JP3735619B2 - Preamble pattern identification method - Google Patents

Description

そして、上記Ｘ(k)(k=0、1、……、N-1)の電力和ｐは式(2)で表わせる。
【００４９】
【数２】

この式(2)の|Ｘ(k)|² は、Ｘ(k)とその複素共役Ｘ^*(k)の積になるので、次の式(3)が成立する。なお、ここで、Ｘ^*(k)に付されている符号＊は複素共役を表わす。ここで、この複素共役とは虚数部の符号が反転されていることをいう。
【００５０】
【数３】

そこで、式(2)の|Ｘ(k)|² に式(3)を代入すると、電力和ｐは、次の式(4)となる。
【００５１】
【数４】

ここで、この式(4)のＳ_n,m は、初項ａ₁＝Ｗ_N ^(n-m)0＝１、公比ｒ＝Ｗ_N ^(n-m)の初項から第Ｎ項までの等比級数の和であり、これは、次の式(5)公式で求められる。
【００５２】
【数５】

ここで、公比ｒが１となるのは、ｎ−ｍがＮの整数倍(ｎ−ｍ＝０、±Ｎ、±２Ｎ、……)の場合であり、式(4)ではｎ＝０、１、……、Ｎ−１、ｍ＝０、１、……、Ｎ−１であるから、|ｎ−ｍ|≦Ｎ−１であり、ｒ＝１となるのは、ｎ−ｍ＝０、すなわちｎ＝ｍの場合だけであり、このｎ＝ｍの場合は、ｒ＝Ｗ_N ⁰＝１であるから、式(5)より次の式(6)が成立する。
【００５３】
【数６】

一方、ｎ≠ｍの場合は、ｒ＝Ｗ_N ⁰≠１であるから、同じく式(5)より、こんどは次の式(7)が成立する。
【００５４】
【数７】

この式(7)において、ｎ、ｍは整数(ｎ≠ｍ)であるから、２π(ｎ−ｍ)は２πの整数倍であり、従って、次の式(8)が成立する。
【００５５】
【数８】

また、このとき、ｎ−ｍ＝±１、±２、……、±(Ｎ−１)であるから、−２π＜２π(ｎ−ｍ)／Ｎ＜２πとなり、且つ、２π(ｎ−ｍ)／Ｎ≠０なので、次の式(9)が成立する。
【００５６】
【数９】

従って、この場合、式(7)の分子＝０、分母≠０となるから、ｎ≠ｍのとき、式(10)が成立する。
【００５７】
【数１０】

そこで、式(6)と式(10)を纏めて表わすと式(11)となる。
【００５８】
【数１１】

そして、式(4)は、式(12)となる。
【００５９】
【数１２】

以上の結果、時間領域の信号ｘ(n)(n=0、1、2、……、N-1)の電力和|ｘ(0)|²＋|ｘ(1)|²＋……＋|ｘ(N-1)|² に、ＤＦＴのポイント数Ｎを乗算してやれば、全周波数成分Ｘ(k)(k=0、1、2、……、N-1)の電力和ｐ＝|Ｘ(0)|²＋|Ｘ(1)|²＋……＋|Ｘ(N-1)|² の演算が得られることが判り、従って、時間領域信号の電力和により周波数領域の電力和が算出できるのである。
【００６０】
次に、プリアンブル信号の自己相関により周波数偏差が検出できる点について説明する。
【００６１】
いま、入力ベースバンド信号の任意のＮサンプルを連続して取り出し、それらをｓ(0)、ｓ(1)、……、ｓ(N-1)とすると、その自己相関は次の式(13)で表わせる。
【００６２】
【数１３】

ここで、ｍは負でない整数、ｓ^*(n)の＊は複素共役を表わし、複素共役とは虚数部の符号を反転することをいう。
【００６３】
このとき、プリアンブルのベースバンド信号は、上述した従来技術にも示されるように(特許文献１の図５)、８シンボル周期の信号で、２シンボル毎に点対称であるため、ｒ(m)の振幅２乗値|ｒ(m)|² は、ｍ＝０、２Ｎ_ov、４Ｎ_ov、……と２シンボル毎に値がピ-クとなる。
【００６４】
ここで、Ｎ_ov はオーバーサンプル比で、１シンボル当りのオーバーサンプル 数である。そして、この信号は、周波数偏差が無い場合、２シンボル前の瞬時値に対する位相差が常にπ／２で、かつｓ(n)＝ｓ(ｎ−２Ｎ_ov)ｅ^{j π /2} であり、従って、式(14)が成立する。
【００６５】
【数１４】

そうすると、式(13)の自己相関の２シンボル毎のピーク値は式(15)で表わすことができ、ｒ(0)の位相は０、ｒ(２Ｎ_ov)の位相はπ／２、ｒ(４Ｎ_ov)の位相は πというように、ピーク値は２シンボル毎にπ／２ずつ位相が回転する。
【００６６】
【数１５】

ここで、受信信号にΔｆ(Hz)の周波数偏差がある場合、１シンボルで２πΔｆ／ｆ_b(ラジアン)の位相回転が受信信号に加わるため、２ｋシンボルでは４ｋπ Δｆ／ｆ_bの位相回転が加わり、式(16)で示すようになる。
【００６７】
【数１６】

そこで、自己相関の２シンボル毎のピーク値は、式(17)に示すようになる。
【００６８】
【数１７】

また、ｒ(２ｋＮ_ov)の位相は、式(18)に示すようになる。
【００６９】
【数１８】
θ＝arg{r(2kNov)}=kπ/2+4kπΔf/f_b ……式(18)
ここで、arg{ｒ(２ｋＮ_ov)}はｒ(２ｋＮ_ov)の位相を表わす。
【００７０】
よって、周波数偏差Δｆは、自己相関のピーク値ｒ(２ｋＮ_ov)の位相θを用いて、式(19)により算出でき、従って、プリアンブル信号の自己相関により周波数偏差が検出できるのである。
【００７１】
【数１９】

次に、本発明によるプリアンブルパターン識別方法及び周波数偏差検出方法の一実施の形態が適用される受信機の本体について、一例として図５により説明する。
【００７２】
この図５に示した受信機は、上述の従来技術にも示されているもので、図示してないアンテナで受信された信号は入力端子１１０１に供給され、高周波回路１１０２に入力される。
【００７３】
そして、この高周波回路１１０２で周波数変換され、無線周波数から中間周波数に変換されてＡ／Ｄ変換器１１０３に供給され、ここでサンプリングされ、量子化されてデジタル信号に変換さた上で乗算器１１０４−１、１１０４−２に並列に入力される。
【００７４】
このとき、正弦波発生回路１１０６は、角周波数ωの正弦波信号cosωt を発生し、これを乗算器１１０４−１にはそのまま供給し、乗算器１１０５には、移相器１１０５を介して位相をπ／２(ラジアン)進め、正弦波信号−cosωt としてから供給する。
【００７５】
そこで、乗算器１１０４−１は、Ａ／Ｄ変換器１１０３の出力と正弦波発生回路１１０６の出力の積を演算し、乗算器１１０４−２では、Ａ／Ｄ変換器１１０３の出力と移相器１１０５の出力の積を演算する。
【００７６】
このとき、これら乗算器１１０４−１、１１０４−２の出力には不要な高周波成分が含まれている。そこで、各々の出力をローパスフィルタ１１０７−１、１１０７−２に入力し、不要成分を除去してから夫々ロールオフフィルタ１１０８−１、１１０８−２に供給する。
【００７７】
これらロールオフフィルタ１１０８−１、１１０８−２では、入力された信号の帯域を制限してベースバンド信号を得、夫々をベースバンド信号出力端子１１０９−１、１１０９−２に供給する。
【００７８】
このとき、一方のベースバンド信号出力端子１１０９−１から出力される信号はベースバンド信号の同相成分を表わし、他方のベースバンド信号出力端子１１０９−１から出力される信号はベースバンド信号の直交成分を表わしており、従って、ベースバンド信号は、同相成分を実数部とし、直交成分を虚数部とする複素数信号である。
【００７９】
そこで、これらベースバンド信号出力端子１１０９−１、１１０９−２に現われたベースバンド信号が図示してない受信機の復調部を含む信号処理系に供給され、伝送されてきた情報が再生されることになるのであるが、このとき、これらベースバンド信号出力端子１１０９−１、１１０９−２に現われたベースバンド信号から、上記したプリアンブルパターンの識別と周波数偏差が検出されることになる。
【００８０】
ここで、図１が本発明の一実施形態で、上記したベースバンド信号出力端子１１０９−１、１１０９−２に現われたベースバンド信号は、この図における入力端子１０１に入力される。
【００８１】
このとき、ベースバンド信号は、上記したように、同相成分と直交成分からなる複素数信号であるが、ここでは、説明を簡単にするため、１個の入力端子１０１だけ示してある。
【００８２】
ここで、この信号は、上記したように、１シンボル当たりＮ_ov 回、サンプリングされたベースバンド信号で、このとき、Ｎ_ov は２以上の正整数である。
【００８３】
なお、この図１において、入力端子１０１とシフトレジスタ１１４、スイッチ１１３、窓掛け回路１０２、プリアンブルパターン識別回路１０７、それにシンボルタイミング検出回路１０９は、上記した従来技術(特許文献１の図１)と同じなので、詳しい説明は省略する。
【００８４】
図１において、まず、スイッチ１１３は、一定の時間間隔Ｎ_step のサンプル周期に１回閉じ、これにより、シフトレジスタ１１４からＮ_win サンプルのデータｓ(0)、ｓ(1)、……、ｓ(Ｎ_win−１)が取り出され、自己相関演算回路１２１と周波数偏差補正回路１２８に入力される。
【００８５】
このとき、例えばオーバーサンプル比Ｎ_ov＝２、Ｎ_step＝８、Ｎ_win＝６４である。ここで、Ｎ_win はシフトレジスタ１１４の段数であり、従って正の整数である。
【００８６】
そして、まず、自己相関演算回路１２１は、スイッチ１１３により取り出されてくるｓ(0)、ｓ(1)、……、ｓ(Ｎ_win-1)のＮ_win 個のサンプルについて、自己相関ｒ(m)の内のｒ(２Ｎ_ov)を、上記した式(13)により演算し、その演算値(複素数)を位相検出回路１２２に入力する。
【００８７】
そこで、位相検出回路１２２は、入力された自己相関ｒ(２Ｎ_ov)の位相θを、上記した式(18)により、−π≦θ＜πの範囲で検出し、減算器１２９の(＋)側入力端子に入力し、ここで、入力された位相θから定数値π／２を減算し、−π≦θ−π／２＜πの範囲で位相θ−π／２を演算し、乗算器１２３の一方の入力端子に供給する。
【００８８】
つまり、この減算器１２９では、位相θからπ／２を減算した結果がθ−π／２＜−πとなったとき、この減算結果に２πを加算して乗算器１２３の一方の入力端子に入力することになる。
【００８９】
乗算器１２３では、入力された位相θ−π／２に定数値ｆ_b／４πを乗算して周波数偏差Δｆを演算し、それを周波数偏差補正回路１２８に入力する。このときの処理は式(19)で説明した通りであり、演算した周波数偏差Δｆは周波数偏差出力端子１１１にも供給され、必要に応じて外部に出力される。
【００９０】
そこで、周波数偏差補正回路１２８は、乗算器１２３から入力された周波数偏差Δｆを用い、スイッチ１１３から入力されるＮ_win サンプルのベースバンド信号ｓ(0)、ｓ(1)、……、ｓ(N_win-1)について、式(20)に示すように、位相を1サンプルにつき−２πΔｆ／Ｎ_ovｆ_b ラジアンずつ回転させ、周波数偏差を補正する。
【００９１】
【数２０】

こうして、周波数偏差補正回路１２８から入力されたＮ_win サンプル補正信号ｓ'(n)(ｎ＝０、１、２、……、Ｎ_win-1)は、窓掛け回路１０２に入力され、ここで窓(例えばハミング窓：Hamming Window)を掛け、窓掛けされたベースバンド信号ｘ(n)(ｎ＝０、１、２、……、Ｎ_win-1)が全電力算出回路１２４とＤＦＴ演算回路１２６、１２７に入力される。
【００９２】
そして、まず、全電力算出回路１２４では、式(12)で説明したように、入力されたＮ_win サンプルのベースバンド信号ｘ(n)(ｎ＝０、１、２、……、Ｎ_win-1)の全サンプルの振幅２乗値の和|ｘ(0)|²＋|ｘ(1)|²＋……＋|ｘ(N_win-1)|² を演算し、乗算器１２５の一方の入力端子に入力し、ここで定数値Ｎ_win を乗じて乗算結果ｐ＝Ｎ_win{|ｘ(0)|²＋|ｘ(1)|²＋……＋|ｘ(N_win-1)|² }をプリアンブルパターン識別回路１０７に入力する。
【００９３】
一方、ＤＦＴ演算回路１２６は、窓掛け回路１０２から入力されるＮ_win サンプルのベースバンド信号ｘ(n)(ｎ＝０、１、２、……、Ｎ_win-1)についてのＤＦＴ Ｘ(0)、Ｘ(1)、……、Ｘ(N_win-1)の中で、−(３／８)ｆ_b の周波数成分であるＸ₁＝Ｘ(N_win−３Ｎ_win／８Ｎ_ov)についてだけ演算し、その電力ｐ₁＝|Ｘ₁|² をプリアンブルパターン識別回路１０７に入力し、このときの周波数成分Ｘ₁ の位相φ₁ はシンボルタイミング検出回路１０９に入力する。
【００９４】
また、ＤＦＴ演算回路１２７は、窓掛け経路１０２から入力されるＮ_win サンプルのベースバンド信号ｘ(n)(ｎ＝０、１、２、……、Ｎ_win-1)についてのＤＦＴ Ｘ(0)、Ｘ(1)、……、Ｘ(N_win-1)の中で、(１／８)ｆ_b の周波数成分であるＸ₂＝Ｘ(N_win／８Ｎ_ov)についてだけ演算し、その電力ｐ₂＝|Ｘ₂|² をプリアンブルパターン識別回路１０７に入力し、このときの周波数成分Ｘ₂ の位相φ₂ はシンボルタイミング検出回路１０９に入力する。
【００９５】
そこで、まず、プリアンブルパターン識別回路１０７は、乗算器１２５から入力される乗算結果ｐと、ＤＦＴ演算回路１２６、１２７から各々入力される電力ｐ₁、ｐ₂ から、従来技術(特許文献１の図１)と同様にしてプリアンブルパターンの識別を行い、その結果をプリアンブルパターン識別出力端子１１０に供給する。
【００９６】
また、シンボルタイミング検出回路１０９は、ＤＦＴ演算回路１２６、１２７から入力される位相φ₁、φ₂ から、これも従来技術(特許文献１の図１)と同様にしてシンボルタイミングを検出し、その結果をシンボルタイミング出力端子１１２に供給する。
【００９７】
従って、この実施形態によれば、プリアンブルパターン識別出力端子１１０の出力により、プリアンブルのパターンが識別されたことが確認でき、シンボルタイミング出力端子１１２の出力により、シンボルタイミングが確認できるので、受信機は、容易に高速同期をとることができる。
【００９８】
ところで、以上の説明では、自己相関演算回路１２１による自己相関の演算がｒ(２Ｎ_ov)の演算によるものになっているが、ｒ(２Ｎ_ov)の代りにｒ(４Ｎ_ov)、ｒ(６Ｎ_ov)、ｒ(８Ｎ_ov)の演算を用いても良い。
【００９９】
このとき、減算回路１２９で減算すべき値は、式(19)から、ｒ(４Ｎ_ov)の場合はπになり、ｒ(６Ｎ_ov)の場合は(３／２)π、ｒ(８Ｎ_ov)の場合は、０(２πなので０でも良い)となり、また、乗算器１２３で乗ずべき値は、ｒ(４Ｎ_ov)の場合、ｆ_b／８π、ｒ(６Ｎ_ov)の場合でｆ_b／１２π、ｒ(８Ｎ_ov)の場合でｆ_b／１６πとなる。
【０１００】
また、以上の説明は、プリアンブルが１００１(2進数)の繰り返しパターンのときの実施形態についてのものであるが、プリアンブルが０１１０(2進数)の繰り返しパターンのときは次の通りになる。
【０１０１】
すなわち、この場合、自己相関ｒ(m)は変り無く、ｒ(0)、ｒ(２Ｎ_ov)、ｒ(４Ｎ_ov)、……の２シンボル毎に振幅２乗値がピークとなるが、この場合は位相が逆になるため、周波数偏差が無い場合、２シンボル前の瞬時値に対する位相差が常に−π／２となる。
【０１０２】
従って、ｒ(0)の位相は０、ｒ(２Ｎ_ov)の位相は−π／２、ｒ(４Ｎ_ov)の位相は−πというように、ピーク値の位相が２シンボル毎に−π／２ずつ回転するため、減算回路１２９で減算すべき値は、ｒ(２Ｎ_ov)の場合−π／２、ｒ(４Ｎ_ov)の場合で−π、ｒ(６Ｎ_ov)の場合で−(３／２)π、ｒ(８Ｎ_ov)の場合で０(ここでも、−２πなので０で良い)となる。そして、乗算器１２３で乗ずべき値は、ｒ(２Ｎ_ov)の場合はｆ_b／４π、ｒ(４Ｎ_ov)の場合はｆ_b／８π、ｒ(６Ｎ_ov)の場合はｆ_b／１２π、ｒ(８Ｎ_ov)の場合ではｆ_b／１６πとなる。
【０１０３】
また、この場合は、プリアンブルパターンの周波数成分が、−(１／８)ｆ_b と(３／８)ｆ_b になる。そこで、ＤＦＴ演算回路１２６では−(１／８)ｆ_b の周波数成分であるＸ(Ｎ_win−Ｎ_win／８Ｎ_ov)を演算し、ＤＦＴ演算回路１２７では、(３／８)ｆ_b の周波数成分であるＸ(３Ｎ_win/８Ｎ_ov)を演算することになる。
【０１０４】
更に、以上の説明は、変調方式がπ／４シフトＱＰＳＫ方式で、プリアンブルが１００１(２進数)の繰り返しパターンのときの実施形態であるが、本発明の実施形態としては、変調方式がＱＰＳＫ方式又はＱＡＭ方式で、上記した従来技術(特許文献１の図４)に示すように、対角する２つの最外角点を１シンボル置きに繰り返すパターンに対応したものも考えられる。
【０１０５】
この場合、プリアンブル信号は２シンボル周期の信号で、自己相関ｒ(m)は、ｒ(0)、ｒ(２Ｎ_ov)、ｒ(４Ｎ_ov)、……の２シンボル毎に振幅２乗値がピークとなり、従って、周波数偏差が無い場合、２シンボル前の瞬時値に対して常に同位相になるため、ｒ(２Ｎ_ov)、ｒ(４Ｎ_ov)、……の位相は０である。
【０１０６】
そこで、この場合、減算回路１２９は不要で、省略することができる。また、この場合、プリアンブルの周波数成分は±(１／２)ｆ_b になる。そこで、このとき、ＤＦＴ演算回路１２６は−(１／２)ｆ_b の周波数成分であるＸ₁＝Ｘ(N_win−N_win／２N_ov)を演算し、ＤＦＴ演算回路１２７は(１／２)ｆ_b の周波数成分であるＸ２＝Ｘ(Ｎ_win／２Ｎ_ov)を演算することになる。
【０１０７】
ところで、上述の実施形態では、自己相関演算回路１２１、位相検出回路１２２、減算器１２９、乗算器１２３による周波数偏差検出と、周波数偏差補正回路１２８による周波数偏差補正が、入力端子１０１に入力される信号がプリアンブル以外の信号のときも行なわれ、この補正結果を用いてプリアンブル識別を行っている。
【０１０８】
そうすると、この場合の周波数偏差補正は正しい結果にならず、プリアンブルパターン識別回路１０７によるプリアンブルパターン識別に影響を及ぼすように考えられるかも知れない。
【０１０９】
しかし、全電力算出回路１２４の演算結果に関しては、周波数偏差補正回路１２８は入力信号の位相を回転させるだけであり、従って、影響を及ぼす虞れは無い(ｐの値には影響しない)。
【０１１０】
また、ＤＦＴ演算回路１２６、１２７の演算に関しては、周波数偏差補正回路１２８により周波数成分がシフトされるので、結果として別の周波数成分の電力が算出されてしまう。
【０１１１】
しかし、プリアンブル以外の信号の場合、帯域内では、ガードタイムを除いた期間で全周波数成分がほぼ均等になり、ガードタイムのとき１種の周波数成分になる(オール０がマッピングされているので)かの何れかである。
【０１１２】
従って、これら２種の周波数成分が算出されても、それらの電力和が全電力の大部分を占めることはなく、従って、誤検出をもたらすような影響が発生する虞れはない。
【０１１３】
ところで、上述した実施形態では、ＦＦＴ出力の全電力算出に代えて窓掛け出力の全電力算出に置き換え、ＦＦＴ演算に代えて２種の周波数成分だけのＤＦＴ演算に置き換えているので、演算量の削減が図られている。
【０１１４】
そこで、以下、この点について説明すると、まず、ＦＦＴ演算の場合、ポイント数をＮ_win とすると、２Ｎ_win log₂Ｎ_win 回の積和演算を行う必要があり、従って、従来技術では、窓長Ｎ_win＝６４の場合、２×６４log₂６４＝７６８回、窓長Ｎ_win＝１２８の場合は、２×１２８log₂１２８＝１７９２回の積和演算を必要とした。
【０１１５】
一方、上記実施形態では、ＦＦＴ演算を２種の周波数成分によるＤＦＴ演算に置き換えた結果、自己相関演算と位相検出処理が追加されている。しかし、このＤＦＴによる１成分の演算はＮ_win 回の積和演算で済むので、自己相関演算は約Ｎ_win 回の積和演算(正確にはＮ_win−２Ｎ_ov 回)になり、位相検出処理は約３００回程度の積和演算に相当する演算量(ＤＳＰによるソフトウェア処理の場合)になる。
【０１１６】
そうすると、上記実施形態の場合、Ｎ_win＝６４のとき、ＤＦＴ演算が２×６４＝１２８回で自己相関演算がＮ_win＝６４回、位相検出処理に３００回、合計６４＋１２８＋３００＝４９２回の積和演算に相当する演算量になり、従来技術の７６８回の演算量に比較して大幅に演算量が低減される。
【０１１７】
また、同じくＮ_win＝１２８の場合、ＤＦＴ演算が２×１２８＝２５６回、自己相関演算がＮ_win＝１２８回、位相検出処理に３００回、合計２５６＋１２８＋３００＝６８４回の積和演算に相当する演算量になるので、従来技術の１７９２回の演算量に比較して、更に大幅に演算量が低減されることになり、このことから、上記実施形態の場合、ＤＦＴ(ＦＦＴ)のポイント数Ｎ_win が多くなる程、演算量削減の効果は大きくなることが判る。
【０１１８】
ところで、上述した実施形態における入力ベースバンド信号の自己相関には、シンボルタイミングの同期状態に関わらず、入力ベースバンド信号の周期性を示すピーク値が存在する。
【０１１９】
従って、上記実施形態においては、自己相関演算回路１２１と位相検出回路１２２、減算器１２９、それに乗算器１２３による周波数偏差の検出を、シンボルタイミングの同期状態に関わらず行うことができる。
【０１２０】
更に、上述の実施形態の場合、周波数偏差の検出精度は位相検出回路１２２の処理方法に依存し、ＤＦＴのポイント数(窓長)Ｎ_win には依存しない。
【０１２１】
このため、従来技術では、周波数偏差の検出精度を上げるために、ＦＦＴ(或いはＤＦＴ)のポイント数を上げるか、ＦＦＴ(或いはＤＦＴ)出力の補間処理を必要としたが、上述の実施形態ではその必要がなく、この結果、演算量が更に削減できる。
【０１２２】
【発明の効果】
本発明のプリアンブルパターン識別方法によれば、全周波数成分の電力和の演算を時間領域信号の電力和に置き換え、周波数スペクトルの演算は２種の周波数成分のみで行なうので、演算量が削減される。
【図面の簡単な説明】
【図１】 本発明によるプリアンブルパターン識別方法の一実施形態を示すブロック図である。
【図２】 同期バーストフレームのフレーム構造の一例を示す説明図である。
【図３】 通信チャネルフレームのフレーム構造の一例を示す説明図である。
【図４】 送信パターンの一例を示す説明図である。
【図５】 本発明が対象とする受信機の一例を示すブロック図である。
【符号の説明】
１０１ 入力端子
１０２ 窓掛け回路
１０７ プリアンブルパターン識別回路
１０９ シンボルタイミング検出回路
１１０ プリアンブルパターン識別出力端子
１１１ 周波数偏差出力端子
１１２ シンボルタイミング出力端子
１１３ スイッチ
１１４ シフトレジスタ
１２１ 自己相関演算回路
１２２ 位相検出回路
１２３ 乗算器
１２５ 乗算器
１２４ 全電力算出回路
１２６ ＤＦＴ演算回路(−(３／８)ｆb 成分用)
１２７ ＤＦＴ演算回路((１／８)ｆb 成分用)
１２８ 周波数偏差補正回路
１２９ 減算器[0001]
BACKGROUND OF THE INVENTION
The present invention relates to a receiving apparatus of a digital wireless communication system, and more particularly to a method for identifying whether or not a known signal pattern is included in a received signal.
[0002]
[Prior art]
In recent years, when digital information is wirelessly transmitted, the information (a bit string of a binary signal) is divided into a predetermined number of bits, and a frame having a predetermined configuration including the divided bit string is formed. For example, there is known a digital wireless communication system in which digital modulation is performed by, for example, the π / 4 shift QPSK method and transmitted as a digital modulated signal.
[0003]
At this time, in this system, the first frame when the transmitter starts transmission for the first time from the state where the transmission has not been performed is set as the first frame. In the following, signals are transmitted sequentially from the second frame, the third frame, and so on.
[0004]
On the other hand, in order to receive information structured in this way, it is necessary to synchronize with the frame of the signal. At this time, in the above system, frame synchronization is performed from the received signal. Therefore, a fixed bit pattern having a prescribed format called a preamble is added to a transmitted frame with a predetermined length at a predetermined position.
[0005]
Then, when the signal of the first frame is received, the receiver uses the preamble included in the first frame, and based on this, performs high-speed synchronization on the signal transmitted thereafter. Here, the high-speed synchronization refers to a process of performing synchronization from an asynchronous state and reflecting the synchronized information in the demodulation (detection) operation of the current frame.
[0006]
On the other hand, synchronization has already been established in the frame before one frame period, and after that, synchronization is performed for the variation between the current frame and the frame before one frame period, and the synchronized information is demodulated for the next frame. The process reflected in is referred to as normal synchronization in contrast to the above high-speed synchronization.
[0007]
Therefore, the first frame is also called a synchronization burst, and this frame includes a preamble. Therefore, the receiver performs high-speed synchronization using the synchronization burst of the first frame, and performs normal synchronization after the second frame.
[0008]
Then, the receiver needs to detect the frequency deviation and the symbol timing from the preamble for high-speed synchronization, and for this purpose, first determine whether or not a synchronization burst including the preamble is received, The preamble pattern must be identified.
[0009]
At this time, normally, one frame to several frames are continuously transmitted as the frames of the synchronization burst. For example, in FIG. 4 to be described later, two frames of the first frame and the second frame are frames of a synchronization burst.
[0010]
Here, FIG. 2 shows the frame structure of a synchronous burst frame according to ARIB STD-T61, which is a standard of the SCPC (Single Channel Per Carrier) system. In this case, as shown in FIG. It comprises a ramp-up unit LP + R, a preamble unit Pb, a communication information channel unit RI, a synchronization word pattern unit SW, a parameter information channel unit PI, and a guard time unit G.
[0011]
Next, FIG. 3 shows a frame structure of a communication channel frame. Similarly, a linearizer preamble / ramp-up unit LP + R and preamble unit Pb, a communication information channel unit RI, a synchronization word pattern unit SW, and a parameter information channel unit PI. Here, a communication channel portion Tch is inserted between the preamble portion Pb and the communication information channel portion RI, and an undefined portion UD and a communication channel portion Tch are added after the synchronization word pattern portion SW. On the other hand, the guard time part G is omitted.
[0012]
2 and 3, the numerical value written below each symbol represents the number of bits constituting each area. At this time, the preamble portion Pb is “1”. , 0, 0, 1 ″ are composed of fixed patterns.
[0013]
On the other hand, FIG. 4 shows an example of a transmission pattern when transmission is performed in the mobile radio communication system, which is a transmission pattern based on the ARIB STD-T61 standard, where SB ₀ and SB ₁ are synchronous burst frames, TCH ₀ , TCH ₁ , TCH ₂ ,..., TCH _M are communication channel frames carrying main data, and the numerical value M is a natural number.
[0014]
In the ARIB STD-T61 standard, when transmission is started, a synchronization burst frame is transmitted m times (m is a positive integer of 2 or more) prior to transmission of a communication channel frame.
[0015]
For example, in FIG. 4, first, two synchronization burst frames are transmitted, then transmission channel frames are transmitted, and transmitted until the transmission is completed. Therefore, in this case, the call is synchronized with two frames. And M frame communication channels. The number of frames M of the communication channel at this time varies depending on the length of the call.
[0016]
By the way, the receiver needs to identify the preamble from the received signal and to detect the frequency deviation and the symbol timing from the preamble because of the high-speed synchronization described above.
[0017]
Here, for the preamble identification method and the frequency deviation and symbol timing detection method at this time, FFT (Fast Fourier Transform) is used for calculation of the frequency spectrum of the received signal, and all frequency components are calculated by FFT. A method for identifying a preamble by performing an operation is known as a prior art (see, for example, Patent Document 1).
[0018]
In this prior art, the identification of the preamble pattern, the calculation of all frequency components by the FFT necessary for the detection of the frequency deviation and the detection of the symbol timing are executed at regular time intervals (for example, every 4 symbols). Yes.
[0019]
[Patent Document 1]
Japanese Patent Laid-Open No. 2002-232503
[Problems to be solved by the invention]
The prior art described above has a problem in the feasibility of the system, taking into account that a huge amount of calculation is required for preamble identification, frequency deviation, and symbol timing detection.
[0021]
As described above, in the prior art, the preamble pattern is identified, the frequency deviation is detected, and the symbol timing is detected while the receiver is in an asynchronous state. Therefore, the calculation of all frequency components by FFT is performed for a certain period of time. Must be done every (eg every 4 symbols).
[0022]
At this time, in order to increase the accuracy of frequency deviation detection, it is necessary to increase the number of FFT points (window length) or add interpolation processing. Therefore, as described above, the processing amount is enormous. As a result, there are some problems in the feasibility of the system.
[0023]
The present invention has been made in view of the circumstances described above, and its object is to provide a preamble pattern identification method requires less calculation amount.
[0024]
[Means for Solving the Problems]
Based on Symbol purpose the preamble is added to the information to be transmitted, receives a digital modulated signal, the method of identifying a preamble pattern by detecting the preamble from the received baseband signals, said received a band signal one symbol per N ov times (N ov is a positive integer of 2 or more) extracts a baseband signal of an arbitrary continuous N win samples oversampling (N win is a positive integer of 2 or more), the extracts The baseband signal multiplied by an arbitrary window function, and the total power sum of the baseband signal x (n) ( n = 0, 1,..., N win-1) multiplied by the window function is multiplied by N win . p = N win {| x (0) | ² + | x (1) | ² + …… + | x ( N win-1) | ² } is calculated, and two of the two frequency components at the time of preamble transmission are calculated . , lower the frequency f1, the frequency of the higher and f2, corresponding to the frequency f 1 week The wavenumber number k 1, a frequency number corresponding to the frequency f 2 as k 2, the baseband signal x (n) (n = 0,1 , ......, N win -1) discrete Fourier transform value X (k ) (k = 0,1, ......, among N win-1), calculates the two components of X (k1) and X (k 2), the power p 1 = the X (k 1) | X ( k 1) | ² and the X (k 2) of the power p 2 = | X (k 2 ) | 2 by calculating a power of each of the power p 1 and the sum p 1 + p 2 of p 2 and the p calculates the ratio (p 1 + p 2) / p , when said power ratio (p 1 + p 2) / p exceeds a predetermined threshold value, achieved as determined to be a preamble portion of the received baseband signal Is done.
[0025]
Here, the present invention can be implemented in the following forms.
[0026]
<Embodiment 1>
Of the two frequency components of the preamble transmission, the frequency of the lower and f _1, the frequency of the higher and f _2, of the frequency spectrum of the received baseband signal, two components of the frequencies f ₁ and f ₂ And the power of the component of frequency f ₁ is p ₁ , the power of the component of frequency f ₂ is p _2, and the average power of the received baseband signal is multiplied by a predetermined constant to obtain the power p of all frequency components of the spectrum. the calculated power ratio of each of the power p ₁ and the power p ₂ of the sum p ₁ + p ₂ and the power p to (p ₁ + p ₂₎ / p is calculated, said power ratio (p ₁ + p ₂₎ / p is given A preamble pattern identification method, wherein when the threshold value is exceeded, it is determined that the received baseband signal is a preamble signal.
[0027]
<Embodiment 2>
_{Oversample the} received baseband signal N _ov times per symbol (N _ov is a positive integer greater than or equal to 2), and extract any continuous N _win samples (N _win is a positive integer greater than or equal to 2) baseband signal, Multiply the extracted baseband signal by an arbitrary window function, and add N _win to the total power sum of the baseband signal x (n) (n = 0, 1,..., N _win −1) multiplied by the window function. Multiplying p = N _win {| x (0) | ² + | x (1) | ² + …… + | x (N _win −1) | ² } and calculating two frequency components at the time of preamble transmission Of these, the lower frequency is f ₁ , the higher frequency is f ₂ , the frequency number corresponding to the frequency f ₁ is k ₁ , the frequency number corresponding to the frequency f ₂ is k _2, and x (n) (n = 0, 1,..., N _win −1) DFT (Discrete Fourier Transform) X (k) (k = 0, 1,..., N _win −1), X (k ₁ ) and Starring two components of X (k ₂₎ And, X (k ₁₎ of the power _{p 1 = | X (k 1} ) | 2 and X (k ₂₎ of the power _{p 2 = | X (k 2} ) | 2 calculates the respective power p ₁ and p power ratio between the sum p ₁ + p ₂ and the p of _{2 (p 1 + p 2)} / p is calculated, said power ratio _{(p 1 + p 2) /} p is received if beyond a predetermined threshold baseband A method for identifying a preamble pattern, wherein the signal is determined to be a preamble signal.
[0028]
<Embodiment 3>
In the preamble pattern identification method according to the second embodiment, the modulation method of the received baseband signal is π / 4 shift QPSK (Quaternary Phase Shift Keying), and the calculation of the DFT is k ₁ = N _win -3N _win / 8N _ov And a preamble signal that is a repetitive pattern of 1001 (binary number), and is performed on frequency components of k ₂ = N _win / 8N _ov .
[0029]
<Embodiment 4>
In the preamble pattern identification method according to the third embodiment, the modulation method of the received baseband signal is π / 4 shift QPSK, and the calculation of the DFT is k ₁ = N _win −N _win / 8N _ov and k ₂ = 3N _win / 8N performed on the frequency components of the _ov, 0110 preamble pattern identification method characterized by identifying the preamble signal is a repeating pattern of (binary).
[0030]
<Embodiment 5>
In the preamble pattern identification method according to the third embodiment, the modulation method of the received baseband signal is QPSK or QAM (Quadrature Amplitude Modulation), and the calculation of the DFT is k ₁ = N _win −N _win / 2N _ov , k ₂ A preamble pattern identification method, which is performed on frequency components of = N _win / 2N _ov , and identifies a preamble signal that is a pattern in which two outermost corner points that are opposite are repeated every other symbol.
[0039]
DETAILED DESCRIPTION OF THE INVENTION
Here, before describing the implementation embodiments of the preamble pattern identification how according to the present invention, the preamble pattern will be described.
[0040]
In the following description, unless otherwise specified, regarding the preamble pattern, it is assumed that the modulation method is a π / 4 shift QPSK method and 4 bits of 1001 (binary number) are repeated.
[0041]
Then, as shown in the above-mentioned Patent Document 1 (FIG. 8 of Patent Document 1), when the roll-off rate α is 0.5 or less, the frequency component of this preamble pattern is X ₁ = − ( The two components 3/8) ω _b and X ₂ = (1/8) ω _b occupy most of the total power. Here, ω _b = 2πf _b , and f _b is a symbol rate (symbol / second).
[0042]
Therefore, in the above prior art, focusing on this point, the X ₁ component power p ₁ = | X ₁ | ² and X ₂ component power p ^₂ = | X ₂ | ₂ total frequency for the sum p ₁ + p ₂ of The preamble pattern is identified by comparing the ratio (p ₁ + p ₂ ) / p of the power sum p of the components with a predetermined threshold value (threshold value).
[0043]
However, if the viewpoint is changed here, the calculation of the power sum p of all frequency components described above can be replaced with the power sum of the time domain signal, as will be described later.
[0044]
Therefore, in the present invention, the FFT operation is not performed, and the power sum p of all frequency components is calculated by the power sum of the time domain signal, and only DFT (Discrete Fourier Transform: Discrete Fourier Transform) is performed for only _two components of the X ₁ component and the X ₂ component. Transform) is performed, electric powers p ₁ and p ₂ are obtained from these, and the ratio (p ₁ + p ₂ ) / p is calculated, thereby reducing the amount of calculation.
[0045]
However, in order to perform the DFT calculation of only the X ₁ component and the X ₂ component in this way, the frequency deviation of the input signal needs to be compensated. For this reason, in the present invention, the frequency deviation is separately detected by autocorrelation, and the DFT calculation is performed after compensating for the frequency deviation.
[0046]
Therefore, first, the point that the power sum in the frequency domain can be calculated from the power sum of the time domain signal described above will be described, and then the point that the frequency deviation can be detected by the autocorrelation of the preamble signal will be described. Embodiments will be described.
[0047]
Now, assuming that the number of points for DFT is N, the DFT of N signals x (0), x (1),... X (N−1) is expressed by equation (1).
[0048]
[Expression 1]

The power sum p of the above X (k) (k = 0, 1,..., N−1) can be expressed by Expression (2).
[0049]
[Expression 2]

Since | X (k) | ^{2 in} this equation (2) is the product of X (k) and its complex conjugate X ^* (k), the following equation (3) is established. Here, the symbol * attached to X ^* (k) represents a complex conjugate. Here, the complex conjugate means that the sign of the imaginary part is inverted.
[0050]
[Equation 3]

Therefore, when the expression (3) is substituted into | X (k) | ² of the expression ( ² ), the power sum p becomes the following expression (4).
[0051]
[Expression 4]

Here, Sn _{, m} in this equation (4) is the geometric series from the first term to the Nth term of the first term a ₁ = W _N ^{(nm) 0} = 1 and the common ratio r = W _N ^(nm). This is calculated by the following formula (5).
[0052]
[Equation 5]

Here, the common ratio r is 1 when n−m is an integer multiple of N (n−m = 0, ± N, ± 2N,...), And n = 0 in equation (4). ,..., N−1, m = 0, 1,..., N−1, so | n−m | ≦ N−1 and r = 1 This is only when 0, that is, n = m, and when n = m, r = W _N ⁰ = 1, so the following equation (6) is established from equation (5).
[0053]
[Formula 6]

On the other hand, in the case of n ≠ m, r = W _N ⁰ ≠ 1, so from the same equation (5), the following equation (7) is established.
[0054]
[Expression 7]

In this formula (7), since n and m are integers (n ≠ m), 2π (n−m) is an integer multiple of 2π, and therefore the following formula (8) is established.
[0055]
[Equation 8]

At this time, since nm = ± 1, ± 2,..., ± (N−1), −2π <2π (n−m) / N <2π and 2π (n−m ) / N ≠ 0, the following equation (9) is established.
[0056]
[Equation 9]

Therefore, in this case, since the numerator of the equation (7) = 0 and the denominator ≠ 0, the equation (10) is established when n ≠ m.
[0057]
[Expression 10]

Therefore, when Expression (6) and Expression (10) are collectively expressed, Expression (11) is obtained.
[0058]
## EQU11 ##

Expression (4) becomes Expression (12).
[0059]
[Expression 12]

As a result, the power sum of the signal x (n) (n = 0, 1, 2,..., N−1) in the time domain | x (0) | ² + | x (1) | ² + …… + Multiply | x (N-1) | ² by the number N of DFT points, and the power sum p = | of all frequency components X (k) (k = 0, 1, 2,..., N-1) X (0) | ² + | X (1) | ² +... + | X (N-1) | ² can be obtained. Therefore, the power sum in the frequency domain is obtained by the power sum of the time domain signals. Can be calculated.
[0060]
Next, the point that the frequency deviation can be detected by the autocorrelation of the preamble signal will be described.
[0061]
Now, if arbitrary N samples of the input baseband signal are taken out continuously and are taken as s (0), s (1),..., S (N-1), the autocorrelation is expressed by the following equation (13). ).
[0062]
[Formula 13]

Here, m is a non-negative integer, ^* in s ^* (n) represents a complex conjugate, and complex conjugate means that the sign of the imaginary part is inverted.
[0063]
At this time, the preamble baseband signal is an 8-symbol cycle signal as shown in the above-described prior art (FIG. 5 of Patent Document 1) and is point-symmetric every two symbols, so r (m) The value of the squared amplitude value | r (m) | ² becomes a peak every two symbols such as m = 0, 2N _ov , 4N _ov ,.
[0064]
Here, N _ov is an oversample ratio, which is the number of oversamples per symbol. When there is no frequency deviation, this signal always has a phase difference of π / 2 with respect to the instantaneous value two symbols before and s (n) = s (n−2N _ov ) e ^{j π / 2} , and therefore Equation (14) is established.
[0065]
[Expression 14]

Then, the peak value of every two symbols of the autocorrelation in the equation (13) can be expressed by the equation (15), the phase of r (0) is 0, the phase of r (2N _ov ) is π / 2, r ( The phase of 4N _ov ) is π, and the peak value rotates by π / 2 every two symbols.
[0066]
[Expression 15]

Here, when the received signal has a frequency deviation of Δf (Hz), a phase rotation of 2πΔf / f _b (radian) is added to the received signal in one symbol, so a phase rotation of 4 kπ Δf / f _b is added to the 2k symbol. As shown in equation (16).
[0067]
[Expression 16]

Therefore, the peak value for every two symbols of autocorrelation is as shown in Equation (17).
[0068]
[Expression 17]

Further, the phase of r (2kN _ov ) is as shown in Expression (18).
[0069]
[Formula 18]
θ = arg {r (2kNov)} = kπ / 2 + 4kπΔf / f _b …… Equation (18)
Here, arg {r (2kN _ov )} represents the phase of r (2kN _ov ).
[0070]
Therefore, the frequency deviation Δf can be calculated by the equation (19) using the phase θ of the autocorrelation peak value r (2 kN _ov ). Therefore, the frequency deviation can be detected by the autocorrelation of the preamble signal.
[0071]
[Equation 19]

Next, a receiver main body to which an embodiment of a preamble pattern identification method and a frequency deviation detection method according to the present invention is applied will be described with reference to FIG.
[0072]
The receiver shown in FIG. 5 is also shown in the above-described prior art, and a signal received by an antenna (not shown) is supplied to an input terminal 1101 and input to a high-frequency circuit 1102.
[0073]
The frequency is converted by the high-frequency circuit 1102, converted from a radio frequency to an intermediate frequency, supplied to the A / D converter 1103, sampled, quantized and converted into a digital signal, and then a multiplier 1104. -1, 1104-2 are input in parallel.
[0074]
At this time, the sine wave generation circuit 1106 generates a sine wave signal cos ωt having an angular frequency ω, and supplies the sine wave signal cos ωt to the multiplier 1104-1 as it is. The phase of the multiplier 1105 is changed via the phase shifter 1105. π / 2 (radian) is advanced and supplied as a sine wave signal −cosωt.
[0075]
Thus, the multiplier 1104-1 calculates the product of the output of the A / D converter 1103 and the output of the sine wave generation circuit 1106, and the multiplier 1104-2 provides the output of the A / D converter 1103 and the phase shifter. The product of the outputs of 1105 is calculated.
[0076]
At this time, the outputs of the multipliers 1104-1 and 1104-2 include unnecessary high frequency components. Therefore, the respective outputs are input to the low-pass filters 1107-1 and 1107-2, and unnecessary components are removed, and then supplied to the roll-off filters 1108-1 and 1108-2, respectively.
[0077]
These roll-off filters 1108-1 and 1108-2 limit the bandwidth of the input signal to obtain baseband signals, and supply the baseband signals to the baseband signal output terminals 1109-1 and 1109-2, respectively.
[0078]
At this time, the signal output from one baseband signal output terminal 1109-1 represents the in-phase component of the baseband signal, and the signal output from the other baseband signal output terminal 1109-1 is the orthogonal component of the baseband signal. Therefore, the baseband signal is a complex signal having an in-phase component as a real part and an orthogonal component as an imaginary part.
[0079]
Therefore, baseband signals appearing at these baseband signal output terminals 1109-1 and 1109-2 are supplied to a signal processing system including a demodulation unit of a receiver (not shown), and transmitted information is reproduced. At this time, the identification of the preamble pattern and the frequency deviation are detected from the baseband signals appearing at these baseband signal output terminals 1109-1 and 1109-2.
[0080]
Here, FIG. 1 shows an embodiment of the present invention, and the baseband signals appearing at the baseband signal output terminals 1109-1 and 1109-2 are input to the input terminal 101 in this figure.
[0081]
At this time, the baseband signal is a complex signal composed of an in-phase component and a quadrature component as described above, but only one input terminal 101 is shown here for the sake of simplicity.
[0082]
Here, as described above, this signal is a baseband signal sampled N _ov times per symbol, and at this time, N _ov is a positive integer of 2 or more.
[0083]
In FIG. 1, the input terminal 101, the shift register 114, the switch 113, the windowing circuit 102, the preamble pattern identification circuit 107, and the symbol timing detection circuit 109 are the same as those in the prior art (FIG. 1 of Patent Document 1). Since it is the same, detailed explanation is omitted.
[0084]
In FIG. 1, first, the switch 113 is closed once in a sample period of a constant time interval N _step , whereby N _win sample data s (0), s (1),. (N _win −1) is extracted and input to the autocorrelation calculation circuit 121 and the frequency deviation correction circuit 128.
[0085]
At this time, for example, the oversample ratio N _ov = 2, N _step = 8, and N _win = 64. Here, N _win is the number of stages of the shift register 114, and is therefore a positive integer.
[0086]
First, the autocorrelation operation circuit 121 calculates autocorrelation r (n) for N _win samples of s (0), s (1),..., S (N _win −1) extracted by the switch 113. r (2N _ov ) in m) is calculated by the above-described equation (13), and the calculated value (complex number) is input to the phase detection circuit 122.
[0087]
Therefore, the phase detection circuit 122 detects the phase θ of the input autocorrelation r (2N _ov ) in the range of −π ≦ θ <π by the above equation (18), and the (+) of the subtractor 129. Input to the side input terminal, where the constant value π / 2 is subtracted from the input phase θ, and the phase θ−π / 2 is calculated in the range of −π ≦ θ−π / 2 <π, and the multiplier 123 to one input terminal.
[0088]
That is, in the subtractor 129, when the result of subtracting π / 2 from the phase θ is θ−π / 2 <−π, 2π is added to the subtraction result, and one of the input terminals of the multiplier 123 is added. Will be input.
[0089]
The multiplier 123 multiplies the input phase θ−π / 2 by a constant value f _b / 4π to calculate a frequency deviation Δf, and inputs it to the frequency deviation correction circuit 128. The processing at this time is as described in Expression (19), and the calculated frequency deviation Δf is also supplied to the frequency deviation output terminal 111 and is output to the outside as necessary.
[0090]
Therefore, the frequency deviation correction circuit 128 uses the frequency deviation Δf input from the multiplier 123 and uses the N _win sample baseband signals s (0), s (1),. For N _win −1), as shown in equation (20), the phase is rotated by −2πΔf / N _ov f _b radians per sample to correct the frequency deviation.
[0091]
[Expression 20]

Thus, the N _win sample correction signal s ′ (n) (n = 0, 1, 2,..., N _win −1) input from the frequency deviation correction circuit 128 is input to the windowing circuit 102, where A window (for example, Hamming Window) is multiplied, and the baseband signal x (n) (n = 0, 1, 2,..., N _win −1) that has been windowed is the total power calculation circuit 124 and the DFT operation circuit. 126 and 127.
[0092]
First, in the total power calculation circuit 124, as described in the equation (12), the input N _win sample baseband signal x (n) (n = 0, 1, 2,..., N _win − 1) the sum of the amplitudes squared value over all samples ^{| x (0) | 2 +} | x (1) | 2 + ...... + | x (N win -1) | 2 is calculated, and one of the multiplier 125 _Is multiplied by a constant value N _win and the multiplication result is p = N _win {| x (0) | ² + | x (1) | ² + …… + | x (N _win −1) ² } is input to the preamble pattern identification circuit 107.
[0093]
On the other hand, the DFT arithmetic circuit 126 performs DFT X (0) on the N _win sample baseband signal x (n) (n = 0, 1, 2,..., N _win −1) input from the windowing circuit 102. ), X (1),..., X (N _win -1), only for X ₁ = X (N _win -3N _win / 8N _ov ) which is the frequency component of-(3/8) f _b The power p ₁ = | X ₁ | ² is input to the preamble pattern identification circuit 107, and the phase φ ₁ of the frequency component X ₁ at this time is input to the symbol timing detection circuit 109.
[0094]
Further, the DFT arithmetic circuit 127 performs DFT X (0) for the N _win sample baseband signal x (n) (n = 0, 1, 2,..., N _win −1) input from the windowing path 102. ), X (1),..., X (N _win -1), only (1/8) f _b frequency component X ₂ = X (N _win / 8N _ov ) The power p ₂ = | X ₂ | ² is input to the preamble pattern identification circuit 107, and the phase φ ₂ of the frequency component X ₂ at this time is input to the symbol timing detection circuit 109.
[0095]
Therefore, first, the preamble pattern identification circuit 107 uses the multiplication result p input from the multiplier 125 and the powers p ₁ and p ₂ input from the DFT operation circuits 126 and 127, respectively. The preamble pattern is identified as in 1), and the result is supplied to the preamble pattern identification output terminal 110.
[0096]
Further, the symbol timing detection circuit 109 detects the symbol timing from the phases φ ₁ and φ ₂ input from the DFT operation circuits 126 and 127 in the same manner as in the prior art (FIG. 1 of Patent Document 1). The result is supplied to the symbol timing output terminal 112.
[0097]
Therefore, according to this embodiment, it can be confirmed that the preamble pattern has been identified by the output of the preamble pattern identification output terminal 110, and the symbol timing can be confirmed by the output of the symbol timing output terminal 112. High-speed synchronization can be easily taken.
[0098]
In the above description, the autocorrelation calculation by the autocorrelation calculation circuit 121 is based on the calculation of r (2N _ov ), but r (4N _ov ), r (6N) instead of r (2N _ov ). _ov), may be using the calculation of r (8N _ov).
[0099]
At this time, the value to be subtracted by the subtracting circuit 129 is π in the case of r (4N _ov ), (3/2) π, r (8N _ov ) in the case of r (6N _ov ), from the equation (19). for), 0 (2 [pi since 0 any good), and also, the value be multiplied in a multiplier 123, in the case of _{r (4N ov), f b} / 8π, f in the case of r (6N _{ov) b} / In the case of 12π and r (8N _ov ), it becomes f _b / 16π.
[0100]
The above description is for the embodiment when the preamble is a repetitive pattern of 1001 (binary number), but is as follows when the preamble is a repetitive pattern of 0110 (binary number).
[0101]
That is, in this case, the autocorrelation r (m) does not change, and the amplitude square value peaks every two symbols of r (0), r (2N _ov ), r (4N _ov ),. In this case, since the phase is reversed, if there is no frequency deviation, the phase difference with respect to the instantaneous value two symbols before is always −π / 2.
[0102]
Therefore, the phase of r (0) is 0, the phase of r (2N _ov ) is −π / 2, the phase of r (4N _ov ) is −π, and the phase of the peak value is −π / every two symbols. Since the rotation is performed by two, the value to be subtracted by the subtraction circuit 129 is −π / 2 for r (2N _ov ), −π for r (4N _ov ), and − (3 for r (6N _ov ). / 2) In the case of π, r (8N _ov ), it is 0 (again, it is 0 because it is −2π). Then, the value be multiplied in the multiplier 123, r (2N _ov) of f _b / 4 [pi] _{If, r (4N ov) f b} / 8π For, r (6N _ov) For f _b / 12π, In the case of r (8N _ov ), it becomes f _b / 16π.
[0103]
In this case, the frequency components of the preamble pattern are − (1/8) f _b and (3/8) f _b . Therefore, the DFT operation circuit 126 calculates X (N _win −N _win / 8N _ov ), which is the frequency component of − (1/8) f _b , and the DFT operation circuit 127 calculates the frequency of (3/8) f _b . The component X (3N _win / 8N _ov ) is calculated.
[0104]
Further, the above description is an embodiment when the modulation method is a π / 4 shift QPSK method and the preamble is a repetitive pattern of 1001 (binary number). However, as an embodiment of the present invention, the modulation method is a QPSK method. Alternatively, the QAM method may correspond to a pattern in which two opposite outermost corner points are repeated every other symbol, as shown in the above-described prior art (FIG. 4 of Patent Document 1).
[0105]
In this case, the preamble signal is a signal of two symbol periods, and the autocorrelation r (m) has an amplitude square value for every two symbols of r (0), r (2N _ov ), r (4N _ov ),. If there is no frequency deviation, the phase is always the same as the instantaneous value two symbols before, so the phases of r (2N _ov ), r (4N _ov ) _,.
[0106]
Therefore, in this case, the subtraction circuit 129 is unnecessary and can be omitted. In this case, the frequency component of the preamble is ± (1/2) f _b . Therefore, at this time, the DFT operation circuit 126 calculates X ₁ = X (N _win −N _win / 2N _ov ), which is a frequency component of − (1/2) f _b , and the DFT operation circuit 127 (1/2) ) X2 = X (N _win / 2N _ov ) which is a frequency component of f _b is calculated.
[0107]
Incidentally, in the above-described embodiment, frequency deviation detection by the autocorrelation calculation circuit 121, phase detection circuit 122, subtractor 129, and multiplier 123 and frequency deviation correction by the frequency deviation correction circuit 128 are input to the input terminal 101. It is also performed when the signal is a signal other than a preamble, and preamble identification is performed using this correction result.
[0108]
Then, the frequency deviation correction in this case does not give a correct result, and it may be considered that the preamble pattern identification circuit 107 influences the preamble pattern identification.
[0109]
However, with respect to the calculation result of the total power calculation circuit 124, the frequency deviation correction circuit 128 only rotates the phase of the input signal, and therefore there is no possibility of influence (does not affect the value of p).
[0110]
In addition, regarding the calculation of the DFT calculation circuits 126 and 127, the frequency component is shifted by the frequency deviation correction circuit 128. As a result, the power of another frequency component is calculated.
[0111]
However, in the case of a signal other than the preamble, in the band, all frequency components are almost equal in the period excluding the guard time, and become one kind of frequency component at the guard time (because all 0s are mapped). It is either.
[0112]
Therefore, even if these two kinds of frequency components are calculated, the sum of their power does not occupy most of the total power, and therefore there is no possibility of causing an erroneous detection.
[0113]
By the way, in the above-described embodiment, instead of calculating the total power of the FFT output, it is replaced by calculating the total power of the windowed output, and instead of the FFT calculation, it is replaced by the DFT calculation of only two types of frequency components. Reductions are being made.
[0114]
Therefore, this point will be described below. First, in the case of FFT calculation, if the number of points is N _win , it is necessary to perform 2N _win log ₂ N _win times of product-sum calculation. When N _win = 64, 2 × 64 log ₂ 64 = 768 times, and when the window length N _win = 128, 2 × 128 log ₂ 128 = 1792 times of product-sum operation is required.
[0115]
On the other hand, in the above embodiment, auto-correlation calculation and phase detection processing are added as a result of replacing FFT calculation with DFT calculation using two types of frequency components. However, since the calculation of one component by DFT is N _win times of product-sum operation, the autocorrelation operation is about N _win times of product-sum operation (more precisely, N _win -2N _ov times), and phase detection processing Is an amount of computation corresponding to about 300 product-sum operations (in the case of software processing by a DSP).
[0116]
Then, in the case of the above embodiment, when N _win = 64, the DFT operation is 2 × 64 = 128 times, the autocorrelation operation is N _win = 64 times, the phase detection processing is 300 times, and a total of 64 + 128 + 300 = 492 times. The amount of calculation corresponds to the calculation, and the amount of calculation is significantly reduced as compared with the amount of calculation of 768 times of the prior art.
[0117]
Similarly, when N _win = 128, the DFT operation is 2 × 128 = 256 times, the autocorrelation operation is N _win = 128 times, the phase detection process is 300 times, and the sum is 256 + 128 + 300 = 684 times. Therefore, the amount of calculation is further reduced as compared with the amount of calculation of 1792 times of the prior art. From this, in the case of the above embodiment, the number of points N _win of DFT (FFT) It can be seen that the effect of reducing the amount of computation increases as the number of increases.
[0118]
Incidentally, the autocorrelation of the input baseband signal in the above-described embodiment has a peak value indicating the periodicity of the input baseband signal regardless of the synchronization state of the symbol timing.
[0119]
Therefore, in the above embodiment, the detection of the frequency deviation by the autocorrelation calculation circuit 121, the phase detection circuit 122, the subtractor 129, and the multiplier 123 can be performed regardless of the symbol timing synchronization state.
[0120]
Furthermore, in the case of the above-described embodiment, the frequency deviation detection accuracy depends on the processing method of the phase detection circuit 122 and does not depend on the number of DFT points (window length) N _win .
[0121]
For this reason, in the prior art, in order to increase the frequency deviation detection accuracy, the number of FFT (or DFT) points is increased, or FFT (or DFT) output interpolation processing is required. There is no need, and as a result, the amount of calculation can be further reduced.
[0122]
【The invention's effect】
According to the preamble pattern identification method of the present invention, the calculation of the power sum of all frequency components is replaced with the power sum of the time domain signal, and the calculation of the frequency spectrum is performed with only two types of frequency components. .
[Brief description of the drawings]
Is a block diagram showing an embodiment of a preamble pattern identification how according to the invention; FIG.
FIG. 2 is an explanatory diagram showing an example of a frame structure of a synchronous burst frame.
FIG. 3 is an explanatory diagram showing an example of a frame structure of a communication channel frame.
FIG. 4 is an explanatory diagram showing an example of a transmission pattern.
FIG. 5 is a block diagram showing an example of a receiver targeted by the present invention.
[Explanation of symbols]
DESCRIPTION OF SYMBOLS 101 Input terminal 102 Windowing circuit 107 Preamble pattern identification circuit 109 Symbol timing detection circuit 110 Preamble pattern identification output terminal 111 Frequency deviation output terminal 112 Symbol timing output terminal 113 Switch 114 Shift register 121 Autocorrelation operation circuit 122 Phase detection circuit 123 Multiplier 125 multiplier 124 total power calculation circuit 126 DFT operation circuit (for-(3/8) fb component)
127 DFT arithmetic circuit (for (1/8) fb component)
128 Frequency deviation correction circuit 129 Subtractor

Claims (1)

In a method in which a preamble is added to information to be transmitted, a digitally modulated signal is received, the preamble portion is detected from the received baseband signal, and a preamble pattern is identified.
Baseband signal one symbol per N ov times that the received (N ov is a positive integer of 2 or more) extracts a baseband signal of an arbitrary continuous N win samples oversampling (N win is a positive integer of 2 or more) And
Multiply the extracted baseband signal by an arbitrary window function, and add N win to the total power sum of the baseband signal x (n) ( n = 0, 1,..., N win-1) multiplied by the window function. obtained by multiplying p = N win {| x ( 0) | 2 + | x (1) | 2 + ...... + | x (N win-1) | 2} is calculated and,
Of the two frequency components of the preamble transmission, the lower the frequency f1, the frequency of the higher and f2, k 1 the frequency number corresponding to the frequency f 1, a frequency number corresponding to the frequency f 2 k as 2, the baseband signal x (n) (n = 0,1 , ......, n win -1) discrete Fourier transform value X (k) of the (k = 0,1, ......, n win-1) Of the two components X (k1) and X ( k2 )
Wherein X (k 1) of the power p 1 = | X (k 1 ) | power p 2 = ² and the X (k 2) | X ( k 2) | 2 by calculating a respective power p 1 power ratio between the sum p 1 + p 2 and the p of p 2 (p 1 + p 2 ) / p is calculated,
A preamble pattern identification method, wherein when the power ratio ( p 1 + p 2) / p exceeds a predetermined threshold, it is determined as a preamble portion of a received baseband signal.

Images