding an interference bit in a weight coefficient generation unit based on the 1 / Q bit method. In order to facilitate comparison with the one-bit method described above, FIG. 46 (a) shows a multiplier of a feedback unit and a linear combination unit for an interference bit in a weight coefficient generation unit (see FIG. 43) using the one-bit method. Show the part.
[0275]
In the 1 / Q bit method, Q multipliers are required before the inverting amplifier. Also, Q multipliers are required in the subsequent stage of the MF. On the other hand, in the 1-bit method, as shown in FIG._iAnd W_iIs provided, the multipliers before and after the MF can be combined into one.
[0276]
The reason for this is that in the 1 / Q bit method, it is necessary to input the multiplication results of different spread signals to the MF at the same time. For example, in the second and third terms on the left side of equation (26), t_kWhen ≤t≤T / 2, the spread signal s_k(T) and this t_kIn a time zone of 0 ≦ t ≦ T / 2 that partially overlaps with ≦ t ≦ T / 2, the spread signal s_kSpread signal s obtained by advancing (t) by time T / 2_kIt is necessary to simultaneously input two spread signals (t + T / 2) to the MF. On the other hand, in the one-bit method, a certain spread signal and a spread signal obtained by shifting the spread signal in time are not input to the MF in the same time zone.
[0277]
With reference to FIG. 44, a more specific description will be given by taking the case of L + 2 ≦ k ≦ L + M as an example. Spread signal s₄To shift (t) to section 3, the spread signal s₄It is necessary to advance the portion of section 4 of (t) to T / 2, and to advance the time of the section 5 of the same signal by T. Thereby, the spread signal s₄(T) is divided into three parts indicated by reference numerals (1) to (3). Since the parts (1) and (2) do not overlap in the time zone, the multipliers at the front and rear stages of the MF can be shared, so that only one multiplier is required for each of the front and rear stages.
[0278]
On the other hand, since the portion (2) occupies the time zone of 0 ≦ t ≦ T / 2, the multiplier used by the other portions (1) and (3) cannot be shared. Therefore, a multiplier for performing the multiplication of the part (2) is separately required. As a result, a total of two multipliers are required: a multiplier shared by the parts (1) and (3) and a multiplier used by the part (2).
[0279]
When generalized to the 1 / Q bit method, in the 1 / Q bit method, a spread signal corresponding to one bit is divided into (Q + 1) parts. Among these parts, the length (time) of the (Q-1) parts is T / Q. The remaining two parts do not overlap in time and together have the length of the furniture T / Q. Therefore, the multiplier that processes these two parts is the multiplier 700 in the uppermost stage in FIG.₁And 800₁Corresponding to Multiplier 700₁Before and 800₁At the subsequent stage, switches are provided, respectively, and two parts are switched and processed. Multiplier 700₂~ 700_QAnd 800₂~ 800_QIs a multiplier that processes (Q-1) parts. Thus, the 1 / Q bit method requires a total of 2Q multipliers.
[0280]
It is also possible to eliminate two switches and provide two multipliers for individually processing two parts whose time does not overlap. In this case, the total number of multipliers is 2Q + 2. Similarly, in the 1-bit method, as shown in FIG. 46B, two switches can be provided without switches.
[0281]
Assuming that a circuit surrounded by a broken line in FIG. 45 is a circuit J, the weighting factor generator based on the 1 / Q bit method has a configuration shown in FIG. This weight coefficient generation unit is composed of K circuits J₁~ J_K, P synthesizers 900₁~ 900_P, And P distributors 901₁~ 901_PHaving. To avoid complicating the drawing, the circuit J₁~ J_KAnd synthesizer 900₁~ 900_PIs shown as a connection network R. Distributor 901₁~ 901_PAnd circuit J₁~ J_KIs also shown as a connection network L.
[0282]
Here, P is the number of divisions of the observation section (the range of N + 1 bits in the case of the observation window length N), and in the example of FIG. 44, P = 6.
[0283]
Circuit J₁~ J_KAs shown in FIG. 45, the output signal to the connection network R is the result of multiplication of the shifted spread signal and the weight coefficient. This output signal is connected to the combiner corresponding to the section to which the spread signal before the shift belongs. Therefore, the connection of the connection network R is uniquely determined if the arrangement of the spread signal is determined.
[0284]
Distributor 901₁~ 901_KHas a synthesizer 900₁~ 900_PIs input as a feedback signal. And the distributor 901₁~ 901_KConnects the output of the synthesizer to the circuit J via the connection network L.₁~ J_KDistribute to The right combiner and the left divider are connected one by one in the same section, and the entire system forms a feedback loop. The connection structure of the connection network L is symmetric with the connection network R.
[0285]
From the modified spread signal section 12, the spread signal g₁(T)-g_LIn addition to (t), a plurality of signals (deformed spread signal g) advanced or delayed with respect to them by an integral multiple of 1 / Q of the integration time of the matched filter.₁(T)-g_L(T) or the modified spread signal Σg_j(T)) is output and input to the network S. The output of the circuit network S is output to a multiplier (in FIG. 43, a multiplier 700₁~ 700_QAnd multiplier 800₁~ 800_Q). Since the length of the spread signal input to the multiplier is 1 / Q bits or less, the modified spread signal s₁(T) -s_K(T) or modified spread signal Σs_j(T), the modified spread signal Σs_j(T) is generated by changing the connection timing and length of the switch provided between the input and output of the circuit network S. Control of the switch is performed by an external control circuit (not shown in FIG. 37).
[0286]
In an asynchronous system, a received signal may be concentrated in a specific section. Therefore, the number of inputs to the synthesizer may be large or small depending on the corresponding section. In that case, the number of input / output terminals of the circuit network R is increased, switches are provided between the input / output terminals, and the number of necessary paths to the synthesizer is secured (not shown in FIG. 37). Is performed by the control circuit. Similar treatment is applied to the circuit network L.
[0287]
Next, an actual operation was confirmed by simulation based on the device configuration shown in the description of the fourth embodiment, and the result will be described.
The simulation model is such that N = 2, L = 2, K = N × L = 4, the bit arrangement and the spread signal pattern are the same as those of the simulation performed according to the first and second modes shown in FIG. Is the same. By applying the 1-bit method described as the fourth embodiment, the section of the desired signal bit (bit number 3) is set to section 2, and the section before the section 1 and the section 3 after the section are set. All the spread signals have been shifted to section 2. The device configuration is as shown in FIG. FIG. 46 shows a state in which each spread signal is repeatedly input to the input of the multiplier (input 2 defined in FIG. 40) at a period T, the feedback loop reaches a steady state, and a correct weighting factor is obtained. The vertical axis represents the value of the weighting coefficient, and the horizontal axis represents the time normalized by the bit length T, where the start of the desired signal bit (bit number 3) is set to t = 0. Since the state of convergence is difficult to understand because the model is simple, the deviation from the exact solution of the weighting factor (obtained as the inverse matrix of the correlation coefficient matrix; see FIG. 22C) is shown in FIG. is there. From this figure, it can be seen that the weight coefficients required for the correlation type reception can be obtained by the techniques shown in the third and fourth embodiments.
[0288]
<Configuration of mobile device>
Next, a mobile device 200 according to an embodiment of the present invention will be described (fifth embodiment). The configuration of the receiving device of the mobile device 200 is basically the same as that of the radio base station device 120, except that only the signal addressed to the own station needs to be decoded. , And the number of decision circuits are reduced. Signals addressed to the own station and other stations sent from the same base station are usually sent synchronously and they are orthogonal (they do not appear as interference signals in the despread signal). The signal becomes an interference signal source only when the transmission path of the signal is different and the delay time is different.
[0289]
Inverse matrix element generator 5₁Can be applied to any of the various modes described for the radio base station apparatus 120, but the extended spread signal q which is the output of the linear combination section in FIG.₁Using (t) for the despreading of the received signal at the same time as obtaining the weighting coefficient simplifies the configuration of the matched filter 3, so that the method will be described.
[0290]
FIG. 48 is a block diagram showing a configuration of the mobile device 200. In this figure, only the part necessary for decoding the number of user signals (number of users) L = 1 is shown. The same components as those of the wireless base station device 120 according to the first to fourth embodiments are denoted by the same reference numerals, and detailed description thereof will be omitted.
[0291]
The mobile device 200 shown in FIG.₁~ 5_KInstead of (N + 1) weight coefficient generation units 19₁, 19_{L + 1}, 19_{2L + 1}, ..., 19_{NL + 1}Are provided, and an extended matched filter bank 13 is provided instead of the matched filter bank 3. In addition, the linear combination 6₁~ 6_KIs omitted.
[0292]
Here, N is the observation window length of the received signal shown in FIG. 4 (N = 2 in FIG. 4A, N = 3 in FIG. 4B), L is the number of user signals (the number of users), and The subscripts “1”, “L + 1”, “NL + 1”, etc. of the code of the unit 4 correspond to the bit numbers in the observation window of each user signal.
[0293]
FIG. 49 is a block diagram showing a configuration of the extended matched filter bank 13. The combining unit 13 includes (N + 1) multipliers 130₁, 130_{L + 1}, 130_{2L + 1}, ..., 130_{NL + 1}, (N + 1) MF131₁, 131_{L + 1}, 131_{2L + 1}, ..., 131_{NL + 1}, (N + 1) SHC 132₁, 132_{L + 1}, 132_{2L + 2}, ..., 132_{NL + 1}, Has.
[0294]
Weight coefficient generator 19₁, 19_{L + 1}, 19_{2L + 1}, ..., 19_{NL + 1}Is the extended spread signal q₁(T), q_{L + 1}(T), q_{2L + 1}(T), ..., q_{NL + 1}(T) is output. Here, the extended spread signal q_i(T) is represented by equation (32).
[0295]
Extended spread signal q₁(T), q_{L + 1}(T), q_{2L + 1}(T), ..., q_{NL + 1}(T) is the multiplier 130 of the synthesis unit 13₁, 130_{L + 1}, 130_{2L + 1}, ..., 130_{NL + 1}Are respectively input to the second input terminals. Multiplier 130₁, 130_{L + 1}, 130_{2L + 1}, ..., 130_{NL + 1}Are the baseband signal Y (t) input to the first input terminal and the extended spread signal q input to the second input terminal, respectively.₁(T)-q_{NL + 1}(T), and multiply the result by MF131₁~ 131_{NL + 1}Give to each. The multiplication result is MF131₁~ 131_{NL + 1}Respectively. The integration time is (N + 1) T, where T is the bit length. The integration result is SHC132₁~ 132_{NL + 1}Given to.
[0296]
SHC132₁~ 132_{NL + 1}Outputs the value sampled at the end of each observation window.
[0297]
The output of the extended matched filter bank 13 is₁Identifies a discrete value or sign.
[0298]
Note that the spread signal used here is used not only for obtaining a weighting coefficient but also for despreading the received signal, so that the received signal is chip-synchronized (with an accuracy of a fraction of the chip length or less). Sync). In the case of repeating a plurality of times in order to increase the calculation accuracy of the weighting coefficient, the last time is synchronized with the received signal. In the example of FIG. 36, in the case of (a), the third time synchronizes the received signal with the chip. However, except for the last round, it is possible to perform the calculation using the time-shifted modified spread signal shown in (b) and repeat the calculation with the convergence value as the initial value of the last round.
[0299]
The first to fifth embodiments have been described above, and modifications common to them will be described.
[0300]
Let the bit number of the desired signal be k and the weighting factor x for the desired bit_kThen, the value obtained by dividing all the other weighting factors is calculated as a new weighting factor x_j´ = x_j/ X_k(J = 1, 2,..., K) and input to the linear combination unit 6, the output is x₁, X₂, ..., x_KIs 1 / x_KIt will be doubled. Therefore, x_kInstead of using the predicted average value to ignore the level change, or if the transmitted modulated signal has no amplitude information (for example, BPSK), x_kCan be a constant (for example, 1). In this case, a device for calculating a weight coefficient for the desired signal bit (the first coefficient generation unit shown in FIG. 13 and the feedback unit 501 for the desired bit shown in FIG. 34)₁) Can be omitted.
[0301]
The error correction (gain correction) according to the gain of the feedback loop shown in FIG. 17 is performed by the matching filter (MF) 51 shown in FIG. 34 (or FIG. 35)._b, 52_b, ..., 5K_bApplying to has the same effect. That is, the matched filter (MF) 51_b, 52_b, ..., 5K_bIs added to each input. At this time, the gain of the local feedback loop formed by the input and output of the filter is adjusted by inserting an amplifier or an attenuator in the path from the output to the input of each matched filter.
[0302]
In the feedback loop formed by the feedback unit and the linear combination unit, the connection order of the inverting amplifier and the LPF (or matched filter) can be changed (in the example of FIG. 13, the connection order of the amplifier 51g and the LPF 51h is changed. In the example, the connection order of the amplifier 52g and the LPF 52h is changed). In addition, the position of the inverting amplifier can be changed within the feedback loop. However, in the feedback loop, it is necessary to change the level and the sign of addition and subtraction, if there is a place to perform subtraction.
[0303]
In the circuit shown in FIG. 34, the feedback unit 501₁~ 501_KAn amplifier or an inverting amplifier is inserted until the input of each feedback section.₁~ 501_KSome or all of the inverting amplifiers in parentheses) can be omitted. However, when an amplifier is inserted, it is necessary to insert an inverter at any point in the feedback loop to operate as a negative feedback loop.
[0304]
Any component in the feedback loop (for example, the multiplier 51d in FIGS. 13 and 34)₁~ 51d_k34, the multipliers 51a to 5Ka and the MFs 51b to MF5Kb) are formed of active elements, and if the gain of the feedback loop can be secured without an inverting amplifier, the inverting amplifier can be replaced with an inverting attenuator.
[0305]
Although the embodiments described so far apply the present invention to wireless communication, the present invention can be applied not only to wireless communication but also to wired communication.
[0306]
【The invention's effect】
In the case of using the method according to the present invention, particularly, the method according to claims 1 to 3, if the number of bits to be subjected to interference removal is assumed to be K, the circuit scale which has conventionally been proportional to the cube of K is K The size can be reduced in proportion to the square.
[0307]
In addition, the time required to obtain K weight coefficients required for decoding one bit of a desired signal can be reduced, and the integration time of a matched filter for determining a correlation coefficient can be reduced.
[0308]
Further, when the number M of data strings to be decoded is smaller than K (this applies to a mobile device), the number of matched filters can be reduced.
[Brief description of the drawings]
FIG. 1 is a block diagram showing an overall configuration of a mobile communication network system according to an embodiment of the present invention.
FIG. 2 shows a configuration of a signal according to the DS-CDMA system in an asynchronous and multi-rate case received by a radio base station apparatus.
3A and 3B show a configuration of a signal according to the DS-CDMA system received by the radio base station apparatus, wherein FIG. 3A shows a configuration of a reception signal in an asynchronous case, and FIG. 3B shows a configuration of a reception signal in a synchronous case. The configuration is shown respectively.
4A and 4B are diagrams showing the concept of an observation window, wherein FIG. 4A shows a bit sequence of observation window length N = 2 and the number of users L, and FIG. 4B shows a bit sequence of observation window length N = 3 and the number of users L. .
FIG. 5 is a block diagram illustrating a configuration of a multi-user receiving device of a wireless base station.
FIG. 6 shows a relationship between a spread signal and a burst spread signal.
FIG. 7 is a block diagram showing a detailed configuration of a matched filter bank.
FIG. 8 is a block diagram showing a configuration of a matched filter bank for obtaining a correlation coefficient.
FIG. 9 is a block diagram showing another configuration of a matched filter bank for obtaining a correlation coefficient.
FIG. 10 is a block diagram illustrating a configuration of a spread signal generation unit.
FIG. 11 is a block diagram illustrating a configuration of a wireless base station device (particularly, a multi-user receiving unit) according to the first embodiment.
FIG. 12 is a block diagram illustrating a configuration of a weight coefficient generation unit according to the first embodiment.
FIG. 13 is a block diagram illustrating a detailed configuration of a first coefficient generation unit of the weight coefficient generation unit.
FIG. 14 is a block diagram illustrating a detailed configuration of a second coefficient generation unit of the weight coefficient generation unit.
FIG. 15 is a block diagram illustrating a configuration of a linear combination unit.
FIG. 16 is a diagram illustrating a configuration of a feedback unit of a first coefficient generation unit that has performed gain correction.
FIG. 17 is a diagram illustrating another configuration of the feedback unit of the first coefficient generation unit that has performed gain correction.
FIG. 18 is a diagram illustrating a configuration of a feedback unit of the first coefficient generation unit in which a position where MMSE correction is performed is changed.
FIG. 19 is a diagram illustrating another configuration of the feedback unit of the first coefficient generation unit in which a position where MMSE correction is performed is changed.
20 is a diagram illustrating another configuration of the feedback unit of the first coefficient generation unit illustrated in FIG.
FIG. 21 is a diagram showing waveforms of a user signal used for the simulation and a spread signal corresponding to the user signal.
FIG. 22 is a diagram illustrating specific numerical values of a chip sequence of a spread signal and an inverse matrix of a correlation coefficient matrix used in the simulation.
FIG. 23 is a diagram illustrating a temporal change of a weight coefficient, which is a result of a simulation.
FIG. 24 is a diagram illustrating a temporal change of a deviation from an exact solution of a weight coefficient, which is a result of a simulation.
FIG. 25 is a diagram illustrating a temporal change of a deviation from an exact solution of a weight coefficient, which is a result of a simulation.
FIG. 26 is a block diagram illustrating a configuration of a wireless base station device according to a first mode (a modification).
FIG. 27 is a block diagram illustrating a configuration of a correlation coefficient generation unit.
FIG. 28 is a block diagram illustrating a configuration of a weight coefficient generation unit.
FIG. 29 shows a configuration in which the sample holding circuit is removed in the first step of conversion of the first coefficient generation unit.
FIG. 30 shows a configuration in which the position multiplied by the weight coefficient is changed from the output of the MF to the input in the second step of the conversion of the first coefficient generator.
FIG. 31 shows a configuration in which the MF and the feedback loop filter are replaced in the third step of the conversion of the first coefficient generation unit.
FIG. 32 shows a configuration in which the order of multiplication by the multiplier is changed in the fourth step of conversion of the first coefficient generation unit.
FIG. 33 is a block diagram illustrating a configuration after conversion of a second coefficient generation unit.
FIG. 34 is a block diagram illustrating a detailed configuration of a weight coefficient generation unit (second embodiment).
FIG. 35 is a block diagram illustrating a detailed configuration of a weight coefficient generation unit (second embodiment).
FIG. 36 is a diagram illustrating two techniques for an iterative operation in the weight coefficient generation unit.
FIG. 37 is a block diagram illustrating a configuration of a wireless base station device according to a third embodiment having a modified spread signal generation unit.
FIG. 38A shows positions of spread signals corresponding to two bits of each of two single-rate user signals received asynchronously, and FIGS. 38B to 38D show these spread signals. The figure shows a shift to a certain section (1 bit length) of a desired signal bit.
FIG. 39 (a) shows the position of a spread signal corresponding to each bit of L user signals; ) Is shown.
FIG. 40 shows a correspondence relationship between two inputs of a multiplier and a section in a feedback unit of a weight coefficient generation unit (second mode).
FIG. 41 is a block diagram illustrating a configuration of a weight coefficient generation unit according to the third embodiment.
FIG. 42 (a) shows the same arrangement of spread signals as in FIG. 38 (a), except that the section is divided into 6 sections every 1/2 bit. (B) to (d) show a state where the spread signal in each section is shifted to section 3 which is the first half of the desired signal bit.
FIG. 43 is a block diagram illustrating a configuration of a multiplier of a feedback unit and a linear combination unit in a weight coefficient generation unit based on the 1 / Q bit method.
FIGS. 44 (a) and (b) show a configuration of a multiplier of a feedback unit and a linear combination unit based on the 1-bit method.
FIG. 45 is a block diagram illustrating a configuration of a weight coefficient generation unit based on the 1 / Q bit method.
FIG. 46 is a graph showing a result of a simulation performed to verify the one-bit method (time change of a weight coefficient).
FIG. 47 is a graph showing the result of the simulation as a deviation from the exact solution of the weight coefficient.
FIG. 48 is a block diagram illustrating a configuration of a mobile device according to an embodiment of the present invention.
FIG. 49 is a diagram illustrating a configuration of a matched filter bank for performing despreading with an extended spread signal.
[Explanation of symbols]
120₁~ 120₄Wireless base station equipment.
200 mobile device
3 Matched filter bank
4 Spread signal generator
5₁~ 5_K, 9₁~ 9_K, 18₁~ 18_K, 19₁~ 19_{NL + 1}Weight coefficient generator
6₁~ 6_K, 512, 522, 11a₁~ 11a_KLinear combination
8 MMSE correction unit
10,511,521 correlation coefficient generator
11 ° weight coefficient calculation unit
12 deformed spread signal generator
13 Extended matched filter bank
14 inverse matrix operation unit
15, 10a₁-10a_KCorrelation coefficient matching filter bank
16 deformation matched filter bank
513, 523, 11b₁~ 11b_KReturn section
3a₁~ 3a_K, 51a₁~ 51a_K, 51d₁~ 51d_KMultiplier
3b₁~ 3b_K, 51b₁~ 51b_KMatched filter (MF)
51g, 52g inverting amplifier
51e, 52e, 70₁~ 70_pSynthesizer
80₁~ 80_p Distributor

Claims (19)

DS-CDMA (and L pieces) a plurality of received signals of the system spread signal _{f i (t) (i =} 1,2, ..., L) to produce a, a spread signal generating section,
A matched filter bank for inputting an output obtained by despreading the received signal with the spread signal to a plurality of matched filters, sampling an output of the matched filter, and outputting the sampled output;
K spread signals s _j (t) (j = 1, 2) corresponding to a plurality (K) of values (y ₁ , y ₂ ,..., Y _K ) in the output of the matched filter bank. ,..., K), a background noise is added to a correlation coefficient generation unit that generates a K × K matrix correlation coefficient matrix R using the correlation coefficient between the spread signals as an element, or a diagonal element of the R. A modified correlation coefficient generation unit that generates a modified correlation coefficient matrix R ′, to which a value proportional to a ratio of a power density and an estimated received power per bit of a received signal corresponding to the diagonal element is added;
A weighting factor generator for calculating _K weighting factors (x ₁ , x ₂ ,..., X _K ) for the K values (y ₁ , y ₂ ,..., YK);
If k is the bit number of the desired signal,
An inverse matrix of the correlation coefficient matrix R or the modified correlation coefficient matrix R ′ is calculated, and the k-th row or the k-th column of the inverse matrix is replaced with the K weighting factors (x ₁ , x ₂ ,. x _K )
The sum of the products of the _K weighting factors (x ₁ , x ₂ ,..., X _K ) and the K values (y ₁ , y ₂ ,..., Y _K ) is y ₁ x ₁ + y ₂ x _2. + ... + y _K x _K
In a multi-user receiver that decodes a desired signal bit by determining a discrete level or a sign with respect to the output of the linear combination unit,
The observation section including the section where the reception signal corresponding to the _K values (y ₁ , y ₂ ,..., Y _K ) exists is evenly divided by a predetermined length τ, and , And are called section 1, section 2, ..., section P,
The spread signal generation unit is configured to generate a plurality of modified spread signals that are advanced or delayed by an integer multiple of τ with respect to the spread signal f _i (t) (i = 1, 2,..., L), or Generating a signal obtained by advancing or delaying the entire modified spread signal by a predetermined time;
One of the P-number of the section and a section p _0,
For a section p where p <p ₀ , a part of the spread signal s _i (t) (j = 1, 2,..., K) belonging to the section p is delayed by a time (p ₀ −p) τ, A part that matches with the modified extended signal is _defined as a spread signal s _ip (t),
For the section p where p> p ₀ , the part belonging to the section p of the spread signal s _i (t) (j = 1, 2,..., K) is advanced by time (p−p ₀ ) τ, A part that matches with the modified extended signal is _defined as a spread signal s _ip (t),
Regarding the section p ₀ , a part of the spread signal s _j (t) (j = 1, 2,..., K) belonging to the section p ₀ is defined as a spread signal s _ip (t),
The weighting factor generator includes:
The calculation of the correlation coefficient or the weighting coefficient is performed by _{dividing the} spread signal s _ip (t) (j = 1, 2,..., K, p = 1, 2,. using multiple matched filters is the same tau, performed in the interval p _0,
Multi-user receiver. A plurality of (L) spread signals are generated from a received signal of the DS-CDMA system. A spread signal generator and an output obtained by despreading the received signal with the spread signal are input to a plurality of matched filters. A matched filter bank that outputs a value obtained by sampling the output of the matched filter;
The spread signal s _j (t) (j = 1, 2,..., K) corresponding to a plurality of (K) values (y ₁ , y ₂ ,..., Y _K ) in the output of the matched filter bank ) To calculate _K weighting factors (x ₁ , x ₂ ,..., X _K );
The sum of the products of the _K weighting factors (x ₁ , x ₂ ,..., X _K ) and the K values (y ₁ , y ₂ ,..., Y _K ) is y ₁ x ₁ + y ₂ x _2. + ... + y _K x _K
In a multi-user receiver that decodes a desired signal bit by determining a discrete level or a sign with respect to the output of the linear combination unit,
The weighting factor generator includes:
The K spread signals s _j (t) (j = 1, 2,..., K) are multiplied by the K weight coefficients x _j (j = 1, 2,. _j (t) x _j , a linear combination unit composed of K multipliers j (j = 1, 2,..., K) that obtains _j , and a combiner that adds all of their outputs;
A second multiplier j (j = 1, 2,..., K) for multiplying the output of the linear combination unit by each of the spread signals;
Provided with K matched filters j (j = 1, 2,..., K);
If k is the bit number of the desired signal,
From the output of the linear combination unit, after subtracting the spread signal s _k corresponding to the desired signal bits _(t), the multiplication by spreading signal s _{k (t)} by the second multiplier k, at its output inverting amplifier The output obtained by adding a predetermined constant to the output passed through the matched filter k is connected as a weight coefficient x _{k to} the multiplier k of the linear combination unit,
The output of the linear combination unit is multiplied by K−1 spread signals s _j (t) different from the spread signal s _k (t) by a second multiplier j, and the output is inverted by an inverting amplifier. the output through the respective matched filter j as x _j, connected to the multiplier j of the linear combination unit configuration,
Alternatively, in addition to the above configuration, the output s _j (t) x _j of the multiplier j (j = 1, 2,..., K) may be added to the background noise power density and the bit number of the received signal corresponding to the bit number j. , Multiplied by a value proportional to the ratio of the estimated received power, and adding the result to the input side of a second multiplier j (j = 1, 2,..., K).
Multi-user receiver. A spread signal generation unit that generates a plurality (L) of spread signals f _j (t) (j = 1, 2,..., L) from a received signal of the DS-CDMA system;
A matched filter bank for inputting an output obtained by despreading the received signal with the spread signal to a plurality of matched filters, sampling an output of the matched filter, and outputting the sampled output;
K spread signals s _j (t) (j = 1, 2) corresponding to a plurality (K) of values (y ₁ , y ₂ ,..., Y _K ) in the output of the matched filter bank. ,..., K) to calculate _K weighting factors (x ₁ , x ₂ ,..., X _K );
The K weighting factors _{(x 1, x 2, ...} , x K) and the K output _{(y 1, y 2, ...} , y K) a sum of products of each of _{y 1 x 1 + y 2 x} 2 + ... + y _K x _K
In a multi-user receiver that decodes a desired signal bit by determining a discrete level or a sign with respect to the output of the linear combination unit,
The observation section including the section where the reception signal corresponding to the _K values (y ₁ , y ₂ ,..., Y _K ) exists is evenly divided by a predetermined length τ, and , And are called section 1, section 2, ..., section P,
The spread signal generation unit is configured to generate a plurality of modified spread signals that are advanced or delayed by an integer multiple of τ with respect to the spread signal f _i (t) (i = 1, 2,..., L), or Generating a signal obtained by advancing or delaying the entire modified spread signal by a predetermined time;
The weighting factor generator includes:
K matched filters j (j = 1, 2,..., K) whose integration time is τ equal to the section length;
P synthesizers p (p = 1, 2,..., P) with multiple inputs and one output;
P distributors p (p = 1, 2,..., P) with one input and multiple outputs;
A first connection network L connecting the P distributors p and the K matched filters j,
A second connection network R connecting the K matched filters j and the P synthesizers p,
Including a plurality of multipliers,
The plurality of multipliers are named as a first multiplier ip and a second multiplier ip in association with a bit number i and a section p,
One of the P-number of the section and a section p _0,
For a section p where p <p ₀ , a part of the spread signal s _i (t) (j = 1, 2,..., K) belonging to the section p is delayed by a time (p ₀ −p) τ, A part that matches with the modified extended signal is _defined as a spread signal s _ip (t),
For the section p where p> p ₀ , the part belonging to the section p of the spread signal s _i (t) (j = 1, 2,..., K) is advanced by time (p−p ₀ ) τ, A part that matches with the modified extended signal is _defined as a spread signal s _ip (t),
Regarding the section p ₀ , a part of the spread signal s _j (t) (j = 1, 2,..., K) belonging to the section p ₀ is defined as a spread signal s _ip (t),
The spread signal s _ip (t) (j = 1, 2,..., K, p = 1, 2,..., P) and the weighting factor x _j (j = 1, 2,. The output obtained by combining the result of the multiplication by the multiplier jp through the second connection network R by the combiner p is
Output through a distributor p and a first connection network L,
If k is the bit number of the desired signal,
If i = k, the output of the first connection network L _{minus the} modified spread signal s _ip (t) is multiplied by the spread signal s _ip (t) by a second multiplier ip,
If i is different from k, the output of the first connection network L and the spread signal s _ip (t) are multiplied by a second multiplier ip,
The output of the second multiplier is input to a matched filter i via an inverting amplifier. If i = k, a predetermined constant is added to the output of the matched filter i.
If i is different from k, a connection is made to the first multiplier ip as it is as a weighting factor x _i for all or a part of p,
Or, in addition to the above configuration,
The output s _jp (t) x _j of the first multiplier ip is multiplied by a value proportional to the ratio between the background noise power density and the estimated received power per bit of the received signal corresponding to the bit number j, and A configuration in which the addition to the input of the multiplier jp on the first circuit network L side is performed for all or a part of p.
Multi-user receiver. In claim 2 or 3,
The weighting factor generator repeats the above operation a plurality of times in a cycle of the integration time of the matched filter of the weighting factor generator;
Multi-user receiver. In claim 3,
The weighting factor generator includes:
When p ≠ q, if one of the spread signals s _ip (t) and s _iq (t) is zero and the other is not zero, the first multiplier ip and the first multiplier The two multipliers of iq are replaced by one multiplier, and the output is connected to a combiner p and a combiner q via a second network R by a one-to-two pole switch. ,
Two multipliers, the second multiplier ip and the second multiplier iq, are replaced by a single multiplier, the input of which is a one-to-two-pole switch via the first network L. And has a configuration connected to the distributor p and the distributor q,
Alternatively, the first multiplier ip and the first multiplier iq are not replaced with one multiplier,
Alternatively, the second multiplier ip and the second multiplier iq are configured not to be replaced with one multiplier.
Multi-user receiver. In claim 3,
The weighting factor generator uses the spread signal s _ip (t) for the section p where the spread signal s _k (t) is zero and p <p _0, where k is the bit number of the desired signal. The weight coefficient (x ₁ , x ₂ ,..., X _K ) is obtained.
Activating a feedback loop when all or some of the K matched filters become zero or forcibly zero,
Multi-user receiver. In claim 3,
The number of input terminals of the P combiners p and the number of output terminals of the P distributors p, which are constituent elements of the weight coefficient generation unit, and the number of input terminals of the first network L and the second network R The number of output terminals is increased, a plurality of switches are provided between the input and output of the first network L and the second network R, and K matched filters are provided even if the received signal is concentrated in a specific section. And an external control device so as to secure a necessary number of paths of a feedback loop formed by the P synthesizers p and the P distributors p.
Multi-user receiver. In claim 3,
A third circuit network S is provided between an output of a component of the weight coefficient generator, an input of the first multiplier jp and an input of the second multiplier jp, and an output of the spread signal generator. , A plurality of switches are provided between the input and output thereof, and a part of the output of the spread signal generator is selectively or commonly connected to the inputs of the first multiplier jp and the second multiplier jp. , Controlled by an external control device.
Multi-user receiver. A plurality of (L) spread signals are generated from the received signals of the DS-CDMA system. A spread signal generator and an output obtained by despreading the received signals by the spread signals are input to a plurality of matched filters. A matched filter bank for outputting a value obtained by sampling the output of the matched filter;
K spread signals s _j (t) (j = 1, 2) corresponding to a plurality (K) of values (y ₁ , y ₂ ,..., Y _K ) in the output of the matched filter bank. ,..., K), the background noise power is added to a correlation coefficient generation unit that generates a K × K matrix of correlation coefficient matrix R using the correlation coefficient between the spread signals as an element, or a diagonal element of R. A modified correlation coefficient generation unit that generates a modified correlation coefficient matrix R ′ by adding a ratio of a density and an estimated received power per bit of a received signal corresponding to the diagonal element;
A weighting factor generator for calculating _K weighting factors (x ₁ , x ₂ ,..., X _K ) for the _K values (y ₁ , y ₂ ,..., Y _K );
The sum of the products of the _K weighting factors (x ₁ , x ₂ ,..., X _K ) and the K values (y ₁ , y ₂ ,..., Y _K ) is y ₁ x ₁ + y ₂ x _2. + ... + y _K x _K
In a multi-user receiver that decodes a desired signal bit by determining a discrete level or a sign with respect to the output of the linear combination unit,
The weighting factor generator includes:
The j-th column or the j-th row (a _1j , a _2j ,..., A _Kj ) of the correlation coefficient matrix R or the modified correlation coefficient matrix R ′, and weight coefficients (x ₁ , x _{2 ,.} )), And K multipliers that are composed of K multipliers and a combiner that adds all of their outputs to produce a _1j x ₁ + a _2j x ₂ +... + A _Kj x _K And a connecting portion j (j = 1, 2,..., K).
If k is the bit number of the desired signal,
The first constant is subtracted from the output of the linear combination part k, inverted by an inverting amplifier, passed through a low-pass filter, and the output obtained by adding the second constant to the output is given as x _k to the K linear combination parts. connection,
The output of the K-1 pieces of linear coupling portion j is j ≠ k inverted by the inverting amplifier output through a low-pass filter, as x _j, having a structure connected to the K linear combination unit,
Multi-user receiver. In claim 9,
The weighting factor generator includes:
The j-th column or the row (a ′ _1j , a ′ _2j ,..., A ′ _Kj ) of the correlation coefficient matrix R or the modified correlation coefficient matrix R ′, and weighting coefficients (x ₁ , x ₂ ,. inner product of _{x K) (a'1j x 1} + a'2j x 2 + ... + a'Kj x K) produce, the K composed of K multipliers and combiner summing all their outputs A linear combination j (j = 1, 2,..., K)
If k is the bit number of the desired signal,
An output obtained by inverting the output of the linear combination unit k by an inverting amplifier and passing through a low-pass filter is _defined as x _k ′.
A multiplier for multiplying y _k and x _k in the linear combination unit k, connected in place of x _k,
an output obtained by adding a constant to the x k _'as x _k, connected to the K-1 pieces of linear coupling portion j is j ≠ k,
The output of the K-1 pieces of linear coupling portion j is j ≠ k inverted by the inverting amplifier output through a low-pass filter as x _j, having a configuration to be connected to the K linear combination unit,
Multi-user receiver. In claim 9,
The weighting factor generator includes:
From the output of the low-pass filter after the K linear combination units, one of the K weight coefficients (x ₁ , x ₂ ,..., X _K ) distributed to the multipliers of all the linear combination units is obtained. Having a configuration for extracting a part from an input of the low-pass filter,
Multi-user receiver. In any one of claims 2 to 11,
The weighting factor generator includes:
A configuration in which part or all of the output of a low-pass filter or a matched filter connected to the linear combiner is added to respective inputs of the filter, and a positive feedback loop is locally formed by the input and output of the filter. Having,
Multi-user receiver. In any one of claims 2 to 12,
The weighting factor generator includes:
When bit number k of the desired signal, configured to the weighting coefficients x _k, given to all of the linear combiner of the multiplier as a constant, to remove components was required when obtaining the weight coefficients x _k as a variable Having,
Multi-user receiver. In any one of claims 2 to 13,
A multi-user receiver in which all or a part of the position of the inverting amplifier, which is a component of the weight coefficient generation unit, is changed to an arbitrary position of a feedback loop including the inverting amplifier as a component. In any one of claims 2 to 14,
A multi-user receiver in which the position of all or a part of the inverting amplifier, which is a component of the weight coefficient generation unit, is changed to a common path of a plurality of feedback loops including the inverting amplifier as a component. In any one of claims 2 to 15,
The inverting amplifier, which is a component of the weight coefficient generation unit, was changed to an inverter or an inverting attenuator, and any circuit forming a feedback loop including the inverter or the inverting attenuator was changed to a circuit having a gain. Having the structure,
Multi-user receiver. In any one of claims 2 to 16,
The weighting factor generator includes:
No circuit for estimating the value of the background noise power density and the received signal power ratio per bit as zero,
Multi-user receiver. A wireless base station device provided in a mobile communication network system,
A wireless base station apparatus comprising the multi-user receiver according to any one of claims 1 to 17. A mobile device that performs wireless communication by connecting to a mobile communication network,
A mobile station comprising the multi-user receiver according to any one of claims 1 to 17.

Images