JP4724979B2 - Multi-carrier transmission channel characteristic estimation method, multi-carrier transmission channel characteristic estimation apparatus, and multi-carrier demodulation apparatus having the same - Google Patents

Description

【００１５】
ここで、キャリアが、ｋ＝０からＮ−１までの全数を使っていないとき、有効シンボルキャリアを全数含むＭ個のキャリアが、ｎ＝０からＮ−１までのＮ個のサンプル点のうちＭ個のサンプル点で復調可能な場合の条件を求める。これを次のように表す。
【数２】

【００１６】
式（２）において、Σは、Ｍ個の整数から成る集合｛k_p｝について取る。また、集合｛n_q｝は、Ｍ個の整数から成る。ここで1≦p,q≦M、0≦k_p,n_q≦N-1、k_p＜k_p+1、n_q＜n_q+1とする。式（２）は、Ｍ個の複素数ｘ(n_q)から成るベクトルが、ｑ行ｐ列がexp(2πjk_pn_q／N)から成る行列とＭ個の複素数Ｘ(k_p)から成るベクトルとの積であることを示している。すると、Ｍ個の複素数ｘ(n_q)から成るベクトルが得られた場合、Ｍ個の複素数Ｘ(k_p)から成るベクトルを算出できるかどうかは、ｑ行ｐ列がexp(2πjk_pn_q／N)から成る行列が逆行列を有するかどうかと等価である。
【００１７】
ｑ行ｐ列がexp(2πjk_pn_q／N)である行列について、集合｛k_p｝、｛n_q｝がどちらも０以上Ｎ−１以下の整数Ｎ個の集合と一致する場合は式（１）に帰着する。逆行列は、Ｗ_N＝exp(-2πj／N)として、ｋ＋１行ｎ＋１列がＷ_N ^knのＮ行Ｎ列の行列である（ただし０≦ｋ≦Ｎ−１、０≦ｎ≦Ｎ−１、ＩＤＦＴ（Ｎ点逆離散フーリエ変換）とＤＦＴ（Ｎ点離散フーリエ変換）の関係）。一方、ｑ行ｐ列がexp(2πjk_pn_q／N)である行列が逆行列を有するよう、０以上Ｎ−１以下の、互いに異なるＭ個の整数から成る集合｛k_p｝、｛n_q｝を選ぶ方法は極めて広範囲である。しかし、次のようにして、当該集合｛k_p｝、｛n_q｝が、ｑ行ｐ列がexp(2πjk_pn_q／N)である行列が逆行列を有しないものを判断すること、即ち、逆行列を有するよう設定することが可能である。
【００１８】
〔逆行列の有無にいて〕
ｑ行ｐ列がexp(2πjk_pn_q／N)（各々Ｍ個の整数から成る集合｛k_p｝、｛n_q｝は、1≦p,q≦M、0≦k_p,n_q≦N-1、k_p＜k_p+1、n_q＜n_q+1）のＭ行Ｍ列の行列が逆行列を有するか（行列式が０でないか）どうかについては、次の変換をした行列の行列式が０かどうかと等価である。即ち、第ｑ'行と第ｐ'列に着目し、全要素をexp(2πjk_p'n_q'／N)で除したのち、第ｐ列（ｐ≠ｐ'）をexp(2πj(k_p-k_p')n_q'／N)で除し第ｑ行（ｑ≠ｑ'）をexp(2πjk_p'(n_q-n_q')／N)で除すことである。これにより、第ｑ'行と第ｐ'列の要素をすべて１にすることができる。ここで、行について、Ｎの約数ｂ（Ｎ＝ａｂ）をとり、ｂによる剰余系で元の数が最も多くなる｛n_q｝の剰余系（その元の数をＢとする）を選ぶ。列については第１列の成分を全て１に、前記ｂにより選んだ｛n_q｝の剰余系に対応する行のうちの１行の成分を全て１にする。
【００１９】
これについて実例を上げる。Ｎ＝１２、ｑ行ｐ列がexp(πjk_pn_q／６)、｛k_p｝＝｛０，１，４，５，８｝、｛n_q｝＝｛０，１，２，３，６｝とする。ｂ＝３（即ち、ａ＝４）による｛n_q｝の剰余系は｛０，３，６｝と｛１｝と｛２｝で、元の最多となるものは｛０，３，６｝。当該行列は、次のものである。
【数３】

【００２０】
数３の行列は、第１列の成分が全て１となっており、｛n_q｝＝｛０，１，２，３，６｝のうち、ｂ＝３による剰余系で元の数が最も多くなる行｛n_q｝＝｛０，３，６｝のうち、n_q＝０となる第１行の成分が全て１となっている。
【００２１】
行列式の０か否かについては、行の交換、列の交換に影響されない。今、ｂによる剰余系で元の数が最も多くなる行を第１行乃至第Ｂ行に置き替える。さらに、第１列（ｐ＝１）がすべて１、第１行（ｑ＝１）がすべて１となるというにする。数３の行列においては、元の最多となる｛０，３，６｝に対応する行を、第１行乃至第３行（即ちＢ＝３）に置き替ればよく、次のとおりとなる。
【数４】

【００２２】
ここまでの作業は、問題となっている行列について、Ｎの約数ｂによる整数の集合｛n_q｝の剰余系を勘案したのち、元の数が最多（その数をＢ）となる剰余系に対応する行を第１行乃至第Ｂ行に置き替え、全要素をexp(2πjk₁n₁／N)で除したのち、第ｐ列（ｐ≠１）をexp(2πj(k_p-k₁)n₁／N)で除し第ｑ行（ｑ≠１）をexp(2πjk₁(n_q-n₁)／N)で除すことで得られる。これらはいわゆる行列の基本変形であり、このような変形の前後の行列について、行列式が０か否かは入れ代わることが無い。
【００２３】
すると、列｛k_p｝について、ａによる剰余系の種類の数をＡとおくと、第１乃至第Ｂ行は、Ａ通りの成分配置しかない。実際、数３の行列において、｛k_p｝＝｛０，１，４，５，８｝のａ＝４についての剰余は、２種類（即ちＡ＝２）で｛０，４，８｝と｛１，５｝である。これを変形した（行列式が０かどうかには関係しない）数４において、第１乃至第３行の成分のパターンは、第１、３、５列と第２、４列の２種類しかない。
【００２４】
このように考えると、次のことが明らかである。
〔１〕Ａ＜Ｂならば、変形後の行列式に対応する行列を下三角行列化した際、対角成分はＡ＋１行Ａ＋１列乃至Ｂ行Ｂ列まで０が並び、そもそも元の行列式が０である。例えばＡ＝２、Ｂ＝３である上記数４の行列を下三角化すると、第３行第３列の成分が０となる。
〔２〕Ａ≧Ｂが、Ｎの１とＮ以外の約数ａ、ｂの全ての対について言える（ただし、行と列を入れ換え、行｛n_q｝の剰余系の種類の数と列｛k_p｝の剰余系で元の数が最多のものも検討した上で）なら、変形後の行列式に対応する行列は必ず対角化可能。即ち、元の行列式は０とならない。
【００２５】
よって、Ｎが素数なら常に逆行列を有する。Ｎが素数でないなら、Ｍ個の整数から成る集合｛k_p｝、｛n_q｝を、Ｎ＝ａｂとなる、Ｎの１とＮ以外の約数ａ、ｂの全ての対に関する剰余系につて検討し、｛k_p｝のａによる剰余系の種類の数Ａが、｛n_q｝のによる剰余系の元の数の最大値Ｂより小さいようなＮの約数の対ａ，ｂがあれば、当該行列式は０。そのようなＮの約数の対ａ，ｂが全くなければ（ただし、行と列を入れ換え、行｛n_q｝の剰余系の種類の数と列｛k_p｝の剰余系で元の数が最多のものも検討した上で）、当該行列式は０でない。
【００２６】
また、ｑ行ｐ列がexp(2πjk_pn_q／N)（ただし、各々Ｍ個の整数から成る集合｛k_p｝、｛n_q｝は、1≦p,q≦M、0≦k_p,n_q≦N-1、k_p＜k_p+1,n_q＜n_q+1とする）のＭ行Ｍ列の行列が逆行列を有するか（行列式が０でないか）どうかは、ｑ'行ｐ'列がexp(2πjk_p'n_q'／N)（ただし1≦p',q'≦N-M、0≦k_p',n_q'≦N-1、k_p'＜k_p'+1,n_q'＜n_q'+1、k_p'≠k_p、n_q'≠n_q）のＮ−Ｍ行Ｎ−Ｍ列の行列が逆行列を有するか（行列式が０でないか）どうかと等価であることも明らかである。当該２つの行列は、Ｎ点ＩＤＦＴの行列（逆行列はＤＦＴの行列）から、互いに異なるＭ行Ｍ列の行列とＮ−Ｍ行Ｎ−Ｍ列の行列を取り出したものの関係にあるからである。
【００２７】
また、Ｍ個の連続した整数から成る集合｛k_p｝に対し、どんなＭ個の互いに異なる整数から成る集合｛n_q｝をとってもｑ行ｐ列がexp(2πjk_pn_q／N)のＭ行Ｍ列の行列は逆行列を有する。逆に、Ｍ個の連続した整数から成る集合｛n_q｝に対し、どんなＭ個の互いに異なる整数から成る集合｛k_p｝をとってもｑ行ｐ列がexp(2πjk_pn_q／N)のＭ行Ｍ列の行列は逆行列を有する。これらは互いに等価である。このうち後者の証明の概略は次のようである。即ち、Ｍ個の連続した整数から成る集合｛n_q｝の上記ｂ（Ｎ＝ａｂ）による剰余系の元の数の最大値Ｂは、−[−Ｍ／ｂ]である。ただし[Ｒ]はガウスの記号で、実数Ｒを越えない最大の整数を表す。Ｍ個の互いに異なる整数から成る集合｛k_p｝のａによる剰余系の種類Ａの最小値を求めると、ａによる剰余系には最大ｂ＝Ｎ／ａ個までしか元が無いのであるから、−[−Ｍ／ｂ]となる。即ち、Ｍ個の連続した整数から成る集合｛n_q｝の上記ｂによる剰余系の元の数の最大値Ｂは、Ｍ個の互いに異なる整数から成る集合｛k_p｝のａによる剰余系の種類Ａの最小値に等しい。このとき、ＢはＡを越えることが無い。
【００２８】
さらに、Ｎ＝ｃ^m（ｃは素数、ｍは２以上の整数）であれば、Ｍ個の連続した整数から成る集合｛k_p｝と、ｃ^m'（ｍ'＝[log_cＭ]、[Ｒ]はガウスの記号で、実数Ｒを越えない最大の整数）による剰余系（剰余の種類と各々の元の個数）が完全一致するＭ個の整数から成る集合｛k_p｝に対し、どんなＭ個の互いに異なる整数から成る集合｛n_q｝をとってもｑ行ｐ列がexp(2πjk_pn_q／N)のＭ行Ｍ列の行列は逆行列を有する。これは｛n_q｝と｛k_p｝を逆にしても成立する。ｃ^m'とは（ｍ'＝[log_cＭ]）、Ｍに等しいか、又はＭを越えない最大の、Ｎの唯一の素因数の階乗である。｛k_p｝と、ｃ^m'による剰余系が完全一致するとは、Ｍ個の連続した整数の任意の各k_pに対し、k_p±ｃ^m'、k_p±２ｃ^m'、k_p±３ｃ^m'、…、k_p±(ｃ^m−ｃ)（ただし、０以上Ｎ−１以下の範囲外にあたるものは一切除外するものとする）のうちのひとつに置き替えた（置き換えによる重複は許されない）、０以上Ｎ−１以下のＭ個の互いに異なる整数の集合を意味する。
【００２９】
〔逆行列についてまとめ〕
１．Ｎが素数であれば当該行列は逆行列を有する。
２．｛n_q｝と｛k_p｝の少なくとも一方が、連続したＭ個の整数から成る集合であれば、ｑ行ｐ列がexp(2πjk_pn_q／N)のＭ行Ｍ列の行列は逆行列を有する。
３．Ｎ＝ｃ^m（ｃは素数、ｍは２以上の整数）であれば、上述の通り、連続したＭ個の整数の集合でなくても、｛n_q｝と｛k_p｝の少なくとも一方が、連続したＭ個の整数の集合とｃ^m'（ｍ'＝[log_cＭ]）による剰余系について剰余の種類と各々の元の個数が一致する集合であれば当該行列は逆行列を有する。
４．その他の｛n_q｝と｛k_p｝であっても、Ｎの１とＮ以外の約数ａ、ｂの全ての対に関する剰余系につて検討し、｛k_p｝のａによる剰余系の種類の数Ａが、｛n_q｝のｂによる剰余系の元の数の最大値Ｂより小さいようなＮの約数の対ａ，ｂが全くなく、且つ、｛n_q｝のａによる剰余系の種類の数Ａが、｛k_p｝のｂによる剰余系の元の数の最大値Ｂより小さいようなＮの約数の対ａ，ｂが全くないならば、当該行列は逆行列を有する。
【００３０】
ところで上記逆行列はＭ行Ｍ列であるが、そこから所望のＬ行を取り出したＬ行Ｍ列の行列（Ｌ≦Ｍ）を用いれば、Ｍ本のサブキャリアから任意のＬ本のキャリアをＭ個のサンプル点から求めるための行列となることは自明である。即ち、本願発明うち復調方法及び復調装置の実施に使われる行列は正方行列に限定されない。
【００３１】
また、ここまでの議論が、Ｍ本のサブキャリアにヌルキャリア含まれていたとてしても成立することは明らかである。すると、Ｍ本のサブキャリアの中に更にＭ−Ｌ本（Ｌ＜Ｍ）のヌルキャリアがあったとして、残りＬ本のキャリアをＭ個のサンプル点から求めるための行列は、上記議論におけるｑ行ｐ列がexp(2πjk_pn_q／N)のＭ行Ｍ列の行列の逆行列であるＭ行Ｍ列の行列から当該Ｌ本のキャリアを復調ために必要なＬ行を取り出したもので良いことは明らかである。即ち、Ｌ本のキャリアに対応するk_pの他に、Ｍ−Ｌ個の任意のk_p（Ｌ本のキャリアに対応するk_pを除く）を仮に用いたＭ行Ｍ列の行列を構成してその逆行列を求め、そこからＬ本のキャリアに対応するＬ行Ｍ列の行列（Ｌ＜Ｍ）を取りだせば良い。簡単な考察から、このようなＬ行Ｍ列の行列は無数に構成することができる。
【００３２】
〔他のマルチキャリアについて〕
上記はＯＦＤＭを代表例に、等周波数間隔のＮ本のサブキャリアから成る通信方法について説明したが、上記説明から明らかなように、本発明の本質は等周波数間隔のＮ本のサブキャリアから成る通信方法に限定されない。即ち、周波数間隔比が整数比であれば、上記逆行列の有無を検討した内容をそのまま考慮することで、Ｍ点のサンプリング点の設計が可能である。この際、Ｍ点のサンプリング点の設計において、「Ｎ本のサブキャリアを復調できるＮ点」が、そもそも等間隔でない場合であっても、上記逆行列の有無を検討した内容をそのまま考慮することで、Ｍ点のサンプリング点の設計が可能である。このことは、周波数間隔比が整数比であるＮ本のサブキャリアから成る通信方法は、等周波数間隔のＮ’本のサブキャリアから成る通信方法であってＮ’−Ｎ本がヌルキャリアであるものと同等と考えることができるからである。
【００３３】
〔実施例〕
図１は、本発明の具体的な一実施例であるマルチキャリア伝送用伝搬路特性推定装置１２０を有するマルチキャリア復調装置１０００の構成を示したブロック図である。本実施例においては、ガードインターバルを有する、Ｎ本のサブキャリアから成るＯＦＤＭ変調信号を受信してアナログ直交復調するマルチキャリア復調装置１０００を示した。
【００３４】
マルチキャリア復調装置１０００は、図示しない外部からの受信波を受け、発振器１０２及び位相変換器１０２Ｉ、乗算器１０３Ｉ及び１０３Ｑ、ＬＰＦ１０４Ｉ、１０４Ｑ、Ａ／Ｄ１０５Ｉ、１０５Ｑ、ガードインターバル除去器１０６Ｉ及び１０６Ｑにより、Ｎ個の複素数に対応するＮ点のサンプリングを行う（図１でＲＣＶと示した部分）。次に、二重選択器１２１、線形演算器１２２、周波数特性演算器１２３から構成されるマルチキャリア伝送用伝搬路特性推定装置１２０においてパイロット信号が復調され、周波数特性が出力される。前記Ｎ個の複素数は、Ｓ／Ｐ１０７Ｉ、１０７Ｑを通してＮポイントＤＦＴ１０８で復調され、パイロットシンボルの周波数特性により等化回路１０９で補正が行われる。
【００３５】
ここで二重選択器１２１及び線形演算器１２２がパイロットシンボルを復調する手段又はパイロットシンボル演算器にあたる。尚、受信波から同相成分Ｉ_Rと直交成分Ｑ_Rを取り出す部分（図１でＲＣＶとした部分）がＮ点のディジタル信号を得るサンプリング装置、又は、直交復調及びサンプリング装置にあたる。
【００３６】
図１のマルチキャリア復調装置１０００の作用は次のとおりである。後述するマルチキャリア伝送用伝搬路特性推定装置１２０の出力するパイロット信号に基づき、同期回路１０１が同期信号を発振器１０２に出力する。受信波は、乗算器１０３Ｉ、１０３Ｑにおいて発振器１０２の出力する正弦波及び位相変換器１０２Ｉを通したπ／２位相のずれた正弦波と乗ぜられ、低域濾波器（ＬＰＦ）１０４Ｉ、１０４Ｑを通ってベース信号Ｉ_A、Ｑ_Aとなる。これをアナログ／ディジタル変換器（Ａ／Ｄ）１０５Ｉ、１０５Ｑによりディジタル信号の同相成分Ｉ_Dとディジタル信号の直交成分Ｑ_Dとなる。これをガードインターバル除去器１０６Ｉ及び１０６Ｑでガードインターバルを除去して、各々１シンボル中にＮ個のディジタル信号を有する同相成分Ｉ_Rと直交成分Ｑ_Rとなる。サンプリングレートは、Ｎ本のサブキャリアの周波数間隔をΔｆとしたとき、１／（ＮΔｆ）である。
【００３７】
直並列変換器（Ｓ／Ｐ）１０７Ｉと１０７Ｑとにより、Ｎ個のディジタル信号Ｉ_R、Ｑ_RがＮ個の複素信号としてＮ点離散フーリエ変換器（ＮポイントＤＦＴ）１０８に入力され、Ｎ個の複素信号が復調され、等化回路１０９、デマッピング回路１１０、並直列変換器（Ｐ／Ｓ）１１１を通して所望の信号列が再生される。ここで、等化回路１０９においては、以下のマルチキャリア伝送用伝搬路特性推定装置１２０の出力により、ＮポイントＤＦＴ１０８の出力するＮ個の複素信号について、伝搬路の周波数特性に応じて補正が行われる。
【００３８】
マルチキャリア伝送用伝搬路特性推定装置１２０は、二重選択器１２１、線形演算器１２２、周波数特性演算器１２３から構成される。これらの作用は次のとおりである。二重選択器１２１は、ガードインターバル除去器１０６Ｉ及び１０６Ｑが出力するＮ個のディジタル信号を有する同相成分Ｉ_Rと直交成分Ｑ_Rとから、各々必要なＭ個の信号を選択し、直並列変換したのち線形演算器１２２に出力する。この際の必要なＭ個の信号は、予め設定されたサンプリング番号に対するものであるが、例えば１シンボル長中の最初から連続したＭ個とすれば良い。線形演算器１２２は、二重選択器１２１が出力する２Ｍ個の信号を、Ｍ個の複素信号として、予め設定された係数によりパイロット信号を復調する。この際、Ｍ個の複素信号から、１乃至複数個のパイロット信号が復調される。この演算は、ソフトウエア的にＭ個の複素信号から線形演算するものである。線形演算、即ち線形結合における係数は、本実施例の記載の前において詳述した逆行列の一部である。なお、パイロット信号のキャリアについてのみ解ければ良いので、当該逆行列が存在しない場合であってもパイロット信号を復調するための線形係数が求められることが有り得る。尚、若干の構成の修正の基に、ハードウエア的にＦＩＲにより構成することも可能である。線形演算器１２２の出力は周波数特性演算器１２３と、同期回路１０１に出力される。同期回路１０１では当該パイロット信号により同期信号を生成することは上述の通りである。
【００３９】
線形演算器１２２が復調した１乃至複数個のパイロット信号を、周波数特性演算器１２３にかけて、伝搬路特性を推定する。その結果は、上述した等化回路１０９に出力され、ＮポイントＤＦＴ１０８の出力するＮ個の複素信号について、伝搬路の周波数特性に応じて補正が行われる。ここにおいて、線形演算器１２２が１乃至複数個のパイロット信号を復調する際、ＮポイントＤＦＴを用いず、且つ１乃至複数個のパイロット信号を復調するのに最低限必要なＬ個以上のＭ個の複素信号のみを用いてそれらの線形演算により当該１乃至複数個のパイロット信号のみを復調しているので、パイロット信号復調をより短時間で行うことができる。
【００４０】
これは図２から理解することができる。即ち、本来の１シンボルの時間軸上の長さＴ（サンプル数Ｎ）よりも、短い時間Ｔ'（サンプル数Ｍ、Ｍ＜Ｎ）でパイロットシンボルについて復調可能である。これは、パイロットシンボルを含むシンボル区間で、パイロットシンボルを含む有効キャリアがＭ本以下のＬ本であるから可能となっている。
【００４１】
〔変形例〕
上述の実施例において、マルチキャリア伝送用伝搬路特性推定装置１２０の各構成要素の働きを次のように変えることで、伝搬路特性の時間変化をより的確に検出することができる。ただし、１乃至複数個のパイロット信号を復調するために必要な複素信号の個数Ｍは、Ｎ／２以下であるとする。
【００４２】
本変形例におけるマルチキャリア伝送用伝搬路特性推定装置１２０を構成する二重選択器１２１と線形演算器１２２の作用は、ガードインターバル除去器１０６Ｉ及び１０６ＱがＮ個のディジタル信号を出力する毎に２回行われる。即ち、ガードインターバル除去器１１Ｉ及び１１Ｑが各々前半のＮ／２個のディジタル信号を出力すると、二重選択器１２１と線形演算器１２２はそこからＭ個の複素信号を選択して１乃至複数個のパイロット信号を復調する。次に、ガードインターバル除去器１０６Ｉ及び１０６Ｑが各々後半のＮ／２個のディジタル信号を出力すると、二重選択器１２１と線形演算器１２２はそこからＭ個の複素信号を選択して１乃至複数個のパイロット信号を復調する。こうして、ガードインターバル除去器１０６Ｉ及び１０６Ｑの出力するディジタル信号の前半と後半とで１乃至複数個のパイロット信号が２度復調される。周波数特性演算器１２３は、それら２度復調された１乃至複数個のパイロット信号の変化をも検出して、等化回路１０９に伝搬路特性及びその時間変化を出力する。このようにして本変形例では、伝搬路特性及びその時間変化を１シンボル長中で検出することができるので、精度の良い等化を行うことができる。
【００４３】
これは図３から理解することができる。即ち、本来の１シンボルの時間軸上の長さＴ（サンプル数Ｎ）の１／２よりも短い時間Ｔ₁'、Ｔ₂'（どちらもサンプル数Ｍ、Ｍ＜Ｎ／２）でパイロットシンボルについて復調可能である。これは、パイロットシンボルを含むシンボル区間で、パイロットシンボルを含む有効キャリアがＭ本以下のＬ本であるから可能となっている。
【００４４】
この変形例においては、１シンボルの時間軸上の長さＴ（サンプル数Ｎ）を前半と後半に分割し、各々サンプル数Ｍ（Ｍ＜Ｎ／２）のサンプル点を取る方法とし、２組のＭ個のサンプル点に重なりが無いようにしているが、２組のＭ個のサンプル点に重なりはあっても構わない。１シンボルのサンプル数Ｎから、連続するＭ点のサンプル点の取り出し方はＮ−Ｍ＋１通り有り、そこから２組以上のＭ個のサンプル点からパイロットシンボルを復調することで、伝搬路特性及びその時間変化を検出することが可能である。尚、Ｍ個のサンプル点の間隔（第１サンプル点から、各サンプル点までの間隔のパターン）が一致すれば、２組以上のサンプル点からパイロットシンボルを復調する際、Ｍ点は連続していなくてもよい。
【００４５】
上記実施例では、ガードインターバルを有するＯＦＤＭによる信号を受信する例を挙げたが、ガードインターバルは本発明の実施に必須ではなく、また、通信方法もＯＦＤＭに限定されない。また、上記実施例ではアナログ直交復調する例を挙げたが、ディジタル直交復調するものでも良く、また、直交復調を行わないものでも良い。
【００４６】
本実施例では、通常の復調のためのＮポイントＤＦＴとＡ／Ｄ変換及び直交復調を合わせて、サンプリングレート１／（ＮΔｆ）にてサンプリングする構成としたが、Ｎ本のサブキャリアの復調を他の方法（例えば逆行列演算）により行うなどの場合、そのサンプリングレートは１／（ＮΔｆ）より小さい（より速いサンプリング）で行うことも可能である。また、ＮポイントＤＦＴとＡ／Ｄ変換及び直交復調を別に有する場合、パイロット信号を復調できるのであれば、サンプリングレートは任意である。
【図面の簡単な説明】
【図１】 本発明の具体的な実施例に係るマルチキャリア復調装置の構成を示すブロック図。
【図２】 実施例のパイロットシンボルを復調する時間を示した説明図。
【図３】 変形例のパイロットシンボルを復調する時間を示した説明図。
【符号の説明】
１０００ マルチキャリア復調装置
１２０ 伝搬路推定装置
１２１ 二重選択器
１２２ 線形演算器
１２３ 周波数特性演算器
１０１ 同期回路
１０６Ｉ、１０６Ｑ ガードインターバル除去器
１０８ ＮポイントＤＦＴ
１０９ 等化回路２[0001]
BACKGROUND OF THE INVENTION
  The present invention relates to a method and apparatus for estimating propagation path characteristics of a multicarrier transmission scheme that modulates a plurality of subcarriers and transmits a plurality of modulated signals with one symbol length. The present invention is particularly effective for a demodulator that receives a signal modulated by, for example, an Orthogonal Frequency Division Multiplex (OFDM) system.
[0002]
[Prior art]
  For example, in the OFDM scheme, in order to estimate the propagation path characteristics, it is common to form pilot symbols using a certain number of subcarriers (arbitrary number), demodulate the reception side, and estimate the propagation path characteristics. is there. That is, the symbol length including the pilot symbol is demodulated in the same manner as the symbol length not including the pilot symbol. At this time, for example, in the OFDM system, modulation is performed using N-point IDFT (N-point inverse discrete Fourier transform) on the modulation side, and therefore demodulation is performed using N-point DFT (N-point discrete Fourier transform) on the demodulation side. It was. After that, the propagation path characteristic is estimated from the difference from the original value of the demodulated pilot symbol.
[0003]
[Problems to be solved by the invention]
  By the way, in general, a lot of null carriers that are always 0 are included in N-point IDFT (N-point inverse discrete Fourier transform) inputs (signals to be placed on subcarriers) on the modulation side. In addition, in the symbol length including the pilot symbol, only the subcarrier on which the pilot symbol is carried may be modulated, and the other subcarrier may be a null carrier.
[0004]
  The present invention pays attention to the above points, and it is not necessary to use an N-point DFT for pilot symbol demodulation on the demodulation side, so that pilot symbol demodulation is possible with a symbol length shorter than one symbol length.
[0005]
[Means for Solving the Problems]
  The invention according to claim 1 is a multicarrier transmission channel for receiving a signal having a pilot symbol using L subcarriers for a specific symbol length and estimating a channel characteristic of the signal in the multicarrier transmission method. In the characteristic estimation method, an N-point digital signal capable of demodulating all N (N> L) subcarriers is obtained,Based on the number of M subcarriers other than NM null carriers and the number of sample points taken out of M subcarriers other than NM null carriers from M sample points A matrix for demodulating desired L subcarriers (L ≦ M) is prepared, and N−M null carriers are obtained from the product of the prepared matrix and a vector composed of M sample points. Demodulate pilot symbols by demodulating desired L subcarriers among M subcarriers other thanThe propagation path characteristic of the received signal is estimated from the demodulated pilot symbol. Here, “a signal having a pilot symbol using L subcarriers for a specific symbol length” means that a signal having a pilot symbol isNM booksOf subcarriers, of which pilot symbolsL booksThis means that subcarriers are used. The “modulation of N subcarriers” may be complex modulation or modulation of only amplitude or phase. Of course, when N subcarriers are complex-modulated, the N-point digital signal should be a complex signal. In any case, it is premised that the synchronization is achieved by some means, but the present invention is not affected by the means for achieving the synchronization.
[0006]
  The invention according to claim 2Multi-carrier transmission methodModulate using inverse discrete Fourier transformThe digital signal is NSampling interval 1 / (NΔf) where Δf is the frequency interval between adjacent subcarriersIt is an N-point complex digital signal that has been orthogonally demodulated by.Here, the quadrature demodulation and sampling are included in the present invention both in the case of analog / digital conversion after analog quadrature demodulation and in the case of digital quadrature after analog / digital conversion.
[0007]
  The invention according to claim 3 is characterized in that the preset M points are M consecutive sampling points. Furthermore, the invention according to claim 4 estimates the propagation path characteristics by demodulating the pilot symbols h times (where h is a natural number 2 ≦ h ≦ N−M + 1) within a specific symbol length having the pilot symbols. Features.
[0008]
  The invention according to claims 5 to 8 is a multi-carrier transmission propagation path characteristic estimation apparatus adopting the multi-carrier transmission propagation path characteristic estimation method according to claims 1 to 4. The same applies to the explanation of terms in the claims. Note that the pilot symbol calculator includes a software based on a so-called matrix calculation or a hardware based on an FIR circuit or the like.
[0009]
  The invention according to claim 9 is the invention described in claims 5 to 8.Any one ofA multi-carrier demodulating apparatus comprising the multi-carrier transmission channel characteristic estimating apparatus described in 1.
[0010]
[Operation and effect of the invention]
  The propagation path characteristic estimation method or propagation path characteristic estimation apparatus according to the present invention uses a symbol that can be demodulated at M points (M <N) of symbols that should be demodulated at N points. Measures to make this possible will be described later. In other words, pilot symbols can be demodulated with a smaller number of sampling points without using a normal N-point sampling demodulation method or demodulator. Therefore, the amount of calculation can be minimized, and the propagation path propagation characteristic can be estimated from the pilot symbol demodulation result and reflected more quickly in the entire demodulation method or the entire demodulator (claims 1, 5, 9). ).
[0011]
  In particular, it is effective as a demodulation method or a propagation path characteristic estimation method in a demodulation device that receives, for example, an OFDM signal modulated using inverse discrete Fourier transform. This is because, for example, an OFDM-based modulated signal modulated using inverse discrete Fourier transform has a uniform design method. That is, it is possible to easily design the required M sampling numbers from the carrier numbers of the subcarriers having pilot symbols and the carrier numbers of other subcarriers that are not null carriers (claims 2, 6, 9). . The design of M number sampling numbers at this time will also be described later.
[0012]
  Furthermore, by using M consecutive sampling points, the pilot symbol can be demodulated regardless of the setting of the carrier number of the subcarrier having the pilot symbol and the carrier number of the subcarrier other than the null carrier. 3, 7, 9). The reason why M consecutive sampling points are sufficient will be described later. Considering how to select M consecutive, pilot symbols can be demodulated at most NM + 1 times within a specific symbol length having pilot symbols. By comparing two or more sets of pilot symbols demodulated in this way, it is possible to calculate the time variation of the transmission path propagation characteristics and more accurately reflect it in the entire demodulation method or the entire demodulation device ( Claims 4, 8, and 9).
[0013]
DETAILED DESCRIPTION OF THE INVENTION
  Hereinafter, the carrier number and sampling point number that enable the present invention will be described using OFDM as an example. As will be described later, the present invention is not limited to OFDM. The following explanation is based on the fact that the rank (rank) is M (there is an inverse matrix) for an M-by-M matrix extracted from an N-by-N matrix showing N-point inverse discrete Fourier transform. The purpose is to confirm without direct calculation such as keratinization.
[0014]
[About carrier number and sampling point number enabling the present invention]
  First, consider an OFDM carrier, that is, a waveform that satisfies the following equation (1) at the nth point (an integer from n = 0 to N−1).
[Expression 1]

[0015]
  Here, when the carrier does not use all the numbers from k = 0 to N−1, the M carriers including all the effective symbol carriers are among the N sample points from n = 0 to N−1. A condition when demodulation is possible with M sample points is obtained. This is expressed as follows.
[Expression 2]

[0016]
  In Equation (2), Σ is a set of M integers {k_p} Take about. Also, the set {n_q} Consists of M integers. Where 1≤p, q≤M, 0≤k_p, n_q≤ N-1, k_p<K_{p + 1}, N_q<N_{q + 1}And Equation (2) is expressed by M complex numbers x (n_q) And q rows and p columns are exp (2πjk_pn_q/ N) and M complex numbers X (k_p) Is a product with a vector consisting of Then, M complex numbers x (n_q), M complex numbers X (k_p) Can calculate a vector of q rows and p columns exp (2πjk_pn_q/ N) is equivalent to whether the matrix has an inverse matrix.
[0017]
  q rows and p columns are exp (2πjk_pn_q/ N) for a matrix {k_p}, {N_q} Both coincide with an integer set of N that is greater than or equal to 0 and less than or equal to N−1, which results in equation (1). The inverse matrix is W_N= Exp (-2πj / N) where k + 1 row n + 1 column is W_N ^kn(Where 0 ≦ k ≦ N−1, 0 ≦ n ≦ N−1, relationship between IDFT (N-point inverse discrete Fourier transform) and DFT (N-point discrete Fourier transform)). On the other hand, q rows and p columns are exp (2πjk_pn_q/ N) is a set of M integers different from each other but not smaller than 0 and not larger than N−1 so that the matrix having the inverse matrix {k_p}, {N_q} Is very broad. However, the set {k_p}, {N_q} Is q rows and p columns is exp (2πjk_pn_qIt is possible to determine that the matrix that is / N) does not have an inverse matrix, that is, it can be set to have an inverse matrix.
[0018]
[With or without inverse matrix]
  q rows and p columns are exp (2πjk_pn_q/ N) (set of M integers each {k_p}, {N_q} Is 1 ≦ p, q ≦ M, 0 ≦ k_p, n_q≤ N-1, k_p<K_{p + 1}, N_q<N_{q + 1}Whether the matrix of M rows and M columns in () has an inverse matrix (whether the determinant is not 0) is equivalent to whether or not the determinant of the matrix after the next transformation is 0. That is, paying attention to the q′th row and the p′th column, exp (2πjk_{p '}n_{q '}/ N) and then pth column (p ≠ p ′) is exp (2πj (k_p-k_{p '}) n_{q '}/ N) divided by q line (q ≠ q ') exp (2πjk_{p '}(n_q-n_{q '}) / N). Thereby, all the elements in the q′-th row and the p′-th column can be set to 1. Here, for a row, a divisor b of N (N = ab) is taken, and the original number is the largest in the remainder system by b {n_q} 'S remainder system (whose original number is B). For the column, all the components in the first column are set to 1 and selected by b above {n_q}, All the components in one row among the rows corresponding to the remainder system are set to 1.
[0019]
  An example is given about this. N = 12, q rows and p columns are exp (πjk_pn_q/ 6), {k_p} = {0, 1, 4, 5, 8}, {n_q} = {0, 1, 2, 3, 6}. {n by b = 3 (ie a = 4)_q} Is {0, 3, 6}, {1}, and {2}, and the original largest is {0, 3, 6}. The matrix is:
[Equation 3]

[0020]
  In the matrix of Equation 3, all the components in the first column are 1, and {n_q} = {0, 1, 2, 3, 6}, the row {n with the largest number of elements in the remainder system with b = 3_q} = {0, 3, 6}, n_qThe components in the first row where = 0 are all 1.
[0021]
  Whether the determinant is 0 or not is not affected by the exchange of rows and exchange of columns. Now, the row with the largest original number in the remainder system by b is replaced with the first row to the Bth row. Further, it is assumed that the first column (p = 1) is all 1 and the first row (q = 1) is all 1. In the matrix of Equation 3, the row corresponding to {0,3,6} that is the original maximum may be replaced with the first row to the third row (that is, B = 3), and is as follows. .
[Expression 4]

[0022]
  Up to this point, the set of integers by the divisor b of N for the matrix in question {n_q}, After replacing the lines corresponding to the residue system with the largest number (the number of which is B) with the first to B rows, all the elements exp (2πjk₁n₁/ N), then pth column (p ≠ 1) is exp (2πj (k_p-k₁) n₁/ N) divided by q line (q ≠ 1) exp (2πjk₁(n_q-n₁) / N). These are so-called basic deformations of a matrix, and the matrix before and after such deformation does not change whether the determinant is 0 or not.
[0023]
  Then the sequence {k_p}, Where A is the number of types of residue systems by a, the first to Bth rows have only A component arrangements. In fact, in the matrix of Equation 3, {k_p} = {0,1,4,5,8}, the remainder for a = 4 is {0,4,8} and {1,5} in two types (ie, A = 2). In Formula 4 (which is not related to whether the determinant is 0), which is a modification of this, there are only two types of patterns of the first to third row components: the first, third and fifth columns and the second and fourth columns. .
[0024]
  Given this, it is clear that:
[1] If A <B, when the matrix corresponding to the modified determinant is converted into a lower triangular matrix, the diagonal components are arranged with 0 from A + 1 row A + 1 column to B row B column, and the original determinant is originally 0. For example, if the matrix of the above formula 4 with A = 2 and B = 3 is triangulated, the component in the third row and the third column becomes zero.
[2] A ≧ B can be said for all pairs of N and divisors a and b other than N (however, the row and the column are interchanged and the row {n_q} Number of residue types and sequence {k_p}, The matrix corresponding to the determinant after transformation can always be diagonalized. That is, the original determinant is not zero.
[0025]
  Thus, if N is a prime number, it always has an inverse matrix. If N is not a prime number, a set of M integers {k_p}, {N_q} For the remainder system for all pairs of divisors a and b other than 1 and N, where N = ab, and {k_p}, The number A of the residue system type by a is {n_q}, If there is a divisor pair a, b of N that is smaller than the maximum value B of the original number of the residue system, the determinant is 0. If there is no such N divisor pair a, b (but replace row and column, row {n_q} Number of residue types and sequence {k_p}, The determinant is not zero (after considering the largest number of remainders).
[0026]
  In addition, q row p column is exp (2πjk_pn_q/ N) (where a set of M integers {k_p}, {N_q} Is 1 ≦ p, q ≦ M, 0 ≦ k_p, n_q≤ N-1, k_p<K_{p + 1}, n_q<N_{q + 1}Whether the matrix of M rows and M columns has an inverse matrix (the determinant is not 0) is expressed as q 'row p' column is exp (2πjk_{p '}n_{q '}/ N) (where 1≤p ', q'≤N-M, 0≤k_{p '}, n_{q '}≤ N-1, k_{p '}<K_{p '+ 1}, n_{q '}<N_{q '+ 1}, K_{p '}≠ k_p, N_{q '}≠ n_qIt is also clear that the NM-by-N-M matrix of) has an inverse matrix (the determinant is not 0). This is because the two matrices are in the relationship of an N-point IDFT matrix (an inverse matrix is a DFT matrix) obtained by extracting different M-row and M-column matrices and NM-row and NM-column matrices. .
[0027]
  In addition, a set of M consecutive integers {k_p} For any set of M different integers {n_q} And q rows and p columns are exp (2πjk_pn_q/ N) M-by-M matrix has an inverse matrix. Conversely, a set of M consecutive integers {n_q}, Any set of M different integers {k_p} And q rows and p columns are exp (2πjk_pn_q/ N) M-by-M matrix has an inverse matrix. These are equivalent to each other. The outline of the latter proof is as follows. That is, a set of M consecutive integers {n_q}, The maximum value B of the original number of the residue system by b (N = ab) is − [− M / b]. However, [R] is a Gaussian symbol and represents the maximum integer not exceeding the real number R. A set of M different integers {k_p}, The minimum value of the type A of the residue system by a is obtained as-[-M / b] because the residue system by a has only up to b = N / a elements. That is, a set of M consecutive integers {n_q} Is the set of M different integers {k_p} Is equal to the minimum value of the type A of the residue system by a. At this time, B does not exceed A.
[0028]
  Furthermore, N = c^m(Where c is a prime number and m is an integer of 2 or more), a set of M consecutive integers {k_p} And c^{m '}(M '= [log_cM] and [R] are Gaussian symbols, and a set of M integers whose residue system (the type of the residue and the number of each element) is completely matched {k_p} For any set of M different integers {n_q} And q rows and p columns are exp (2πjk_pn_q/ N) M-by-M matrix has an inverse matrix. This is {n_q} And {k_p} Is also true. c^{m '}Is (m '= [log_cM]), the factorial of the only prime factor of N that is equal to or does not exceed M. {K_p} And c^{m '}That the residue system by is completely coincidental is any k of M consecutive integers_pAgainst k_p± c^{m '}, K_p± 2c^{m '}, K_p± 3c^{m '}, ..., k_p± (c^m-C) (however, anything outside the range of 0 or more and N-1 or less will be excluded) Means a set of different integers.
[0029]
[Summary of inverse matrix]
1. If N is a prime number, the matrix has an inverse matrix.
2. {N_q} And {k_p} Is a set of consecutive M integers, q rows and p columns are exp (2πjk_pn_q/ N) M-by-M matrix has an inverse matrix.
3. N = c^m(Where c is a prime number and m is an integer of 2 or more), as described above, even if it is not a set of consecutive M integers, {n_q} And {k_p} Is a continuous set of M integers and c^{m '}(M '= [log_cIf the set of the residue type and the number of each element in the residue system according to M]) is the same, the matrix has an inverse matrix.
4). Other {n_q} And {k_p}, Consider the remainder system for all pairs of N and divisors a and b other than N, and {k_p}, The number A of the residue system type by a is {n_q}, There is no divisor pair a, b of N that is smaller than the maximum value B of the original number of the residue system by b, and {n_q}, The number A of the type of residue system by a is {k_p}, If there is no divisor pair a, b of N that is smaller than the maximum value B of the original number of residue systems by b, the matrix has an inverse matrix.
[0030]
  By the way, although the inverse matrix has M rows and M columns, if an L row and M column matrix (L ≦ M) obtained from the desired L rows is used, any L carriers from M subcarriers can be obtained. It is obvious that the matrix is obtained from M sample points. That is, the matrix used for implementing the demodulation method and the demodulation device of the present invention is not limited to a square matrix.
[0031]
  In addition, it is clear that the discussion so far holds even if a null carrier is included in M subcarriers. Then, assuming that there are M−L (L <M) null carriers in M subcarriers, the matrix for obtaining the remaining L carriers from M sample points is q in the above discussion. Row p column is exp (2πjk_pn_qIt is obvious that the L rows necessary for demodulating the L carriers from the M rows and M columns matrix, which is the inverse matrix of the (N) M rows and M columns matrix, may be used. That is, k corresponding to L carriers_pBesides, M-L arbitrary k_p(K corresponding to L carriers_pThe matrix of M rows and M columns is used temporarily, the inverse matrix is obtained, and the matrix of L rows and M columns (L <M) corresponding to L carriers is obtained therefrom. From simple considerations, an infinite number of such L rows and M columns can be constructed.
[0032]
[About other multi-carriers]
  In the above description, a communication method including N subcarriers with equal frequency intervals has been described using OFDM as a representative example. As is apparent from the above description, the essence of the present invention is composed of N subcarriers with equal frequency intervals. The communication method is not limited. That is, if the frequency interval ratio is an integer ratio, it is possible to design M sampling points by directly considering the contents of the existence of the inverse matrix. At this time, in the design of the sampling points of M points, even if the “N points where N subcarriers can be demodulated” are not equally spaced in the first place, the content of examining the presence or absence of the inverse matrix should be taken into consideration. Thus, it is possible to design M sampling points. This is because a communication method composed of N subcarriers with a frequency interval ratio of an integer ratio is a communication method composed of N ′ subcarriers with equal frequency intervals, and N′−N are null carriers. This is because it can be considered equivalent to a thing.
[0033]
〔Example〕
  FIG. 1 is a block diagram showing a configuration of a multicarrier demodulator 1000 having a channel characteristic estimation device 120 for multicarrier transmission which is a specific embodiment of the present invention. In the present embodiment, a multi-carrier demodulator 1000 that receives an OFDM-modulated signal composed of N subcarriers having a guard interval and performs analog orthogonal demodulation is shown.
[0034]
  The multicarrier demodulator 1000 receives an externally received wave (not shown), and includes an oscillator 102 and a phase converter 102I, multipliers 103I and 103Q, LPFs 104I and 104Q, A / D 105I and 105Q, and guard interval removers 106I and 106Q. N points corresponding to N complex numbers are sampled (portion indicated as RCV in FIG. 1). Next, the double selector 121,linearThe pilot signal is demodulated in the multi-carrier transmission channel characteristic estimation device 120 including the arithmetic unit 122 and the frequency characteristic arithmetic unit 123, and the frequency characteristic is output. The N complex numbers are demodulated by the N-point DFT 108 through the S / Ps 107I and 107Q, and corrected by the equalization circuit 109 according to the frequency characteristics of the pilot symbols.
[0035]
  Here, the double selector 121 and the linear calculator 122 correspond to means for demodulating pilot symbols or a pilot symbol calculator. The in-phase component I from the received wave_RAnd orthogonal component Q_RThe portion from which the signal is taken out (the portion designated as RCV in FIG. 1) corresponds to a sampling device for obtaining an N-point digital signal or a quadrature demodulation and sampling device.
[0036]
  The operation of the multicarrier demodulator 1000 of FIG. 1 is as follows. The synchronization circuit 101 outputs a synchronization signal to the oscillator 102 based on a pilot signal output from a multi-carrier transmission channel characteristic estimation device 120 described later. The received wave is multiplied by the sine wave output from the oscillator 102 in the multipliers 103I and 103Q and the sine wave having a phase shift of π / 2 through the phase converter 102I, and passes through the low-pass filters (LPF) 104I and 104Q. Base signal I_A, Q_AIt becomes. This is converted into an in-phase component I of a digital signal by analog / digital converters (A / D) 105I and 105Q._DAnd the quadrature component Q of the digital signal_DIt becomes. The guard interval is removed by the guard interval removers 106I and 106Q, and an in-phase component I having N digital signals in one symbol each._RAnd orthogonal component Q_RIt becomes. The sampling rate is 1 / (NΔf), where Δf is the frequency interval of N subcarriers.
[0037]
  Serial digital-to-parallel converters (S / P) 107I and 107Q allow N digital signals I_R, Q_RIs input to an N-point discrete Fourier transformer (N-point DFT) 108 as N complex signals, and the N complex signals are demodulated, and an equalization circuit 109, a demapping circuit 110, and a parallel-serial converter (P / S) ) 111, a desired signal sequence is reproduced. Here, in the equalization circuit 109, the N complex signals output from the N-point DFT 108 are corrected according to the frequency characteristics of the propagation path by the following output of the propagation characteristic estimation apparatus 120 for multicarrier transmission. Is called.
[0038]
  The multi-carrier transmission channel characteristic estimation apparatus 120 includes a double selector 121,linearIt comprises a calculator 122 and a frequency characteristic calculator 123. These actions are as follows. The double selector 121 has an in-phase component I having N digital signals output from the guard interval removers 106I and 106Q._RAnd orthogonal component Q_RFrom the above, each of the necessary M signals is selected, subjected to serial / parallel conversion, and then output to the linear arithmetic unit 122. The necessary M signals at this time are for a preset sampling number, but may be M consecutive from the beginning in one symbol length, for example. The linear arithmetic unit 122 demodulates the pilot signal with a preset coefficient, using the 2M signals output from the double selector 121 as M complex signals. At this time, one to a plurality of pilot signals are demodulated from the M complex signals. This calculation is a linear calculation from M complex signals in terms of software. The coefficients in the linear operation, that is, the linear combination, are part of the inverse matrix described in detail before the description of this embodiment. Since only the pilot signal carrier needs to be solved, a linear coefficient for demodulating the pilot signal may be obtained even when the inverse matrix does not exist. It should be noted that, based on a slight modification of the configuration, it is also possible to configure by hardware FIR. The output of the linear calculator 122 is output to the frequency characteristic calculator 123 and the synchronization circuit 101. As described above, the synchronization circuit 101 generates a synchronization signal using the pilot signal.
[0039]
  One or more pilot signals demodulated by the linear calculator 122 are applied to the frequency characteristic calculator 123 to estimate the propagation path characteristics. The result is output to the above-described equalization circuit 109, and the N complex signals output from the N-point DFT 108 are corrected according to the frequency characteristics of the propagation path. Here, when the linear arithmetic unit 122 demodulates one or a plurality of pilot signals, it does not use an N-point DFT, and is at least L M or more necessary for demodulating one or a plurality of pilot signals. Since only one or a plurality of pilot signals are demodulated by using only these complex signals by their linear operation, pilot signal demodulation can be performed in a shorter time.
[0040]
  This can be seen from FIG. That is, the pilot symbol can be demodulated in a time T ′ (sample number M, M <N) shorter than the original length T (sample number N) of one symbol. This is possible because there are L or less effective carriers including pilot symbols in a symbol interval including pilot symbols.
[0041]
[Modification]
  In the above-described embodiment, by changing the function of each component of the propagation path characteristic estimation device 120 for multicarrier transmission as follows, it is possible to more accurately detect the time variation of the propagation path characteristic. However, it is assumed that the number M of complex signals necessary for demodulating one or more pilot signals is N / 2 or less.
[0042]
  The operation of the double selector 121 and the linear arithmetic unit 122 constituting the channel characteristic estimation apparatus 120 for multicarrier transmission in this modification is 2 every time the guard interval removers 106I and 106Q output N digital signals. Performed once. That is, when the guard interval removers 11I and 11Q each output N / 2 digital signals in the first half, the double selector 121 and the linear arithmetic unit 122 select M complex signals therefrom to select one or more. The pilot signal is demodulated. Next, when the guard interval removers 106I and 106Q each output N / 2 digital signals in the latter half, the double selector 121 and the linear arithmetic unit 122 select M complex signals therefrom to select one or more. The pilot signals are demodulated. Thus, one or a plurality of pilot signals are demodulated twice in the first half and the second half of the digital signals output from the guard interval removers 106I and 106Q. The frequency characteristic calculator 123 also detects a change in one or a plurality of pilot signals demodulated twice, and outputs the propagation path characteristic and its time change to the equalization circuit 109. In this way, in this modification, the propagation path characteristic and its time change can be detected within one symbol length, so that equalization with high accuracy can be performed.
[0043]
  This can be seen from FIG. That is, the time T shorter than 1/2 of the original length T (number of samples N) of one symbol.₁', T₂'(Both are the number of samples M, M <N / 2), and the pilot symbols can be demodulated. This is possible because there are L or less effective carriers including pilot symbols in a symbol interval including pilot symbols.
[0044]
  In this modification, the length T (number of samples N) of one symbol on the time axis is divided into the first half and the second half, and each sample has a sample number M (M <N / 2). Although the M sample points are not overlapped, the two sets of M sample points may be overlapped. There are NM + 1 ways of extracting M consecutive sample points from the number N of samples of one symbol, and by demodulating pilot symbols from two or more sets of M sample points, propagation path characteristics and It is possible to detect time changes. If the intervals of the M sample points (interval pattern from the first sample point to each sample point) match, the M points are continuous when demodulating pilot symbols from two or more sets of sample points. It does not have to be.
[0045]
  In the above embodiment, an example in which a signal by OFDM having a guard interval is received has been described. However, the guard interval is not essential for implementing the present invention, and the communication method is not limited to OFDM. In the above-described embodiment, an example of performing analog quadrature demodulation has been described. However, digital quadrature demodulation may be performed, and quadrature demodulation may not be performed.
[0046]
  In this embodiment, the N-point DFT for normal demodulation, A / D conversion, and quadrature demodulation are combined and sampled at the sampling rate 1 / (NΔf). However, N subcarriers are demodulated. In the case of performing by other methods (for example, inverse matrix operation), the sampling rate may be smaller than 1 / (NΔf) (faster sampling). In addition, when the N-point DFT, A / D conversion, and orthogonal demodulation are separately provided, the sampling rate is arbitrary as long as the pilot signal can be demodulated.
[Brief description of the drawings]
FIG. 1 is a block diagram showing a configuration of a multicarrier demodulator according to a specific embodiment of the present invention.
FIG. 2 is an explanatory diagram showing a time for demodulating pilot symbols in the embodiment.
FIG. 3 is an explanatory diagram showing a time for demodulating pilot symbols according to a modification.
[Explanation of symbols]
1000 Multi-carrier demodulator
120 propagation path estimation apparatus
121 Double selector
122 linear operator
123 Frequency characteristics calculator
101 Synchronous circuit
106I, 106Q Guard interval remover
108 N point DFT
109 Equalization circuit 2

Claims (9)

In the multi-carrier transmission method, in the multi-carrier transmission channel characteristic estimation method for receiving a signal having a pilot symbol using L subcarriers for a specific symbol length and estimating the channel characteristic of the signal,
Obtain N digital signals that can demodulate all N (N> L) subcarriers,
Based on the number of M subcarriers other than the NM null carriers and the number of the extracted sample points, the M number of samples other than the NM null carriers from the M sample points. Prepare a matrix for demodulating desired L subcarriers (L ≦ M) among the subcarriers,
The pilot is demodulated by demodulating desired L subcarriers among M subcarriers other than the NM null carriers from the product of the prepared matrix and a vector composed of M sample points. Demodulate the symbol,
A channel characteristic estimation method for multicarrier transmission, characterized by estimating a channel characteristic of a received signal from the demodulated pilot symbol. The multi-carrier transmission method is a method of modulating using inverse discrete Fourier transform ,
2. The multi- point digital signal according to claim 1, wherein the digital signal is an N-point complex digital signal that is orthogonally demodulated at a sampling interval 1 / (NΔf) where Δf is an adjacent frequency interval of N subcarriers. Carrier transmission channel characteristic estimation method.   3. The method of estimating channel characteristics for multicarrier transmission according to claim 2, wherein the preset M points are M consecutive sampling points.   4. The propagation path characteristic is estimated by demodulating a pilot symbol h times (where h is a natural number 2 ≦ h ≦ N−M + 1) within a specific symbol length having a pilot symbol. Channel characteristics estimation method for multicarrier transmission. In a multicarrier transmission system receiver, a multicarrier transmission channel characteristic estimation apparatus that receives a signal having a pilot symbol using L subcarriers for a specific symbol length and estimates a channel characteristic of the signal,
A sampling device for obtaining N digital signals capable of demodulating all N (N> L) subcarriers;
An extraction device for extracting M points from the digital signal of N points;
Based on the number of M subcarriers other than the NM null carriers and the number of sample points taken out by the extraction device, the NM points from the M sample points taken out by the extraction device. Of the N subcarriers, the matrix for demodulating desired L subcarriers (L ≦ M), and a vector composed of M sample points, and the N−M A pilot symbol calculator that demodulates the pilot symbols by demodulating desired L subcarriers (L ≦ M) among M subcarriers other than null carriers ;
A channel characteristic estimation apparatus for multicarrier transmission, which estimates channel characteristics of a received signal from the demodulated pilot symbols. The multi-carrier transmission method is a method of modulating using inverse discrete Fourier transform ,
6. The multi- point digital signal according to claim 5, wherein the digital signal is an N-point complex digital signal that is orthogonally demodulated at a sampling interval 1 / (NΔf) where Δf is an adjacent frequency interval of N subcarriers. Carrier transmission channel characteristic estimation device.   The channel characteristic estimation apparatus for multicarrier transmission according to claim 6, wherein the preset M points are M consecutive sampling points.   8. The propagation path characteristic is estimated by demodulating a pilot symbol h times (where h is a natural number 2 ≦ h ≦ N−M + 1) within a specific symbol length having a pilot symbol. Multi-carrier transmission channel characteristic estimation device. A multicarrier demodulation apparatus comprising the multicarrier transmission channel characteristic estimation apparatus according to any one of claims 5 to 8.

Images